Fax: (011) 86-571-87951676
A study of the volatile organic compounds exhaled by lung cancer cells in vitro for breath diagnosis
Article first published online: 28 JUN 2007
Copyright © 2007 American Cancer Society
Volume 110, Issue 4, pages 835–844, 15 August 2007
How to Cite
Chen, X., Xu, F., Wang, Y., Pan, Y., Lu, D., Wang, P., Ying, K., Chen, E. and Zhang, W. (2007), A study of the volatile organic compounds exhaled by lung cancer cells in vitro for breath diagnosis. Cancer, 110: 835–844. doi: 10.1002/cncr.22844
- Issue published online: 2 AUG 2007
- Article first published online: 28 JUN 2007
- Manuscript Accepted: 29 MAR 2007
- Manuscript Revised: 23 MAR 2007
- Manuscript Received: 10 OCT 2006
- National Natural Science Fund of China. Grant Number: 30411120602
- Zhejiang Province Science and Technology Department. Grant Number: 2002C33004
- lung cancer;
- early detection;
- breath diagnosis;
- cell microenvironment;
- electronic nose
The specific volatile organic compounds (VOCs) exhaled by lung cancer cells in the microenvironment are the source biomarkers of lung cancer and also serve as direct evidence that the diagnosis of lung cancer by breath is possible. However, to the authors' knowledge, few articles published to date have provided accurate VOCs in the microenvironment, thereby leading to different points of view with regard to searching for biomarkers in the breath from lung cancer patients In this article, an innovative pathologic analysis method of lung cancer and the early diagnosis of lung cancer at the cellular level were introduced for this purpose.
Solid-phase microextraction combined with gas chromatography is used as the detection system to determine the VOCs in the culture medium of several target cells, including different kinds of lung cancer cells, bronchial epithelial cells, tastebud cells, osteogenic cells, and lipocytes. As a result, each kind of cells has a unique chromatogram. There are 4 special VOCs that were found to exist in all culture mediums of lung cancer cells, which are the metabolic products of lung cancer cells and can be viewed as markers of lung cancer.
The authors were able to determine a correlation between VOCs in the metabolic products of lung cancer cells and VOCs in the breath of lung cancer patients, some of whom had stage I and II disease, and eventually hope to certify the biomarkers in the breath of lung cancer patients.
This research is significant and provides the basis for the noninvasive detection and the breath diagnosis of lung cancer using an electronic nose. Cancer 2007; 110:835–44. © 2007 American Cancer Society.
Human breath reportedly contains >200 VOCs, which can revel a great deal about the human body.1–3 In individuals with lung cancer, diabetes, nephropathy, or metabolic disturbance, the VOCs in the breath can indicate organ dysfunction.4 Many recent articles have supported this concept. In 1999, Phillips et al5 first published their research results concerning the correlation between breath odors and lung cancer and indicated that 22 VOCs had been regarded as markers of lung cancer. Many scientists also are concerned with the microenvironment in which the cells live. Cells communicate with each other through information that is transmitted in the microenvironment and certain types of diseased cells can spread the sickness throughout the microenvironment and, in some cases, the blood.6–10
However, to our knowledge, there are few articles published to date regarding the correlation between the statistical results of the biomarkers in the breath and biomarkers in the microenvironment, especially in the case of lung cancer. Therefore, the noninvasive method of breath diagnosis for lung cancer cannot be supported by cellular evidence. Since 1994, we have been conducting research on an electronic nose with which to diagnose diseases. In 1997, we published what to our knowledge is the first article11 concerning the diagnosis of diabetes using a gas sensors array. We then tried to develop a noninvasive and more convenient method with which to diagnose diabetes and lung cancer12 and in 2005 published an article13 regarding the noninvasive diagnosis of lung cancer based on a virtual surface acoustic wave (SAW) gas sensors array and imaging recognition. These results and conclusions all depend on pathologic statistical methods with which to determine the biomarkers.
Where do these VOCs come from? Why can just these VOCs be considered as biomarkers and how do they reach the lung and be exhaled in the breath? The answers to these questions remain unknown. In the current study, an innovative method of lung cancer pathologic analysis and early diagnosis at the cellular level is introduced. We used the solid-phase microextraction combined with gas chromatography (SPME-GC) detection system to analyze the culture medium of several target cells, including different kinds of lung cancer cells, bronchial epithelial cells, tastebud cells, osteogenic cells, and lipocytes. Compared with the results of the analysis, 4 VOCs can be observed in the culture medium of all lung cancer cells, which are the metabolic product of lung cancer cells and can be viewed as markers of lung cancer at the cellular level. This provide us with a basis with which the electronic nose can monitor health and even perform a noninvasive diagnosis.14–17 Fortunately, we also found evidence that the breath diagnostic method is suitable for those lung cancer patients with stage I and stage II disease.
METHODS AND MATERIALS
We cooperated with the Run Run Shaw Hospital in Hangzhou, China and obtained a total of 49 breath samples and 16 lung tissues from 65 volunteers.
The 49 breath samples came from 13 healthy persons, 29 lung cancer patients, and 7 patients with chronic bronchitis. All 29 lung cancer patients were clinically diagnosed using various diagnostic methods including bronchoscope biopsy, computed tomography (CT) scan (pulmonary puncture), lymph node biopsy, ultrasound (pulmonary puncture), and surgeryThe results of the clinical diagnosis demonstrated that 19 patients had adenocarcinoma, 6 patients had squamous cell carcinoma, and 4 patients had small cell carcinoma. Breath samples were collected in 10 patients prior to surgery. The lung tissue of these 10 patients was then used for the cell culture experiments. Nine lung cancer patients had received chemotherapy before their breath samples were collected. All patients and healthy subjects ate nothing 2 hours before the breath test. For the test, they followed the 3 steps to breathe into the Tedlar bags: 1) take a deep breath, 2) hold it for 1 second, and 3) exhale into the bag. The entire hospital is under constant temperature and humidity control. There are no flowers and plants in the rooms, so the VOCs in the breath bags could be considered as originating from the subjects themselves.13
The 16 lung tissues were obtained from patients undergoing thoracic surgery. The patients' information is listed in Table 1. We selected the edge of each tumor tissue for cultivation because it is believed to grow faster and the cells have more energy for cleavage compared with the center part of the tumor tissue. The excised tissue, from which excess connective tissue had been removed, was then rinsed in phosphate-buffered saline (PBS) several times and cut into pieces using ocular scissors. The PBS (pH of 7.4 ± 0.4) contained 8.00g/L of NaCl, 0.20g/L of KCl, 1.38g/L of Na2HPO4, and 0.20g/L of KH2PO4. The pieces, which measured approximately 1 mm3 when prepared for cultivating, were rinsed in PBS 3 times and digested with 0.25% trypsin (Sigma Chemical Company, St. Louis, MO) for 45 minutes at 37°C. The cells were freed with 3 attempts at gentle agitation, for 15 minutes each time. Ten percent fetal bovine serum was added to neutralize the trypsin. The solution was filtered (with a 200-mesh filter) and collected. The collected cells were washed with PBS, suspended, and counted with a hematometer to achieve a solution (containing Dulbecco modified Eagle medium-H [DMEM-H from Gibco BRL, Inc.] added to 10% fetal calf serum, 100 U/mL Penicillin, and 100 U/mL streptomycin) at a density of 4-5*105 cells/mL. Each 10mL of this solution was plated into 3 dishes measuring 70 mm in greatest dimension for parallel cultivation. The dishes were sterilized for 20 minutes at 120 to 160°C and prepared by coating 0.01% Poly-L-Lysine solution (Sigma Chemical Company) at the bottom and then drying at 37° and 5% carbon dioxide. The cells in the dishes were cultivated at 37°C in a humidified environment containing 5% carbon dioxide for 10 to 12 days. After that, the culture medium was then removed and collected in 3 sterilized glass bottles for the test and 10 mL each of fresh culture medium was added into the 3 dishes for another 8 to 10 days of cultivation. The cells were surviving in the culture medium for a total of 18 to 22 days. During this period of time, we used a microscope to examine the cells each day and the culture medium in each dish was replaced once, collected, and tested twice.18, 19
|Sample no.||Gender (no.)||Age, y||Category, description|
|1–15 (lung cancer)||M (11)||41–67||Squamous cell carcinoma (8 patients), adenocarcinoma (2 patients), bronchioloalveolar carcinoma (1 patient)|
|F (4)||52–65||Squamous cell carcinoma (1 patient), nonsmall cell carcinoma (1 patient), adenocarcinoma (1 patient), bronchioloalveolar carcinoma (1 patient)|
|16 (Healthy)||F||53||Chronic bronchitis and emphysema|
All the other normal cells including human bronchial epithelial cells, rat tastebud cells, osteogenic cells, and lipocytes were prepared for a control experiment.20–22 These cells were obtained from Zhejiang Run Run Shaw Hospital and washed, resuspended, and counted to obtain a single cell suspension at a density of 2-5*105 cells/mL. Approximately 10 mL of this suspension was plated into a dish measuring 70 mm in greatest dimension, as mentioned above. At approximately 50% confluence, the culture medium was removed and 10mL of fresh culture medium, which contained DMEM-H, 10% fetal calf serum, 100 U/ml of penicillin, and 100 U/ml of streptomycin was added to the dish for an additional 8 to 10 days of cultivation. The medium was then collected and tested as the control sample.
The SPME technique23, 24 is used as the preconcentration step to adsorb the VOCs in samples. Because there are 2 different types of samples (the breath sample and cellular culture medium sample), we used different methods to preconcentrate.
Two-bag system and SPME for breath samples
Two Tedlar bags and SPME techniques are used for preconcentration, as shown in Figure 1. A Tedlar bag with 5L of breath and another vacuum Tedlar bag are connected to each side of the pump and all stainless steel tubes are filled with nitrogen gas before preconcentration. A manual SPME holder with an extraction fiber coated with polydimethylsiloxane (PDMS) is then inserted into the tube and the pump works to bring the breath into the vacuum bag. The VOCs extracted by SPME will get into partial equilibrium with the VOCs in the flow. To improve extraction efficiency, the 2-bag system is held at a constant temperature (40°C) for approximately 1 hour.
Headspace SPME adsorption of the cell culture medium
An optimized headspace SPME (HS-SPME) method is used to adsorb the VOCs in the cell culture medium. Approximately 30 mL of fresh cell culture medium and a 10-mm long magnetic rotor are put into a 100-mL clean glass bottle with a cover on. The glass bottle with the SPME is then inserted in and placed on the working platform for 40 minutes of adsorption. The working platform is used to control the rotating velocity at approximately 1100 revolutions per minute (rpm) and provide the constant adsorption temperature at 37°C, as shown in Figure 2.
Detected by GC
After preconcentration, the extracted fiber in the manual SPME holder is inserted into the injector of GC-17A, which is set to 260°C for 10 minutes of desorption and works at the splitless model for 10 minutes. The oven temperature was programmed for 1 minute at 40°C, then 5°C per minute to 250°C, and finally at 250°C for 1 minute. Therefore, the VOCs can go through a DB-1 fused-silica capillary column (measuring 0.25-μm thick, 0.25-mm inner diameter × 30 minutes; purchased from Sigma–Aldrich, Inc) for separation and reach the flame ionization detector (FID) 1 by 1.25, 26 The column pressure is set to 160 kilopascals (kPa) and the flow rate is 35mL/minute. After comparison with the standard modular set, the molecular structure of the VOC is determined.
Experiment system optimizing
Extraction condition determination
Several conditions, including the type of fiber coatings, extraction time, humidity, and temperature, can influence the efficiency of SPME extraction.27 Because the VOCs in human breath are nonpolar, PDMS is used as the fiber coating. We compared the extraction results under a temperature of 37°C and 18°C for 40 minutes each, as shown in Figure 3a. The temperature of 37°C was found to be better. Because it was the culture temperature, 37°C was chosen as the final extraction temperature. The extraction time is then determined by the results of compared experiments, under the temperature of 37°C ± 0.5°C, using the extraction times of 20 minutes, 40 minutes, and 60 minutes for 3 times each, as shown in Figure 3b. The result demonstrates that the equilibrium between the sample and fiber was established within 40 minutes. With additional extraction time, there was a small increase noted in the peak area. The efficiency of the extraction time was considered and 40 minutes was chosen as the extraction time. Because high humidity causes a decrease in the SPME adsorption, the distance between the fiber and the surface of the solution is carefully calculated to achieve good repeatability.
FID, purchased from Sigma–Aldrich Inc, is equipped as the current detector. For its selectivity, it can cut the response of some components such as nitrogen, air, etc that we do not want. Three concentrations (400 μg/L, 800 μg/L, and 1200 μg/L) of each component were prepared for SPME-GC calibration to determine the retention time and coefficient k. The calibration results are shown in Table 2. These 11 VOCs were used as the biomarkers of lung cancer in breath diagnosis and were proven by the previous statistic study.13
|Chemical names of VOCs||Mean retention time ± 0.020, min||k = Concentration/peak area||Breath diagnosis cutoff (g/mL) (peak area = 200)|
Reproducibility of Cell Culture
We tested the cell medium for each of the 3 groups (dishes) to monitor the reproducibility of VOCs in cell culture and found that there was no obvious difference in these 3 tests. When comparing the first 10 to 12 days of cultivation with the total 18 to 22 days of cultivation of each group (dish), the density of cells/cm2 had obviously increased (as observed microscopically). The HS-SPME-GC test results also showed that there was an increase in the amount of VOCs. The first 10 to 12 days'' cultivated medium contained fewer VOCs than the following 8 to 10 days' cultivated medium. For this reason, the culture time was controlled and modified to achieve good reproducibility of the cell culture.
Pathologic study of VOCs in the microenvironment in vitro
Among the collection samples, we found a total of 4 types of lung cancer cells: squamous cell carcinoma, adenocarcinoma, bronchioloalveolar carcinoma, and nonsmall cell carcinoma. After the cultivation, the cells' medium was detected using the method mentioned above. Because the 11 VOCs we expected to find in culture medium could quickly pass through the capillary column, the chromatograms in the first 15 minutes were chosen. Although a type of cancer cell came from different tissue samples, the VOCs in the culture medium were nearly the same. Compared with Figure 4, we were able to conclude that the VOCs in the culture medium of lung cancer cells obviously differ from that in the virgin culture medium. Figure 5 shows the results of the experiment involving the lung cancer cells and that of the control cells, including tastebud cells, osteogenic cells, lipocytes, and bronchial epithelial cells. According to Figures 4 and 5, it could be proven that the lung cancer cells had the unique VOCs, which were the metabolic products of the lung cells and could be regarded as the biomarkers.
During the experiments we found that there were different VOCs within different types of lung cancer cells. However, some common VOCs in these 4 lung cancer cells were observed, as shown in Figure 4. They were marked 1, 2, 3, and 4 and are believed to be the metabolic products of the lung cancer cells existing in the microenvironment and viewed as the general biomarkers of lung cancer at the cellular level. Conversely, the different VOCs (the retention times of which were between 1.7 minutes and 3.6 minutes) might be the unique biomarkers of each type of lung cancer.
We obtained evidence to prove that lung cancer tissues from patients with stage I and stage II disease had the same VOCs in the microenvironment as observed in tissue from stage III and stage IV lung cancer. This provided the opportunity to detect lung cancer in patients with stage I and stage II disease using breath diagnostic methods. A normal lung tissue that was undetectable on macroscopic examination was prepared using the method mentioned above to obtain the control results. However, the comparison result in Figure 6 shows that the tissue appears to contain lung cancer cells. The VOCs were nearly the same whereas only the concentrations demonstrated any difference. After 2 weeks of cultivation, cancer cells were observed by microscope, as shown in Figure 7. This demonstrated that the microenvironment changes a great deal whereas only a few cancer cells exist in the tissue that was proven surgically to be healthy. This phenomenon provides evidence at the cellular level that can provide the possibility of using the study data from lung cancer patients with stage III and stage IV disease to detect those patients with stage I and stage II disease.
Pathologic study of VOCs in the breath
We collected breath samples from study participants at Run Run Shaw Hospital. At the time of last follow-up, we had a total of 49 volunteers, including 13 healthy persons, 29 lung cancer patients, and 7 patients with chronic bronchitis. All the patients were age >50 years with the exception of 1patient aged 47 years. Only 3 of the lung cancer patients had nothing detected in their breath sample. As a control experiment, the breath of 7 patients with chronic bronchitis and 13 healthy persons was examined as well. The amounts of the 11 characteristic VOCs (listed in Table 1) in their breath were less than those noted in the patients with lung cancer.13
Table 3 shows a statistical study of lung cancer patients compared with healthy persons and those patients with chronic bronchitis. Using the 11 characteristic VOCs as the breath biomarkers to separate lung cancer from noncancerous conditions and comparing the results with the clinical results, we detected a sensitivity of 86.2% for patients with lung cancer, as well as a specificity of 69.2% in healthy persons and a specificity of 71.4% in patients with chronic bronchitis. The accuracy of the test was approximately 79.6%. We assumed the incidence of lung cancer was 59.2% according to this test. The positive predictive value was 80.6% and the negative predictive value was 77.8%.
|No.||Clinical diagnosis results (no.)||Breath diagnosis results (no.)||Sensitivity and specificity||Positive predictive value||Negative predictive value|
|1||Adenocarcinoma (19)||Lung cancer (17)||Lung cancer sensitivity, 86.2% (25/29)||80.6% (25/31)||77.8% (14/18)|
|Small cell carcinoma (4)||Lung cancer (4)|
|Squamous cell carcinoma (6)||Lung cancer (4)|
|2||Healthy (13)||Noncancerous (9)||Specificity, 70% (14/20)|
|Lung cancer (4)|
|Chronic bronchitis (7)||Noncancerous (5)|
|Lung cancer (2)|
The rule of breath diagnosis is that if a person's breath contains ≥1 of the 11VOCs with a concentration that is higher than the breath diagnostic cutoff listed in Table 2 (the peak area of the VOCs is >200), the patient is regarded as a lung cancer patient. Otherwise, the patient is considered to have a noncancerous condition.
There were 6 lung cancer patients (including 3 patients with stage I and stage II disease) whose VOCs were near the breath diagnostic cutoff point shown in Table 2 (The peak area between 50 and 400). According to the breath diagnosis rule, 5 of these patients were diagnosed correctly whereas only 1 patient was diagnosed incorrectly. There were 6 healthy individuals whose VOCs were within this range. Only 2 of these subjects were diagnosed correctly while the diagnosis of 4 patients was found to be wrong. In addition, there were 6 patients with chronic bronchitis whose VOCs were also within this range; 4 were diagnosed correctly while 2 patients received an incorrect diagnosis. Using the breath diagnosis rule to diagnose those persons whose VOCs were within this range, the total correct rate dropped to 61.1% (11 of 18 patients) and the specificity dropped to 50% (6 of 12 patients). Therefore, principal component analysis (PCA) was performed to study the differentiation between the lung cancer patients and healthy individuals at this range and to prove that it is possible to improve the sensitivity and specificity using an optimized diagnosis algorithm, which is commonly used with the electronic nose.
Figure 8 shows the PCA results of these 18 samples. The 11 characteristic VOCs in the breath from these individuals were calculated using PCA. PC1 and PC2 were used to differentiate between lung cancer patients, healthy individuals, and patients with chronic bronchitis. The results in Figure 8 demonstrate that it is easy to distinguish healthy individuals from patients with lung disease (lung cancer and chronic bronchitis). Because the breath sample from some of the lung cancer patients was found to contain some of the same VOCs as the breath from chronic bronchitis patients, 1 patient with lung cancer was regarded as a chronic bronchitis patient according to the PCA results of this small sample set. In this way, we cannot only distinguish healthy individuals from those with lung diseases but we also now have a tool with which to compensate for the shortage of the breath diagnosis rule when it is applied at the range of 50 to 400 peak area.
Correlation between VOCs in the Microenvironment In Vitro and VOCs in Patients' Breath
When the retention time of the VOCs in culture medium is compared with that of VOCs in patient' breath, we find the Peak 1 (Fig. 4; retention time of 1.257 minutes) in the culture medium of squamous cell carcinoma cells appears similar to isoprene (retention time of 1.290 minutes) and Peak 4 (Fig. 4; retention time of 10.248 minutes) in the culture medium of nonsmall cell carcinoma cells is close to that of undecane (retention time of 10.560 minutes). The peaks for which the retention time is shown as being between 1.7 minutes and 3.6 minutes in Figure 4 are the characteristic VOCs of various lung cancer cells and, to our knowledge, their modular structures have not yet been determined. Three kinds of the lung cancers and the virgin culture medium are shown in Figure 4 as having Peak 3 (retention time of 8.557 minutes) and, because the healthy cells shown in Figure 5 do not have Peak 3, it appears that this component in the virgin culture medium is a nutriment that is required for growth by healthy cells but is not so urgently needed by cancer cells. Unfortunately, to our knowledge, the modular structure of Peak 3 is not known. Isoprene can be found in the majority of breath samples from lung cancer patients, as well as in breath samples from healthy individuals, although at a lower concentration in the healthy person. This phenomenon also can be found in the detection results of microenvironment, as shown in Figures 4 and 5 with Peak 1.
In this article, an innovative method of lung cancer pathologic analysis and early diagnosis at the cellular level was introduced. An optimum HS-SPME-GC method was used to detect the odors of some cells in human breath, including squamous cell carcinoma cells, adenocarcinoma cells, bronchioloalveolar carcinoma cells, nonsmall cell carcinoma cells, bronchial epithelial cells, tastebud cells, osteogenic cells, and lipocytes, and we were able to obtain an odor chromatogram of these cells. The results of the current study demonstrated that each type of cell has a unique chromatogram. There were 4 components in all 4 types of carcinoma that we believe could be regarded as the biomarkers of lung cancer at the cellular level. Conversely, the breath of 49 individuals, including 29 lung cancer patients, 7 patients with chronic bronchitis, and 13 healthy individuals, also were submitted for pathologic analysis. Eleven VOCs were selected as biomarkers of lung cancer in breath diagnosis. When the 4 components were compared with the 11 calibrated VOCs, 1 component appeared to be similar to isoprene, whereas another resembled undecane. These 11 VOCs could be found in the breath of lung cancer patients but were rarely detected in breath samples from healthy persons.
However, the differences between the VOCs present in the breath samples from lung cancer patients and the culture medium of lung cancer cells demonstrated the complexity of the transmission mechanism. Further study should be performed to determine the exact modular structure of the 4 components and the characteristic VOCs of various lung cancer cells for which the retention time is shown as being between 1.7 minutes and 3.6 minutes (Fig. 4) and to discover the mechanism of transmission.
- 172000; Paper No. 2aEA4. Available at http://www.aromako.co.kr/aroma/tech/pdf/asa2000.pdf.. The zNose: A New Electronic Nose Using Acoustic Technology. The Journal of the Acoustical Society of America Paper No: 2aEA4. City of Publication: Publishing organization;
- 22Culture of Animal Cells. A Manual of Basic Techniques. 4th ed. New York: Wiley-Liss; 2000..
- 23Solid Phase Micro Extraction. Theory and Practice. New York: John Wiley & Sons; 1997..