- Email: [email protected]
Evaluation of FTIR spectroscopy as diagnostic tool for colorectal cancer using spectral analysis
Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 122 (2014) 288–294
Contents lists available at ScienceDirect
Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy journal homepage: www.elsevier.com/locate/saa
Evaluation of FTIR spectroscopy as diagnostic tool for colorectal cancer using spectral analysis Liu Dong a, Xuejun Sun a,⇑, Zhang Chao a, Shiyun Zhang a, Jianbao Zheng a, Rajendra Gurung a, Junkai Du a, Jingsen Shi a, Yizhuang Xu b, Yuanfu Zhang b, Jinguang Wu b a b
Department of General Surgery, First Affiliated Hospital of Medical College of Xi0 an Jiaotong University, Xi0 an, Shaanxi, China College of Chemistry and Molecular Engineering, Peking University, Beijing, China
h i g h l i g h t s
g r a p h i c a l a b s t r a c t
Colorectal cancer tissues and normal
Of 180 samples colorectal cancer and colorectal tissues, the majority of majority peaks were in the 1000– 1800 cm1 region. FTIR mean spectra from the two groups.
colorectal tissue was studied by FTIR spectroscopy. IR spectra of tissue were investigated. The FTIR spectrum of each sample identified 12 bands from 1000 to 4000 cm1. FTIR spectroscopy can identify colorectal cancer and colorectal tissues.
a r t i c l e
i n f o
Article history: Received 5 April 2013 Received in revised form 9 October 2013 Accepted 2 November 2013 Available online 16 November 2013 Keywords: FTIR spectroscopy Colorectal cancerl Principal component analysis Spectrum
a b s t r a c t The aim of this study is to confirm FTIR spectroscopy as a diagnostic tool for colorectal cancer. 180 freshly removed colorectal samples were collected from 90 patients for spectrum analysis. The ratios of spectral intensity and relative intensity (/I1460) were calculated. Principal component analysis (PCA) and Fisher’s discriminant analysis (FDA) were applied to distinguish the malignant from normal. The FTIR parameters of colorectal cancer and normal tissues were distinguished due to the contents or configurations of nucleic acids, proteins, lipids and carbohydrates. Related to nitrogen containing, water, protein and nucleic acid were increased significantly in the malignant group. Six parameters were selected as independent factors to perform discriminant functions. The sensitivity for FTIR in diagnosing colorectal cancer was 96.6% by discriminant analysis. Our study demonstrates that FTIR can be a useful technique for detection of colorectal cancer and may be applied in clinical colorectal cancer diagnosis. Ó 2013 Published by Elsevier B.V.
Introduction
⇑ Corresponding author. Address: Department of General Surgery, First Affiliated Hospital of Medical College, Xi0 an Jiaotong University, 277 West Yanta Road, Xi0 an 710061, Shaanxi Province, China. Tel.: +86 029 85323879; fax: +86 02985323215. E-mail address: [email protected] (X. Sun). 1386-1425/$ - see front matter Ó 2013 Published by Elsevier B.V. http://dx.doi.org/10.1016/j.saa.2013.11.031
Colorectal cancer is a common disease that leads millions of people to death annually (6,08,0000 deaths per year) worldwide [1] It is the fifth leading cause of cancer-related mortality in China [2]. The survival rates of colorectal cancer patients are strongly associated with the tumor stage at the time of diagnosis. Although progress has been made to the quality of existing diagnostics and
289
L. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 122 (2014) 288–294
screening programs for colorectal cancer, the sensitivity and specificity of non-invasive tests still need to be improved [3,4]. Existing diagnostic strategies such as colorectaloscopy are invasive, whereas molecular biological detection only has limited applications because of the requirement for repeated measurements and interference with various factors [5]. Therefore, newer and more effective methods are needed. Fourier transform infrared (FTIR) spectroscopy is an effective and non-invasive diagnostic tool for the investigation of chemical changes to tissue or cell on a molecular level [6–14]. When pathological changes occur to a tissue, other tissue components including carbohydrate, protein, lipid, nucleic acid etc. will also experience changes. Previously we have demonstrated that FTIR spectroscopy can reliably distinguish multiple types of carcinoma from normal tissue [14–17]. The results suggest that FTIR spectroscopy can be a potential useful tool for screening a variety of human tumors. The aim of this pilot study is to determine whether malignant and normal human colorectal tissues can be distinguished by FTIR in a short time. Using 12 IR absorption bands between 4000 and 900 cm1 as markers for biochemical changes, we performed PCA and FDA, two multivariate analytical methods, to differentiate the two sample groups. Materials and methods The ethical committee of Xi’an Jiaotong University, First Hospital, approved this project. Patients were enrolled after being informed about the study and asked to provide written consent. For the study, 180 freshly removed colorectal samples (1 cm 1 cm 1 cm) were collected from 90 consent patients undergoing right or left hemi-colectomy. Before the FTIR spectrometry analysis, patients’ samples were collected from both tumor tissues and the respected margins (5 cm away from the tumor) of histological normal tissues. The samples were washed by distilled water for 3 times at the room temperature of 26 °C. Cool air was gently blown onto the mucosal surface for approximately 1 min to remove ambient moisture and thus reduce the effect of water absorption in the collected spectra. Fourier transform infrared spectra collection WQF-660 FTIR spectrometer instrument linked with a modified attenuated total reflectance (ATR) fiber probe (Beijing Rayleigh Analytical Instrument Corporation, Beijing, China) was used to obtain spectra. The FTIR spectrometer was equipped with a Mercury Cadmium Telluride (MCT) detector. The MCT detector had to be cooled using liquid nitrogen. The ATR fiber probe detected freshly removed tissue samples. The fiber attachment is comprised of an incident mid-infrared fiber, an exiting fiber, and a ZnSe ATR probe crystal. The incident and exiting fiber are connected by the ATR probe. The infrared fiber is hollow fiber with a metal coating layer on the internal surface, and their lengths can be arbitrarily set according to clinical requirements. Spectral measurements The fresh surgical resection specimens were immediately analyzed with the mobile WQD-660 spectrometer. The ATR crystal was cleaned with ethanol, then scanned, resulting in a spectrum of ambient air acquired by the ATR probe as a background spectrum; The ATR probe was placed in tight contact at 90° on the tissue mucosal surface for spectrum acquisition [18,19]. To collect the data for each spectrum, we performed 32 scans to achieve a good signal-to-noise ratio in the mid-infrared range (wave number
4000–900 cm1) at a resolution of 8 cm1. The procedure took approximately 1–2 min. Data processing and analysis We used OMNIC 5.0 as the data-processing software. The 15point moving average smoothing method was adopted for each spectrum to reduce the random noise in the data. The intensity and peak position of each band were measured using Spapro version 2.2 software (College of Chemistry and Molecular Engineering, Peking University, Beijing, China). The relative intensity ratios were calculated. All results were expressed as mean ± SE. Comparisons between malignant and normal tissue FTIR parameters were performed with the t-test using IBM SPSS Statistics 21 (New York, NY, USA). P < 0.05 was considered statistically significant. Subsequently, multivariate analysis approaches (PCA) followed by Fisher’s discriminant analysis (FDA) was used. Spectrum intensity data were exported from OMNIC format to the Unscrambler software suite (7.5 CAMO, Norway) for PCA, which allowed visualization and interpretation of spectral groupings. Upon receiving the variables and samples, the software generated an output known as a loadings plot. The loadings plot provides a projection view of the inter-variable relationships among the parameters of interest [20]. The FTIR spectra of the two groups were investigated with FDA (IBM SPSS 21.0). The algorithm was derived from the equation: L = b1x1 + b2x2 + b3x3 + + bnxn + c. Results The clinical characteristics of the patients are summarized in Table 1.Of the 90 samples from colorectal cancer tissues, the majority of majority peaks are in the 1800–900 cm1 region. FTIR mean spectra from the two groups are shown in Fig. 1b and c show the median spectra for cancer and normal sputum, highlighting key regions of wave number absorbance where differences exist between the two groups. Twelve peaks were identified and given a preliminary assignment as nucleic acid, protein, lipid or carbohydrate (Table 2). The spectrum intensity data at 1700–900 cm1 and 4000–1700 cm1 regions subjected to PCA are shown in Fig. 2. Spectral malignant and normal relative intensity ratios scores scatter plot based on PC1 and PC2 showing the distribution of the two different tissues (Fig. 3). Some FTIR parameters in the spectra of malignant tissues differed considerably from normal tissues, as shown in Fig. 4. The relative intensity ratios of 1460 cm1 (related to lipid); 3260 (related to nitrogen containing and water); 1640
Table 1 Clinical characteristics of the 90 patients with colorectal cancer. Classification
N
Median (range) age (years) Sex ratio (M:F) Histopathological grading Well differentiated Moderately differentiated Poorly differentiated Dukes stage A B C D Pathological type Papilary adenocarcinoma Tubular adenocarcinoma Mucinous adenocarcinoma Surgical procedures Right hemicolectomy Left hemicolectomy
32–75(53.5) 3:2 20 46 24 4 28 36 22 42 36 12 66 24
290
L. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 122 (2014) 288–294
Fig. 1. Mean Fourier transform infrared spectra of 90 cases colorectal cancer and 90 cases colorectal tissues. (a) Entire spectra with wave number ranging from 4000 to 900 cm1; (b and c) fingerprint region (the most sensitive region) at (b) 1800–1350 cm1 and (c) 1350–900 cm1.
L. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 122 (2014) 288–294 Table 2 Partly assignments of FTIR bands in normal colorectal tissues. Peak numbers
Frequency (cm1)
Assignments
1 2 3 4 5 6 7 8 9 10 11
3260 2920 2850 1740 1640 1550 1460 1400 1300 1240 1160
mOAH, mNAH (water, protein) mas, CH3 ms, CH2
12
1080
C@O stretching bands (lipids) Amide I (protein) Amide II(protein) dCAH dCAH, dCAOAH dCAH, dCAOAH Amide III band, etc. Asymmetric PO 2 stretching in RNA, etal dCAO, dCAOAH (carbohydrate or glycoprotein), etc. das PO2 (nucleic acid), etc.
Table 3 Parameters selected by Fisher’s discriminant analysis and their coefficients for classification of normal colorectal samples and colorectal cancer samples.
X1 X2 X3 X4 X5 X6 Constant
Selected parameters
Coefficient
P1080 P1300 I1080/I1460 I1640/I1460 I3260/I1460 I11740/I1460
0.27 0.083 0.42 0.774 0.008 0.17 184.153
Group centroid M
N
1.158
1.603
Table 4 Colorectal samples from 180 cases, comparison of the result between ATR-FTIR spectroscopy and histologic examination (gold standard*). Statistic result of discrimination models for prediction. Compared with histologic diagnosis, the sensitivity for FTIR in diagnosing colorectal cancer was 96.6%. Classification resultsa Result
Original
ATR-FTIR spectroscopy %
1.00 2.00 1.00 2.00
Histologic diagnosis 1.00
2.00
87 10 96.6 10.1
3 80 3.4 89.9
Total
90 90 100.0 100.0
a 93.8% of original grouped cases correctly classified by ATR-FTIR spectroscopy with FDA. 1: colorectal cancer samples. 2: colorectal samples.
and 1550 cm1 (amide groups related to protein); 1080 cm1 (related to nucleic acid) increased significantly in the malignant group. In contrast, those of 2925, 1400 and 1740 cm1 (related to lipid) and 1160 cm1 (related to carbohydrate) decreased noticeably. Six parameters were selected as independent factors from PCA and t-test (Fig. 4). Therefore the linear discriminant function was: L ¼ 0:27 1 0:083 2 þ 0:42 3 0:774 4 þ 0:008 5 þ0:17 6 184:153.After executing the above equation with the new data, sex determination could be made with the help of the adjusted canonical centroids of 1.603 to–1.158 i.e [21,22]. If the product obtained is close to –1.158 then the proposed spectrum of tissue is malignant and if the other centroid is close to 1.603 then the proposed spectrum of tissue is normal (Table 4). The sensitivity for FTIR in diagnosing colorectal cancer was 96.6%, specificity was 89.9% and the accuracy was 93.8%. Discussion The theoretical basis for this present study is that FTIR spectroscopy identifies different chemical ‘signatures’ in response to change in the intracellular environment [23]. It is a well
291
established and constantly evolving analytical technique, which allows for rapid high-throughput, non-destructive analysis of a wide range of sample types. Our results show that ATR-FTIR spectroscopy can be a good tool to distinguish between malignant and normal colorectal tissues based on the spectral. As shown in Fig. 2(a) ‘‘X-expl: 62%, 31%’’ meant that the first two principal components, PC1 and PC2, accounted for 93% (62% + 31%) of the total variance in the data obtained from normal and malignant[20], PC-1, which explained 62% variance, is mainly due to 1640 cm1 and 1550 cm1 of Amide group (protein) levels. PC-2, which explained 31% variance, is mainly due to 1080 cm1 relate to PO2 (nucleic acid). As shown in Fig. 2(b), samples with high score values for PC-1, which explained 59% variance, is mainly due to 3260 cm-1 of mOAH, mNAH (water, nitrogen containing) levels. PC-2, which explained 39% variance, is mainly due to 1740 cm-1 of C@O stretching bands (lipids). Analysis by PCA scores scatter plot of PC-1, which explained 95% variance, and PC-2, which explained 4% variance, based on PCA, normal and malignant relative intensity ratios spectra (n = 180) can give a clear classification of tissues for discrimination Fig. 3. There is a degree of overlap between them, but 82% of separation between normal and malignant spectra can be achieved by line perpendicular to FDA. According to Fig. 4a, which illustrates the relative intensity ratios of 1460 cm1 (related to lipid), the following observations were made: (1) 3260(related to nitrogen content and water): Remarkable changes are observed in 2500–4000 cm1 region for malignant tissues as compared to their normal counterparts. Malignant tissues have more water and mucus than normal tissues, with vibrational bands at 3260 cm1 arising mainly due to NAH and OH stretching vibrations of protein and water. (2) 1640 and 1550 cm1 (amide groups related to protein): Significant difference between the normal and malignant colon tissue is shown in the region of 1500–1700 cm1, This region denotes amide I and II bands of proteins. Vibrational bands at 1640 and 1550 cm1 arise from NAH bending vibrations, attribute to colon adenocarcinoma secrete a large amount of mucus. Moreover, malignant tissues have more beta structure than alpha compare to normal tissues [24]. (3) 1080 cm1 (related to nucleic acid) increased significantly in the malignant group. The difference is that malignant tissue content endless replication of DNA in cancerous cells. Cancerous tissue contains greater amounts of nucleic acids, collagen, and some amino acids compared to the normal ones [23,25]. (4) 1300 cm1 bands be considered as CAH deformation vibration or Amide III band. This study shows that the peak intensity of cancer was higher compare with normal tissues, as additional malignant tissues peak positions move to higher wave number (Fig. 4b). One possibility is that the associated protein of tumor structure has changed, but this requires further study. In contrast, the intensity ratios for carbohydrates (1160 cm1 related to carbohydrate) and lipids(2925, 1740 and 1400 cm1 related to lipid) decreased noticeably owing to the increased consumption by the Malignant tissues [26]. Vibrational bands decrease were mainly due to the growth of cancer cells that needed to consume large amounts of fat and sugar as nutrients compare to normal cells, resulting in reduced CAH and C@O stretching vibrations. The implications of alterations peak position and FWHM are still not clear (Fig. 4c). These parameters could indicate changes
292
L. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 122 (2014) 288–294
Fig. 2. (a) PCA correlation loadings discriminating colorectal cancer samples from normal samples based on eight characteristics peaks (1700–900 cm1). (b) Four characteristics peak (4000–1700 cm1). N: normal colorectal tissues. M: colorectal cancer tissues.
Fig. 3. PCA of relative intensity ratios for colorectal cancer tissues and normal colon tissues, clear classification of tissues for discrimination. N: normal colorectal tissues. M: colorectal cancer tissues.
L. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 122 (2014) 288–294
in the conformation of functional groups, the order of chemical bonds, the amount of hydrogen bonding or the secondary protein structure. Based on the t-test (Fig. 4) and PCA, six parameters were selected (Table 3) by FDA and our investigation demonstrated that colorectal cancer can be diagnosed by the FTIR technique. Based on this technique, The sensitivity for FTIR in diagnosing colorectal cancer was 96.6% by discriminant analysis. 93.5% of separation
293
between normal and malignant spectra can be classified correctly by FDA (Table 4). The results suggest a potential use of FTIR spectroscopy in detecting occult colorectal cancer. Colonoscopy is a commonly performed procedure for the diagnosis and treatment of colorectal neoplasia. But it is an invasive procedure, which is often associated with serious complications such as perforation, hemorrhage and spleen rupture [27,28]. It is subjective and time consuming [29]. Our results show that FTIR
Fig. 4. Some FTIR parameters in the spectra of malignant tissues differed considerably from normal tissues (a) peak intensity (I)/at 1460 cm1 = relative intensity ratio (1460 is the most stable band in the intensity parameters); (b) peak position. (c) FWHM (full width at half maximum) t test for data with a normal distribution and Mann-Whitney U test for other data. Colorectal cancer (n = 90); colorectal tissues (n = 90). Stand for P < 0.05. Stand for P < 0.01.
294
L. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 122 (2014) 288–294
spectroscopy, combined with appropriate pattern recognition algorithms, can distinguish normal from malignant colorectal disease. Compared with colorectal colonoscopy diagnosis, FTIR converted biochemical information into spectral data, and provided digitalized criteria between normal and malignant tissues. Moreover, it is a computer-based system, which helps standardize interpretation and avoids subjective factors. Additionally, FTIR is non-invasive, with easy patient preparation, and cost-effective. The FTIR technique needs, however, to improve in sensitivity and accuracy compared with histology. The aim of future investigations with FTIR spectroscopy is to detect cancer in vivo using intra-operative or pre-operative diagnosis to avoid unnecessary dissection to minimize surgical trauma. In future studies, an optic fiber attached to an ATR probe will be used to detect colorectal cancer in operating room. Potentially, FTIR could initiate important advancement in colorectal cancer detection. Acknowledgements We thank the National Nature Science Foundation of China (Grant Nos. 81172362 and 81101847) and the Fundamental Research Funds for the Central Universities for supporting this work. We also thank Miss Zhu Qing and Dr Shifu Weng for their assistance and advice during collection and processing of samples. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.saa.2013.11.031. References [1] Cancer, 2012.
[5] J.M. Yi, M. Dhir, A.A. Guzzetta, C.A. Iacobuzio-Donahue, K. Heo, K.M. Yang, H. Suzuki, M. Toyota, H.M. Kim, N. Ahuja, Tumour Biol. 33 (2012) 363–372. [6] L. Di Giambattista, D. Pozzi, P. Grimaldi, S. Gaudenzi, S. Morrone, A.C. Castellano, Anal. Bioanal. Chem. 399 (2011) 2771–2778. [7] A. Derenne, R. Gasper, E. Goormaghtigh, Analyst 136 (2011) 1134–1141. [8] C. Chonanant, N. Jearanaikoon, C. Leelayuwat, T. Limpaiboon, M.J. Tobin, P. Jearanaikoon, P. Heraud, Analyst 136 (2011) 2542–2551. [9] N.S. Ozek, S. Tuna, A.E. Erson-Bensan, F. Severcan, Analyst 135 (2010) 3094– 3102. [10] C. Petibois, M. Cestelli-Guidi, M. Piccinini, M. Moenner, Anal. Bioanal. Chem. 397 (2010) 2123–2129. [11] R. Gasper, E. Goormaghtigh, Analyst 135 (2010) 3048–3051. [12] D. Sheng, X. Liu, W. Li, Y. Wang, X. Chen, X. Wang, Spectrochim. Acta A 101 (2013) 228–232. [13] A.A. Isab, M.N. Shaikh, M. Monim-ul-Mehboob, B.A. Al-Maythalony, M.I. Wazeer, S. Altuwaijri, Synth. Spectrochim. Acta A 79 (2011) 1196–1201. [14] X.J. Sun, Y.L. Su, R.D. Soloway, L. Zhang, J.S. Wang, Y. Ren, L.M. Yang, A.G. Zheng, Y.F. Zhang, Y.Z. Xu, S.F. Weng, J.S. Shi, D.F. Xu, J.G. Wu, Gastroenterology 124 (2003) A420–A421. [15] J.K. Du, J.S. Shi, X.J. Sun, J.S. Wang, Y.Z. Xu, J.G. Wu, Y.F. Zhang, S.F. Weng, Hepatobiliary Pancreat Dis. Int. 8 (2009) 75–78. [16] J.K. Du, J.S. Shi, Y.Z. Xu, Y.F. Zhang, X.J. Sun, J.S. Wang, S.F. Weng, J.G. Wu, Guang Pu, Xue Yu, Guang Pu, Fen Xi 28 (2008) 51–54. [17] X.J. Sun, F.L. Sun, J.K. Du, Q.H. Pan, Y.F. Zhang, Y.Z. Xu, J.S. Shi, J.G. Wu, Guang Pu, Xue Yu, Guang Pu, Fen Xi. 29 (2009) 1750–1753. [18] M.J. Walsh, A. Kajdacsy-Balla, S.E. Holton, R. Bhargava, Vib. Spectrosc. 60 (2012) 23–28. [19] T.D. Wang, G. Triadafilopoulos, J.M. Crawford, L.R. Dixon, T. Bhandari, P. Sahbaie, S. Friedland, R. Soetikno, C.H. Contag, Proc. Natl. Acad. Sci. USA 104 (2007) 15864–15869. [20] Z.H. Loh, B.Y. Sia, P.W. Heng, C.C. Lee, C.V. Liew, AAPS Pharm. Sci. Technol. 12 (2011) 1374–1383. [21] A. Farzan, S. Mashohor, R. Ramli, R. Mahmud, Diagn. Pathol. 6 (2011) 105. [22] S.T. Bharath, G.R. Kumar, R. Dhanapal, T. Saraswathi, J. Forensic Dent. Sci. 3 (2011) 58–62. [23] C. Krafft, S.B. Sobottka, G. Schackert, R. Salzer, Analyst 129 (2004) 921–925. [24] T. Yamada, N. Miyoshi, T. Ogawa, K. Akao, M. Fukuda, T. Ogasawara, Y. Kitagawa, K. Sano, Clin. Cancer Res. 8 (2002) 2010–2014. [25] N. Fujioka, Y. Morimoto, T. Arai, M. Kikuchi, Cancer Detect Prev. 28 (2004) 32– 36. [26] D.K. Nomura, J.Z. Long, S. Niessen, H.S. Hoover, S.W. Ng, Cell. 140 (2010) 49–61. [27] T. Rasul, E. Leung, K. McArdle, R. Pathak, S. Dalmia, Dig. Endosc. 22 (2010) 351– 353. [28] D.H. Kim, P.J. Pickhardt, A.J. Taylor, C.O. Menias, Curr. Probl. Diagn. Radiol. 37 (2008) 165–177. [29] C. Kendall, M. Isabelle, F. Bazant-Hegemark, J. Hutchings, L. Orr, J. Babrah, R. Baker, N. Analyst. 134 (2009) 1029–1045.
Contents lists available at ScienceDirect
Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy journal homepage: www.elsevier.com/locate/saa
Evaluation of FTIR spectroscopy as diagnostic tool for colorectal cancer using spectral analysis Liu Dong a, Xuejun Sun a,⇑, Zhang Chao a, Shiyun Zhang a, Jianbao Zheng a, Rajendra Gurung a, Junkai Du a, Jingsen Shi a, Yizhuang Xu b, Yuanfu Zhang b, Jinguang Wu b a b
Department of General Surgery, First Affiliated Hospital of Medical College of Xi0 an Jiaotong University, Xi0 an, Shaanxi, China College of Chemistry and Molecular Engineering, Peking University, Beijing, China
h i g h l i g h t s
g r a p h i c a l a b s t r a c t
Colorectal cancer tissues and normal
Of 180 samples colorectal cancer and colorectal tissues, the majority of majority peaks were in the 1000– 1800 cm1 region. FTIR mean spectra from the two groups.
colorectal tissue was studied by FTIR spectroscopy. IR spectra of tissue were investigated. The FTIR spectrum of each sample identified 12 bands from 1000 to 4000 cm1. FTIR spectroscopy can identify colorectal cancer and colorectal tissues.
a r t i c l e
i n f o
Article history: Received 5 April 2013 Received in revised form 9 October 2013 Accepted 2 November 2013 Available online 16 November 2013 Keywords: FTIR spectroscopy Colorectal cancerl Principal component analysis Spectrum
a b s t r a c t The aim of this study is to confirm FTIR spectroscopy as a diagnostic tool for colorectal cancer. 180 freshly removed colorectal samples were collected from 90 patients for spectrum analysis. The ratios of spectral intensity and relative intensity (/I1460) were calculated. Principal component analysis (PCA) and Fisher’s discriminant analysis (FDA) were applied to distinguish the malignant from normal. The FTIR parameters of colorectal cancer and normal tissues were distinguished due to the contents or configurations of nucleic acids, proteins, lipids and carbohydrates. Related to nitrogen containing, water, protein and nucleic acid were increased significantly in the malignant group. Six parameters were selected as independent factors to perform discriminant functions. The sensitivity for FTIR in diagnosing colorectal cancer was 96.6% by discriminant analysis. Our study demonstrates that FTIR can be a useful technique for detection of colorectal cancer and may be applied in clinical colorectal cancer diagnosis. Ó 2013 Published by Elsevier B.V.
Introduction
⇑ Corresponding author. Address: Department of General Surgery, First Affiliated Hospital of Medical College, Xi0 an Jiaotong University, 277 West Yanta Road, Xi0 an 710061, Shaanxi Province, China. Tel.: +86 029 85323879; fax: +86 02985323215. E-mail address: [email protected] (X. Sun). 1386-1425/$ - see front matter Ó 2013 Published by Elsevier B.V. http://dx.doi.org/10.1016/j.saa.2013.11.031
Colorectal cancer is a common disease that leads millions of people to death annually (6,08,0000 deaths per year) worldwide [1] It is the fifth leading cause of cancer-related mortality in China [2]. The survival rates of colorectal cancer patients are strongly associated with the tumor stage at the time of diagnosis. Although progress has been made to the quality of existing diagnostics and
289
L. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 122 (2014) 288–294
screening programs for colorectal cancer, the sensitivity and specificity of non-invasive tests still need to be improved [3,4]. Existing diagnostic strategies such as colorectaloscopy are invasive, whereas molecular biological detection only has limited applications because of the requirement for repeated measurements and interference with various factors [5]. Therefore, newer and more effective methods are needed. Fourier transform infrared (FTIR) spectroscopy is an effective and non-invasive diagnostic tool for the investigation of chemical changes to tissue or cell on a molecular level [6–14]. When pathological changes occur to a tissue, other tissue components including carbohydrate, protein, lipid, nucleic acid etc. will also experience changes. Previously we have demonstrated that FTIR spectroscopy can reliably distinguish multiple types of carcinoma from normal tissue [14–17]. The results suggest that FTIR spectroscopy can be a potential useful tool for screening a variety of human tumors. The aim of this pilot study is to determine whether malignant and normal human colorectal tissues can be distinguished by FTIR in a short time. Using 12 IR absorption bands between 4000 and 900 cm1 as markers for biochemical changes, we performed PCA and FDA, two multivariate analytical methods, to differentiate the two sample groups. Materials and methods The ethical committee of Xi’an Jiaotong University, First Hospital, approved this project. Patients were enrolled after being informed about the study and asked to provide written consent. For the study, 180 freshly removed colorectal samples (1 cm 1 cm 1 cm) were collected from 90 consent patients undergoing right or left hemi-colectomy. Before the FTIR spectrometry analysis, patients’ samples were collected from both tumor tissues and the respected margins (5 cm away from the tumor) of histological normal tissues. The samples were washed by distilled water for 3 times at the room temperature of 26 °C. Cool air was gently blown onto the mucosal surface for approximately 1 min to remove ambient moisture and thus reduce the effect of water absorption in the collected spectra. Fourier transform infrared spectra collection WQF-660 FTIR spectrometer instrument linked with a modified attenuated total reflectance (ATR) fiber probe (Beijing Rayleigh Analytical Instrument Corporation, Beijing, China) was used to obtain spectra. The FTIR spectrometer was equipped with a Mercury Cadmium Telluride (MCT) detector. The MCT detector had to be cooled using liquid nitrogen. The ATR fiber probe detected freshly removed tissue samples. The fiber attachment is comprised of an incident mid-infrared fiber, an exiting fiber, and a ZnSe ATR probe crystal. The incident and exiting fiber are connected by the ATR probe. The infrared fiber is hollow fiber with a metal coating layer on the internal surface, and their lengths can be arbitrarily set according to clinical requirements. Spectral measurements The fresh surgical resection specimens were immediately analyzed with the mobile WQD-660 spectrometer. The ATR crystal was cleaned with ethanol, then scanned, resulting in a spectrum of ambient air acquired by the ATR probe as a background spectrum; The ATR probe was placed in tight contact at 90° on the tissue mucosal surface for spectrum acquisition [18,19]. To collect the data for each spectrum, we performed 32 scans to achieve a good signal-to-noise ratio in the mid-infrared range (wave number
4000–900 cm1) at a resolution of 8 cm1. The procedure took approximately 1–2 min. Data processing and analysis We used OMNIC 5.0 as the data-processing software. The 15point moving average smoothing method was adopted for each spectrum to reduce the random noise in the data. The intensity and peak position of each band were measured using Spapro version 2.2 software (College of Chemistry and Molecular Engineering, Peking University, Beijing, China). The relative intensity ratios were calculated. All results were expressed as mean ± SE. Comparisons between malignant and normal tissue FTIR parameters were performed with the t-test using IBM SPSS Statistics 21 (New York, NY, USA). P < 0.05 was considered statistically significant. Subsequently, multivariate analysis approaches (PCA) followed by Fisher’s discriminant analysis (FDA) was used. Spectrum intensity data were exported from OMNIC format to the Unscrambler software suite (7.5 CAMO, Norway) for PCA, which allowed visualization and interpretation of spectral groupings. Upon receiving the variables and samples, the software generated an output known as a loadings plot. The loadings plot provides a projection view of the inter-variable relationships among the parameters of interest [20]. The FTIR spectra of the two groups were investigated with FDA (IBM SPSS 21.0). The algorithm was derived from the equation: L = b1x1 + b2x2 + b3x3 + + bnxn + c. Results The clinical characteristics of the patients are summarized in Table 1.Of the 90 samples from colorectal cancer tissues, the majority of majority peaks are in the 1800–900 cm1 region. FTIR mean spectra from the two groups are shown in Fig. 1b and c show the median spectra for cancer and normal sputum, highlighting key regions of wave number absorbance where differences exist between the two groups. Twelve peaks were identified and given a preliminary assignment as nucleic acid, protein, lipid or carbohydrate (Table 2). The spectrum intensity data at 1700–900 cm1 and 4000–1700 cm1 regions subjected to PCA are shown in Fig. 2. Spectral malignant and normal relative intensity ratios scores scatter plot based on PC1 and PC2 showing the distribution of the two different tissues (Fig. 3). Some FTIR parameters in the spectra of malignant tissues differed considerably from normal tissues, as shown in Fig. 4. The relative intensity ratios of 1460 cm1 (related to lipid); 3260 (related to nitrogen containing and water); 1640
Table 1 Clinical characteristics of the 90 patients with colorectal cancer. Classification
N
Median (range) age (years) Sex ratio (M:F) Histopathological grading Well differentiated Moderately differentiated Poorly differentiated Dukes stage A B C D Pathological type Papilary adenocarcinoma Tubular adenocarcinoma Mucinous adenocarcinoma Surgical procedures Right hemicolectomy Left hemicolectomy
32–75(53.5) 3:2 20 46 24 4 28 36 22 42 36 12 66 24
290
L. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 122 (2014) 288–294
Fig. 1. Mean Fourier transform infrared spectra of 90 cases colorectal cancer and 90 cases colorectal tissues. (a) Entire spectra with wave number ranging from 4000 to 900 cm1; (b and c) fingerprint region (the most sensitive region) at (b) 1800–1350 cm1 and (c) 1350–900 cm1.
L. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 122 (2014) 288–294 Table 2 Partly assignments of FTIR bands in normal colorectal tissues. Peak numbers
Frequency (cm1)
Assignments
1 2 3 4 5 6 7 8 9 10 11
3260 2920 2850 1740 1640 1550 1460 1400 1300 1240 1160
mOAH, mNAH (water, protein) mas, CH3 ms, CH2
12
1080
C@O stretching bands (lipids) Amide I (protein) Amide II(protein) dCAH dCAH, dCAOAH dCAH, dCAOAH Amide III band, etc. Asymmetric PO 2 stretching in RNA, etal dCAO, dCAOAH (carbohydrate or glycoprotein), etc. das PO2 (nucleic acid), etc.
Table 3 Parameters selected by Fisher’s discriminant analysis and their coefficients for classification of normal colorectal samples and colorectal cancer samples.
X1 X2 X3 X4 X5 X6 Constant
Selected parameters
Coefficient
P1080 P1300 I1080/I1460 I1640/I1460 I3260/I1460 I11740/I1460
0.27 0.083 0.42 0.774 0.008 0.17 184.153
Group centroid M
N
1.158
1.603
Table 4 Colorectal samples from 180 cases, comparison of the result between ATR-FTIR spectroscopy and histologic examination (gold standard*). Statistic result of discrimination models for prediction. Compared with histologic diagnosis, the sensitivity for FTIR in diagnosing colorectal cancer was 96.6%. Classification resultsa Result
Original
ATR-FTIR spectroscopy %
1.00 2.00 1.00 2.00
Histologic diagnosis 1.00
2.00
87 10 96.6 10.1
3 80 3.4 89.9
Total
90 90 100.0 100.0
a 93.8% of original grouped cases correctly classified by ATR-FTIR spectroscopy with FDA. 1: colorectal cancer samples. 2: colorectal samples.
and 1550 cm1 (amide groups related to protein); 1080 cm1 (related to nucleic acid) increased significantly in the malignant group. In contrast, those of 2925, 1400 and 1740 cm1 (related to lipid) and 1160 cm1 (related to carbohydrate) decreased noticeably. Six parameters were selected as independent factors from PCA and t-test (Fig. 4). Therefore the linear discriminant function was: L ¼ 0:27 1 0:083 2 þ 0:42 3 0:774 4 þ 0:008 5 þ0:17 6 184:153.After executing the above equation with the new data, sex determination could be made with the help of the adjusted canonical centroids of 1.603 to–1.158 i.e [21,22]. If the product obtained is close to –1.158 then the proposed spectrum of tissue is malignant and if the other centroid is close to 1.603 then the proposed spectrum of tissue is normal (Table 4). The sensitivity for FTIR in diagnosing colorectal cancer was 96.6%, specificity was 89.9% and the accuracy was 93.8%. Discussion The theoretical basis for this present study is that FTIR spectroscopy identifies different chemical ‘signatures’ in response to change in the intracellular environment [23]. It is a well
291
established and constantly evolving analytical technique, which allows for rapid high-throughput, non-destructive analysis of a wide range of sample types. Our results show that ATR-FTIR spectroscopy can be a good tool to distinguish between malignant and normal colorectal tissues based on the spectral. As shown in Fig. 2(a) ‘‘X-expl: 62%, 31%’’ meant that the first two principal components, PC1 and PC2, accounted for 93% (62% + 31%) of the total variance in the data obtained from normal and malignant[20], PC-1, which explained 62% variance, is mainly due to 1640 cm1 and 1550 cm1 of Amide group (protein) levels. PC-2, which explained 31% variance, is mainly due to 1080 cm1 relate to PO2 (nucleic acid). As shown in Fig. 2(b), samples with high score values for PC-1, which explained 59% variance, is mainly due to 3260 cm-1 of mOAH, mNAH (water, nitrogen containing) levels. PC-2, which explained 39% variance, is mainly due to 1740 cm-1 of C@O stretching bands (lipids). Analysis by PCA scores scatter plot of PC-1, which explained 95% variance, and PC-2, which explained 4% variance, based on PCA, normal and malignant relative intensity ratios spectra (n = 180) can give a clear classification of tissues for discrimination Fig. 3. There is a degree of overlap between them, but 82% of separation between normal and malignant spectra can be achieved by line perpendicular to FDA. According to Fig. 4a, which illustrates the relative intensity ratios of 1460 cm1 (related to lipid), the following observations were made: (1) 3260(related to nitrogen content and water): Remarkable changes are observed in 2500–4000 cm1 region for malignant tissues as compared to their normal counterparts. Malignant tissues have more water and mucus than normal tissues, with vibrational bands at 3260 cm1 arising mainly due to NAH and OH stretching vibrations of protein and water. (2) 1640 and 1550 cm1 (amide groups related to protein): Significant difference between the normal and malignant colon tissue is shown in the region of 1500–1700 cm1, This region denotes amide I and II bands of proteins. Vibrational bands at 1640 and 1550 cm1 arise from NAH bending vibrations, attribute to colon adenocarcinoma secrete a large amount of mucus. Moreover, malignant tissues have more beta structure than alpha compare to normal tissues [24]. (3) 1080 cm1 (related to nucleic acid) increased significantly in the malignant group. The difference is that malignant tissue content endless replication of DNA in cancerous cells. Cancerous tissue contains greater amounts of nucleic acids, collagen, and some amino acids compared to the normal ones [23,25]. (4) 1300 cm1 bands be considered as CAH deformation vibration or Amide III band. This study shows that the peak intensity of cancer was higher compare with normal tissues, as additional malignant tissues peak positions move to higher wave number (Fig. 4b). One possibility is that the associated protein of tumor structure has changed, but this requires further study. In contrast, the intensity ratios for carbohydrates (1160 cm1 related to carbohydrate) and lipids(2925, 1740 and 1400 cm1 related to lipid) decreased noticeably owing to the increased consumption by the Malignant tissues [26]. Vibrational bands decrease were mainly due to the growth of cancer cells that needed to consume large amounts of fat and sugar as nutrients compare to normal cells, resulting in reduced CAH and C@O stretching vibrations. The implications of alterations peak position and FWHM are still not clear (Fig. 4c). These parameters could indicate changes
292
L. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 122 (2014) 288–294
Fig. 2. (a) PCA correlation loadings discriminating colorectal cancer samples from normal samples based on eight characteristics peaks (1700–900 cm1). (b) Four characteristics peak (4000–1700 cm1). N: normal colorectal tissues. M: colorectal cancer tissues.
Fig. 3. PCA of relative intensity ratios for colorectal cancer tissues and normal colon tissues, clear classification of tissues for discrimination. N: normal colorectal tissues. M: colorectal cancer tissues.
L. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 122 (2014) 288–294
in the conformation of functional groups, the order of chemical bonds, the amount of hydrogen bonding or the secondary protein structure. Based on the t-test (Fig. 4) and PCA, six parameters were selected (Table 3) by FDA and our investigation demonstrated that colorectal cancer can be diagnosed by the FTIR technique. Based on this technique, The sensitivity for FTIR in diagnosing colorectal cancer was 96.6% by discriminant analysis. 93.5% of separation
293
between normal and malignant spectra can be classified correctly by FDA (Table 4). The results suggest a potential use of FTIR spectroscopy in detecting occult colorectal cancer. Colonoscopy is a commonly performed procedure for the diagnosis and treatment of colorectal neoplasia. But it is an invasive procedure, which is often associated with serious complications such as perforation, hemorrhage and spleen rupture [27,28]. It is subjective and time consuming [29]. Our results show that FTIR
Fig. 4. Some FTIR parameters in the spectra of malignant tissues differed considerably from normal tissues (a) peak intensity (I)/at 1460 cm1 = relative intensity ratio (1460 is the most stable band in the intensity parameters); (b) peak position. (c) FWHM (full width at half maximum) t test for data with a normal distribution and Mann-Whitney U test for other data. Colorectal cancer (n = 90); colorectal tissues (n = 90). Stand for P < 0.05. Stand for P < 0.01.
294
L. Dong et al. / Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 122 (2014) 288–294
spectroscopy, combined with appropriate pattern recognition algorithms, can distinguish normal from malignant colorectal disease. Compared with colorectal colonoscopy diagnosis, FTIR converted biochemical information into spectral data, and provided digitalized criteria between normal and malignant tissues. Moreover, it is a computer-based system, which helps standardize interpretation and avoids subjective factors. Additionally, FTIR is non-invasive, with easy patient preparation, and cost-effective. The FTIR technique needs, however, to improve in sensitivity and accuracy compared with histology. The aim of future investigations with FTIR spectroscopy is to detect cancer in vivo using intra-operative or pre-operative diagnosis to avoid unnecessary dissection to minimize surgical trauma. In future studies, an optic fiber attached to an ATR probe will be used to detect colorectal cancer in operating room. Potentially, FTIR could initiate important advancement in colorectal cancer detection. Acknowledgements We thank the National Nature Science Foundation of China (Grant Nos. 81172362 and 81101847) and the Fundamental Research Funds for the Central Universities for supporting this work. We also thank Miss Zhu Qing and Dr Shifu Weng for their assistance and advice during collection and processing of samples. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.saa.2013.11.031. References [1] Cancer, 2012.
[5] J.M. Yi, M. Dhir, A.A. Guzzetta, C.A. Iacobuzio-Donahue, K. Heo, K.M. Yang, H. Suzuki, M. Toyota, H.M. Kim, N. Ahuja, Tumour Biol. 33 (2012) 363–372. [6] L. Di Giambattista, D. Pozzi, P. Grimaldi, S. Gaudenzi, S. Morrone, A.C. Castellano, Anal. Bioanal. Chem. 399 (2011) 2771–2778. [7] A. Derenne, R. Gasper, E. Goormaghtigh, Analyst 136 (2011) 1134–1141. [8] C. Chonanant, N. Jearanaikoon, C. Leelayuwat, T. Limpaiboon, M.J. Tobin, P. Jearanaikoon, P. Heraud, Analyst 136 (2011) 2542–2551. [9] N.S. Ozek, S. Tuna, A.E. Erson-Bensan, F. Severcan, Analyst 135 (2010) 3094– 3102. [10] C. Petibois, M. Cestelli-Guidi, M. Piccinini, M. Moenner, Anal. Bioanal. Chem. 397 (2010) 2123–2129. [11] R. Gasper, E. Goormaghtigh, Analyst 135 (2010) 3048–3051. [12] D. Sheng, X. Liu, W. Li, Y. Wang, X. Chen, X. Wang, Spectrochim. Acta A 101 (2013) 228–232. [13] A.A. Isab, M.N. Shaikh, M. Monim-ul-Mehboob, B.A. Al-Maythalony, M.I. Wazeer, S. Altuwaijri, Synth. Spectrochim. Acta A 79 (2011) 1196–1201. [14] X.J. Sun, Y.L. Su, R.D. Soloway, L. Zhang, J.S. Wang, Y. Ren, L.M. Yang, A.G. Zheng, Y.F. Zhang, Y.Z. Xu, S.F. Weng, J.S. Shi, D.F. Xu, J.G. Wu, Gastroenterology 124 (2003) A420–A421. [15] J.K. Du, J.S. Shi, X.J. Sun, J.S. Wang, Y.Z. Xu, J.G. Wu, Y.F. Zhang, S.F. Weng, Hepatobiliary Pancreat Dis. Int. 8 (2009) 75–78. [16] J.K. Du, J.S. Shi, Y.Z. Xu, Y.F. Zhang, X.J. Sun, J.S. Wang, S.F. Weng, J.G. Wu, Guang Pu, Xue Yu, Guang Pu, Fen Xi 28 (2008) 51–54. [17] X.J. Sun, F.L. Sun, J.K. Du, Q.H. Pan, Y.F. Zhang, Y.Z. Xu, J.S. Shi, J.G. Wu, Guang Pu, Xue Yu, Guang Pu, Fen Xi. 29 (2009) 1750–1753. [18] M.J. Walsh, A. Kajdacsy-Balla, S.E. Holton, R. Bhargava, Vib. Spectrosc. 60 (2012) 23–28. [19] T.D. Wang, G. Triadafilopoulos, J.M. Crawford, L.R. Dixon, T. Bhandari, P. Sahbaie, S. Friedland, R. Soetikno, C.H. Contag, Proc. Natl. Acad. Sci. USA 104 (2007) 15864–15869. [20] Z.H. Loh, B.Y. Sia, P.W. Heng, C.C. Lee, C.V. Liew, AAPS Pharm. Sci. Technol. 12 (2011) 1374–1383. [21] A. Farzan, S. Mashohor, R. Ramli, R. Mahmud, Diagn. Pathol. 6 (2011) 105. [22] S.T. Bharath, G.R. Kumar, R. Dhanapal, T. Saraswathi, J. Forensic Dent. Sci. 3 (2011) 58–62. [23] C. Krafft, S.B. Sobottka, G. Schackert, R. Salzer, Analyst 129 (2004) 921–925. [24] T. Yamada, N. Miyoshi, T. Ogawa, K. Akao, M. Fukuda, T. Ogasawara, Y. Kitagawa, K. Sano, Clin. Cancer Res. 8 (2002) 2010–2014. [25] N. Fujioka, Y. Morimoto, T. Arai, M. Kikuchi, Cancer Detect Prev. 28 (2004) 32– 36. [26] D.K. Nomura, J.Z. Long, S. Niessen, H.S. Hoover, S.W. Ng, Cell. 140 (2010) 49–61. [27] T. Rasul, E. Leung, K. McArdle, R. Pathak, S. Dalmia, Dig. Endosc. 22 (2010) 351– 353. [28] D.H. Kim, P.J. Pickhardt, A.J. Taylor, C.O. Menias, Curr. Probl. Diagn. Radiol. 37 (2008) 165–177. [29] C. Kendall, M. Isabelle, F. Bazant-Hegemark, J. Hutchings, L. Orr, J. Babrah, R. Baker, N. Analyst. 134 (2009) 1029–1045.