Beruflich Dokumente
Kultur Dokumente
NeuroImage
j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / y n i m g
Sniffing out cancer using the JPL electronic nose: A pilot study of a novel approach to
detection and differentiation of brain cancer
Babak Kateb a,b,d,⁎, M.A. Ryan c, M.L. Homer c, L.M. Lara c, Yufang Yin d, Kerin Higa d, Mike Y. Chen d
a
International Brain Mapping and Intraoperative Surgical Planning Society (IBMISPS.org), USA
b
Brain Mapping Foundation, 8159 Santa Monica Blvd. #200, West Hollywood, CA 90046, USA
c
Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Dr. Pasadena, CA 91109, USA
d
City of Hope Cancer Center, 1500 East Duarte Road Duarte, CA 91010, USA
a r t i c l e i n f o a b s t r a c t
Article history: A proof-of-concept study was done to determine whether an electronic nose developed for air quality
Received 10 February 2009 monitoring at the Jet Propulsion Laboratory (JPL) could be used to distinguish between the odors of organ
Revised 1 April 2009 and tumor tissues, with an eye to using such a device as one of several modes in multi-modal imaging and
Accepted 2 April 2009
tumor differentiation during surgery.
Available online 9 April 2009
Hypothesis: We hypothesized that the JPL electronic nose (ENose) would be able to distinguish between the
Keywords:
odors of various organ and tumor tissues.
Electronic nose Materials and methods: The odor signatures, or array response, of two organs, chicken heart and chicken liver,
Sensor array and cultured glioblastoma and melanoma tumor cell lines were recorded using the JPL Electronic Nose. The
Brain tumor overall array responses were compared to determine whether they were sufficiently different to allow the
Sniffing cancer organs and cell lines to be identified by their array responses.
Cancer odor Results: The ENose was able to distinguish between the two types of organ tissue and between the two types
Glioblastoma of tumor cell lines. The variation in array response for the organ tissues was 19% and between the two types
Melanoma
of cultured cell lines was 22%.
Conclusion: This study shows that it is possible to use an electronic nose to distinguish between two types of
tumor cells and between two types of organ tissue. As we conducted the experiment with a sensor array built
for air quality monitoring rather than for medical purposes, it may be possible to select an array that is
optimized to distinguish between different types of cells and organ tissues. Further focused studies are
needed to investigate the odor signatures of different cells as well as cellular proliferation, growth,
differentiation and infiltration.
© 2009 Elsevier Inc. All rights reserved.
1053-8119/$ – see front matter © 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.neuroimage.2009.04.015
T6 B. Kateb et al. / NeuroImage 47 (2009) T5–T9
2008; Thaler and Hanson, 2005, Pearce et al., 2003). Currently, the JPL Table 1
ENose technology is being used by NASA on board the International Coatings placed on polymer-carbon black composite films.
Space Station to monitor breathing air for leaks and spills of targeted Poly-4-vinyl phenol (composite A)
species (Ryan et al., 2004a; Ryan et al., 2008). Poly-4-vinyl phenol (composite B)
As electronic noses have developed and their availability has Poly-2,4,6-tribromostyrene (composite A)
Poly-2,4,6-tribromostyrene composite B)
increased, use of these devices over a broad range of medical
Polystyrene-co-maleic acid
disciplines has become more common. Researchers have used Poly-2,2,2-trifluoroethyl methacrylate (composite A)
similar technologies to analyze the breath or urine of diabetic Poly-2,2,2-trifluoroethyl methacrylate composite B)
patients and have showed significant differences compared to Poly(t-butylaminoethyl methacrylate (composite A)
Polyethylene-co-acrylic acid (composite A)
controls (Yu et al., 2005; Mohamed et al., 2002). Others have
Polystyrene (composite A)
used electronic noses for detecting infections in the ear, nose, and Soluble polyimide (composite A)
throat with more than 88% accuracy (Shykhon et al., 2004). Ethyl cellulose (composite A)
Concordantly, Hanson et al. successfully correlated pneumonia to Polyamide resin (composite A)
electronic nose data over a 4 month period in 38 patients (Hanson Styrene/isoprene, 14/86 ABA block polymer(composite A)
Vinyl alcohol/vinyl butyral, 20/80 (composite A)
and Thaler, 2005). Likewise, Hockstein et al. correlated pneumonia
Ethylene-propylene diene terpolymer (composite A)
with electronic nose patterns in 44 patients and concluded that the
Different coatings on each element of the sensor array results in elements that have
technology had promise as a diagnostic adjunct in the management
overlapping affinities for volatile compounds.
of ventilator-associated pneumonia (Hockstein et al., 2005). Dutta et
al. were able to identify two species of Staphylococcus aureus in the
hospital environment with 99.96% accuracy (Dutta et al., 2005). The
technology has also been used for assessment of renal function array which had been optimized to detect targeted species on board
(Voss et al., 2005), oral malodor (Tanaka et al., 2004), body odor the International Space Station.
differences between genders (Penn et al., 2007), unpleasant odor in
coated pharmaceutical tablets (Ohmori et al., 2005), body odor of Samples
schizophrenic patients (Di Natale et al., 2005), skin odor sampling,
monthly variation of volatile compounds in female sweat (Mantini Two types of cell lines were used for the experiments, and each
et al., 2000) and distinguishing cerebrospinal fluid (CSF) from cell line was used to create two different types of samples: pellet
serum (Aronzon et al., 2005). samples and dish samples. The first type of sample is a cell sample
Another medical application that is being explored is the use of centrifuged into pellets. The second type of sample is plated cells in
electronic noses to detect cancer. Interestingly, the ability to detect a culture dish. A human glioblastoma (U251) and a human
a smell related to cancer has been described in canine studies melanoma (A2058) cell line were used to create the different
(Gordon et al., 2008). Willis, Pickel and McCulloch demonstrated in types of samples.
separate studies that dogs could be trained to detect bladder, skin In the first type of cell sample, the cells were centrifuged into
and lung cancers, respectively (Willis et al., 2004; Pickel et al., pellets and then transferred into 50 mL glass test tubes. Cell lines were
2004; McCulloch et al., 2006) Correspondingly, D'Amico showed maintained in 10-mL flasks in medium (Dulbecco's Modified Eagle's
that an electronic nose has the potential to distinguish melano- Medium, DMEM). Cells were plated onto 100 mm dishes 4 days prior
matous lesions from benign skin lesions (D'Amico et al., 2008). to the analysis. The cells were then trypsinized, neutralized with
Breath analysis using an electronic nose has also been used to medium, centrifuged into pellets and transferred into 50-mL test
identify chemicals in human breath related to lung cancer (Chen et tubes, stored for 2 hours at 37 °C, and then transported at room
al., 2007; Di Francesco et al., 2005; Machado et al., 2005; Di Natale temperature while kept in medium. Four preparations, each in
et al., 2003). Recently, investigators have demonstrated that tumor triplicate, were used: 3 × 105 A2058 cells, 1 × 106 A2058 cells, 3 × 105
cell lines, including adenocarcinoma, squamous cell carcinoma, and U251 cells, and 1 × 106 U251 cells. This gave us a total of 12 pellets for
mesothelioma, have distinct response patterns using an electronic testing.
nose, and that they may be distinguished from each other and from The second type of cell samples were made under the same
the normal fibroblast and smooth muscle cells (Gendron et al., conditions as the first except that plated cells were left adherent to the
2007). culture dishes. In addition to the cell line samples there were organ
Thus far, there are no published studies on the use of this samples. Chicken livers and hearts were ground into a paste and put
technology for detection and differentiation of brain cancers. We into culture dishes.
hypothesized that the Jet Propulsion Laboratory's (JPL) electronic nose
(ENose) would be able to distinguish between the odors of various ENose experimental configuration
organ and tumor tissues. In this preliminary, proof-of-concept study,
we investigated the odor signatures of individual organs of chicken We used two experimental configurations to test the “odor” of the
such as liver and heart as well as human glioblastoma and human different cells. In both configurations the air in the head space above
melanoma tumor cell lines. the cells was pulled into the ENose sensing chamber by a pump
incorporated into the ENose instrument. A Teflon-lined Tygon tube
Materials and methods was connected to the air inlet of ENose and held over the head space. A
sketch of the ENose is shown in Fig. 1.
Sensors In the pellet-test tube configuration (condition 1), the air tube
was inserted into the test tube 3 cm from the cell pellet and held
The JPL ENose detector is comprised of 16 sensors with uniquely in place with Parafilm. The film was not held tightly over the top of
coated polymer-carbon black composite films (Table 1). As the the test tube, so room air could replace the air removed by the
vapor environment across the sensors changes and vapors sorb into ENose pump.
and out of the polymer-carbon composite films, the resistance In the second configuration the head space from culture dishes was
across the film changes. The JPL ENose sensing array and methods of sampled by connecting the Tygon tubing to a dish cover to which a
fabricating sensors has been discussed in detail elsewhere (Ryan et piece of glass tubing had been attached (Fig. 2). In this way the head
al., 2004a; Ryan et al., 2004b). For this study, we used a sensing space was sampled at a fixed distance from the samples in all
B. Kateb et al. / NeuroImage 47 (2009) T5–T9 T7
Fig. 1. The JPL electronic nose. The white line describes the path of the air sample
through the unit. Air is pulled in from the sample, passes through a filter (pink area in
figure), goes through a valve, and is dispersed in the sensing chamber (central blue
area). There are four ceramic sensor substrates in the sensing chamber, and the unit
may be operated with 16–32 sensors.
taken first with a layer of medium covering the sample and then with
the medium decanted. Array fingerprint distinguishability was determined as the percent
variation from one pattern to another. Variation is computed by
Data acquisition and analysis summing the magnitude of the fractional difference for each sensor
response and dividing by the number of sensors:
ENose data are recorded as the DC resistance of each sensor
every 20 s. Array fingerprints for each sample (the “odor signature”) Σð j RnA − RnB j =RnA Þ
were constructed by plotting the data as a histogram, where the :
N
height of the bars in the histogram correspond to the steady-state,
normalized change in resistance for each sensor, (R − R0) / R0. R0 is
the resistance of the sensor averaged over the 5 min before a RnA is the response of sensor n to material A, RnB is the response of
reading is taken, and R is the resistance of the sensor averaged over sensor n to material B and N is the total number of sensors. Variation
5 min after it has reached steady state upon being exposed to a less than 10% is not sufficient to distinguish one array signal from
sample. In all experiments described below data were taken on an another, because the 5% error in each sensor response could result in a
array of 16 sensors; one element was not working, so data were variation of 5% for two samples of the same material.
collected from 15 of the 16 elements. For multiple samples of the
same material, the response for each sensor is averaged to construct Results
the corresponding bar in the fingerprint histogram.
The first experiment examined whether cultured cells could be
distinguished from the medium in which they are cultured. The
fingerprint pattern of culture medium with A2058 or U251 (3 × 105
cells each) were compared with the fingerprint pattern of the cells
with medium decanted. Comparison of the array patterns indicates
that there is a significant difference between the signatures of the
medium alone and the medium with the cells. For the A2058
melanoma cells, the array fingerprint distinguishability of the medium
and cells varied by 77%. For U251 cells, the variation was 44%. Three
measurements were averaged for each sample. Error on each sensor
response in the histogram was 5%. There were not enough trials to
compute a standard deviation, so error is reported as the fluctuation in
sensor response.
Next we determined whether the number of cells initially plated in
culture influences the odor fingerprint or array pattern. The total
Fig. 2. The cover used on the culture plates to sample the headspace above the cultured
cells. Air is pulled out through the center tube, through a Teflon-lined Tygon tube to the variation across a 15 element array between 3 × 105 and 1 × 106 A2058
ENose. Air pressure remains static as air enters through the other tubes. cells was 8%, which was not a significant difference. For U251 cells, the
T8 B. Kateb et al. / NeuroImage 47 (2009) T5–T9
difference in array patterns between the 3 × 105 vs. 1 × 106 cells was speculation that there may be other odorants and receptor mechan-
49%, which was significant. isms influencing general cellular trafficking.
We then determined the variance in the fingerprints between
A2058 and U251 cells. The head space of three samples of 1 × 106 cells Conclusion
of each cell line was analyzed. A comparison of the fingerprint for each
cell line is shown in Fig. 3. The lower portion of Fig. 3 is an expansion This study shows that it is possible to use an electronic nose to
of the portion of the array fingerprint histogram; as can be seen in the distinguish between two types of tumor cells and between two types
figure, sensors 4–10 contribute most to the variation in the response of organ tissue. As we conducted the experiment with a sensor array
pattern between the two cell lines. built for JPL rather than medical purposes, it may be possible to select
While the patterns of the two cell lines in Fig. 3 look similar, (e.g. an array which is optimized to distinguish between different types of
sensor five has a negative response for both cell types), there are very cells and organ tissues. While our work has shown that there is a
distinct differences. Looking at ratios between sensors can help difference in chemical composition of the headspace above cells in
illustrate that. If you compare the ratio of responses from sensor 3 to vitro, we did not isolate specific factors that caused these differences.
sensor 4 you will see that the response to U251 is roughly 1:1 whereas As the data boundaries are established for medical applications,
for A2058 sensor 3:sensor 4 is about 1:2. When the differences are researchers will be able to optimize sensing arrays for specific tasks,
calculated, as described earlier, we see that the variation in these two color code signals and possibly turn them into images that could be
cell lines is 22%. used as part of multi-modality intraoperative approach for detection
The fourth experiment compared the sensor array response to two and treatment of brain cancers. Ultimately, once a database of diseases
types of uncooked organ tissue, chicken liver and chicken heart. The and their related smells is established, we hope to have a device that
ENose was able to distinguish between the two types of tissue; the can assist diagnosis in the clinical setting while requiring minimal
variation in array response to the two types of tissue was 19%. specialized training.
Machado, R.F., Laskowski, D., Deffenderfer, O., et al., 2005. Detection of lung cancer by Ryan, M.A., Shevade, A.V., Zhou, H., et al., 2004b. Polymer-carbon-composite sensors for
sensor array analyses of exhaled breath. Am. J. Respir. Crit. Care Med. 171 (11), an electronic nose air quality monitor. MRS Bull. 29 (10), 714–719.
1286–1291. Ryan, M.A., Shevade, A.V., Kisor, A.K., et al., 2008. Ground validation of the third
Mantini, A., Di Natale, C., Macagnano, A., Paolesse, R., Finazzi-Agro, A., D amico , A., generation JPL electronic nose. Proc. 38th Int. Conf. Environ. Syst. 2008–01-2044.
2000. Biomedical application of an electronic nose. Crit. Rev. Biomed. Eng. 28 (3–4), Shykhon, M.E., Morgan, D.W., Dutta, R., Hines, E.L., Gardner, J.W., 2004. Clinical
481–485. evaluation of the electronic nose in the diagnosis of ear, nose and throat infection: a
McCulloch, M., Jezierski, T., Broffman, M., Hubbard, A., Turner, K., Janecki, T., 2006. preliminary study. J. Laryngol. Otol. 118 (9), 706–709.
Diagnostic accuracy of canine scent detection in early- and late-stage lung and Tanaka, M., Anguri, H., Nonaka, A., Kataoka, K., Nagata, H., Kita, J., Shizukuishi, S., 2004.
breast cancers. Integr. Cancer Ther. 5 (1), 30–39. Clinical assessment of oral malodor by the electronic nose system. J. Dent. Res. 83
Mohamed, E.I., Linder, R., Perriello, G., et al., 2002. Predicting Type 2 diabetes using an (4), 317–321.
electronic nose-base artificial neural network analysis. Diabetes Nutr. Metab. 15 (4), Thaler, E.R., Hanson, C.W., 2005. Medical applications of electronic nose technology.
215–221. Exp. Rev. Med. Devices 2 (5), 559–566.
Ohmori, S., Ohno, Y., Makino, T., Kashihara, T., 2005. Application of an electronic nose Spehr, M., Gisselmann, G., Poplawski, A., Riffell, J.A., Wetzel, C.H., Zimmer, R.K., Hatt, H.,
system for evaluation of unpleasant odor in coated tables. Eur. J. Pharm. Biopharm. 2003. Identification of a testicular odorant receptor mediating human sperm
59, 289–297. chemotaxis. Science 299 (5615), 2054–2058 Mar 28.
Parmentier, M., Libert, F., Schurmans, S., Schiffmann, S., Lefort, A., Eggerickx, D., Vanderhaeghen, P., Schurmans, S., Vassart, G., Parmentier, M., 1997a. Specific repertoire
Ledent, C., Mollereau, C., Gérard, C., Perret, J., 1992. Expression of members of of olfactory receptor genes in the male germ cells of several mammalian species.
the putative olfactory receptor gene family in mammalian germ cells. Nature Genomics 39 (3), 239–246 Feb 1.
355 (6359), 453–455. Vanderhaeghen, P., Schurmans, S., Vassart, G., Parmentier, M., 1997b. Molecular cloning
Pearce, T.C., Schiffman, S.S., Nagle, H.T., Gardner, J.W. (Eds.), 2003. Handbook of Machine and chromosomal mapping of olfactory receptor genes expressed in the male germ
Olfaction: Electronic Nose Technology. Weinheim, Wiley VCH. line: evidence for their wide distribution in the human genome. Biochem. Biophys.
Penn, D.J., Oberzaucher, E., Grammer, K., et al., 2007. Individual and gender fingerprints Res. Commun. 237 (2), 283–287 Aug 18.
in human body odour. J. R. Soc. Interface 4 (13), 331–340. Voss, A., Baier, V., Reisch, R., et al., 2005. Smelling renal dysfunction via electronic nose.
Pickel, D., Manucy, G.P., Walker, D.B., Hall, S.B., Walker, J.C., 2004. Evidence for canine Ann. Biomed. Eng. 33 (5), 656–660.
olfactory detection of melanoma. Appl. Anim. Behav. Sci. 89 (1–2), 107–116. Willis, C.M., Church, S.M., Guest, C.M., Cook, W.A., Mccarthy, N., Bransbury, A.J., Church,
Reid, L.M., O Donnell , C.P., Downey, G., 2006. Recent technological advances for the M.R.T., Church, J.C.T., 2004. Olfactory detection of human bladder cancer by dogs:
determination of food authenticity. Trends Food Sci. Technol. 17 (7), 344–353. proof of principle study. Br. Med. J. 329 (7468), 712–714A.
Ryan, M.A., Zhou, H., Buehler, M.G., et al., 2004a. Monitoring space shuttle air quality Yu, J.B., Byun, H.G., So, M.S., Huh, J.S., 2005. Analysis of diabetic patient's breath with
using the JPL electronic nose. IEEE Sens. J. 4 (3), 337–347. conducting polymer sensor array. Sens. Actuators, B,-Chem. 108, 305–308.