Non-blood circulating tumor DNA detection in cancer

Tumor DNA contains specific somatic alterations that are crucial for the diagnosis and treatment of cancer. Due to the spatial and temporal intra-tumor heterogeneity, multi-sampling is needed to adequately characterize the somatic alterations. Tissue biopsy, however, is limited by the restricted access to sample and the challenges to recapitulate the tumor clonal diversity. Non-blood circulating tumor DNA are tumor DNA fragments presents in non-blood body fluids, such as urine, saliva, sputum, stool, pleural fluid, and cerebrospinal fluid (CSF). Recent studies have demonstrated the presence of tumor DNA in these non-blood body fluids and their application to the diagnosis, screening, and monitoring of cancers. Non-blood circulating tumor DNA has an enormous potential for large-scale screening of local neoplasms because of its non-invasive nature, close proximity to the tumors, easiness and it is an economically viable option. It permits longitudinal assessments and allows sequential monitoring of response and progression. Enrichment of tumor DNA of local cancers in non-blood body fluids may help to archive a higher sensitivity than in plasma ctDNA. The direct contact of cancerous cells and body fluid may facilitate the detection of tumor DNA. Furthermore, normal DNA always dilutes the plasma ctDNA, which may be aggravated by inflammation and injury when very high amounts of normal DNA are released into the circulation. Altogether, our review indicate that non-blood circulating tumor DNA presents an option where the disease can be tracked in a simple and less-invasive manner, allowing for serial sampling informing of the tumor heterogeneity and response to treatment.


INTRODUCTION
Gene mutation and methylation play essential roles in tumorigenesis and metastasis [1][2][3]. Tumor DNA contains specific somatic alterations that are crucial for the diagnosis and treatment of cancer. Due to the spatial and temporal intra-tumor heterogeneity, multi-sampling is needed to adequately characterize the somatic alterations [4][5][6]. Tissue biopsy, primarily the "gold standard" for diagnosis, however, is limited by the restricted access to sample and the challenges to recapitulate the tumor clonal diversity [7,8]. Plasma circulating tumor DNA (ctDNA) is DNA fragments that contain the tumor-specific somatic alterations in the blood. Recent studies have demonstrated the application of plasma ctDNA detection in the tumor diagnosis and monitoring [9][10][11]. It is of non-invasive and can be collected repeatedly with minimal discomfort to the patient. It also reflects the total tumor burden and genetic heterogeneity.
Non-blood circulating tumor DNA are tumor DNA fragments presents in other body fluids, such as urine, saliva, sputum, stool, pleural fluid, and cerebrospinal fluid (CSF). Recent studies have demonstrated the presence of tumor DNA in these non-blood body fluids and their application to the diagnosis, screening, and monitoring of cancers [12] (Figure 1). The collection of these body fluids, such as urine, saliva, sputum, and stool, are relatively safe, non-invasive, economic and can be performed at home, without professional help [13,14]. It also applies perfectly to patients with anemia, which is quite common in Review advanced stage cancer patients. Moreover, enrichment of tumor DNA of local cancers in non-blood body fluids may help to archive a higher sensitivity than in plasma. Nonblood body fluids are a viable alternative to blood samples as a source of DNA for tumor diagnosis and monitoring.

Sources of non-blood tumor DNA
There are 2 types of non-blood tumor DNA: genomic DNA from local tumor cells that shed into the body fluid (cellular tumor DNA) and cell-free tumor DNA (cfDNA) from plasma cell-free DNA or from neighboring tumor cells due to necrosis or apoptosis ( Figure 2).
Urinary cfDNA mainly originates from plasma cfDNA that pass the kidney barrier [15]. Circulating tumor cfDNA is highly fragmented and primarily present in the blood as part of super molecule complexes, such as nucleosomes. During circulation, cfDNA was filtrated from the blood into the primary urine through the kidney barrier, which has been proved to be permeable to DNA molecules [16]. However, it is not known of the mechanism of DNA translocation from the bloodstream into the urine. Limited by the basal membrane and slit membranes between podocytes pedicles, only complexes smaller than 6.4 nm in diameter and with a molecular weight no greater than 70 kDa can pass through the kidney barrier and enter the nephron. It corresponds to DNA of about 100 bp in size, which is smaller than mononucleosome [15]. Moreover, the negatively charged cfDNA might face an additional barrier because of the negative charge of the glomerular basement membrane. It might be due to the non-globular shape or by the deformability of the DNA complexes. Another explanation is that cfDNA may be covered by liposomes, which make their penetration through the kidney barrier theoretically possible [17,18]. Besides, renal permeability might increase for some physical and pathological conditions, such as pregnancy, cancer, and inflammation.
Ying et al. reported two size categories of urinary DNA: low molecular weight (MW) urine DNA and high MW urine DNA. Low MW class of urine DNA is between 150 to 250 bp and derived from the circulation, while the high MW urine DNA is greater than 1 kb and mostly from the cells shed into the urinary tract [13]. Similar findings were also reported in CSF. Size distribution of CSF cfDNA peaks at 160 and 340bp indicates an apoptotic source. CSF cellular DNA is of larger size and may origin from blood cells and tumor cells in CSF [19]. Saliva DNA can also be locally generated by cell necrosis, apoptosis, and exfoliation, or by active transport, passive diffusion or ultrafiltration from the plasma cfDNA [20]. The possible origins of pleural DNA also include ultrafiltration from the plasma (cfDNA) and local dying or apoptotic cells (cellular DNA) [21].
Urinary cfDNA can be detected in non-urothelial carcinoma patients. The p53 mutation was found by Selena et al. in the urine of hepatocellular carcinoma patients and most of the p53 mutation was detected in the low MW urine DNA fraction [57]. Chen et al. detected EGFR mutations in urinary cfDNA in non-small cell lung cancer (NSCLC) patients with a concordance of 88% to their primary tumors [55].
Su et al. demonstrated that DNA methylation in urine is better in predicting recurrence than cytology and cystoscopy in bladder cancer patients after transurethral resection. By using a three-marker panel (SOX1, IRAK3, and L1-MET), they could predict tumor recurrence in 80% of patients, which is superior to cytology (35%) and cystoscopy (15%) [58]. Reckamp et al. detected the EGFR activating mutations and the T790M resistance mutation in urine and plasma of NSCLC patients by short footprint mutation enrichment next generation sequencing assays. They found that with a recommended specimen volume (90-100 mL), the sensitivity of urine and plasma are comparable [54]. Chen et al. conducted a serial monitoring trial to detect urinary cfDNA of EFGR mutation in NSCLC patients receiving EGFR-TKIs. They found a concordance in the quantity of urinary cfDNA and plasma cfDNA at baseline. During the treatment, a concordance in the decline of the quantity of cfDNA was also observed. Nevertheless, a more significant decrease in urinary cfDNA than plasma cfDNA was found during the early phase of monitoring as a result of treatment, which indicated that urinary cfDNA might potentially be of higher sensitivity [55].

Saliva tumor DNA
Saliva provides good-quality genomic DNA, which is comparable to blood as a template for genotyping [59,60]. Salivary DNA has been used in germline mutations detection for various cancer screening, such as breast cancer [61] and braintumor [62].   showed a statistically significant correlation in loss of heterozygosity (LOH) between saliva and tumor with some sets of markers [64]. Salivary DNA promoter hypermethylation analysis has also been found to be an efficient tool for diagnosis of HNSCC [65][66][67] (  [14]. It is therefore suggested that there is a link between the peripheral circulatory system and the salivary glands that translocate cfDNA from the bloodstream into saliva [14].

Sputum tumor DNA
Sputum contains cells from the lungs and lower respiratory tract and provides sufficient tumor DNA for detection. Numerous studies have shown that sputum tumor DNA could be a promising tool for early detection of lung cancer (Table 3). In 1994, Mao et al. detected the K-ras mutation and p53 mutation in sputum samples of lung cancer patients. Using a PCR-based assay, they detected mutant DNA in the sputum of 8 of 10 patients with oncogene mutations in their primary tumor prior to clinical diagnosis [88]. DNA hypermethylation in sputum also helps the diagnosis of lung cancer [89]. Wang et al. carried out a meta-analysis to comprehensively review the evidence for using sputum aberrant methylation DNA to detect NSCLC. They found that the combined sensitivity was 62% (95% CI: 0.59-0.65), and specificity was 73% (95% CI: 0.70-0.75) [90]. Miglio et al. demonstrated that MGMT promoter methylation was present in small cell lung cancer and cytological samples were perfectly adequate for methylation analysis [91]. Flow cytometric DNA analysis of sputum cells also showed good sensitivity in the diagnosis of lung cancer. Compared with cytologic morphology of sputum cells, sensitivity of sputum DNA heteroploidy analysis was significantly higher (82.8% vs 27.6%, P < 0.005) [92].
DNA analysis from sputum was consistent with that from plasma in patients with lung cancer. By using a fluorescent PCR-based approach, Castagnaro et al. were able to assess the consistency of DNA microsatellite analysis of induced sputum. They demonstrated a significant trend in the percentage of the genetic alterations, found both in induced sputum and in blood samples, from healthy subjects to heavy smokers and lung cancer patients [93]. LOH and microsatellite instability (MSI) in at least one locus was observed in 55% of patients, in 18% of smokers, and in 4.5% of healthy subjects. These results showed that sputum DNA provided data that were consistent with those from plasma [93].

Stool tumor DNA
Stool is the fecal discharge from the bowels. It is biologically rational to use stool as a non-invasive sample for colorectal cancer (CRC). Stool DNA detection has a marked improvement of the sensitivity when compared to fecal blood tests. Compared to colonoscopy, which is currently the dominant screening test, stool DNA detection is patient-friendly and free from unpleasant cathartic bowel preparations and diet or medication restrictions. Furthermore, colonoscopy is operator dependent and has been shown to be not as effective detecting proximal lesions [94]. Stool DNA testing detects proximal and distal colorectal neoplasms equally well.
According to a meta-analysis involving 7524 patients, the pool sensitivities of stool DNA testing for CRC were 48% for single-gene and 77.8% for multiple-gene assays, and the specificities were 97% and 92.7%, respectively [95]. In a cross-sectional study including 9989 participants, Thomas et al. evaluated the effectiveness of multitarget stool DNA (mt-sDNA) test in colorectal-cancer screening. The mt-sDNA test included quantitative molecular assays for K-ras mutations, aberrant NDRG4, and BMP3 methylation, and β-actin, plus a hemoglobin immunoassay. The sensitivity for detecting colorectal cancer was significantly higher with mt-sDNA test than immunochemical test (FIT) (92.3% vs. 73.8%, p = 0.002) [96].
Besides CRC, stool tumor DNA can also be detected in other digestive system neoplasms. In 1994, Caldas demonstrated the presence of K-ras mutation in the stool of pancreatic cancer patients [97]. In a more recent work, Kisiel et al found a sensitivity of 67% and a specificity of 90% in detecting pancreatic cancer with a combination of stool mutated K-ras and methylated BMP3 detection [98].
Plasma based DNA tests, especially marking the aberrant methylation of SEPT9 gene, have been evaluated as a potential screening tool for CRC and advanced adenomas [99]. Ahlquist et al. conducted a case-control study to compare the sensitivities of multimarker stool DNA test and plasma methylated SEPT9 test in identifying patients with large adenomas or CRC. Their results demonstrated that mt-sDNA test had a significantly greater level of sensitivity than the plasma methylated SEPT9 test for detection of both CRC and large adenomas (87% vs. 60% CRC sensitivity and 82% vs. 14% adenoma sensitivity) [100]. The high sensitivity of stool DNA test may be related to disproportionately copious exfoliation of cancer cells, remarkably large functional surface area of neoplasms, enhanced survival of shed dysplastic cells, and relative stability and informativeness of tumor-associated DNA changes [100].   and lung cancer. They found that the methylation ratios for the three genes were significantly higher in lung cancer than in MPM, which suggested that pleural fluid DNA could be a possible marker for differentiating MPM from lung cancer [104]. Using high resolution melting (HRM) analysis, Lin et al. assessed the pleural cfDNA and pleural cellular DNA of 13 known NSCLC cancer patients with EGFR mutation in matched biopsy tumor tissues, and found mutated EGFR in 12 and 9 of 13 cases, respectively [105]. Similar results were reported by Liu et al. using amplification refractory mutation system. Higher sensitivity of pleural cfDNA might be due to the tumor cells damaged under high speed of centrifugation, and DNA fragments were released from the nucleus, making up the dominant components of the supernatant [106].

CSF tumor DNA
Circulating tumor DNA has been detected in a variety of cancers. However, it is rarely found in patients with isolated brain tumors, presumably owing to the blood-brain barrier [107]. CSF is a clear, colorless body fluid that bathes the brain and spinal cord. It circulates nutrients and chemicals filtered from the blood and removes waste products from the brain. Examining the fluid can be useful in diagnosing many diseases of the nervous system, including brain tumors.
CSF tumor DNA provides a minimally invasive method to assess the genomic alterations of the tumor and monitor the therapy effect that helps both diagnosis and treatment. Using next generation sequencing approach, Pentsova et al. sequenced 341 cancer-associated genes in CSF of 53 patients with suspected or known CNS cancers. They detected high-confidence somatic alterations in 63% (20 of 32) of patients with CNS metastases of solid tumors, 50% (6 of 12) of patients with primary brain tumors, and 0% (0 of 9) of patients without CNS involvement by cancer [108]. Wang et al. found that all medulloblastomas, ependymomas, and high-grade gliomas that abutted a CSF space or cortical surface were detectable (100% of 21 cases; 95% CI = 88-100%), whereas no CSF tumor DNA was detected in patients whose tumors were not directly adjacent to a CSF reservoir [109].
In De Mattos-Arruda et al.'s seminal work, CSF ctDNA was identified in brain primary and metastatic tumors and represented private mutations from CNS lesions. Furthermore, sensitivity of ctDNA for somatic mutations of the CNS was higher than plasma DNA in patients with a CNS-restricted disease (58% vs. 0%, p = 0.0006). While in patients with the abundant visceral disease, the sensitivity of CNS DNA and plasma DNA was comparable (60.5% vs. 55.5%). The investigators further monitored the change of mutant allelic frequency (MAF) of CSF DNA and plasma DNA in a serial study. MAFs of CSF DNA decreased with surgical resection and/or responses to systemic therapy and increased with tumor progression [7]. Similar findings were also presented by Pan et al. The median concentration of cfDNA in CSF is lower than that in plasma (2.1 ng/mL vs. 7.7 ng/mL). However, the ability to detect mutations in CSF is stronger than in plasma in brain tumor patients with low systemic metastatic burden [19]. EGFR mutation in CSF was also detected in a case with suspected leptomeningeal metastasis from EGFR mutant lung adenocarcinoma, which indicates the characterization of brain tumor genomic aberrations through CSF DNA analysis is possible. Very few cells are present in CSF under routine conditions (0-5 cells/L). The scarcity of cells in CSF may reduce the background noise from normal DNA when detecting mutations [19].

CONCLUSIONS
Non-blood circulating tumor DNA has an enormous potential for large-scale screening of local neoplasms because of its non-invasive nature, close proximity to the tumors, easiness and it is an economically viable option. It permits longitudinal assessments and allows sequential monitoring of response and progression [110]. The direct contact of cancerous cells and body fluid may facilitate the detection of tumor DNA, while vascular invasion likely happens at a later stage in tumorigenesis, which may explain the low sensitivity of plasma-based tests [111]. Furthermore, normal DNA always dilutes the ctDNA, which may be aggravated by inflammation and injury when very high amounts of normal DNA are released into the circulation [112]. Altogether, our review indicate that non-blood circulating tumor DNA presents an option where the disease can be tracked in a simple and less-invasive manner, allowing for serial sampling informing of the tumor heterogeneity and response to treatment.

ACKNOWLEDGMENTS AND FUNDING
The authors greatly appreciate Dr. Shiyu Chen, for her kindly help with the figures of the manuscript.

CONFLICTS OF INTEREST
There are not any financial/commercial conflicts of interests involving this study.