OxyHb - Wiley Online Library

6 downloads 77 Views 722KB Size Report
Jan 16, 2013 - the CaHm group and pyrrole ring of OxyHb of liver cancer patients. In this paper, principal component analysis (PCA) combined.
Research article Received: 6 May 2012

Revised: 25 September 2012

Accepted: 17 October 2012

Published online in Wiley Online Library: 16 January 2013

(wileyonlinelibrary.com) DOI 10.1002/jrs.4216

Near-infrared surface-enhanced Raman spectroscopy (NIR-SERS) studies on oxyheamoglobin (OxyHb) of liver cancer based on PVA-Ag nanofilm Renming Liu,a*† Yang Xiong,b† Weiyue Tang,b Yan Guo,c Xinhui Yand and Minzhen Sia A near-infrared surface-enhanced Raman spectroscopy (NIR-SERS) method was employed for oxyheamoglobin (OxyHb) detection to develop a simple blood test for liver cancer detection. Polyvinyl alcohol protected silver nanofilm (PVA-Ag nanofilm) used as the NIR-SERS active substrate to enhance the Raman scattering signals of OxyHb. High quality NIR-SERS spectrum from OxyHb adsorbed on PVA-Ag nanofilm can be obtained within 16 s using a portable Raman spectrometer. NIR-SERS measurements were performed on OxyHb samples of healthy volunteers (control subjects, n = 30), patients (n = 40) with confirmed liver cancer (stage I, II and III) and the liver cancer patients after surgery (n = 30). Meanwhile, the tentative assignments of the Raman bands in the measured NIR-SERS spectra were performed, and the results suggested cancer specific changes on molecule level, including a decrease in the relative concentrations and the percentage of aromatic amino acids of OxyHb, changes of the vibration modes of the CaHm group and pyrrole ring of OxyHb of liver cancer patients. In this paper, principal component analysis (PCA) combined with independent sample T test analysis of the measured NIR-SERS spectra separated the spectral features of the two groups into two distinct clusters with the sensitivity of 95.0% and the specificity of 85.7%. Meanwhile, the recovery situations of the liver cancer patients after surgery were also assessed using the method of discriminant analysis-predicting group membership based on PCA. The results show that 26.7% surgeried liver cancer patients were distinguished as the normal subjects and 63.3% were distinguished into the cancer. Our study demonstrated great potentials for developing NIR-SERS OxyHb analysis into a novel clinical tool for non-invasive detection of liver cancers. Copyright © 2013 John Wiley & Sons, Ltd. Supporting information can be found in the online version of this article. Keywords: PVA-Ag nanofilm; NIR-SERS; liver cancer; oxyheamoglobin; principal component analysis

Introduction

362

Liver cancer is the fourth leading cause of cancer related death in the world and the second most common cancer in China.[1,2] Now, patients suffering from liver cancer still have a poor prognosis.[2] There are considerable epidemiologic differences in the incidence of hepatocellular carcinoma. High incidence regions are, for example, eastern Asia, with age-adjusted rates for liver cancer of approximately 30 per 100 000 in Japan, 50 per 100 000 in Korea and 55 per 100 000 in China, whereas in Europe and the United States, the rate is consistently below 10 per 100 000.[2,3] Meanwhile, there is also an increasing incidence all around the world, especially in western countries and China, with growing oncologic relevance. It was reported that, screening methods are available which can reduce the incidence by removal of adenomas and can reduce deaths in diagnosed cancer cases by earlier detection.[4] Screening for liver cancer, the currently accepted screening indicators include serum alpha-fetoprotein (AFP) and liver ultrasound. However, AFP increased can also be seen in chronic liver disease, pregnancy and embryonic gonad tumors and benign familial AFP increased, making the AFP in hepatocellular carcinoma and early diagnosis have a higher false negative rate of up to 40%, which limits its general application in the diagnosis of liver cancer.[5] Undoubtedly,

J. Raman Spectrosc. 2013, 44, 362–369

B-type ultrasound examination is convenient, inexpensive and non-invasive, but there are also many disadvantages in B-type ultrasound examination for liver cancer, the main problem is the existence of the ultrasonic testing blind. Thus, a fast, simple and accurate screening of liver cancer is urgent and important. Raman spectroscopy is a powerful tool that provides fingerprinting information of biomacromolecules including proteins,

* Correspondence to: Renming Liu, Application Institute of Spectroscopy Technology, Chuxiong Normal University, Chuxiong 675000, China. E-mail: [email protected]

The first two authors contributed equally to this paper.

a Application Institute of Spectroscopy Technology, Chuxiong Normal University, Chuxiong 675000, China b Department of Physics, School of Physics and Engineering, Zhengzhou University, Zhengzhou 450052, China c Tumor Hospital of Zhengzhou University, Zhengzhou 45000, China d People’s Hospital of Zhengzhou University, Zhengzhou 45000, China

Copyright © 2013 John Wiley & Sons, Ltd.

NIR-SERS studies on OxyHb of liver cancer lipids and nucleic acids.[6] Thus, Raman spectroscopy has been employed as a novel nondestructive diagnostic tool for cancer detection and identification in recent years. For instance, several groups have already studied the applications of Raman spectroscopy in discriminating normal and malignant tissues in various body sites, such as lung, breast, bladder, prostate, cervix, skin and so on.[7–10] However, the Raman scattering cross section of most biological macromolecules is extremely small, which generally limits its potential uses.[11] This inefficient scattering requires longer signal collection times and the use of higher laser power for the acquisition of Raman spectra, which may result in damages to the biological samples.[12] Since its discovery in 1974,[13] surface-enhanced Raman scattering (SERS) has received more and more attention from researchers around the world, not only because of its high sensitivity and the small volume of sample needed,[14] but also due to the possible wide applicability.[15] Recently, some groups have already studied the applications of SERS for blood plasma analysis for disease including cancer detections and obtained many meaningful results.[4,16,17] However, in SERS studies of blood plasma, the SERS spectra of blood plasma will be influenced obviously due to the presence of anticoagulants according to our previous studies.[18] On the other hand, visible excitations usually cause photodecompositions as well as strong fluorescence background of the biological macromolecules. A way to avoid these questions is the employment of near-infrared (NIR) excitations. NIR excitation has a basic advantage in SERS detection and spectroscopy for the background, including fluorescence and Raman scattering of the surrounding medium or solvent, is able to be extremely decreased.[19] In addition, NIR excitation is nonresonant for most molecules which can be employed high excitation intensities up to saturation without photobleaching.[19] In this paper, we explored the applications of polyvinyl alcohol protected silver (PVA-Ag) nanofilm-based NIR-SERS for biochemical analysis of oxyheamoglobin (OxyHb) for liver cancer. Principal component analysis (PCA) combined with independent sample T test analysis were used to analyze and classify the OxyHb NIR-SERS spectra acquired from normal subjects and liver cancer patients. Meanwhile, the recovery situations of the liver cancer patients after surgery were assessed using the method of discriminant analysis-predicting group membership based on PCA. To the best of our knowledge, this is the first report on NIR-SERS analyses of OxyHb for liver cancer detection.

Materials and methods Subjects and methods Three groups were studied in our work. The first group consisted of 40 patients with confirmed clinical and histopathological diagnosis of liver cancers. There are 9 cases of liver cancer in stage I, 20 cases in stage II and 11 cases in stage III. The second group consists of 30 healthy volunteers. The third group consists of 30 liver cancer patients after surgery about 1–2 months. For more detailed information on these patients, please see the Supplementary Information (Table S1). All patients were from the Tumor Hospital and the People’s Hospital of Zhengzhou University, respectively. These patients had similar ethnic and socioeconomic backgrounds, and the ethical approval has been obtained in order to study the human OxyHb samples. The mean age for the liver cancer group was 51  12 years, for the surgeried cancer group was 49  11 years and for the control healthy group was 35  9 years. Preparation of PVA-Ag nanofilms Briefly, a mixture of 200 ml contained 20 mg silver nitrate and 100 mg PVA was prepared with deionized water. Meanwhile, two ultrasonically cleaned silver poles were used as two working electrodes. Electrochemical reduction was carried out for 1 h at 30 V (at 22  C). Then, PVA-Ag nanoparticles (NPs) were obtained. Based on the electrophoresis experiment, it was found that the PVA-Ag NPs were adsorbed with positive charges. Then, glass slides were ultrasonically cleaned in acetone, ethanol, and deionized water for 30 min, respectively, and dried in the atmosphere of high purity N2 (99.99%). After that, these dried glass substrates were dipped into the solution of Piranha (the volume ratio of 98% H2SO4 and 30% H2O2 is 7:3) for 30 min at 70  C. These treated glass slides were washed with deionized water for three times and dried in the atmosphere of high purity N2 again. Next, these dried glass slides were dipped into a mixture of 30% H2O2, 28% NH3 and deionized water in a volume ratio of 1:1:5 for 24 h. Then, the surfaces of these treated glass slides would take on the electronegative groups of [–O–Si–O–]. Last, these surface-modified glasses slides were dipped into the pre-prepared silver colloid (contained PVA-Ag NPs) for electrostatic self-assembly about 24 h, and then PVA-Ag nanofilms were obtained on the surfaces of the glass slides. Figure 1 shows the TEM image and UV–vis absorption spectrum of typical morphology of PVA-Ag NPs, the average particle size of PVA-Ag NPs is 45  8 nm, and their absorption

J. Raman Spectrosc. 2013, 44, 362–369

Copyright © 2013 John Wiley & Sons, Ltd.

wileyonlinelibrary.com/journal/jrs

363

Figure 1. TEM image of the colloidal Ag nanoparticles (a), SEM image of the PVA-Ag nanofilm prepared by using electrostatic self-assembly (b).

R. Liu et al. 2.5

600

2.0

(b)

Abs

1.5

290

420

1.0

0.5

(a) 0.0

-0.5 400

600

800

wavelength / nm Figure 2. UV–vis absorption spectra of the colloidal Ag nanoparticles (a) and the PVA-Ag nanofilm (b).

maximum locate at 418 nm. Figure 2(a) shows the representative SEM image of PVA-Ag nanofilm, the average size of the aggregated NPs on its surface is up to 200  50 nm. Especially, lots of nanoscale caves with the average size of ~ 300  50 nm are formed between the adjacent particles. Preparation of human OxyHb and NIR-SERS measurements To obtain human OxyHb, human blood samples obtained from the study subjects between 7: 00 and 8: 00 A.M were placed in glass tubes containing 200 ml of 3.2% sodium citrate aqueous solution act as anticoagulant. The plasma and white cells were removed after a low-speed centrifugation (2000 rpm, at 4  C for 10 min). The red cells were washed in phosphate buffered saline three times and centrifuged again (2000 rpm, at 4  C for 10 min), and 100 ml of concentrated red cells were lysed by adding deionized water in a volume ratio of 20:1 (water/packed cells). Then, the dilute OxyHb solution was centrifuged at 6000 rpm for 15 min to remove cell debris and stored at 4  C for experimental needed.[20] Then, 50 ml OxyHb was dropped onto the surfaces of the PVA-Ag nanofilms with a diameter of 1 cm and dried at 4  C for NIR-SERS analysis. The NIR-SERS spectra were recorded with a portable Raman spectrometer (R-3000TM, Ocean Optics Co., USA) in the range of 200–2000 cm 1 under a 785-nm diode laser excitation. The 785-nm radiation from a 65-mW air-cooled diode laser was used as the excitation source. The laser light was vertically projected onto the samples with a resultant beam intensity of ~103 W.cm 2, and the integration time was 16 s. Statistical analysis

364

PCA is an effective tool to simplify complex data sets and determine the key variables in a multidimensional data set that best explain the differences in the observations.[16] Independent sample T test determines the discriminant function line that maximizes the variance in the data between groups while minimizing the variance between members of the same group. To test the capability of OxyHb NIR-SERS spectra for distinguishing liver cancer from normal, we employed PCA combined with independent sample T test to classify the measured NIR-SERS spectra. To obtain more

wileyonlinelibrary.com/journal/jrs

accurate results, the fluorescence background of the original NIR-SERS data was removed using the method employed by Feng et al.[16] and Zhao et al.,[21] then each Raman band area was normalized by the total Raman band areas of the NIR-SERS spectrum, and then data in the spectral range of 470–1600 cm 1 was analyzed by the PCA method using SPSS software. PCA was performed on the normalized spectral data matrices to generate PCs comprising a reduced number of orthogonal variables that accounted for most of the total variance in the original spectra. Each PC was related to the original normalized Raman band area by a parameter called the PC score, representing the weight of that particular component against the basis. Meanwhile, independent sample T test was employed to identify diagnostically significant PC scores for each case by using the probability value of P < 0.05. The results show that three significant PC scores (PC1, PC2 and PC7) were able to show the most significant differences between the healthy subjects and the liver cancer group. By using the method of discriminate analysis, we got the specificity and sensitivity with the value is 85.7% and 95.0%, respectively, and the total correct rate is up to 91.4%. In addition, we also predicted the group membership of the 30 liver cancer patients after surgery and estimated the recovery situation of these patients using the method of discriminant analysis-predicting group membership.

Results Results of TEM, SEM and UV–vis absorption spectra The TEM image and UV–vis absorption spectrum of the silver colloid prepared using electrolysis is shown in Fig. 1 (a), the average size of the colloidal Ag NPs is ~50  8 nm. The absorption maximum of these Ag NPs is located at ~420 nm as shown in Fig. 2 (a). When the excitation wavelength is near 420 nm, the resonance Raman scattering would be expected to occur.[22] The previous work has reported that the intense band at ~395 nm in pure Ag colloids is a characteristic of the plasma resonance absorption for Ag spheres in water,[23] and the band at ~400 nm in the UV–vis absorption spectrum will undergo red shift with the increasing of the particles size.[24] Figure 1 (b) shows the SEM images of PVA-Ag nanofilm prepared by using the method of electrostatic self-assembly. The PVA-Ag NPs assembled densely and formed different layers on the surface of the glass slide. The average size of the aggregated Ag NPs on the surface layer is ~200 50 nm. Important, lots of nano-scale caves with an average size of 300 50 nm are formed between the adjacent PVA-Ag NPs. Thus, ‘hot spots’[25] will be formed, and the local electromagnetic fields in these hot spots will present a strong enhancement. The UV–vis adsorption spectrum of the PVA-Ag nanofilm has also been studied, as shown in Fig. 2 (b). One can see that, the plasmonic resonance band of this PVA-Ag nanofilm is broad (400–900 nm), which extends to the infrared optical region (780–2526 nm). This plasmonic characteristic works well with the excitation wavelength laser employed in this work (785 nm). In addition, the 785-nm laser source is able to avoid the photodissociation and fluorescence excitation of the biological macromolecules effectively.[26] Results of SERS measurements The regular Raman spectrum and NIR-SERS spectra of OxyHb samples from the same healthy subject were recorded in order to assess the PVA-Ag nanofilm enhancement effects on the Raman scattering of human OxyHb. The three spectra shown in

Copyright © 2013 John Wiley & Sons, Ltd.

J. Raman Spectrosc. 2013, 44, 362–369

NIR-SERS studies on OxyHb of liver cancer

1586

1335 1430

1124 1211

(c)

(b) (a) 300

600

900

1200

1500

1800

Wavenumber/cm-1 Figure 3. The regular Raman spectrum of the OxyHb sample from a healthy person (a), SERS spectrum of the OxyHb sample from the same person obtained by mixing the OxyHb sample with Ag colloid at a 1:1 proportion (b) and NIR-SERS spectrum of the same OxyHb sample adsorbed on PVA-Ag nanofilm (c).

1588

1336 1428

816 913 950 1006

660

1124 1213

476

0.8

304

Raman Intensity

(A)

and the OxyHb molecules. Because of this interaction, OxyHb molecules can closely adsorbed on the Ag NP surface; meanwhile, OxyHb molecules will be embedded into the nano-scale caves via some groups and adsorbed onto the hot spots induced by these nano-scale caves where there existed strong local electromagnetic fields, which leads to an extraordinary enhancement in the intensity of the Raman scattering.[27] To verity the spectroscopy reproducibility of the PVA-Ag nanofilms, the batch-to-batch variation has been evaluated using the reporter molecule of OxyHb. Four batches of PVA-Ag nanofilm were prepared using the same method. For each batch, three substrates were selected randomly for the NIR-SERS detection of OxyHb. Figure 4A displays the typical NIR-SERS spectra of OxyHb (1.5%) adsorbed on the different PVA-Ag nanofilms obtained from different batches. The error bars of area ratio for Raman bands at ~816/476, 1213/476, 1336/476, 1428/476 and 1588/476 cm 1 for the OxyHb molecules adsorbed on the different PVA-Ag nanofilms are shown in Fig. 4B. It is found that the maximum error of these area rations is lower than 20%. Thus, the spectroscopy reproducibility of these PVA-Ag nanofilms is perfect. Figure 5A displays all the NIR-SERS spectra of OxyHb samples obtained from 30 normal, 30 liver cancer patient after surgery and 40 liver cancer patients. Shown on the top are the NIR-SERS spectra for normal subject OxyHb (green), in the mid are the NIR-SERS spectra for surgeried liver cancer patient OxyHb (black) and the NIR-SERS spectra of liver cancer patient are shown on the bottom (red). It can be seen that while mean NIR-SERS spectral differences exist between normal, cancer and cancer after surgery, as shown in Fig. 5B. NIR-SERS peaks at ~473, 817, 1123, 1212, 1330, 1430 and 1581 cm 1 can all be observed in both normal and cancer OxyHbs, with the strongest signals at 473, 817 and 1581 cm 1. The areas of the NIR-SERS peaks at 473, 817 and 1581 cm 1 increases in the order normal OxyHb > liver cancer after surgery OxyHb > liver cancer OxyHb. Especially, the NIR-SERS band at 817 cm 1 is almost disappeared in liver cancer OxyHb samples. Figure 6 is a plot of the area values of select NIR-SERS bands from individual spectra, and their mean value with associated standard deviations. All these selected NIR-SERS bands have significantly different (p < 0.05) mean areas as determined by Student’s t-test analysis. The obvious differences between normal and liver cancer OxyHb can be found in the peaks at ~817, 960, 1123, 1212, 1230 and 1430 cm 1.

Values of the area ratios / a.u.

660

816 915 950 1006

308

Raman Intensity

473

Fig. 3(a)–(c) were measured under the same instrumentation setup. The laser light was vertically projected onto the samples with a resultant beam intensity of 103 W.cm 2. The integration time was 16 s. Figure 3(a) shows the regular Raman spectrum of the OxyHb solution, Fig. 3(b) is the SERS spectrum of the same OxyHb solution mixed with Ag colloid (the volume ration is 1: 1). One can see that, it is difficult to obtain effective Raman signals from OxyHb solution and mixed with Ag colloid. As we all know, the Raman scattering cross section of OxyHb molecule is relative small while the autofluorescence background of this molecule is strong. Thus, it is difficult to obtain effective Raman signals. However, strong Raman scattering signals from OxyHb sample adsorbed on PVA-Ag nanofilm is observed in the region of 400–1600 cm 1, which demonstrates that the intensity of many dominant vibrational bands have been increased dramatically by NIR-SERS, indicating strong interactions between PVA-Ag nanofilm

(B)

A

/A 816 476

0.7

A /A 1213 476

0.6

A

A /A 1336 476 /A 1428 476

A /A 1588 476

0.5 0.4

0.2

0.1

0.0 200 400 600 800 10001200140016001800

1

Wavenumber/cm-1

2

3

4

Batches

J. Raman Spectrosc. 2013, 44, 362–369

Copyright © 2013 John Wiley & Sons, Ltd.

wileyonlinelibrary.com/journal/jrs

365

Figure 4. (A) Typical NIR-SERS spectra of OxyHb (1.5%) adsorbed on the different PVA-Ag nanofilms of different batches 1–4. (B) Error bars of area ratio for Raman bands at ~816/476, 1213/476, 1336/476, 1428/476 and 1588/476 cm 1 in the NIR-SERS spectra of OxyHb (1.5%) adsorbed on the different PVA-Ag nanofilms.

(A)

(B)

473

R. Liu et al. (a)

200 400 600 800 100012001400160018002000

Wavenumber / cm-1

200

400

600

1581 1580

1212

1330 1430 1337 1423

1123 1122 1208

663 720 817 915 960 1006 658 716 816

474

308 309

Raman Intensity

Raman Intensity

(c)

800 1000 1200 1400 1600

Wavenumber / cm-1

Figure 5. (A) NIR-SERS spectra of OxyHb for 30 normal persons (green, top), 30 liver cancers after surgery (black, mid) and 40 liver cancer patients (red, bottom). The shaded areas represent the standard deviations of the means. (B) Mean NIR-SERS spectra of OxyHb for normal (a), liver cancer after surgery (b) and liver cancer (c). This figure is available in colour online at wileyonlinelibrary.com/journal/jrs

Raman Band Area (a.u.)

140000 120000 100000 80000 60000 40000 20000 0 -20000 400

600

800

1000

1200

1400

1600

Wavenumber / cm-1 Figure 6. The area values of the selected bands with the most distinguishable differences between normal human OxyHb (green square) and liver cancer OxyHb (red star). The corresponding mean values and standard deviations are also displayed on the right side. This figure is available in colour online at wileyonlinelibrary.com/journal/jrs

The 663, 1212 and 1430 cm 1 NIR-SERS bands of normal OxyHb undergo red shifts to 658, 1208 and 1423 cm 1, respectively, while 1330 cm 1 NIR-SERS band has a blue shift to 1337 cm 1 in the mean NIR-SERS spectrum of liver cancer OxyHb. Result of statistical analysis

366

To test the capability of OxyHb NIR-SERS spectra for discriminating liver cancer from normal, the method of PCA combined with independent sample T test was performed on the measured OxyHb NIR-SERS spectra. PCA is a statistical technique for simplifying complex data sets and determining the key variables in a multidimensional data set that best explain the differences in the observations.[16] First, the fluorescence background of the original NIR-SERS data was removed according to the method used by Zhao et al.,[21] and then each Raman band area was normalized by the total Raman band areas of the NIR-SERS spectrum for each OxyHb sample. Finally, all the normalized Raman band areas were fed into the SPSS software package for PCA analysis, and seven PCs in total were obtained in this work.

wileyonlinelibrary.com/journal/jrs

Independent sample T test on all PCs comparing normal and cancerous groups showed that the PCs 1, 2 and 7 were most diagnostically significant (p < 0.0001, p < 0.0001 and p < 0.0001) for discriminating normal and cancerous groups. Figure 6 illustrates the employment of PC scores for diagnostic classification, direct comparisons between normal and liver cancer groups. The scatter plot of the PC1 versus the PC7 of the NIR-SERS spectra for the normal and liver cancer groups is shown in Fig. 7A, and the dotted line (PC1 = 0.667PC7 - 0.128) as diagnostic algorithm separates the two groups with sensitivity of 97.5% and diagnostic specificity of 94.3%, the overall diagnostic accuracy was 95.4%. In this plot, 30 normal OxyHb NIR-SERS spectra were compared with 40 liver cancers OxyHb NIR-SERS spectra. Another comparison of the PCs 2 and 7 of the NIR-SERS spectra for the normal and liver cancer groups is shown in Fig. 7B, and the dotted line (PC2 = 1.00PC7 + 0.338) as diagnostic algorithm separates the two groups with sensitivity of 95.0% and diagnostic specificity of 83.3%, the overall diagnostic accuracy was 90.0%. Figure 7C shows the 3D scatter plot of the PCs 1, 2 and 7 of the NIR-SERS spectra for the normal and liver cancer groups. The result showed that the OxyHb NIR-SERS spectra can be employed for liver cancer detection with high sensitivity and specificity. In addition, we also assessed the recovery situation of the surgeried liver cancer patients by using the method of discriminant analysis combined with predicting group membership, according to the PC scores of PC1, PC2 and PC7. There are 26.7% surgeried liver cancers distinguished as the normal subjects and 63.3% surgeried liver cancers distinguished into the cancer group.

Discussion Our study demonstrated that there are specific differences in NIR-SERS spectra between cancerous and normal OxyHb, indicating great potential for OxyHb NIR-SERS in liver cancer detection and screening applications. One can see that, the NIR-SERS spectra of human OxyHb are dominated by a number of characteristic vibration modes in the region of 400–1600 cm–1, which may be changed in intensity or location associated with liver cancer developments. Table 1 shows the preliminary assignments for the observed

Copyright © 2013 John Wiley & Sons, Ltd.

J. Raman Spectrosc. 2013, 44, 362–369

NIR-SERS studies on OxyHb of liver cancer

Figure 7. (A) Plots of PC1 versus PC7 for normal group versus liver cancer group. The dotted line of PC1 = 0.667PC7 0.128 as diagnostic algorithm can separate the two groups very well. (B) Plot of PC2 versus PC7 for normal group versus liver cancer group. The dotted line of PC2 = 1.00PC7 + 0.338 as diagnostic algorithm can separate the two groups very well. (C) The 3D mapping of PC1, PC2 and PC7 for the normal group (green square) and the liver cancer group (red star). This figure is available in colour online at wileyonlinelibrary.com/journal/jrs Table 1. Tentative assignment for the mean NIR-SERS bands of OxyHb (based on Refs.28–31) Normal 473 660 720 817 915 960 1006 1123 1212 1330 1430 1581

Liver cancer

Tentative assignments

474 658 716 816

pyr. Fold ds (pyr, deformation) das (pyr, deformation) n (pyr, breathing) globin modes globin modes Phenyl modes of Phe, Trp and Tyr nas(pyr, half ring) d(CmH) ns(pyr, half ring) ns (CaCm) nas (CaCm)

1122 1208 1337 1423 1580

J. Raman Spectrosc. 2013, 44, 362–369

Copyright © 2013 John Wiley & Sons, Ltd.

wileyonlinelibrary.com/journal/jrs

367

NIR-SERS bands according to the literature data,[28–31] in order to better understand the molecular basis for the observed NIR-SERS of human OxyHb. We know that OxyHb is a kind of protein that undergoes conformational or domain changes on experiencing a chemical reaction.[32] Hence, ligand binding OxyHb, the main component of the cytoplasm of the red blood cell, is a globular protein that facilitates oxygen transport via binding to heme groups. The bands at 1581 cm 1 can be assigned to the asymmetric

stretching vibrations between the Ca atoms of the pyrrole rings and the Cm atom connecting the rings. The band at 1430 cm 1 is assigned to the symmetric stretching vibration between the Ca and the Cm atoms. The 1290 and 1212 cm 1 bands are attributed to the deformation vibrations of the Cm–H atoms between the pyrrole rings. The spectra also show bands at 1330 and 1123 cm 1 from symmetric and asymmetric stretching vibrations of the pyrrole rings. The 663 cm 1 band belong to the ring deformation of pyrrole, and the 817 cm 1 band is due to the ring breathing of pyrrole. The weaker band at 1006 cm 1 is correlated to the phenyl modes of the aromatic aminoacid residues, namely, Phe, Trp and Tyr in the globin part of oxyhemoglobin.[31] This band is unobserved in the NIR-SERS spectra of liver cancers. Meanwhile, we also observed another two Raman bands at 915 and 965 cm 1, which can be also correlated to globin modes, but its nature could not be decided.[31] These two weaker bands were also absent in the NIR-SERS spectra of liver cancers. However, there are some disagreements of the spectroscopic features including the number and peak position of the Raman bands between the NIR-SERS data and the previous Raman spectra.[31] The authors hold that the selection rules[33] for SERS enhancement and the interactions, such as charge transfer[34,35] (though we do not know accurately how this process occurred at present) between the OxyHb molecules and the silver atoms play a key role. Distinctive NIR-SERS spectral features and intensity differences for liver cancer and normal groups could reflect molecular

R. Liu et al.

Figure 8. Plots of discriminate scores of the normal control group (A) and the liver cancer group (B).

368

changes associated with malignant transformation. For instance, the NIR-SERS bands at ~663, 817 and 1581 cm 1 of OxyHb are lower percentage signal in cancer group, suggesting that there is abnormal metabolism of OxyHb in the blood plasma of liver cancer patients. In liver cancer OxyHb, both of the two NIR-SERS bands at 663 and 817 cm 1 are much reduced in the normalized area, indicating that liver cancer may be associated with a decrease in the relative amounts of the symmetric bending and stretching vibration modes of pyrrole ring. Similarly, the normalized area of 1581 cm 1 band is also reduced obviously in liver cancer OxyHb (Fig. 5B), which indicates that the relative amount of the antisymmetric stretching vibration mode of the CaCm group is decreased in cancer OxyHb. In addition, the NIR-SERS bands in the spectral region of 1200–1300 cm 1 of liver cancer OxyHb are different from those of normal OxyHb (Fig. 5B), including the 663, 1212 and 1430 cm 1 NIR-SERS bands of normal OxyHb red shift to 658, 1208 and 1423 cm 1, respectively, while the 1330 cm 1 NIR-SERS band has a blue shift to 1337 cm 1 in the mean NIR-SERS spectrum of liver cancer OxyHb. These Raman bands are mostly associated with the vibrations of the pyrrole ring and the CaCm group of OxyHb. Thus, compared to normal control subjects, the vibrational modes and the conformations for some groups, especially for the pyrrole ring and the CaCm group of OxyHb underwent some changes in the development process of liver cancer. Meanwhile, the bands at ~915, 965 and 1006 cm 1 were disappeared in the NIR-SERS spectra of liver cancers, indicating a decrease in the percentage of aromatic amino acids such as Phe, Trp and Tyr in the in the globin part of OxyHb of liver cancer patients, according to the preliminary assignments[31] of the observed Raman bands. This result matches with the previous results reported on blood plasma of nasopharyngeal cancer patients by Feng et al.[16] In recent years, some research groups all around the world have developed simple but effective diagnostic methods based on the analysis of Raman spectra in terms of peak intensity or peak intensity ratio measurements, which were used to classify normal and cervix,[36] breast[37] and colorectal[4] cancers. For example, the ratio of Raman band intensities at 725 cm 1 and 638 cm 1 was considered as an important fingerprint for diabetes mellitus diagnosis at the serum level by Han et al.[38] and Lin et al.[4] In this present work, PCA was employed to reduce the large amount of data contained in the measured NIR-SERS spectra into a few important PCs, and it is shown that the PC scores of 1, 2 and 7 stand for the normalized NIR-SERS bands at 473, 817 and

wileyonlinelibrary.com/journal/jrs

1581 cm 1, which were explored to classify liver cancer and normal OxyHb samples (Fig. 7(C)). One can see that, the three PC scores for normal and liver cancer were distributed in separate areas, indicating that they are able to be used to discriminate between the NIR-SERS spectra of the liver cancer group and the normal group. The specificity and sensitivity of this method for identifying liver cancer from the OxyHb NIR-SERS analysis are calculated to be 85.7% and 95.0%, respectively. Figure 8 shows the plots of discriminate scores for the normal control group (Fig. 8A) and the liver cancer group (Fig. 8B) based on the PC1, PC2 and PC7, using the method of discriminate analysis. One can see that the discriminate scores of the healthy volunteers are mainly in the region of 3.5–0.0, while the discriminate scores of the liver cancer patients are mainly in the region of 0.0–3.0. Thus, the discriminate score of zero is able to distinguish the liver cancer from the normal. Meanwhile, the standard deviation of the discriminate scores of the healthy control group (1.185) is greater than that (0.837) of the liver cancer group, indicating the distribution of the normal control group is wider than that for the liver cancer group, according with the result shown in Fig. 7. This is may be explainable. For the normal group, the concentration of the OxyHb will be different from each other to some extent, which may be caused by the different individual characteristics among them such as different constitutions, dietary structures and so on. However, for

Figure 9. PCA score plots corresponding to principal components (PCs) 1, 2 and 7 for the NIR-SERS spectra acquired from the OxyHb samples of the healthy volunteers, the liver cancer patients (red star) and the liver cancer patients after surgery (blue triangle). This figure is available in colour online at wileyonlinelibrary.com/journal/jrs

Copyright © 2013 John Wiley & Sons, Ltd.

J. Raman Spectrosc. 2013, 44, 362–369

NIR-SERS studies on OxyHb of liver cancer the liver cancer patients, the abnormal growth of the tumor tissue requires a lot of nutrients to the epithelial mucosa due to microvascular proliferation, which will induce the oxygen content to greatly reduce, resulting in the decrease of OxyHb;[39] at the same time, the disordered metabolism of liver cancer patients will also induce the OxyHb in the tumor tissue to significantly lower.[39] The OxyHb concentration of the liver patients is usually lower than that of normal group and has a relative narrower distribution in spite of different individual characteristics. Thus, the NIR-SERS spectrum distribution of the normal control group will be wider than that of the liver cancer group, which is consistent with the experimental results (Fig. 5A). Based on the PCA results above, we also differentiated the recovery situation of 30 liver cancer patients after surgery using the method of discriminant analysis-predicting group membership. The result shows that about 26.7% surgeried liver cancer patients were distinguished as the normal subjects, and 63.3% were distinguished into the cancer group (Fig. 9), indicating the recovery situation of most surgeried liver cancer patients is not very good. However, the recovery situations of the liver cancer patients after surgery were not analyzed compared with other diagnostic methods in this article, which is a limitation of our paper.

Conclusions In summary, PVA-Ag nanofilm based NIR-SERS was employed to analyze the OxyHb obtained from liver cancer patients, surgeried liver cancer patients and healthy volunteers in this paper. Using PCA multivariate analysis combined with independent sample T test, the liver cancer was able to be differentiated from normal with high diagnostic specificity (85.7%) and sensitivity (95.0%). Meanwhile, the tentative assignments of the Raman bands in the measured NIR-SERS spectra of OxyHb were performed, and the results displayed cancer specific changes on molecule level, including a decrease in the relative concentrations and the percentage of aromatic amino acids of OxyHb, changes of the vibration modes of the CaCm group and pyrrole ring of OxyHb of liver cancer patients compared to those of the healthy subjects. These alterations may be caused by the body; especially, the liver organ of the cancer patient underwent some metabolic changes. The results show that the method of PCA combined with independent sample T test can discriminate the OxyHb samples of liver cancer patients from those of healthy volunteers, which is able to separate these samples into two distinct clusters. Meanwhile, the recovery situation of the surgeried liver cancer patients can also be assessed based on this method. The results from this study demonstrated tremendous promise for the development of NIR-SERS OxyHb analysis into a clinical tool for non-invasive detection and screening of liver cancers. Acknowledgements This work was supported by the National Natural Science Foundations of China (Nos. 11064001 and 10864001), and the Science and Technology Project of Yunnan Province (Nos. 2008ZC159M). Supporting information Supporting information can be found in the online version of this article.

References [1] D. M. Parkin, Lancet Oncol. 2001, 2, 533. [2] D. Habermehl, K. Haase, S. Rieken, J. Debus, S. E. Combs, Tumori 2011, 97, 609. [3] D. M. Parkin, F. Bray, J. Ferlay, P. Pisani, Int. J. Cancer 2001, 94, 153. [4] D. Lin, S. Y. Feng, J. J. Pan, Y. P. Chen, J. Q. Lin, G. N. Chen, S. S Xie, H. S. Zeng, R. Chen, Opt. Express 2011, 19, 13565. [5] G. Malaguarnera, M. Giordano, I. Paladina, M. Berretta, A. Cappellani, M. Malaguarnera, Dij. Dis. Sci. 2010, 55(10), 2744. [6] A. Kudelski, Talanta 2008, 76(1), 1. [7] Z. Huang, A. McWilliams, H. Lui, D. I. McLean, S. Lam, H. Zeng, Int. J. Cancer 2003, 107(6), 1047. [8] K. E. Shafer-Peltier, A. S. Haka, M. Fitzmaurice, J. Crowe, J. Myles, R. R. Dasari, M. S. Feld, J. Raman Spectrosc. 2002, 33(7), 552. [9] P. Crow, A. Molckovsky, N. Stone, J. Uff, B. Wilson, L. M. WongKeeSong, Urology 2005, 65(6), 1126. [10] U. Utzinger, D. Heintzelman, A. Mahadevan-Jansen, A. Malpica, M. Follen, R. R. Kortum, Appl. Spectrosc. 2001, 55(8), 955. [11] R. Kumar, H. Zhou, S. B. Cronin, Appl. Phys. Lett. 2005, 91, 223105. [12] M. Kahraman, M. M. Yazıcı, F. Şahin, M. Ģulha, Langmuir 2008, 24, 894. [13] M. Fleischmann, P. J. Hendra, A. J. McQuillan, Chem. Phys. Lett. 1974, 26, 163. [14] P. Etchegoin, R. C. Maher, L. F. Cohen, H. Hartigan, R. J. C. Brown, M. J. T. Milton, J. C. Gallop, Chem. Phys. Lett. 2003, 375, 84. [15] K. Kneipp, Y. Wang, H. Kneipp, L. T. Perelman, I. Itzkan, R. R. Dasari, M.S. Feld, Phys. Rev. Lets. 1997, 78, 1667. [16] S. Y. Feng, R. Chen, J. Q. Lin, J. J Pan, Y. N. Wu, Y. Z. Li, J. X. Chen, H. S. Zeng, Biosens. Bioelectron. 2011, 26, 3167. [17] H. Han, X. Yan, R. Dong, G. Ban, K. Li, Appl. Phys. B 2009, 94(4), 667. [18] R. M. Liu, X. F. Zi, Y. P. Kang, M. Z. Si, Y. C. Wu, J. Raman. Spectrosc. 2011, 42, 137. [19] K. Kneipp, H. Kneipp, V. B. Kartha, R. Manoharan, G. Deinum, I. Itzkan, Phys. Rev. E 1998, 57, 6281–6284. [20] E. Demoll, D. J. Cox, E. Daniel, A. F. Riggs, Anal. Biochem. 2007, 363, 196. [21] J. Zhao, H. Lui, D. I. Mclean, H. Zeng, Appl. Spectrosc. 2007, 61(11), 1225. [22] G. L. Liu, L. P. Lee, Appl. Phys. Lett. 2005, 87, 074101. [23] S. S. Cortes, J. V. G. Ramos, G. Morcillo, A. Tinti, J. Colloid Interface Sci. 1995, 175, 358. [24] Y. Du, Y. Fang, Spectrochim. Acta. A. 2004, 60, 535. [25] H. X. Xu, M Käll, Phys. Rev. Lett. 2002, 89, 246802. [26] G. J. Puppels, J. H. F. Olminkhof, G. M. J. Segers-Nolten, C. Otto, F. F. M. Demul, J. Greve, Exp. Cell Res. 1991, 195, 361. [27] R. M. Liu, M. Z. Si, Y. P. Kang, X. F. Zi, Z. Q. Liu, D. Q. Zhang, J. Colloid Interface Sci. 2010, 34, 52. [28] M. Abe, T. Kitagawa, Y. Kyogoku, J. Chem. Phys. 1978, 69, 4526. [29] S. Z. Hu, K. M. Smith, T. G. Spiro, J. Am. Chem. Soc. 1996, 118, 12638. [30] R. Gessner, C. Winter, P. Rösch, M. Schmitt, R. Petry, W. Kiefer, M. Lankers, J. Popp, Chemphyschem 2004, 5, 1159. [31] B. Venkatesh, S. Ramasamy, M. Mylrajan, R. Asokan, P. T. Manoharan a, J. M. Rifkind, Spectrochim. Acta A. 1999, 55, 1691. [32] S. Nie, S. R. Emory, Science 1997, 275, 1102. [33] J. F. Li, S. Duan, P. P. Fang. Proceedings of the XXI-st International Conference on Raman Spectroscopy. Brunel University, West London, United Kingdom: IM, 2008, 65–66. [34] S. Schlücker, R. K. Singh, B. P. Asthana, J. Popp, W. Kiefer, J. Phys. Chem. A 2001, 105(43), 9983. [35] M. T. Sun, S. B. Wan, Y. Y. Liu, Y. Jia, H. X. Xu, J. Raman Spectrosc. 2008, 39, 402. [36] U. Utzinger, D. Heintzelman, A. M. Jansen, A. Malpica, M. Follen, R. Richards-Kortum, Appl. Spectrosc. 2001, 55(8), 955. [37] J. L. P. Molina, C. F. Reyes, O. B. García, R. H. Franco, J. L. G. Trujillo, C. A. R. Alvarado, G. G. Juárez, C. M. Gutiérrez, Lasers Med. Sci. 2007, 22(4), 229. [38] H. W. Han, X. L. Yan, R. X. Dong, G. Ban, K. Li, Appl. Phys. B 2009, 94(4), 667. [39] Jain R. K., Cancer Res. 1988, 48(10), 2641.

369

J. Raman Spectrosc. 2013, 44, 362–369

Copyright © 2013 John Wiley & Sons, Ltd.

wileyonlinelibrary.com/journal/jrs