Molecular sputum analysis for the diagnosis of lung cancer

Abstract

Lung cancer is the leading cause of cancer mortality rate worldwide, mainly because of the presence of metastatic disease at the time of diagnosis. Early detection of lung cancer improves prognosis, and towards this end, large screening trials in high-risk individuals have been conducted since the past century. Despite all efforts, the need for novel (complementary) lung cancer diagnostic and screening methods still exists. In this review, we focus on the assessment of lung cancer-related biomarkers in sputum in the past decennium. Besides cytology, mutation and microRNA analysis, special attention has been paid to DNA promoter hypermethylation, of which all available literature is summarised without time restriction. A model is proposed to aid in the distinction between diagnostic and risk markers. Research on the use of sputum for non-invasive detection of early-stage lung cancer has brought new insights and advanced molecular techniques. The sputum shows a promising potential for routine diagnostic and possibly screening purposes.

Lung cancer is the leading cause of cancer mortality worldwide (Ferlay et al, 2010). Despite large-scale investments in research and optimisation of treatment strategies, lung cancer is mostly detected at an advanced stage, resulting in a general 5-year survival of 15% (Siegel et al, 2012). Prognosis greatly improves if lung cancer is detected at an early stage (Patz et al, 2000).

Lung cancer development evolves in approximately 10 to 30 years before it becomes clinically manifest (Hirsch et al, 2001). This latency period offers an opportunity to identify individuals at risk. In the past century (in the 1970s), thorax X-ray screening studies have been conducted for the detection of early-stage lung cancer, in which cytological examination of sputum was part of the diagnostic procedure (Melamed et al, 1984). Sputum cytology turned out neither to be of additive value in enhancing lung cancer detection nor in reducing lung cancer mortality. The average survival time increased after thorax X-ray screening because of lead time and sampling bias. The outcome of a recent low-dose spiral CT (LDCT) screening study seems promising as it reduces lung cancer mortality (National Lung Screening Trial Research Team et al, 2011).

In theory, a biomarker sputum test for early detection may be developed for three possible applications: (i) identification of at-risk individuals, who may be screened with LDCT after a positive biomarker test; (ii) after the first LDCT screen shows a solid lesion, a sputum biomarker may be developed as a diagnostic test for malignancy; and (iii) after the first LDCT screen shows a ground glass lesion, we can determine whether the lesion has a high or low chance of becoming malignant. In the setting of patients with symptomatic lung cancer, a sputum test may be useful for diagnostic workup of malignancy and if diagnosed with lung cancer to perform predictive analysis.

Biomarker screening may be categorised into (i) risk markers, which identify individuals at high risk of developing lung cancer, and (ii) diagnostic markers, which uncover invasive lung cancer. The biomarker must meet several conditions, such as being superior to conventional detection methods in terms of sensitivity and specificity, before it is considered suitable for clinical implementation (Box 1). In this review, diagnostic markers are defined as markers recognising (the transition to) invasive lung cancer. At this stage, the disease may be measurable but still asymptomatic. A risk marker is able to identify subjects at risk without measurable disease. In pathobiological terms, this marker may be associated with several conditions, such as measure of exposure to carcinogen and development of carcinoma in situ (Selamat et al, 2011).

Box 1. Diagnostic vs risk marker.

In context of evaluating the performance of a certain biomarker, we used the following approach for distinction between risk and diagnostic markers. As some asymptomatic lung cancer cases exist in a control population, an estimate for the expected number of undiagnosed lung cancer cases in the general/control population was made based on the following assumptions. A diagnostic biomarker test is able to detect lung cancer (arbitrary) 2 years before becoming symptomatic. Examining a high-risk population (e.g. heavy smokers), with a relative risk on lung cancer of approximately 12, 2.4% of the cases in the control population will be positive (2 (years) × 0.1% (approximate incidence (Siegel et al, 2012)) × 12 (relative risk)). To be on the safe side with this simplified approach, in this review the threshold was set at 4%. Thus, if >4% of the control patients had a positive test, this biomarker was regarded as risk marker.

In 2003, a review summarised the status of mutation analysis and initial methylation findings in sputum (Thunnissen, 2003). This manuscript provides an overview of developments in sputum analysis for lung cancer diagnosis in the past 10 years. The PubMed terms ‘lung cancer' and ‘sputum' were used. In addition, special attention is paid to DNA hypermethylation.

Sputum cytology

By means of cytology, tumour cells can be identified in sputum through aberrant cell morphology. Status of the diagnostic value of sputum cytology has not changed in the past decennium. In the clinical diagnostic setting, the sensitivity of sputum cytology is ∼60%, which also depends on the number of sputum samples examined (Risse et al, 1985). Although in developed countries the procurement of tumour biopsies/tumour cytology replaced the use of sputum cytology as standard for lung cancer diagnosis (Rivera and Mehta, 2007), in lower budget countries sputum cytology is an affordable diagnostic instrument and still clinically implemented (Ammanagi et al, 2012).

Molecular analysis of sputum

DNA mutation analysis

For DNA mutation analysis, currently the relevant part of the gene of interest is amplified, usually using polymerase chain reaction (PCR) technology. This is a very sensitive, low-cost, rapid and simple method. Disadvantages are that contamination may be an issue, as well as that the enzyme DNA polymerase has a small error. In about 0.1% of amplicons, an incorrect nucleotide may be incorporated (Eckert and Kunkel, 1991). If this error happens early in the PCR procedure, it will propagate and may lead to a false-positive signal, thus reducing specificity.

Mutations of tumour-suppressor gene p53 and oncogene KRAS have been identified to have a role in lung carcinogenesis (Hanahan and Weinberg, 2011). In 50% of lung cancer cases, mutations or deletions are present in the p53 gene (Greenblatt et al, 1994). KRAS mutations mostly occur in adenocarcinomas (20–30% in western countries and 10% in eastern countries) (Shigematsu et al, 2005).

Various KRAS mutation detection techniques have been investigated on sputum specimens (Table 1). Peptic nucleic acid–PCR–restriction fragment length polymorphism (PNA–PCR–RFLP) and Point-EXACCT were described as methods of choice (Thunnissen, 2003).

Table 1. Studies on KRAS and p53 mutation analysis in sputum samples.

	Subjects					Molecular alterations
Study	Cases	Controls	Gene	Method	PCR cycles	Cases	Controls	Remarks
Baryshnikova et al (2008)		Smokers	KRAS	PCR–RFLP	40		0/506 (0%)	Follow-up 2–6 years; 18 patients developed lung cancer without molecular alterations at baseline
			p53	PCR and SSCP	40		15/803 (2%)	One patient with p53 mutation at baseline developed SqCC, but was not confirmed in resected tumour tissue. Also DNA promoter hypermethylation tested of p16, NORE1A and RASSF1A (Supplementary Table 1)
Destro et al (2004)	NSCLC	Smokers	KRAS	PCR–RFLP	40	11/50 (22%)	0/100 (0%)	Fourteen of 50 tumour tissue samples tested KRAS mutation positive. In three cases, concomitant p16 hypermethylation (Supplementary Table 1)
Keohavong et al (2003)	All		KRAS	MAE, PCR and DGGE	47	XW: 23/102 (23%) BH: 7/50 (14%)		Data of both tumour and sputum were presented together. Two study populations: Xuan Wei County (XW) and Beijing and Henan (BH), respectively. XW subjects were exposed to coal smoke
Keohavong et al (2004)	Lung cancer NS		KRAS	Cell centrifugation, laser capture microdissection, PCR and DGGE (KRAS)/SSCP (p53)	47	2/15 (13%)		Subjects were exposed to coal smoke. KRAS mutation status of primary tumour unknown
			p53		42	6/15 (40%)
Keohavong et al (2005)		(Non) Smokers	KRAS	Cell centrifugation, laser capture microdissection, PCR and DGGE (KRAS)/SSCP (p53)	30		2/92 (2%)	Subjects were exposed to coal smoke
			p53		42		14/92 (15%)
Zhang et al (2003)	NSCLC		KRAS	MAE, PCR and DGGE	55	10/22 (46%)		In 12 out of 22 matched tumour–sputum samples, KRAS mutation was identified using the same method (κ=0.64, 95% confidence interval: 0.32–0.95, P<0.01). One patient tested negative in tumour, but positive in sputum

Abbreviations: All=all types of lung cancer included; DGGE=denaturing gradient gel electrophoresis; MAE=mutant allele enrichment; NS=not specified; NSCLC=non-small-cell lung cancer; PCR=polymerase chain reaction; RFLP=restriction fragment length polymorphism; SqCC=squamous cell carcinoma; SSCP=single-strand conformational polymorphism.

Destro et al (2004) confirmed KRAS mutation in 79% of the sputum samples from lung cancer patients with a KRAS mutation in their tumour tissue (n=14). In controls, none tested positive. Keohavong and co-workers (2004, 2005) conducted studies in Xuan Wei County (China), where lung cancer rates were fivefold higher than the Chinese national average. Mutation detection was optimised by application of cell cytocentrifugation and laser capture microdissection, enabling detection of low fraction mutations, even in morphologically benign bronchial epithelial cells (Keohavong et al, 2004, 2005). With this approach, examination of cytology is still needed for dissecting abnormal or benign epithelial cells for enrichment. In a cancer-free population, mutations in both genes were identified (15 out of 92) (Keohavong et al, 2005). These mutations occurred in none of the matched buccal epithelial cells, indicating that the latter cells are not suitable as a surrogate marker for lung cancer (risk).

Until recently, most sputum studies have been performed on patients with symptomatic lung cancer. Research conducted before 2003 show that KRAS mutations may be detected in sputum at least 1 year before clinical diagnosis of lung cancer (Somers et al, 1998). Baryshnikova et al (2008) were the first to investigate sputum from a large LDCT screening cohort (n=803) consisting of asymptomatic heavy smokers, assessing frequency of KRAS and p53 mutations, next to DNA promoter hypermethylation of p16, NORE1A and RASSF1A (Supplementary Table 1). KRAS mutation analysis was performed by restriction endonuclease-mediated selective PCR with a reported sensitivity of one mutant per 1000 wild-type genes. No KRAS mutation was identified, especially in the 18 subjects who developed lung cancer during the follow-up period. None of these patients had molecular alterations at baseline. In 15 out of 803 (2%) participants, a p53 mutation was found, of whom one patient was diagnosed with early-stage lung cancer in follow-up without confirmation of the p53 mutation in the tumour.

These studies suggested that KRAS might be more suitable as a diagnostic marker than for risk assessment in precancerous stages. Future studies with further follow-up of participants are needed to elucidate whether molecular alterations of KRAS and p53 are indeed suggestive for lung cancer development.

Mutations in the tyrosine kinase domain of the epidermal growth factor receptor (EGFR) have been identified in parts of lung adenocarcinomas, and are associated with high response rates to treatment with EGFR tyrosine kinase inhibitors (Sharma et al, 2007). Epidermal growth factor receptor mutation analysis has been performed in some sputum samples as part of larger series of other cytological samples, mostly without detailed information and not compared with the original tumour (Boldrini et al, 2007; Takano et al, 2007; Tanaka et al, 2010). In a total of three publications, 3 out of 25 sputum samples were positive in cases with cytologically proven malignant cells.

EML4-ALK is a lung cancer fusion oncogene that is estimated to be expressed in 3–6% of lung adenocarcinomas (Takeuchi et al, 2008), showing marked response to treatment with ALK inhibitors (Kwak et al, 2010). Recently, Soda et al (2012) reported the development of a multiplex RT–PCR system that was able to detect EML4-ALK mutations in 4 out of 35 sputum samples, which were part of a prospective screening cohort of NSCLC patients.

Optimisation of EGFR and EML4-ALK mutation detection in sputum may, in the future, contribute to minimise the use of invasive bronchoscopy or transthoracic needle biopsies to secure tumour biopsies for mutation testing, a clinical need in monitoring personalised treatment.

DNA hypermethylation

Aberrant DNA promoter methylation is a cell control mechanism in lung carcinogenesis (Selamat et al, 2011), involving the addition of a methyl group at the carbon 5 position of cytosines at CpG sites in DNA. A widely used approach to distinguish methylated DNA from unmethylated DNA is exposing genomic DNA to bisulphite before PCR. In this process, unmethylated cytosine is converted into uracil, whereas methylated cytosine remains unchanged. The templates are next subjected to methylation-specific PCR (MSP). It has the same (dis)advantages as PCR in mutation analysis (Shaw et al, 2006), with an additional disadvantage that bisulphite conversion may be incomplete. In that case, not all unmethylated cytosines are converted to uracil, leading to false-positive results, whereas some controls do not correct for this. In some methods like pyrosequencing (Hwang et al, 2011), this incomplete conversion may be detected, but this is not the case with MSP.

To examine the effect of a high number of cycles in MSP, a small interlaboratory study was performed between The Canisius Wilhelmina Hospital in Nijmegen, the Netherlands (CP, ET) and the Fondazione IRCCS Istituto Nazionale Tumouri in Milan, Italy (GZ) with identical set-up (DNA samples, primers and protocols) (Field et al, 2009). Promoter DNA hypermethylation of p16 and β-retinoic acid receptor (RAR-B) (Martinet et al, 2000) was investigated using nested MSP analysis with a number of different PCR cycles (50–80 cycles in total). Consistent results between the laboratories were present up to 55 cycles (in total). In experiments with >55 PCR cycles, a greater number of samples became positive with the loss of reproducibility between and within laboratories. These results point towards false positivity and indicate that caution must be exercised when interpreting data derived from studies in which nested MSP with more than 55 PCR cycles were applied.

Based on this knowledge, we divided literature studies on gene methylation in sputum published up to now into three categories: studies in which <55 PCR cycles were applied, >55 PCR cycles or unknown number of PCR cycles, respectively. The latter category consisted of those studies of which the publications did not provide explicitly the number of PCR cycles used.

Hypermethylation frequency in sputum of mostly symptomatic lung cancer patients and controls was investigated for 54 genes (Supplementary Table 1). Not for all genes information could be specified: some genes were described with incomplete data. Furthermore, inclusion and exclusion criteria for both cases and controls differed between the studies. Some studies included only non-small-cell lung cancer, others also examined small-cell lung cancer and unspecified lung cancer cases. The examined populations consisted usually of more male subjects than female subjects (average 75% in cases, 70% in controls, respectively).

From the summary table, it is apparent that only for a limited number of genes published data were available for categories >55 or ⩽55 PCR cycles, respectively. Concerning p16 gene, five studies (Belinsky et al, 1998; Destro et al, 2004; Olaussen et al, 2005; Cirincione et al, 2006; Shivapurkar et al, 2007) with ⩽55 PCR cycles and nine studies with >55 PCR cycles (Kersting et al, 2000; Palmisano et al, 2000; Konno et al, 2004; Belinsky et al, 2006; Hsu et al, 2007; Liu et al, 2008; Guzmán et al, 2012; Leng et al, 2012; Shin et al, 2012) with sensitivity and specificity data were available for bivariate analysis (Reitsma et al, 2005). Interestingly, mean specificity was shown to be significantly lower in the group of studies with >55 PCR cycles (74% vs 87%, P<0.001), whereas sensitivity was higher, but not significantly different (49% vs 33%, P=0.13). This literature analysis supports the above-mentioned theoretical notion that a high number of PCR cycles leads to a higher chance of false-positive results. It is not excluded that a diagnostic marker may be looked upon as a risk marker (as defined in Box 1), when >55 PCR cycles are run with possible induced false positivity. Moreover, when comparing the number of PCR cycles (>55 or ⩽55) with marker classification (diagnostic vs risk), a biomarker is more likely to be classified as a risk marker if >55 PCR cycles were applied compared with at most 55 PCR cycles (85% vs 58%, P=0.002).

In 10 studies (Kersting et al, 2000; Palmisano et al, 2000; Chen et al, 2002; Liu et al, 2003; Wang et al, 2003; Destro et al, 2004; Olaussen et al, 2005; Cirincione et al, 2006; Belinsky et al, 2007; Hsu et al, 2007) matched tumour and sputum samples were examined. The median frequency of gene hypermethylation was higher in tumour than in sputum samples: 48% (interquartile range 36–64%) vs 38% (interquartile range 31–57%), respectively. A meta-analysis on exact data (Kersting et al, 2000; Liu et al, 2003; Wang et al, 2003; Olaussen et al, 2005; Shivapurkar et al, 2007; Shin et al, 2012) showed that this observed tendency was not significant (P=0.09; Durkalski et al, 2003). Median concordance of methylation between tumour and matched sputum, calculated from the same studies, is 78% (interquartile range 73–91%), indicating that the use of sputum as non-invasive biological fluid for detection of aberrant methylation is representative of the methylation status of primary tumour tissue.

Still, none of the biomarkers yield 100% sensitivity. The multidimensional character of lung cancer, in which various genes might be involved (Hansen et al, 2011), requires a panel of markers that can complement each other in lung cancer detection. Several studies calculated combined sensitivity and specificity for hypermethylated genes (Zöchbauer-Müller et al, 2003; Belinsky et al, 2005, 2006, 2007; Hsu et al, 2007), revealing higher performance when compared with the markers individually. These algorithms seem promising, but are scarcely validated in independent study cohorts. Interestingly, one study (Leng et al, 2012) replicated a panel of previously published hypermethylation markers (Belinsky et al, 2007) in two independent slightly different cohorts: case–control vs asymptomatic stage I lung cancer patients. They showed a slightly higher sensitivity and specificity in the second cohort. However, the methylation panels were not exactly similar between the study cohorts. Also, as sputum samples were stored in Saccomanno after collection without further treatment, DNA quality may be reduced, possibly affecting the study data. Therefore, at this point in time, it is difficult to define an unambiguous biomarker signature panel for lung cancer risk based on these results.

Patient selection, sputum collection and procedure methods might explain the differences in rates of methylation between studies investigating the same biomarker.

Research into additional novel markers remains necessary.

Loss of heterozygosity

Microsatellite alterations present as loss of heterozygosity (LOH), or as microsatellite instability (MSI). Conceptually, LOH is essentially different from previous markers, because it explores the absence of the allele that is present in the normal situation, whereas the other above-mentioned biomarkers look for the presence of a specific abnormality. Because the fraction of tumour cells in sputum is usually <1%, the majority of the cells will not have LOH. Therefore, looking for tumour-related LOH has a disadvantage: requiring a difference that is higher than the threshold of the test based on signal-to-noise ratio. For example, when LOH is present in 1% of tumour cells, the proportion of missing alleles is 0.5%. To demonstrate this, a test is required that is able to make a distinction between 100% (normal reference DNA; e.g. lymphocytes) and 99.5% (mixed sample with 99% normal and 1% heterozygous tumour DNA). It is difficult to perceive a clinical assay with such a low variation coefficient that this small difference can be reliably detected in sputum.

Using polymorphic DNA markers in PCR-based assays, LOH and MSI has been reported in sputum of lung cancer patients. These polymorphic DNA markers are non-informative in individuals who are homozygotic for these markers. Therefore, several markers need to be examined to cover the general population.

Four studies have been conducted on LOH and lung cancer, of which the most recent ones were published in 2007 (Arvanitis et al, 2003; Wang et al, 2003; Castagnaro et al, 2007; Hsu et al, 2007). No studies have followed since. All studies report comparable results with LOH in 26–55% in cases and 0–11% in cancer-free controls. Prevalence of MSI was low in all studies, ranging from 4 to 35% in cases and 0 to 5% in controls. Arvanitis et al (2003) tested 48 markers in sputum and bronchial washings (analysed together), in which non-cancer-specific markers were also included. Looking at informative loci, they calculated fractional allele loss values. Significant variations were observed for the markers, which may be related to non-neoplastic genetic alterations. This kind of results needs to be confirmed by others. Taking these data and the technical considerations into account, there is room for debate whether LOH and MSI by themselves are suitable as sputum biomarkers for lung cancer.

MicroRNA

MicroRNAs (miRNAs) are a class of small non-coding RNA molecules, which are associated with a spectrum of biological and pathological processes.

In a small feasibility study, Xie et al (2010) demonstrated that endogenous miRNAs are stably present in sputum specimens. Using real-time RT–PCR, miR-21 and miR-155 were detected, of which miR-21 was significantly overexpressed in sputum of lung cancer patients as compared with cancer-free subjects. Furthermore, elevated miR-21 expression was more sensitive (70%) than conventional sputum cytology (48%) in diagnosing lung cancer.

The same research group defined miRNA signatures for different histologic types of lung cancer in studies of similar design (Xing et al, 2010; Yu et al, 2010). Sensitivity increased when complementary miRNAs were combined in a panel as compared with single miRNAs. For the diagnosis of squamous cell lung cancer, the combination of miR-205, miR-210 and miR-708 yielded 73% sensitivity and 96% specificity. A panel consisting of miR-21, miR-200b, miR-375 and miR-486 produced 81% sensitivity and 92% specificity in discriminating sputum of lung adenocarcinoma patients from controls. The authors found no association between miRNA expression and stage of lung cancer, suggesting that the miRNA signatures can be used as a tool in the detection of early lung cancer. Overall, miRNA analysis has recently become available and more studies in sputum seem useful.

Messenger RNA

From practical point of view, there is a disadvantage using messenger RNA (mRNA). In contrast to miRNA (see above), mRNA is rapidly degraded in the sputum. Therefore, it is necessary to process sputum after collection as soon as possible.

Several studies investigated aberrant mRNA profiles in sputum (Jheon et al, 2004; Mecklenburg et al, 2004; Sun et al, 2009) (Table 2). Reverse-transcriptase quantitative PCR (RTQ–PCR) was more sensitive than sputum cytology (14%) and immunocytochemistry (16%). In short, two studies revealed high specificity and reasonable sensitivity (Jheon et al, 2004; Sun et al, 2009). Confirmation of these results is needed.

Table 2. Studies investigating presence and/or expression of RNA and tumour-related proteins in sputum.

Study	Cases	Controls	Method	Protein/gene	Results cases	Results controls	Se (%)	Sp (%)	Positive cytology	Remarks
Sun et al (2009)	All	Benign pulmonary disease	RTQ–PCR, immunocytochemistry	APRILa	58/71 (82%)	2/62 (3%)	82	97	Cases 10/71 (14%): all SCC Immunocytol: 11/71 (16%)	Healthy subjects: 1/65 (2%). Cutoff value: mean±2 s.d. of mRNA expression in healthy subjects
Mecklenburg et al (2004)b	All	Benign pulmonary disease	RT–PCR	MAGE-1	2/14 (14%)	0/2 (0%)	14	100	cases 1/8 (13%)	Positive cytology sample was not tested with RT–PCR. Cytology of remaining samples not performed
				MAGE-2	1/14 (7%)	0/2 (0%)	7	100
				MAGE-3/6	0/14 (0%)	0/2 (0%)	0	100
				MAGE-4	2/14 (14%)	0/2 (0%)	14	100
				MAGE-12	2/14 (14%)	0/2 (0%)	14	100
				All combined	5/14 (36%)	0/2 (0%)	36	100
Jheon et al (2004)b	All	Benign pulmonary disease	RT–N-PCR	MAGE A1–6	72/134 (54%)	3/140 (2%)	54	98	Cases 6/31 (19%)	Also spontaneous sputum collected. Data of lung cancer patients from group I (collection at the day of thoracotomy) and II (lung cancer in clinical workup) combined. Follow-up (1 year): no cancer in controls
			TRAP method	Telomerase	8/27 (30%)		30
Pasrija et al (2007)b	All	Cancer-free subjects	TRAP method	Telomerase	23/34 (68%)	3/30 (10%)	68	90
Pio et al (2010)b	All	Cancer-free subjects	Anti-factor H antibodies	Complement factor Ha			80	88		Also spontaneous sputum collected. Se and Sp based on cutoff ROC curve
Kalomenidis et al (2004)	All	Benign pulmonary disease	IRMA	CEAa			57	95		Se and Sp based on cutoff ROC curves
			IRMA	NSEa			19	95
			IRMA	CYFRA 21-1a			36	95
Hillas et al (2008)b	All	COPD	IRMA	CEA			NS	NS	cases 4/50 (8%)	CEA median concentration. Cases: 713 ng ml−1, controls 518 ng ml−1
			IRMA	NSE			NS	NS		NSE median concentration. Cases: 12 ng ml−1, controls 13.7 ng ml−1
			RIA	CYFRA 21-1a			86	75

Abbreviations: All=all types of lung cancer included; CEA=carcinoembryonic antigen; COPD=chronic obstructive pulmonary disease; IRMA=immunoradiometric assay; NSE=neuron-specific enolase; NS=not specified; RIA=radioimmunoassay; ROC=receiver operating characteristic; RT–(Q)(N)–PCR=reverse transcriptase (quantitative) (nested)–polymerase chain reaction; Se=sensitivity; Sp=specificity; TRAP=telomeric repeat amplification protocol.

^aP<0.05 significance level between cases and controls.

^bInduced sputum.

Protein

Several studies explored the presence and/or expression of tumour-related proteins in sputum of lung cancer patients and controls (Table 2). Sun et al (2009) reported significantly elevated expression of a proliferation-inducing ligand (APRIL) in sputum of lung cancer patients compared with controls (82% vs 3%, respectively).

Pio et al (2010) demonstrated increased levels of complement factor H in sputum of lung cancer patients, and suggested that large plasma proteins as factor H reflects hyperpermeability in tumour circulation. Factor H quantification may aid in improving sensitivity of sputum cytology for lung cancer diagnosis, but is not proof of malignancy similar to hemoptysis.

Fluorescence in situ hybridisation

Fluorescence in situ hybridisation (FISH) assay allows detection of chromosomal aneusomy, rearrangements and copy number changes in interphase cells, but usually requires the cytological or automated detection of abnormal cells. Fluorescence in situ hybridisation by itself is not superior to sputum cytology, but can improve sensitivity of lung cancer detection when used in conjunction with sputum cytology or as confirmatory test (Romeo et al, 2003; Katz et al, 2008). Li et al (2007) showed that FISH analysis of both HYAL2 and FHIT deletions was more sensitive than cytology alone (sensitivity: 76% specificity: 92%). Kettunen et al (2006) did not find significant differences in copy number gain between high-risk subjects and healthy never-smokers, indicating that copy number gain is not useful as a risk marker. Qiu et al (2008) used enrichment procedure based on anti-CD14 and anti-CD16 antibody beads before FISH and cytology. However, sensitivity of FISH and cytology results remained comparable (58% vs 53%). No internationally standardised method exists for cytometry (Thunnissen et al, 1996). So far, the data are useful for analysis on group level, but its relevance is questionable for the individual patient.

Other markers

Free DNA exists in higher concentration in the serum of lung cancer patients than in the serum of controls (Sozzi et al, 2003). Van der Drift et al (2008) found that the amount of free DNA in sputum was related to severity of inflammation, but not in the presence of lung cancer.

In a small study, sequence variants in mitochondrial DNA (mtDNA) were investigated in specimens (no sputum) of lung, bladder and kidney cancer patients, and sputum from 12 cancer-free heavy smokers (Jakupciak et al, 2008). Tumours were found to contain significantly more mtDNA mutations compared with matched body fluids and blood, and sputum of controls. Biological relevance of mitochondrial mutations yet needs to be clarified.

Fourier transform infrared (FTIR) spectroscopy is a non-invasive method that visualises biochemical changes in sputum by determination of absorbance levels of infrared wavenumbers. In a small feasibility study, Lewis et al (2010) reported that a panel of wavenumbers was able to distinguish cancer sputum from healthy control sputum. Fourier transform infrared might have the potential as a high-throughput method for screening.

Black matter deposition (anthracosis) was assessed in sputum by Konno et al (2004), next to DNA hypermethylation (Supplementary Table 1). Mean anthracotic index of lung cancer patients was significantly higher than that in controls and might thus be suitable for identifying a population at risk for lung cancer development. Remarkably, this index was not correlated with smoking or with detection of lung cancer cells in the sputum samples.

Conclusion

Ten years of additional research on the use of sputum in risk assessment or the early detection of lung cancer has brought new insights and more advanced molecular techniques. Polymerase chain reaction-based assays made detection of low fraction mutations feasible in sputum, although one has to be cautious for false-positivity induced by high number of PCR cycles. More biomarkers have been identified in sputum, such as DNA hypermethylation markers, miRNAs and tumour-related proteins, which show the potential for screening purposes. A rational for the distinction of a risk from a diagnostic marker was provided.

Although in recent years many markers have been examined in sputum, they are currently not sufficiently validated for clinical application. These studies, comparing sensitivity and specificity of cytology with molecular analysis, respecting technical limitations, should be reported in future studies.