Characterizing Variability of In Vivo Raman Spectroscopic Properties

Dec 15, 2014 - Khay Guan Yeoh,. § and Zhiwei Huang*. ,†. †. Optical Bioimaging Laboratory, Department of Biomedical Engineering, Faculty of Engin...
0 downloads 0 Views 1MB Size
Subscriber access provided by ANADOLU UNIVERSITY

Article

Characterizing variability of in vivo Raman spectroscopic properties of different anatomical sites of normal colorectal tissue toward cancer diagnosis at colonoscopy Mads Sylvest Bergholt, Wei Zheng, Kan Lin, Jianfeng Wang, Hongzhi Xu, Jian-lin Ren, Khek Yu Ho, Ming Teh, Khay Guan Yeoh, and Zhiwei Huang Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/ac503287u • Publication Date (Web): 15 Dec 2014 Downloaded from http://pubs.acs.org on December 22, 2014

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Analytical Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 32

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Characterizing variability of in vivo Raman spectroscopic properties of different anatomical sites of normal colorectal tissue toward cancer diagnosis at colonoscopy Mads Sylvest Bergholt1, Wei Zheng1, Kan Lin1, Jianfeng Wang1, Hongzhi Xu2, Jian-lin Ren2, Khek Yu Ho3, Ming Teh4, Khay Guan Yeoh3, Zhiwei Huang1* 1

Optical Bioimaging Laboratory, Department of Biomedical Engineering, Faculty of Engineering, National University of Singapore, Singapore 117576

2

Institute of Digestive Disease, Department of Gastroenterology, Zhongshan Hospital affiliated to Xiamen University, Xiamen 361004 3

Department of Medicine, Yong Loo Lin School of Medicine, National University of Singapore and National University Health System, Singapore 119260

4

Department of Pathology, Yong Loo Lin School of Medicine, National University of Singapore and National University Health System, Singapore 119074

*Correspondence to: Dr. Zhiwei Huang Optical Bioimaging Laboratory Department of Biomedical Engineering Faculty of Engineering National University of Singapore 9, Engineering Drive 1 Singapore 117576 Tel: +65- 6516-8856 Fax: +65- 6872-3069 Email: [email protected] Running title: In vivo Raman spectroscopy for colorectal cancer diagnosis Keywords: Raman spectroscopy, in vivo diagnosis, colorectal cancer

1

ACS Paragon Plus Environment

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ABSTRACT: This study aims to characterize the in vivo Raman spectroscopic properties of normal colorectal tissues and to assess distinctive biomolecular variations of different anatomical locations in the colorectum for cancer diagnosis. We have developed a novel 785 nm excitation fiber-optic Raman endoscope that can simultaneously acquire in vivo fingerprint (FP) spectra (800−1800 cm-1) and highwavenumber (HW) Raman spectra (2800−3600 cm-1) from the subsurface of colorectal tissue. We applied the FP/HW Raman endoscope for in vivo tissue Raman measurements of various normal colorectal anatomical locations (i.e. ascending colon (n=182), transverse colon (n=249), descending colon (n=124), sigmoid (n=212), and rectum (n=362)) in 50 subjects. Partial least squares (PLS) - discriminant analysis (DA) was employed to evaluate the inter-anatomical variability. The normal colorectal tissue showed a subtle inter-anatomical variability in molecular constituents (i.e., proteins, lipids and water content) and could be divided into three major clusterings: (1) ascending colon, transverse colon, (2) descending colon, and (3) sigmoid and rectum. The PLS-DA multiclass algorithms were able to identify different tissue sites with varying sensitivities (SE) and specificities (SP) (ascending colon: SE: 1.10%, SP: 91.02, transverse colon: SE: 14.06%, SP: 78.78, descending colon: SE: 40.32%, SP: 81.99, sigmoid: SE: 19.34%, SP: 87.90, rectum: SE: 71.55%, SP: 77.84). The inter-anatomical molecular variability was orders of magnitude less than neoplastic tissue transformation. Further PLS-DA modeling on in vivo FP/HW tissue Raman spectra yielded a diagnostic accuracy of 88.8% (sensitivity: 93.9% (93/99); specificity 88.3% (997/1129) for colorectal cancer detection. This work discloses that inter-anatomical Raman spectral variability of normal colorectal tissue is subtle compared to cancer tissue; and the simultaneous FP/HW Raman endoscopic technique has promising potential for realtime, in vivo diagnosis of colorectal cancer at the molecular level.

2

ACS Paragon Plus Environment

Page 2 of 32

Page 3 of 32

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

INTRODUCTION Colorectal cancer is the malignancies with high mortality and morbidity1, which is the third most commonly diagnosed cancer in males and the second in females with over 1.2 million new cancer cases and 608,700 deaths estimated to occur annually.2 Early identification and eradication of precancerous polyps (i.e., adenoma) and cancers in the curable stages are the critical measures to reducing colorectal fatality.3 Modern flexible video colonoscopy remains the standard diagnostic approach for the clinicians. Although colonoscopic screening has significantly increased the survival of colorectal patients, it still suffers from fundamental clinical limitations.4 In particular, it remains clinically unambiguous challenges to distinguish adenoma and early adenocarcinomas from benign hyperplastic polyps in vivo. This is because conventional white light reflectance (WLR) colonoscopy heavily relies on the subjective visual assessments of colorectal polyps (e.g., polyp morphology, surface pit patterns etc.). Current clinical guidelines therefore recommend resection of all polyps identified during colorectal examinations. This approach is exceedingly labor intensive, and the histopathological analysis of small polyps (5 s) that was impractical for routine clinical endoscopic examinations. The diagnostic efficiency of FP Raman spectroscopy could be compromised in patients owing to intrinsically very weak tissue Raman signals and overwhelming tissue autofluorescence (AF) of the internal organs. Recent attentions have been directed toward the use of high-wavenumber (HW) regime (e.g., 2800–3600 cm-1), as HW spectral range exhibits stronger tissue Raman signals while having less background interference from both tissue AF and fiber-optic probes. There are multiple rationales for combining the FP and HW Raman techniques for in vivo tissue Raman measurements: (i) For tissues that could exhibit intense AF overwhelming the tissue FP Raman signals, the HW range could still contain intense tissue Raman peaks for tissue diagnosis; (ii) The FP and HW Raman spectra offer complementary biomolecular information, and combining FP/HW Raman technique could improve tissue characterization and diagnosis. Very recently, we have developed a novel second-generation 785-nm excitation Raman endoscope that can simultaneously measure both FP Raman spectra (i.e., 800−1800 cm-1) and high-wavenumber (HW) (i.e., 2800−3600 cm-1) Raman spectra in real-time (< 0.5 s) for improving in vivo tissue characterization and diagnosis. We have integrated the simultaneous FP/HW Raman technique with a 1.8-mm endoscope-compatible fiber-optic Raman probe coupled with a sapphire ball lens for probing biomolecular signatures from the subsurface of colorectal tissue associated with carcinogenic onset.21 Since different anatomical location of the

4

ACS Paragon Plus Environment

Page 4 of 32

Page 5 of 32

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

colorectum (e.g., ascending colon, transverse colon, descending colon, sigmoid and rectum) are functionally different, and the disease genetic and epigenetic features could also differ by locations in the colorectum22 , there is a need to understand whether the inter-anatomical Raman variability among different locations of normal rectum is significant, and if one should account for inter-anatomical variability in the interpretation and rendering of Raman diagnostic algorithms for in vivo colorectal tissue diagnosis and characterization. Hence, this study aims to characterize the variability of in vivo FP/HW Raman spectra of different anatomical locations of normal colorectal tissue for cancer diagnosis. We applied the simultaneous FP/HW Raman endoscopic technique developed for in vivo measurements of various colorectal anatomical sites (i.e. ascending colon, transverse colon, descending colon, sigmoid, and rectum) during colonoscopy. We employed multivariate statistical analysis (e.g., partial least squares (PLS)-discriminant analysis (DA)) to extract the molecular information for characterizing the molecular compositions of normal colorectal tissue, and compared the magnitude of inter-anatomical variability with that of cancerous tissue transformation for assessing the clinical implications in colorectal cancer diagnosis.

MATERIAL AND METHODS Raman Instrumentation. The fiber-optic Raman spectroscopic system consists of a NIR diode laser (λex = 785 nm), a high-throughput reflective imaging spectrograph (LS-785, Princeton Instruments Inc.,) equipped with a customized 830 gr/mm goldcoated reflective grating which has the diffraction efficiency of ~85% in the NIR region (800 − 1200 nm) and allows the integrated Raman endoscopy system to cover both the FP (800 to 1800 cm−1) and HW (2800 to 3600 cm−1) simultaneously for tissue Raman measurements. A thermo electric-cooled (-70°C), deep depletion CCD camera (PIXIS

5

ACS Paragon Plus Environment

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 32

400BR, Princeton Instruments Inc.,) with extended sensitivity in the NIR range (800 − 1200 nm) was used to detect the FP/HW Raman signals. To compensate for the image aberration and the decline in resolution owing to the broad spectral coverage, a customized parabolic-aligned fiber bundle (64 × 100 µm fibers, NA = 0.22) was coupled into the entrance slit of the spectrograph for hardware binning to significantly improve the signal-to-noise ratio (SNR) (20-fold improvement) as well as the spectral resolution of the Raman system as compared to a conventional straight slit imaging spectrograph.23 This permits in vivo FP Raman spectra (800 to 1800 cm−1) and HW Raman spectra (2800 to 3600 cm−1) to be measured simultaneously with spectral resolution of ∼9 cm-1. We have manufactured a 1.8-mm (in outer diameter) depthselective fiber-optic Raman probe compatible with conventional colonoscopes.

21

The

compact Raman endoscopic probe comprises 9 × 200 µm filter-coated bevelled collection fibers (NA = 0.22) surrounding the central light delivery fiber (200 µm in diameter, NA = 0.22). A miniature 1.0 mm sapphire ball lens (NA = 1.78) is coupled to the fiber tip of the probe to focus the excitation light onto the tissue surface, enabling efficient collection of Raman photons from subsurface molecular structures of colorectal tissues.24 Monte Carlo simulations of photon propagation in gastrointestinal tissue shows that the fiber-optic probe collects ~85% of the signal within ~200 µm in tissue depth.21 Each colorectal tissue Raman spectrum can be acquired within 0.5 sec with the 785-nm laser excitation power of ~12 mW on the tissue (equivalent to ~1.5 W/cm2 within the spot size of ~500 µm), which is less than the American National Standards Institute (ANSI) maximum permissible skin exposure limit set out for a 785-nm laser beam

25

.

Further finite difference thermal modeling26,27 based on the optical properties of colorectal tissue28, 29 shows that even without consideration of other cooling effects (e.g., perfusion and evaporation in tissue), the maximum tissue temperature rise is only about

6

ACS Paragon Plus Environment

Page 7 of 32

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

0.06°C after 1-min of 785-nm laser radiation with an incident power density of ~1.5 W/cm2 during tissue Raman measurements. This temperature rise estimated is far below the level to cause photothermal damage to tissue and cells 30, suggesting that the laser power used in this study is safe for in vivo tissue Raman measurements. The atomic emission lines of mercury−argon spectral calibration lamps (HG-1 and AR-1, Ocean Optics, Inc., Dunedin, FL) in the FP range and the Raman spectrum of 4acetamidophenol that exhibits strong well-defined peaks in the HW region at 2931 cm−1 and 3064 cm−1 (ASTM E1840 standard) are used for wavelength calibration. All wavelength-calibrated spectra are corrected for the wavelength dependence of the system, using a tungsten calibration lamp (RS-10, EG&G Gamma Scientific, San Diego, CA). The clinical Raman endoscopic system is controlled using a foot pedal in an intuitive parallel data framework that processes the FP and HW Raman spectra simultaneously with auditory probabilistic feedback to the gastroenterologist, pushing the frontier into clinical endoscopic examinations.31 Real-time Raman data processing. The in vivo FP/HW Raman spectra were preprocessed prior to multivariate statistical analysis. The raw Raman spectra measured from in vivo tissues represent a combination of weak tissue Raman signals, intense autofluorescence background, fiber-probe background and noise. The fiber probe background is first subtracted from the raw spectra. The tissue spectra are subsequently preprocessed by a first-order Savitzky Golay smoothing filter (window width of 3 pixels) to reduce the spectral noise. In the FP region (800−1800 cm), a fifth-order polynomial is fitted to the autofluorescence background in the noise-smoothed spectrum and this polynomial is then subtracted from the calibrated FP spectrum to yield the tissue Raman spectrum alone.31 In the HW range (2800−3600 cm-1), a 1st order linear fit was found to be optimal for removing the weak autofluorescence base-line.32 The two Raman spectra

7

ACS Paragon Plus Environment

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

are spliced into a complete broadband Raman spectrum covering the FP and HW ranges and subsequently normalized to the integrated area under the curve to reduce spectral measurement variations. Clinical trial protocol. This report represents part of the ongoing study aiming to define a cost effective screening for gastrointestinal cancers. This work was approved by the Institutional Review Board (IRB) of National Health Group of Singapore. Prior to Raman measurements, all patients signed an informed consent permitting the in vivo Raman spectroscopic measurements during colorectal examinations. The trial was performed in accordance with International Conference on Harmonization (ICH) for Good Clinical Practice (GCP) guidelines, Declaration of Helsinki (2000). Exclusion criteria for Raman examination included patients with comorbid illness, insufficient bowel preparation, or where biopsies were contraindicated. A total of 50 patients were enrolled in the lower gastrointestinal Raman endoscopic examinations for various colorectal indications (e.g., anemia, bleeding etc.). As this study was specifically focused on normal colorectal patients, we only selected and analyzed the normal tissue data from the 50 patients. For comparison purposes, we also analyzed Raman spectra from cancerous tissues (i.e., colorectal adenocarcinoma). Colorectal Raman endoscopy. Prior to colonoscopy, the patients were administered polyethylene glycol electrolytes (PEG) for bowel preparation. Sedation was performed using intravenous administered midazolam (2–5 mg). Standard examination of the colorectum was performed with high resolution WLR endoscopy. We defined the five major anatomical locations of the colorectum, that is: ascending colon (cecum to hepatic flexure), transverse colon (hepatic flexure to splenic flexure), descending colon (splenic flexure to sigmoid), sigmoid, and rectum, respectively. Before the Raman measurements, the colorectal tissues were flushed with physiological saline

8

ACS Paragon Plus Environment

Page 8 of 32

Page 9 of 32

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

solution to reduce confounding factors (e.g., stool residual). The fiber-optic Raman probe was placed in gentle contact with the apparently normal tissue, and multiple spectra (∼10) for each tissue site were measured with scanning times of 0.1 to 0.5 sec. After the Raman scans, biopsies were taken and each tissue specimen was fixed in formalin, and sent for histopathology assessment among three independent gastrointestinal pathologists that were blinded to the Raman results. Normal tissues were defined as absence of clinical significant pathology (i.e., inflammatory bowel disease, hyperplasia, adenoma, adenocarcinoma). Multivariate statistical analysis. Mean-centering was performed prior to multivariate statistical analysis to remove common variance from the dataset of in vivo colorectal tissue Raman spectra. Probabilistic partial least squares (PLS) discriminant analysis (DA) using one-against-all multiclass classification was applied to evaluate the inter-anatomical variability.7 Leave-one patient-out, cross-validation was used to validate and optimize the PLS-DA model complexity to the first local minimum in classification error.24 We extracted and visualized the distinct molecular features for each anatomical site through loadings and scores. Statistical significance among the PLS scores for the five anatomical locations (ascending colon, transverse colon, descending colon, sigmoid, and rectum) was calculated using one-way ANOVA and Fisher's least significant difference (LSD) with p values of less than 0.05. Multivariate statistical analysis was performed using in-house written scripts in the Matlab programming environment (Mathworks Inc, Natick, MA).

RESULTS Figure 1A shows the mean in vivo Raman spectra ±1 standard deviation (SD) measured from different normal colorectal tissues (ascending colon (n = 182), transverse

9

ACS Paragon Plus Environment

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

colon (n = 249), descending colon (n = 124), sigmoid (n = 212) and rectum (n = 362)) of 50 patients undergoing clinical colonoscopy. For comparison purposes, we have also plotted the mean in vivo Raman spectra ±1 SD of colorectal cancer (n = 99). The prominent tissue Raman peaks in the FP range with tentative vibrational mode assignments24,33,34 are observed at 956 cm-1 (v(C-C) proteins), 1004 cm-1 (νs(C-C) ring breathing of phenylalanine), 1078 cm-1 (ν(C-C) of lipids), 1265 cm-1 (amide III v(C-N) and δ(N-H) of proteins), 1302 cm-1 (CH2 twisting and wagging of lipids), 1335 cm-1 (nucleic acids, adenine, guanine), 1445 cm-1 (δ(CH2) deformation of proteins and lipids), 1618 cm-1 (v(C=C) of porphyrins), and 1655 cm-1 (amide I v(C=O) of proteins). Intense Raman peaks are also seen in the HW spectral range such as 2850 and 2885 cm−1 (symmetric and asymmetric CH2 stretching of lipids), 2940 cm−1 (CH2 stretching of proteins), 3009 cm−1 (asymmetric =CH stretching of lipids), ~3329 cm−1 (Amide A (NH stretching of proteins)) as well as the broad Raman band of water assigned to symmetric and asymmetric OH stretching vibrations peaking at ~3250 and ~3400 cm−1 that reflects the local structure of hydrogen-bonded networks in the epithelium.35,36 The complete Raman–based biomolecular characterization of the entire colorectal lumen uncovers the subtle inter-anatomical variability. Figure 1B shows agglomerative hierarchical Euclidean clustering dendrogram of the mean Raman spectra (i.e., ascending colon, transverse colon, descending colon, sigmoid and rectum) as well as the comparison with that of cancerous tissue transformation. The normal anatomical sites could be divided into three major clusterings: i.e., (1) ascending colon, transverse colon, (2) descending colon, and (3) sigmoid and rectum. To further explore the specific molecular features among different colorectal tissue, we also plotted the mean difference spectra ± 1SD (Figure 2A-2B). The subtle inter-anatomical variability (i.e., relative intensity changes and band-broadening) can be discerned. We then compared the spectral variation of

10

ACS Paragon Plus Environment

Page 10 of 32

Page 11 of 32

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

normal tissues with that of cancerous tissue transformation (Figure 2C). These results shows that the inter-anatomical variability was several orders of magnitude less than that of cancerous tissue transformation (over the entire FP/HW spectrum), suggesting that the FP/HW Raman endoscope has promising potential to be a clinical diagnostic tool for colorectal cancer diagnosis. To further elucidate the magnitude of inter-anatomical variability, we applied PLS-DA to the dataset of in vivo colorectal tissue Raman spectra (n=1228 spectra). Figure 3 shows the PLS-DA classification error rate as a function of model complexity (i.e., retained number of LVs) using leave-one-patient-out cross validation. The curves generally show the first local error minima at 2 LVs, which is the model complexity that is chosen in this study to best characterize the entire colorectal lumen. Figure 4A shows the loadings on the LVs (LV1 captures 41.00% of the spectral variance accounting for 5.60% of the group affinity, and LV2 captures 8.47% of the spectral variance accounting for 10.00% of the group affinity). Figure 4B-4C depicts data points and box charts of the two LV scores (i.e., LV1 and LV2), revealing the contribution of the loading on each colorectal tissue site. The line within each notch box represents the median, while the lower and upper boundaries of the box indicate the first (25% percentile) and the third (75% percentile) quartiles, respectively. Error bars (whiskers) represent the 1.5-fold interquartile range. We also assessed the classification capability among different anatomical locations along the colorectal lumen. The PLS-DA multiclass algorithms were able to identify different anatomical sites with varying sensitivities (ascending colon: 1.10%, transverse colon: 14.06%, descending colon: 40.32%, sigmoid: 19.34%, rectum: 71.55%), and specificities (ascending colon: 91.02%, transverse colon: 78.78%, descending colon: 81.99%, sigmoid: 87.90%, rectum: 77.84%). Table 1 shows the confusion matrix of the PLS-DA classifications. The rectum can be identified with a

11

ACS Paragon Plus Environment

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

high sensitivity among other anatomical locations. The above results suggest that the proximal parts of the large intestine, namely rectum has a unique molecular profile compared to that of the mid and distal colon. To evaluate the clinical implications of inter-anatomical variability, we also employed PLS-DA for cancer diagnosis, whereby a diagnostic accuracy of 88.8% (sensitivity: 93.9% (93/99); specificity 88.3% (997/1129), can be achieved across the different anatomical sites based on the leave-one patient-out, cross validation method. The Raman spectral changes associated with cancerous tissue transformation (Figures 1A-1B and 2A-2C) were several orders of magnitude larger than inter-anatomical variability of normal colorectal tissue across the FP/HW Raman ranges, further confirming the promising potential of FP/HW Raman endoscopy technique for early diagnosis of colorectal cancer in vivo.

DISCUSSION We presented a simultaneous FP and HW Raman endoscopic technique for molecular characterization of colorectal tissue in vivo. The simultaneous FP/HW Raman endoscopy holds a number of advantages over conventional FP Raman technique for colonoscopy.20,24,37 Firstly, background tissue autofluorescence is reduced in the HW range which enables intense Raman signals to be extracted even in those patients that exhibit intense autofluorescence in the colorectum. Secondly, the molecular information extracted from the FP and HW Raman spectra are complementary for improving tissue characterization. For instance, the changes in gastrointestinal epithelial water contents and structures are manifested in the HW region (3100 - 3600 cm-1), but not in the FP region. Synergizing the FP and HW Raman techniques therefore offers new prospects to study the molecular biology (including proteins, lipids and water structures) of colorectal tissues in the native environments that are not readily obtainable by other

12

ACS Paragon Plus Environment

Page 12 of 32

Page 13 of 32

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

techniques at colonoscopy. High quality FP/HW Raman spectra can routinely be measured within 0.5 s in different anatomical locations of the colorectum in vivo during endoscopic examination (Figure 1A). The overall Raman-active tissue constituents were comparable among different colorectal anatomical sites, but the subtle (while highly molecular specific) inter-anatomical variations were observed (e.g., relative tissue Raman differences (spectral shape, bandwidth, peak position and intensity)). For instance, the HW peaks at 2850 and 2885 cm−1 (symmetric and asymmetric CH2 stretching of lipids) were generally reduced in the rectum, suggesting a lesser amount of lipid content compared to the colon. To extract more specific molecular information, we performed PLS-DA on the spectral dataset of normal colorectal tissues. Loading on LV1 generally contained the specific FP Raman peaks from lipid moieties (i.e., 1078 cm-1 (ν(C-C)), 1302 cm-1 (CH2 twisting and wagging), 1445 cm-1 (δ(CH2) deformation), 1650 cm-1 (v(C=C)), 1745 cm-1 (v(C=O)), and HW Raman peaks at 2850 and 2885 cm−1 (symmetric and asymmetric CH2 stretching) and 3009 cm-1 (asymmetric =CH stretching of lipids) as well as the symmetric and asymmetric OH vibrations in the HW spectral range. One notes that loading on LV1 largely represents the water/lipid ratio of the colorectal tissues. The loading on LV2 captured Raman peaks mainly associated with proteins (i.e., 936 cm-1 (v(C-C)), 1004 cm-1 (νs(C-C)), 1265 cm-1 (amide III v(C-N) and δ(N-H)), 1335 cm-1 (nucleic acids), 1445 cm-1 (δ(CH2) deformation), 1655 cm-1 (amide I v(C=O)), 2940 cm−1 (CH3 stretching), ~3329 cm−1 (Amide A (NH stretching))). Analysis of the PLS scores shows that LV1 essentially distinguishes rectum from colon. This is likely attributed to the fact that the rectum is functionally specialized (essentially nonabsorbing) as compared to the rest of the large intestine which contains a specialized acidic mucous secreting variety of columnar intestinal epithelium.38 On the other hand,

13

ACS Paragon Plus Environment

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

LV2 containing protein associated Raman peaks could differentiate sigmoid and rectum from the remaining colon. This study supports the hypothesis that molecular features vary along the colorectum according to anatomical location (i.e., distal colon, proximal colon and rectum). It is known that the basic cellular composition (i.e., undifferentiated, mucous and columnar cells) of intestinal epithelium varies along the colorectum, which could partially explain the subtle spectral differences observed.38 It is also plausible that there could be a influence of bowel contents on the epithelium (including microbiome) which change gradually along the length (~1.5 m) of the large bowel.22 Our further analysis showed that the inter-anatomical spectral variability of normal tissue was subtle compared to cancerous tissue transformation (Figures 1A-1B and Figures 2A-2C). FP/HW Raman technique can separate cancerous tissue from normal tissues with 88.8% accuracy, suggesting a promising role of simultaneous FP/HW Raman technique in providing effective biomolecular characterization for colorectal cancer diagnosis. Besides detecting colorectal cancer, we are currently also exploring the feasibility of applying simultaneous FP/HW Raman endoscopic technique for differentiating colorectal precancer (i.e., adenoma) and early adenocarcinomas from hyperplastic polyps in vivo during clinical colonoscopy. This work is the first report on in vivo characterization of normal colorectal tissue using simultaneous FP/HW Raman endoscopic technique. Compared to the volume-type Raman technique which has been explored for in vivo tissue Raman measurements,37,39 the depth selective fiber-optic Raman probe holds unparalleled clinical advantages as follows: (i) selective targeting of the epithelial lining associated with early onset of colorectal carcinogenesis is superior to conventional volume-type fiber-optic Raman probes that interrogate with a larger tissue volume (∼1 mm3); (ii) the subsurface interrogation capability reduces autofluorescence interferences and signal

14

ACS Paragon Plus Environment

Page 14 of 32

Page 15 of 32

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

dilution from deeper bulky tissues; (iii) Raman signals can only be collected in contact mode, thereby improving reproducibility and repeatability of clinical Raman measurements. We will further extend the FP/HW Raman technique into multi-center studies aiming to further evaluate Raman differences between ethnicity, sex, ages and diet, etc., to increase the generalizability of our study findings. Another, since the gold standard (i.e., histopathology) can only grossly explain the cellular and morphological discrepancies between different colorectal sub-sites, correlating FP/HW tissue Raman spectroscopy with biomolecular sciences (e.g. immunochemistry, proteomic, lipidomics or even genetics) in a larger cohort of patients could be one of the directions to explore for further understanding of onset of colorectal carcinogenesis. In summary, we have applied a novel simultaneous FP/HW Raman endoscopic technique developed for real-time in vivo colorectal tissue Raman measurements during clinical colonoscopy. Multivariate analysis was utilized to characterize the molecular profiles (e.g., proteins, lipids and water composition) of normal colorectal tissue in different anatomical locations (ascending colon, transverse colon, descending colon, sigmoid and the rectum). The subtle inter-anatomical FP/HW Raman variation among different anatomical locations of normal colorectum was found, which was orders of magnitude less than the spectral variability associated with cancerous tissue transformation. This work suggests a promising potential of simultaneous FP/HW Raman endoscopic technique for early diagnosis of colorectal cancer in vivo at the molecular level.

15

ACS Paragon Plus Environment

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACKNOWLEDGMENTS This work was supported by the National Research Foundation-Proof-of-Concept (NRFPOC), the National Medical Research Council (NMRC), and the Academic Research Fund (AcRF)-Tier 2 from the Ministry of Education (MOE), Singapore.

References (1) Jemal, A.; Siegel, R.; Xu, J.; Ward, E. Cancer Journal For Clinicians 2010, 60, 277300. (2) Jemal, A.; Bray, F.; Center, M. M.; Ferlay, J.; Ward, E.; Forman, D. CA: A Cancer Journal for Clinicians 2011, 61, 69-90. (3) Quintero, E.; Hassan, C.; Senore, C.; Saito, Y. Gastroenterology Research and Practice 2012, 2012, 8. (4) Wallace, M. B.; Kiesslich, R. Gastroenterology 2010, 138, 2140-2150. (5) Bergholt, M. S.; Zheng, W.; Lin, K.; Ho, K. Y.; Teh, M.; Yeoh, K. G.; Yan So, J. B.; Huang, Z. Journal of Biomedical Optics 2011, 16, 037003. (6) Bergholt, M. S.; Zheng, W.; Lin, K.; Ho, K. Y.; Teh, M.; Yeoh, K. G.; Yan So, J. B.; Huang, Z. International journal of cancer 2011, 128, 2673-2680. (7) Bergholt, M. S.; Zheng, W.; Lin, K.; Ho, K. Y.; Teh, M.; Yeoh, K. G.; Yan So, J. B.; Huang, Z. Analyst 2010, 135, 3162-3168. (8) Bergholt, M. S.; Zheng, W.; Lin, K.; Ho, K. Y.; Teh, M.; Yeoh, K. G.; Yan So, J. B.; Huang, Z. Technology in cancer research & treatment 2011, 10, 103-112. (9) Shim, M. G.; Song, L. M. W. K.; Marcon, N. E.; Hassaram, S.; Wilson, B. C. Proc. SPIE 2000, 3918, 114-119. (10) Bergholt, M. S.; Zheng, W.; Huang, Z. Journal of Raman Spectroscopy 2012, 43, 255-262.

16

ACS Paragon Plus Environment

Page 16 of 32

Page 17 of 32

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

(11) Lau, D. P.; Huang, Z.; Lui, H.; Man, C. S.; Berean, K.; Morrison, M. D.; Zeng, H. Lasers in Surgery and Medicine 2003, 32, 210-214. (12) Teh, S. K.; Zheng, W.; Lau, D. P. C.; Huang, Z. Analyst 2009, 134, 1232-1239. (13) Short, M. A.; Lam, S.; McWilliams, A. M.; Ionescu, D. N.; Zeng, H. Journal of Thoracic Oncology 2011, 6, 1206-1214. (14) Vargis, E.; Brown, N.; Williams, K.; Al-Hendy, A.; Paria, B. C.; Reese, J.; Mahadevan-Jansen, A. Annals of Biomedical Engineering 2012, 40, 1814-1824. (15) Draga, R. O. P.; Grimbergen, M. C. M.; Vijverberg, P. L. M.; van Swol, C. F. P.; Jonges, T. G. N.; Kummer, J. A.; Ruud Bosch, J. L. H. Analytical Chemistry 2010, 82, 5993-5999. (16) Huang, Z.; Zheng, W.; Colin, S. Proc. SPIE 2005, 5862. (17) Widjaja, E.; Zheng, W.; Huang, Z. International Journal of Oncology 2008, 32, 653-662. (18) Short, M. A.; Tai, I. T.; Owen, D.; Zeng, H. Optics Express 2013, 21, 5025-5034. (19) Scalfi-Happ, C.; Udart, M.; Hauser, C.; Rück, A. Medical Laser Application 2011, 26, 152-157. (20) Molckovsky, A.; Song, L. M. W. K.; Shim, M. G.; Marcon, N. E.; Wilson, B. C. Gastrointestinal endoscopy 2003, 57, 396-402. (21) Wang, J.; Bergholt, M. S.; Zheng, W.; Huang, Z. Optics Letters 2013, 38, 23212323. (22) Yamauchi, M.; Morikawa, T.; Kuchiba, A.; Imamura, Y.; Qian, Z. R.; Nishihara, R.; Liao, X.; Waldron, L.; Hoshida, Y.; Huttenhower, C.; Chan, A. T.; Giovannucci, E.; Fuchs, C.; Ogino, S. Gut 2012. (23) Huang, Z.; Teh, S. K.; Zheng, W.; Mo, J.; Lin, K.; Shao, X.; Ho, K. Y.; Teh, M.; Yeoh, K. G. Optics Letters 2009, 34, 758-760.

17

ACS Paragon Plus Environment

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(24) Bergholt, M. S.; Zheng, W.; Ho, K. Y.; Teh, M.; Yeoh, K. G.; Yan So, J. B.; Shabbir, A.; Huang, Z. Gastroenterology 2014, 146, 27-32. (25) “American National Standard for the Safe Use of Lasers,” ANSI Standard 2136.11986, American National Standards Institute, Washington, D.C., 1986. (26) Torres, J. H.; Motamedi, M.; Pearce, J. A.; Welch, A. J. Appl. Opt. 1993, 32, 597– 606. (27) Incropera, F. P.; Witt, D. P. D. Fundamentals of Heat and Mass Transfer, John Wiley and Sons, New York, 1990. (28) Marchesini, R.; Pignoli, E.; Tomatis, S.; Fumagalli, S.; Sichirollo, A.E.; Di Palma, S.; Dal Fante, M.; Spinelli, P.; Croce, A. C.; Bottiroli, G. Lasers Surg Med. 1994, 15, 351-357. (29) Bashkatov, A. N.; Genina, E. A.; Kochubey, V. I.; Rubtsov, V. S.; Kolesnikova, E. A.; Tuchin, V. V. Quantum Electronics 2014, 44, 779 – 784. (30) Thomsen, S. Photochem. Photobiol. 1991, 53, 825–835. (31) Duraipandian, S.; Bergholt, M. S.; Zheng, W.; Ho, K. Y.; Teh, M.; Yeoh, K. G.; Yan So, J. B.; Huang, Z. Journal of Biomedical Optics 2012, 17, 081418. (32) Bergholt, M. S.; Zheng, W.; Huang, Z. Journal of Biomedical Optics 2013, 18, 030502. (33) Huang, Z.; McWilliams, A.; Lui, H.; McLean, D. I.; Lam, S.; Zeng, H. International Journal of Cancer 2003, 107, 1047-1052. (34) Stone, N.; Kendall, C. A.; Shepherd, N.; Crow, P.; Barr, H. Journal of Raman Spectroscopy 2002, 33, 564-573. (35) Koljenovic, S.; Schut, T. C. B.; Wolthuis, R.; de Jong, B. W. D.; Santos, L.; Caspers, P. J.; Kros, J. M.; Puppels, G. J. Journal of Biomedical Optics 2005, 10, 031116.

18

ACS Paragon Plus Environment

Page 18 of 32

Page 19 of 32

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

(36) Leikin, S.; Parsegian, V. A.; Yang, W.-H.; Walrafen, G. E. Proceedings of the National Academy of Sciences 1997, 94, 11312-11317. (37) Bergholt, M. S.; Zheng, W.; Ho, K. Y.; Teh, M.; Yeoh, K. G.; So, J. B. Y.; Shabbir, A.; Huang, Z. Journal of Biophotonics 2013, 6, 49-59. (38) Moyer, M. Colon cancer cells; Academic press limited, 1990. (39) Shim, M. G.; Song, L. M. W. K.; Marcon, N. E.; Wilson, B. C. Photochemistry and Photobiology 2000, 72, 146-150.

19

ACS Paragon Plus Environment

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Figure Captions

Figure 1 (A) In vivo FP/HW Raman spectra ± 1 standard deviation (s.d.) of different colorectal tissue sites (ascending (n=182), transverse (n=249), descending (n=124), sigmoid (n=212), and rectum (n=362) and colorectal adenocarcinoma (n = 99) as confirmed by consensus histopathology examination. (B) Agglomerative hierarchical clustering dendrogram (Euclidean distances) of the mean Raman spectra of different anatomical sites and comparison of the spectral variation with cancerous tissue transformation.

Figure 2 Comparison of mean difference spectra ±1 standard deviation (SD) of the distinct anatomical tissue sites: (A) ascending–transverse, ascending–descending, ascending–-sigmoid, ascending–rectum and transverse–descending; (B) transverse– sigmoid, transverse–rectum, descending–sigmoid, sigmoid–rectum, and (C) normal colon–cancer.

Figure 3. Classification error rate as a function of model complexity (i.e., retained number of LVs) using partial least squares discriminant analysis together with leaveone-patient-out cross validation. The error rates show the first local minima at 2 LVs, which is the model complexity that was chosen in this study.

Fig. 4. Partial least squares (PLS) –discriminant analysis (DA): (A) The first two significant LVs Raman spectral loadings for classification between ascending colon, transverse colon, descending colon, sigmoid and rectum; (B) Box charts of the LV1 scores. The line within each notch box represents the median, while the lower and upper boundaries of the box indicate first (25% percentile) and third (75% percentile) quartiles, 20

ACS Paragon Plus Environment

Page 20 of 32

Page 21 of 32

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

respectively. Error bars (whiskers) represent the 1.5-fold interquartile range. All pairwise comparisons of LV1, (except ascending and transverse colon) were statistically significant (One-way ANOVA and Fisher's least significant difference (LSD) with p values less than 0.05); (C) Box charts of the LV2 scores. All pairwise comparisons of LV2, (except ascending and transverse colon) were statistically significant (One-way ANOVA and Fisher's least significant difference (LSD) with p values of less than 0.05).

21

ACS Paragon Plus Environment

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 22 of 32

Table 1 Confusion matrix of the multi-class classification results of Raman spectra of different colorectal tissues using PLS-DA together with leave-one patient -out, crossvalidation.

Predicted anatomical site by Raman Ascending

Transverse

Ascending

2

105

32

16

27

Actual

Transverse

51

35

73

46

44

anatomical

Descending

12

24

50

12

26

site

Sigmoid

21

39

38

41

73

Rectum

1

27

38

37

259

Sensitivity (%)

1.10

14.06

40.32

19.34

71.55

Specificity (%)

91.02

78.78

81.99

87.90

77.84

22

Descending Sigmoid

ACS Paragon Plus Environment

Rectum

Page 23 of 32

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Table of Contents Graphic

23

ACS Paragon Plus Environment

Analytical Chemistry

3250 3400

2850 2885 2940

1245 1302 1335 1445 1618 1655

A

956 1004 1078

Figure 1A

Rectum normal (n=362) Sigmoid normal (n=212)

Norm alized Intensity (a.u.)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 24 of 32

Descending normal (n=124) Transverse normal (n=249) Ascending normal (n=182) Rectum cancer (n=99)

800 1000120014001600 2800

3200 -1

Wavenumber (cm )

24

ACS Paragon Plus Environment

3600

Page 25 of 32

Figure 1B

B 0.014 0.012 0.010 0.008 0.006 0.004 0.002

Normal

25

ACS Paragon Plus Environment

er nc Ca

Re

ct

um

d oi

gm

ng Si

di

De

sc

en

nv as Tr

ce

nd

er

in

g

se

0.000

As

Euclidean distance

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Analytical Chemistry

Figure 2A

A

ascending-transverse (normal) ascending-descending (normal)

Intensity (a.u.)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 26 of 32

ascending-sigmoid (normal) ascending-rectum (normal)

transverse-descending (normal)

800 1000 1200 1400 1600 2800

Raman shift (cm -1)

26

ACS Paragon Plus Environment

3200

3600

Page 27 of 32

Figure 2B

B

transverse-sigmoid (normal)

transverse-rectum (normal)

Intensity (a.u.)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

descending-sigmoid (normal)

descending-rectum (normal)

sigmoid-rectum (normal)

800 1000 1200 1400 1600 2800

Raman shift (cm -1)

27

ACS Paragon Plus Environment

3200

3600

Analytical Chemistry

Figure 2C

C Intensity (a.u.)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 28 of 32

normal-cancer

800 1000 1200 1400 1600 2800

Raman shift (cm -1)

28

ACS Paragon Plus Environment

3200

3600

Page 29 of 32

Figure 3

Error rate (% )

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Rectum vs. rest Sigmoid vs. rest Descending vs. rest Transverse vs. rest Ascending vs. rest

70 65 60 55 50 45 40 35 30 25 20 15 10 5 0 0

2

4

6

8

10 12 14 16 18 20

Latent variable

29

ACS Paragon Plus Environment

Analytical Chemistry

3400

0.05

3250

1745

0.10

2940

1308

0.15

LV1 X: 41.00% Y: 5.60% LV2 X: 8.47% Y: 10.00% 1648

A

1445

Figure 4A

3009

1655

2850 2880

-0.15

1440

-0.10

1027

-0.05

1302

0.00

876

Loading on LV

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 30 of 32

800 1000 1200 1400 1600 2800

Raman shift cm -1

30

ACS Paragon Plus Environment

3200

3600

Page 31 of 32

Figure 4B

0.010 0.005 0.000 -0.005 -0.010

um ct Re

oi gm

Si

sc De 31

ACS Paragon Plus Environment

d

ng di en

sv an Tr

ce

nd

er

in

se

g

-0.015

As

B Score on LV1

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Analytical Chemistry

Figure 4C

C

0.02 0.01 0.00 -0.01 -0.02

32

ACS Paragon Plus Environment

um ct Re

d oi gm

Si

sc De

Tr

an

sv

en

er

di

se

g in nd ce

ng

-0.03

As

Score on LV2

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 32 of 32