Circulating nucleic acids in plasma and serum as a noninvasive investigation for cancer: Time for large‐scale clinical studies?

A noninvasive blood test to detect sporadic cancer has been seen as somewhat of a holy grail by clinicians with an interest in cancer, and its delivery as a quest for many researchers. For this reason, detection of cell-free circulating DNA in the plasma and serum of cancer patients, which has genetic characteristics identical to those of the primary tumour, has resulted in substantial interest and over 200 publications in the medical literature. Interest stems not only from the fact that a blood-based diagnostic and screening test for cancer is an elegant and attractive concept in its own right but also from the fact that conventional diagnostic cancer tests tend to be imperfect.1 As an example, colorectal cancer screening presently relies on faecal occult blood testing, which is both insensitive and nonspecific. In contrast, flexible sigmoidoscopy is sensitive and specific for early distal disease but both invasive and insensitive for proximal disease. Furthermore, barium enema is relatively sensitive and specific but requires colonic preparation, radiation and a day off work, while total colonoscopy is highly sensitive and specific but also invasive and expensive. The situation appears little better for other cancers. No reliable test is available for early detection of lung cancer, with computerised tomography being the most reliable tool. In addition, although several studies indicate that mammographic screening might be a useful strategy for reducing breast cancer mortality, there remains considerable controversy regarding the value of population screening programs.2 Finally, development of conventional tumour markers, e.g., CEA, AFP and the widely used PSA, was driven largely by the introduction of new methods for quantifying small amounts of circulating proteins. However, sensitivity and specificity shortcomings with these assays remain to be overcome.3 The introduction of PCR-based technology in the late 1980s and refinements over the past 10 years have allowed us to detect and quantify extremely small amounts of nucleic acids. This has led to the identification of large numbers of novel molecular targets that may eventually become clinically useful cancer markers. Additionally, many of these markers have been detected in tumour-derived nucleic acids (DNA and RNA) extracted from serum and plasma samples. However, it is over 30 years since initial studies indicated that plasma-based nucleic acids might assist in cancer diagnosis. Studies performed in the early 1970s initially showed that increased quantities of DNA could be found in the plasma of patients suffering from different malignancies,4 but it was not until the 1990s that this circulating DNA was shown to exhibit tumour-related alterations, including decreased strand stability,5 Ras and p53 mutations, microsatellite alterations, aberrant promoter hypermethylation of several genes, rearranged immunoglobulin heavy chain DNA, mitochondrial DNA mutations and tumour-related viral DNA.6, 7 Over this period, it was also shown that tumour-related circulating DNA was not confined to any particular cancer type but appeared to be a ubiquitous finding across the cancer spectrum. Thus, mutant plasma DNA has been found in colorectal, pancreatic, biliary tree, skin, head-and-neck, lung, breast, kidney, ovarian, nasopharyngeal, liver, bladder, gastric, prostate and cervical cancers as well as in haematologic malignancies including lymphomas. The results obtained in plasma/serum DNA in many cancers are opening new research areas and indicate that plasma/serum may eventually be a suitable source for the development of noninvasive diagnostic, prognostic and follow-up tests for cancer. The diagnostic value of plasma DNA testing appears promising in a number of cancers including melanoma,8 B-cell malignancies9 and NPC.10 However, perhaps the closest to achieving clinical significance is an assay for EBV DNA, which is closely associated with NPC in southern Asia. Using real-time PCR, EBV DNA is detectable in 95% of NPC cases compared 5% of healthy controls. In addition, following diagnosis, the test appears to be useful for determining prognosis and monitoring disease response to treatment. Overall, the absolute levels of EBV DNA at presentation are of considerable prognostic value as are levels following treatment, in so far as a high level is suggestive of the presence of residual disease. It appears likely that in the next few years estimation of EBV DNA will become a routine part of the staging procedure for NPC and will directly influence therapeutic options for this tumour. The prognostic value of plasma/serum tumor DNA has also been established for other cancers, with high levels also indicative of a poor prognosis.6, 7 Surprisingly, using essentially similar methodology, cell-free mRNA can also be detected in plasma and should, at least in theory, permit plasma-based expression profiling.11 Studies with RNA markers are particularly promising due to their close association with malignancy. In this short review, we emphasise studies on the most widespread malignancies, colorectal, pancreatic, lung, breast and prostate cancers, which need a simple test that could become clinically available in the not too distant future. AFP, alpha-fetoprotein; APC, adenomatous polyposis coli; CEA, carcinoembryonic antigen; CK, cytokeratin; EBV, Epstein-Barr virus; GSTP1, glutathione-S-transferase P1; hTERT, human telomerase reverse transcriptase; hTR, human telomerase RNA; MSP, methylation-specific PCR; NPC, nasopharyngeal carcinoma; PSA, prostate-specific antigen; SCLC, small cell lung cancer. k-ras alterations occur in about 50% of colorectal cancers, and relatively simple assays to detect these point mutations have been available for some time. Therefore, it is not surprising that some of the first studies to be performed on plasma/serum DNA involved detection of these mutations. These studies showed that k-ras mutations could be detected in the plasma of colorectal cancer patients and that there was little association with tumor stage. These initial studies have been reviewed by Sorenson.12 A large study has expanded these early results and found k-ras mutations in 50% of colorectal cancer cases.13 Mutant serum k-ras was detected preoperatively in a large group of patients with all stages of colorectal neoplasia, including adenomas. Furthermore, a prospective follow-up study was performed on 94 patients who underwent apparently curative resection for colorectal cancer to ascertain if serum mutant k-ras could be used postoperatively as a disease marker. Preoperative sera from 78 patients (group A) and sera from 94 patients obtained 6 months to 2 years postoperatively (group B) were analysed. Codon 12 and 13 mutations were analysed in matched tumour and serum samples in group A. K-ras mutations were found in 41 of 78 tumours (53%) and in 32 of 78 of the preoperative sera (41%). Of the 41 tumour k-ras mutation-positive cases, an identical serum mutation was detected in 31 (76%). In group B, 60 of the 94 cases (64%) were primary tumour k-ras mutation-positive. Of these 60, 16 (27%) became persistently serum mutant k-ras+ postoperatively and 10 of these (63%) developed a recurrence. In contrast, only 1 of 44 patients (2.3%) who remained serum mutant-negative developed recurrent disease (relative risk of developing recurrence if k-ras+ 27.1, 95% confidence interval 8.1–52.7; p < 0.0001). None of 34 tumour mutation-negative cases became serum mutant k-ras+ postoperatively, despite recurrence in 8/34. Overall, postoperative serum mutant k-ras was associated with a significantly lower disease-free survival (p < 0.0001), and the authors concluded that serum mutant k-ras detected postoperatively was a strong predictor of disease recurrence and had potential for use in clinical practice as a marker of preclinical cancer recurrence. Other research also indicates that detection of k-ras mutation in plasma/serum might be used in conjunction with colonoscopy to detect cancer.14 In a prospective study of 240 patients undergoing colonoscopy, k-ras mutations were detected in the plasma of 83% of those who had similar mutations in tumour biopsy tissues. Among those who had no biopsy available, about one-quarter had k-ras alterations and most of these had risk factors for colorectal cancer. Overall, 39% of patients with k-ras mutations in plasma had colorectal neoplasm compared to only 3% of those without mutations. In this respect, this study detected k-ras not only in plasma of carcinoma but also in that of adenoma, suggesting that plasma k-ras may enable detection of early neoplasms. The authors considered that this test might ultimately play a role as a screening test for colorectal cancer. An alternative would be to combine plasma k-ras analysis with conventional markers such as CA19-9. Similar approaches are being investigated using other frequently mutated genes. As an example, mutated APC gene sequences have occasionally been detected in the plasma of patients with sporadic colorectal cancer.15 Work on circulating RNA markers is particularly promising since the percentage of tumours detectable with this assay appears to be higher than that found with DNA markers. A small study on the telomerase RNA component hTERT has produced promising results indicating that such a test might be coupled with k-ras mutation detection.16 Eukaryotic chromosomal ends consist of repeating DNA sequences (TTAGG) termed telomeres. These stabilizing terminal sequences become progressively shortened during each cell cycle, eventually resulting in cellular death. Telomerase is a ribonucleoprotein enzyme that adds telomeric repeats onto chromosomal ends, thereby replacing the lost DNA. Telomerase is composed of 2 core enzymatic subunits, the hTR template and the hTERT protein. hTR and hTERT expression is related to telomerase activity. Thus, normal somatic cells have low or undetectable telomerase levels, while cancer cells have detectable telomerase activity in 85–100% of cases. Using real-time PCR, Dasi et al.16 studied the telomerase marker hTERT and identified 8/9 colorectal cancers using the plasma assay and 9/9 in the plasma of lymphoma patients. In contrast, all 10 healthy controls were negative for hTERT. Interestingly, 2 colorectal cancer patients were T1N0M0 and both plasma samples were telomerase-positive. Moreover, only a single colorectal cancer patient had metastatic disease, indicating that this assay has the ability to detect early disease. Other RNA markers, CK19 and CEA, have also been studied in the plasma of colorectal cancer patients. The presence of RNA from epithelial tumours in plasma is related to advanced stages and the presence of circulating tumour cells.18 Pancreatic cancer differs from colorectal cancer in a number of important respects when it comes to detecting gene alterations in plasma. Firstly, pancreatic cancer can be difficult to differentiate from other conditions on the basis of clinical features and imaging investigations, so a noninvasive test would be especially valuable in this condition. Secondly, the k-ras gene is mutated in approximately 90% of pancreatic adenocarcinomas, suggesting that a comprehensive analysis of many genes would be unnecessary to detect the majority of cases. In the most successful study,18 plasma DNA was isolated from 21 pancreatic cancer patients and k-ras alterations were detected by restriction fragment length polymorphism-PCR assay and subsequent product sequencing. Patients were followed up to determine their clinical outcome. k-ras mutations were found in the plasma of 17 patients (81%). In cases for which both plasma and pancreatic tissue were available, DNA mutations were similar in corresponding plasma and tissue samples. Plasma DNA alterations were found 5–14 months before clinical diagnosis in 4 patients. Mutant DNA was not found in the plasma of 2 patients with chronic pancreatitis or in 5 healthy controls. Microsatellites are repetitive DNA sequences of unknown function that form variable-length stretches of DNA. Using appropriate primers, it is possible to amplify DNA fragments that can be used as microsatellite markers and, with a panel of such markers, to profile tumours. Characteristic genetic changes in the tumour tissue, in this case microsatellite alterations, are detectable in plasma DNA. The first reports of this approach involved patients with SCLC19 and head-and-neck cancer.20 In the former case, microsatellite alterations were present in three-quarters of the tissues and in a similar fraction of plasma; similar results have been described in other histologic subtypes of lung cancer.21 A larger follow-up study was undertaken to prospectively investigate the correlation of abnormal plasma DNA with patient survival.22 Thirty-five patients with SCLC were selected after histologic diagnosis. Polymorphic markers (ACTBP, UT762 and AR) were chosen for their reported high rate of alterations in SCLC and analysed in tumor tissue, normal blood cells and plasma DNA. Moreover, mutations of the TP53 gene in tumour and plasma DNA were also assayed. In 25 patients (71%), at least 1 molecular change, precisely matching that of the primary tumour, was detected in the plasma DNA. A significant difference (p = 0.02) in survival was observed between patients having both microsatellite alterations and TP53 mutations at diagnosis compared to patients bearing only 1 or neither of these alterations. In 17 cases, (68%), there was a correlation in the clinical course of the disease between the tumour response to treatment and the presence or absence of abnormal plasma DNA. A study combining quantitative and qualitative analyses of circulating DNA has also been performed on non-SCLC.23 Circulating plasma DNA was quantitated to determine if levels could help distinguish lung cancer patients from healthy individuals and to determine the association between plasma levels and clinical progression. Furthermore, the kinetics of plasma DNA in disease-free, surgically resected patients was analysed. Plasma DNA quantification and analysis of microsatellite alterations were performed in a consecutive series of 84 patients with non-SCLC, who were followed-up and compared to 43 healthy controls. Mean values of plasma DNA were higher in patients, even those with stage Ia disease, than in controls. Sensitivity and specificity estimates were calculated using receiver operating characteristic curve analysis, producing an area under the curve of 0.844. Variations in DNA level and in microsatellite changes correlated with clinical status in 38 patients monitored during follow-up. These data suggest that quantification and microsatellite characterisation of plasma DNA in lung cancer patients may be a valuable noninvasive diagnostic tool for discriminating patients from unaffected individuals and for detecting early recurrence during follow-up. Aberrant methylation of the p16 gene is also common in lung cancer, and hypermethylated genes have been found in a high proportion (73%) of serum samples taken from these patients.24 RNA markers may eventually improve the early detection rate of lung cancer patients. Fleischhacker et al.25 studied a panel of 5 different tumour-related genes using RT-PCR on plasma/serum samples of lung cancer patients. Expression of these genes was restricted to lung tissue and overexpression associated with malignancy. Two groups of 18 patients with lung cancer were analysed before and during chemotherapy. The β-actin message, which acted as an RNA control, was detected in all sera from the control group and patients with lung cancer. Analysis of CK19 expression was positive in the majority of tumour patients, but positive results were also shown in all of the control sera. Expression of the MAGE-2 and TTF-1 genes was not observed in any of the patients in either the lymphocyte preparations or serum samples. Expression of the PGP 9.5 gene was observed in the cells of all 18 patients, but serum mRNA was detectable in only a single case. hnRNP-B1 mRNA was detectable in 14/18 sera, and Her2/neu-specific mRNA was amplified from the serum of 7/18 cases. Interestingly, it was possible to identify all cancer cases when hnRNP-B1 and Her2/neu-specific mRNA markers were combined. Breast cancer is an interesting target for molecular diagnosis since large quantities of money are presently spent on mass screening programs for this disease. Several studies have been performed using microsatellite analysis with relatively high percentages of sensitivity.26, 27 Silva et al.28 examined a group of breast cancer patients at diagnosis using a variety of molecular techniques, including p53 gene mutation and aberrant methylation at exon 1 of the p16INK4a gene. The clinicopathologic spectrum of this subgroup was analysed and compared to patients without DNA bearing tumour characteristics. A group of 62 patients with breast cancer, who were selected sequentially after mastectomy and diagnosis of breast carcinomas, were analysed for 13 distinct clinicopathologic parameters: 56 cases (90%) with at least 1 molecular event in tumour DNA and 41 cases (66%) with a similar alteration in plasma DNA were identified. Comparison of clinicopathologic parameters between patients with and without plasma DNA revealed significant differences in relation to axillary involvement, rate of invasive ductal carcinoma, proliferative index, lymph node metastases, histologic grade and peritumoural vessel involvement. A high proportion of breast cancer patients at diagnosis exhibited high plasma DNA levels, and their presence correlated significantly with pathologic prognostic factors. Silva et al.17 also investigated DNA alterations before and after mastectomy in the plasma of breast cancer patients without disseminated disease as well as the relationship of these changes to specific tumour parameters. DNA was extracted from tumour tissues, normal breast tissues, mononuclear blood cells and plasma in 41 patients. Alterations in the microsatellite markers D17S855, D17S654, D16S421, TH2, D10S197 and D 9S161 as well as point mutations in the p53 gene and aberrant methylation of p16INK4a were used to identify and characterise tumour and plasma DNA. Eighteen of 27 patients with alterations in tumour DNA (44%) had identical plasma DNA alterations before mastectomy, and persistence of the same molecular features was detected in plasma DNA 4–6 weeks after mastectomy in 8 patients (20%). The presence of vascular invasion, lymph node metastases and high histologic grade at diagnosis were associated with alteration in plasma DNA following mastectomy; and the authors concluded that persistence of mutant plasma DNA was related to a poor prognosis, suggesting the presence of undetectable micrometastatic disease. As previously observed with colorectal and lung cancers, detection of circulating tumour RNA might also be a valuable method to improve the detection rate of breast cancer. Investigators used 6 microsatellite markers, alterations in p53 and methylation patterns of the first exon of p16 to detect plasma DNA mutations in 66% of breast cancer patients bearing a tumour-related DNA alteration.28 However, RNA markers appeared more sensitive, with 27/45 cases (60%) positive for mammaglobin and 22/45 (49%) positive for CK19. Moreover, tumour size and proliferative index were associated with the presence of mammaglobin, CK19 or both RNAs in plasma.18 Other RNA markers have also been found in the plasma and serum of breast cancer patients. Using the 2 telomerase mRNA markers, hTR and hTERT, RNA expression was detected in the plasma of 44% of breast cancer patients.30 Overall, telomerase RNA appears to be a promising breast cancer marker since it can be found in the serum of both small, localised and highly differentiated breast tumours. The above study used a semiquantitative method to detect hTR and hTERT, in contrast to the real-time RT-PCR method used by Dasi et al.16 on colorectal cancer and lymphoma cases. Thus, it is possible that further breast cancer studies using real-time assays could increase the sensitivity of detection. PSA is widely used as a screening and follow-up test for prostate cancer. However, like other antigenic markers, it results in a high rate of and a reliable blood-based assay for would be molecular markers, promoter hypermethylation of the gene is the most DNA alteration in this DNA alteration can be detected by MSP, this has been to promoter hypermethylation has been investigated in DNA isolated from and following prostate as well as from prostate tissues of patients with cancer and patients with were analysed on an gene hypermethylation was undetectable in tissue and from patients with it was found in of tumours of plasma/serum samples 50% of samples and of samples of prostate cancer patients. Additionally, identified circulating tumour cells in of prostate cancer patients. Analysis of promoter hypermethylation by appears to be a relatively specific tool for the molecular diagnosis of prostate cancer in The from molecular studies on plasma nucleic acids to suggest that it would be possible to a simple blood test to help each of the 4 cancers in the to the present it might be possible to detect over of patients with these using a of appropriate RNA and DNA markers. However, further is to present semiquantitative and quantitative identify DNA and RNA tumour markers and present for nucleic acids from In addition, the promising data suggest that it would be to clinical These would molecular assays with the available conventional would not only to molecular diagnosis and follow-up but might also us with the and of these common cancers.