Role of quantitative and qualitative characteristics of free circulating DNA in the management of patients with non-small cell lung cancer

Abstract

Background

The release of DNA into peripheral blood is a common event in cancer patients, occurring as a consequence of necrotic and apoptotic processes typical of tumor cells. However, free circulating DNA (fcDNA) is also present in patients with benign diseases and in healthy individuals. Both quantitative and qualitative aspects of fcDNA have been studied as potential biomarkers in a number of tumor types. In particular, quantitative analysis of fcDNA has been shown to play an important role in the diagnosis of non-small cell lung cancer (NSCLC), because of its ability to discriminate between healthy subjects and individuals with NSCLC. Additionally, fcDNA in cancer patients derives predominantly from tumor tissue and, as such, it can be used for the molecular characterization of the primary tumor. Targeted therapies in NSCLC have, in recent years, produced promising results, highlighting the importance of molecular profiling of the primary cancer lesions. Considering that little or no tumor material is available for at least some of the patients, the possibility of using fcDNA for molecular analysis becomes increasingly important. In the present review we evaluated several quantitative and qualitative aspects of fcDNA that could be instrumental for the differential diagnosis of lung disease.

Conclusions

There is ample evidence in the literature to support the possible use of peripheral blood-derived fcDNA in the early diagnosis and molecular characterization of lung cancer. This non-invasive method may also turn out to be valuable in monitoring drug response and in identifying induced mechanisms of drug resistance. Before it can be implemented in routine clinical practice, however, additional efforts are needed to standardize the methodology.

Origin of fcDNA

Free circulating DNA (fcDNA) is extracellular DNA detectable in blood. It is present in the plasma and serum of healthy individuals and in patients with benign lesions, inflammatory diseases or tissue trauma, and can reach high levels in patients with different tumor types, including lung cancer [1]. These findings have led to the hypothesis that fcDNA may serve as a diagnostic biomarker [2]. fcDNA in healthy individuals is predominantly derived from hematopoietic cells [3] and originates from apoptotic rather than necrotic processes [4, 5]. Thus, DNA fragments in these individuals are usually larger than those derived from cancer cells. In cancer patients the release of fcDNA may be the result of both apoptotic and necrotic processes, which are characteristic of tumors with a high cellular turnover [6, 7]. Both low and high molecular weight DNA fragments, typical of necrotic and apoptotic processes, respectively, are found in serum and plasma of these patients. In the early stages of cancer, a close correlation has also been observed between tumor cell growth and the release of DNA from adjacent non-tumor cells, suggesting a strong interaction between tumor cells and adjacent non-tumor tissue. Moreover, studies on experimental cancer models have revealed a phase-dependent release of fcDNA from cancer cells during tumor progression [8].

fcDNA in cancer patients

Increased fcDNA levels in the plasma or serum of cancer patients compared to healthy individuals suggests that this variable may have clinical relevance for the diagnosis and prognosis of cancer, as well as for disease monitoring during treatment or follow-up [9]. The fcDNA concentration appears to be influenced by cancer-dependent variables such as tumor stage, grade, size, location, aggressiveness, and the presence of metastases. In prostate cancer, however, contradictory results have been reported on the correlation between fcDNA levels and tumor stage, Gleason grade, metastatic spread and prostate-specific antigen levels [10–13]. In breast cancer, conflicting results have also been reported on the association between fcDNA levels and lymph node status or distant metastases [12, 14–16]. The primary aim of studying fcDNA levels is to assess its relevance as a marker for the early detection of cancer, and several studies have been carried out on different cancer types [10, 17–20]. A complicating factor, however, is that fcDNA is also present in patients suffering from benign diseases, infections, diabetes and hepatic disorders [21–23], indicating that the marker is not cancer-specific. In addition to a role as a marker for early cancer detection, fcDNA may also represent a prognostic marker. In several cancer types, high fcDNA concentrations have proven to be an independent indicator of poor outcome in terms of disease-free interval and survival [13, 24–27]. Furthermore, surgically treated patients with cancer of the breast [28], colon [29], esophagus [30], kidney [31] or lung [32] have been found to exhibit a decrease in pretreatment fcDNA levels when the treatment was successful. Conversely, patients with persistently increasing fcDNA levels after surgery have shown an incomplete response to treatment or developed systemic disease [30].

Both the amount of fcDNA and its integrity have been extensively studied. Considering the fact that different sizes of DNA fragments derive from either necrotic or apoptotic processes, the assessment of fcDNA lengths could provide a clue about the source of the DNA. This approach has been used for both diagnostic and prognostic purposes in patients with several cancer types. In one of the earliest studies [33], in which the diagnostic relevance of fcDNA integrity was analyzed in patients with gynecological and breast cancers, a diagnostic accuracy of over 90 % was observed. Although additional promising results have been obtained in patients with colon [34], breast [35], prostate [36], esophagus [37] and head and neck [38] cancers, contradictory findings have also been reported in the literature [39, 40].

In addition to its immediate clinical application, fcDNA could potentially be used for other scientific purposes. As fcDNA in cancer patients appears to be derived predominantly from tumor tissue, it is likely to carry molecular tumor characteristics. Several studies investigating the presence of molecular alterations in fcDNA, such as p53 and KRAS mutations, revealed variable concordances with the alterations present in the corresponding tumor tissue [41–45]. Very similar results were obtained for gene methylation alterations [46]. Alterations in DNA methylation patterns are part of a variety of epigenetic changes that are frequently observed in cancer cells. In particular, in the early stages of carcinogenesis promoter regions of tumor suppressor genes are often hypermethylated, resulting in silencing of their expression. Numerous genes including APC, DAPK, GSTP1, MGMT, p16, RASSF1A, and RARβ have been analyzed in fcDNA for their diagnostic relevance [46]. To date, the most significant limit of these analyses has been low fcDNA sensitivity, due mainly to the limited amount of tumor fcDNA in blood and to the heterogeneity of tumor fcDNA in both blood and primary tumor tissue.

Role of fcDNA in lung diseases

Considering the fact that the lung parenchyma is highly vascularized, the interconnection between lung cells and peripheral blood is very close. Both the levels and the molecular characteristics of fcDNA may play a role in lung cancer diagnosis. Several studies have reported higher levels of fcDNA in the plasma or serum of lung cancer patients compared to those of healthy donors [17, 18, 32, 47–57], albeit with different levels obtained with different methodologies (Table 1), suggesting that fcDNA may be a potential diagnostic marker for lung cancer.

Table 1.

Principal studies on the role of fcDNA in discriminating between lung cancer patients and healthy donors

Studies	Material	Methods	No. of patients/controls	DNA (ng/ml) in patients/controls	Cut off (ng/ml)	Sensitivity (%)	Specificity (%)
Sozzi et al. 2001 [32]	Plasma	DNAdipstick	84/43	318/18	25	75	86
Sozzi et al. 2003 [17]	Plasma	qPCR	100/100	24/3.1	9	90	86
Xie et al. 2004 [47]	Plasma	Pico green	67/44	110.7/11.6	53.8	70	80
Herrera et al. 2005 [48]	Plasma	qPCR	25/11	14.6/10.6	14	48	100
Ludovini et al. 2008 [49]	Plasma	qPCR	76/66	60/5	3.25	80	61
Ulivi et al. 2008 [18]	Serum	qPCR	128/103	48/8.8	25	83	92
Paci et al. 2009 [50]	Plasma	qPCR	151/79	12.8/2.9	2	86	47
Yoon et al. 2009 [51]	Plasma	qPCR	102/105	22.6/10.4	11	91	57
Szpechcinski et al. 2009 [52]	Plasma	qPCR	30/16	12.0/2.65	2.8	86	75
Kumar et al. 2010 [53]	Plasma	qPCR	100/100	122.7/74.0	104.5	52	95
Szpechcinski et al. 2012 [54]	Plasma	qPCR	50/50	8.02/2.27	–	–	–
Catarino et al. 2012 [55]	Plasma	qPCR	104/205	270.0/122.7	15	88	75
Ulivi et al. 2013 [56]	Serum	qPCR	100/100	47.2/9.2	25	82	91

qPCR quantitative polymerase chain reaction

One of the first studies by Sozzi et al. [32], in which a DipStick method was used, reported significantly higher levels of fcDNA in cancer patients than in healthy donors (345 versus 34 ng/ml, P <0.0001). Subsequently, other authors confirmed these results in larger case series using different methodologies, mainly Quantitative Real Time PCR, with a diagnostic accuracy of over 80 % in the majority of studies and with a good sensitivity and specificity at different cut-off values (Table 1). In the attempt to further increase the diagnostic accuracy, fcDNA was analyzed in combination with other markers [18, 56]. A 91 % sensitivity and a 92 % specificity were obtained using a combination of fcDNA level and cycloxygenase-2 mRNA expression level, with a 96 % sensitivity in detecting stage I tumors [18]. Similarly, combination of the fcDNA level with other markers, such as pro-platelet basic protein (PPBP) and peptidylarginine deiminase type 4 (PAD4/PADI4) expression levels, increased the diagnostic accuracy of fcDNA, reachinig sensitivity and specificity values of over 90 %. These latter marker combinations also showed the highest diagnostic accuracy for early-stage tumors, reaching a 93 % and a 100 % sensitivity for stage I and II tumors, respectively [56].

The five-year survival rate is only ~16 % for patients diagnosed with advanced lung cancer, compared to 70–90 % when the disease is diagnosed and treated at earlier stages [58, 59]. Thus, early diagnosis could be a promising strategy to reduce lung cancer mortality. The availability of a non-invasive test performed on peripheral blood, and capable of discriminating between subjects with and without lung cancer, could have two potential implications: first, it could be used as a preliminary screening method to select individuals at high risk of NSCLC who require further investigation with spiral CT and, second, it could be used to discriminate between neoplastic and non-neoplastic diseases in subjects with suspicious nodules detected by CT scans, thereby eliminating the need for serial CTs or invasive biopsies. fcDNA seems to represent a potential biomarker for such purposes. In 2009, however, Sozzi et al. [60] showed that baseline assessment of plasma fcDNA levels did not improve the accuracy of lung cancer screening by spiral CT in heavy smokers. The relevance of the combination of fcDNA with other markers in this context remains to be demonstrated.

Other studies aimed at analyzing fcDNA levels in NSCLC patients and in individuals with non-neoplastic lung diseases have shown that the median fcDNA levels in individual benign diseases are lower than those in cancer patients, but higher than those of healthy donors [47, 61]. In a study by Xie et al. [47], different median values of fcDNA levels were observed in healthy donors, individuals with benign lung diseases and lung cancer patients (11.6, 45.5 and 110.7 ng/ml, respectively). Additionally, our previous results showed that fcDNA levels are very high in individuals with idiopathic pulmonary fibrosis (IPF) [61].

In addition to its diagnostic role, there is evidence to suggest that fcDNA may be used as a prognostic marker. Sozzi et al. [60] observed a correlation between baseline fcDNA and aggressive disease in patients with stage Ib—IV NSCLC, suggesting that pre-surgery fcDNA levels could be indicative of a patient’s prognosis. Other studies have subsequently confirmed this correlation. One of the earliest studies conducted by Gautschi et al. [26] showed a significant association between increased plasma DNA levels and a poor survival rate (P <0.001), and a marked correlation between tumor progression after chemotherapy and increasing plasma DNA concentrations (P = 0.006). It was also found that high pre-treatment fcDNA levels acted as independent prognostic markers for short survival rates [62–64]. In a large series of patients with stage IIIB—IV NSCLC, Sirera et al. [62] showed that pre-treatment high fcDNA levels were associated with a shorter time to progression (P = 0.001) and overall survival rate (P = 0.012), whereas no correlation was found with stage, histology or number of metastatic lesions. Another study with a median follow-up of 6.5 years showed that median fcDNA levels were higher in patients who died than in those who survived (P = 0.02) [63]. In contrast, others did not find any correlation between fcDNA levels and survival [32, 65]. A direct relation was also observed between fcDNA levels and aggressive clinical behavior in the phase III First-SIGNAL trial, [66] the risk of death increasing as baseline fcDNA levels increased, independently of the type of therapy administered.

Molecular alterations in fcDNA

Several studies have shown good concordances between alterations in oncogenes and tumor suppressor genes in lung tumor tissue DNA and its corresponding fcDNA. This could have important implications, as specific molecular alterations in cancers detected in fcDNA could be used for diagnostic purposes and for characterizing the primary tumor using non-invasive methods. The most widely investigated molecular alterations are changes in gene methylation and mutations in oncogenes.

Changes in gene methylation

p16/INK4A is one of the most frequently methylated genes in lung cancer, and several studies have been carried out to investigate whether alterations in the methylation of this gene in fcDNA can facilitate lung cancer diagnosis [67–70]. Different studies have reported sensitivity values ranging from 14 % to 72 % [67–71]. Other studies have aimed at analyzing whether the use of a panel of potentially methylated genes may be of help to increase diagnostic accuracy. Hsu et al. [72] observed similar methylation patterns of the BLU, CDH13, FHIT, p16, RARβ, and RASSF1A genes in tumor tissue and corresponding plasma, with a 73 % sensitivity and a 82 % specificity and, consequently, a similar diagnostic accuracy. In another study, a 73 % sensitivity and a 71 % specificity were obtained by analyzing the methylation status of 4 genes, DCC, Kif1a, NISCH and RARβ [73]. Zhang et al. [74] showed that, of the 9 genes found to be significantly more methylated in the fcDNA of cancer patients than in that of healthy donors, 5 genes (APC, RASSF1A, CDH13, KLK10 and DLEC1) were capable of distinguishing cancer patients from healthy donors, with a 84 % sensitivity and a 74 % specificity. In another study, serum methylation levels of a panel of 6 genes (APC, CDH1, MGMT, DCC, RASSF1A and AIM1) also showed good results in discriminating between cancer patients and healthy individuals [75]. Conversely, hypomethylation of the p53 gene, evaluated in peripheral blood, was associated with an increased risk of lung cancer in male smokers [76].

Mutations in oncogenes

In recent years, targeted therapy of lung cancer has become the most effective therapeutic strategy for subsets of patients carrying specific DNA alterations. Gefitinib and erlotinib were the first drugs to show good responses in patients with specific alterations of the EGFR gene [77, 78], making molecular characterization of the tumor essential before any therapeutic decision can be made. Considering that lung cancer is mainly diagnosed at advanced stages, the only biological material available for molecular analyses often derives from biopsy or needle aspirates and is, consequently, scant. Although several studies have reported the possibility of using very little material (as few as 20–30 cells) derived from fine needle aspirates [79–81], even this amount of material may not be available for some patients. The potential usefulness of fcDNA for the molecular characterization of the primary tumors in this subset of patients has thus been taken into consideration.

One of the first groups to analyze the EGFR gene mutation status in fcDNA was that of Maheswaran et al. [82], who observed a 39 % correlation between EGFR mutations found in tumor tissue and plasma fcDNA, whereas a correlation of 94 % was found when DNA was extracted from circulating tumor cells (CTCs). Numerous other studies have evaluated various possibilities of using plasma or serum DNA for the molecular characterization of primary tumors [82–96], and the best sensitivities were obtained using mutant enriched PCR [86, 89], droplet-based digital PCR [84] or PNA-LNA PCR [92] (Table 2).

Table 2.

Correlations between EGFR mutation status in paired plasma/serum and tumor samples

Studies	No. of EGFR-mutated tumors	Biological material	Methods	Concordance between paired samples
Kimura et al. 2007 [83]	8	Serum	SARMS	75 %
Maheswaran et al. 2008 [82]	18	Plasma	SARMS	39 %
		CTCs	SARMS	94 %
Yung et al. 2009 [84]	12	Plasma	Digital PCR	92 %
Kuang et al. 2009 [85]	30	Plasma	SARMS and WAVE/Surveyor	70 %
He et al. 2009 [86]	18	Plasma	Mutant-enriched PCR	94.4 %
Bai et al. 2009 [87]	77	Plasma	DHPLC	82 %
Mack et al. 2009 [88]	7	Plasma	SARMS	71 %
Jiang et al. 2011 [89]	18	Serum	Mutant enriched PCR	78 %
Brevet et al. 2011 [90]	31	Plasma	Mass spectrometry	61 %
Taniguchi et al. 2011 [91]	44	Plasma	BEAMing	73 %
Chen et al. 2012 [92]	30	Plasma	PNA-LNA PCR	83 %
Nakamura et al. 2012 [93]	39	Plasma	WIP-QP	39 %
Goto et al. 2012 [94]	51	Serum	SARMS	43 %
Punnoose et al. 2012 [95]	8	CTCs	SARMS	13 %
	4	Plasma		100 %
Zhao et al. 2012 [96]	45	Plasma	Mutant enriched PCR	35.6 %

CTC circulating tumor cells, SARMS scorpion amplification refractory mutation system, DHPLC denaturing high-performance liquid chromatography, BEAMing beads, emulsion, amplification and magnetics, WIP-QP wild inhibiting polymerase chain reaction and quenched probe system, PNA-LNA-PCR peptide nucleic acid–locked nucleic acid polymerase chain reaction

Two studies in which EGFR mutations determined in DNA extracted from CTCs and in fcDNA from plasma were compared did not show concordant results. In the study by Mahasweran et al. [82], EGFR mutations found in tumor tissue were more concordant with those observed in CTCs than in those in fcDNA. Punnoose et al. [95] reported a higher sensitivity for mutations detected in fcDNA than in CTCs. These discordant results could be ascribed to the different efficiency of the CTC isolation methods used, i.e., the chip-based method used by Mahasweran et al. and the CellSearch platform, characterized by a low sensitivity and specificity in detecting CTCs, used by Punnoose et al. It must be noted that fcDNA isolation and characterization are usually inexpensive and simple to perform, whereas CTC analysis requires expensive equipment and kits, making this approach less feasible and more problematic in clinical practice.

Other molecular alterations have also been compared in fcDNA and tumor tissue. In a study by Wang et al. [97], a 76.7 % correlation was found between KRAS mutations detected in plasma and corresponding tumor tissue samples, while Punnoose et al. [95] observed an absolute concordance in the same type of analysis performed on a small number of cases. Moreover, a concordance of 73 % was observed between p53 mutations detected in plasma and those found in corresponding tumor tissue [98]. In other studies, microsatellite instability in fcDNA and the corresponding tumor tissue was analyzed, revealing a wide-ranging concordance (40–85 %) between the two biological samples [65, 98–100].

Methodological aspects for fcDNA analysis

Several factors at both the pre-analytical and the analytical level can influence the analysis of fcDNA. First, the starting material used for fcDNA extraction, either serum or plasma, can influence DNA recovery. Serum DNA concentrations are 3- to 24-fold higher compared to those in plasma [13, 101–103] due to leukocyte lysis that occurs during coagulation [103]. Moreover, several aspects of serum and plasma manipulation may interfere with DNA recovery, including anticoagulants used for plasma collection, time interval between collection and centrifugation, centrifugation forces and modalities of sample conservation [48, 101–108]. With regard to DNA extraction, various protocols are available that can result in different DNA extraction efficiencies and, consequently, different final DNA quantities [107, 109–112]. As fcDNA is composed of different sized fragments, it is of paramount importance to use DNA isolation methods that are capable of capturing all fragments of any dimension. In fact, as cancer-derived fcDNA is known to be composed of shorter fragments than that derived from non-cancer cells, it is fundamental to use a procedure that ensures the capture of short fragments in order to circumvent the possibility of selecting one specific type of fcDNA. Another major issue for fcDNA assessment is that different methodologies are used for DNA quantification. In the majority of studies real time PCR-based methods are used, in which single copy genes such as the human telomerase reverse transcriptase gene (hTERT) or the glyceraldehyde-3-phosphate dehydrogenase gene (GAPDH) are amplified (Table 1). When using PCR assays, it is fundamental that the target sequences be as short as possible, thereby permitting the amplification of shorter fragments. Standardized PCR-based methodologies are needed in order to allow a comparison of results from different studies [113].

With regard to molecular analyses performed on fcDNA, different procedures with different sensitivity are used in laboratories. One of the most widely used methods for DNA methylation analysis is methylation-specific PCR (MSP), which has proven to be a very effective tool in diagnostic procedures requiring a high sensitivity [114]. The addition of fluorogenic probes, e.g. in MethyLight procedures, makes MSP more informative, quantitative, and suitable for fcDNA evaluation [115]. The combined bisulfite restriction analysis (COBRA) procedure can be used as an alternative to MSP, its main limitation being that it is only useful to assess the DNA methylation status of CpGs that are located within restriction enzyme sites [116].

Several methodologies for mutation analysis are currently available, among which those with high sensitivity (SARMS, digital PCR, mutant-enriched PCR, and peptide nucleic acidlocked nucleic acid PCR [PNA-LNA-PCR]) are the most appropriate for fcDNA mutation analysis. All of these methodologies are characterized by a sensitivity of at least 1 % [117–119]. Next generation sequencing approaches have also been used for fcDNA characterization, and Narayan et al. [120] recently showed that it is possible to identify rare cancer-associated mutations using a modified deep sequencing method.

Conclusions and perspectives

Plasma or serum fcDNA represents a potentially useful source for the diagnosis and prognosis of NSCLC. Although the fcDNA quantity seems to be able to distinguish lung cancer patients from healthy individuals, its ability to differentiate malignant from non-malignant diseases is modest, highlighting its low specificity for lung cancer. On the other hand, the high levels of fcDNA in idiopathic pulmonary fibrosis (IPF) patients, apart from suggesting potential similarities with lung cancer, could facilitate the differential diagnosis between this serious condition and other non-specific interstitial diseases, such as non-specific interstitial pneumonia (NSIP), which are characterized by a very different prognosis and which are often misdiagnosed.

The molecular characterization of fcDNA seems particularly interesting as it may be applicable to clinical practice. As targeted therapies are increasingly being acknowledged as fundamental approaches in the treatment of lung cancer and many other tumor types, the molecular characterization of the primary tumor has become mandatory to ensure correct therapeutic decision-making. In this context, fcDNA could play two important roles. First, as lung cancer is predominantly diagnosed at a late stage, thus generally precluding the possibility of surgical treatment, the only biological material available often derives from small biopsies or fine needle aspirations. Thus, for some patients, little or no tumor material is available for molecular analysis and, in these cases, fcDNA could represent a valid alternative for the determination of clinically relevant molecular tumor characteristics. Although the sensitivity of this approach is still not very high, it is highly specific. Furthermore, even though the absence of a certain mutation in fcDNA cannot exclude the possibility of the same mutation being present in the tumor, the detection of a specific alteration in fcDNA could guide physicians in their choice of the most appropriate therapy. A second important advantage of fcDNA is that it enables the characterization of induced mechanisms of resistance to therapy via a non-invasive approach. It is known that targeted therapies induce resistance mechanisms, which has important prognostic and predictive implications. A more in-depth biological characterization of tumor material is, therefore, needed to offer the most appropriate tailored treatment options to patients.

In conclusion, although fcDNA analyses are simpler and less expensive to perform than CTC analyses, further efforts are needed to standardize the methodology before it can be implemented in routine clinical practice.