1H HR-MAS NMR Spectroscopy of Tumor-Induced Local Metabolic

Dec 16, 2012 - “Field-Effects” Enables Colorectal Cancer Staging and. Prognostication ... metabolite-based biomarkers for CRC diagnosis, staging a...
0 downloads 0 Views 3MB Size
Article pubs.acs.org/jpr

1

H HR-MAS NMR Spectroscopy of Tumor-Induced Local Metabolic “Field-Effects” Enables Colorectal Cancer Staging and Prognostication Beatriz Jiménez,† Reza Mirnezami,‡ James Kinross,‡ Olivier Cloarec,†,§ Hector C. Keun,† Elaine Holmes,† Robert D. Goldin,∥ Paul Ziprin,‡ Ara Darzi,‡ and Jeremy K. Nicholson*,† †

Section of Biomolecular Medicine, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London, SW7 2AZ London, United Kingdom ‡ Section of Biosurgery and Surgical Technology, Department of Surgery and Cancer, Faculty of Medicine, St Mary’s Hospital, Imperial College London, W2 1NY London, United Kingdom § Korrigan Sciences Ltd., Imperial Place, SL6 2GN Maidenhead, United Kingdom ∥ Centre for Pathology, Department of Medicine, Faculty of Medicine, St Mary’s Hospital, Imperial College London, W2 1NY London, United Kingdom S Supporting Information *

ABSTRACT: Colorectal cancer (CRC) is a major cause of morbidity and mortality in developed countries. Despite operative advances and the widespread adoption of combined-modality treatment, the 5-year survival rarely exceeds 60%. Improving our understanding of the biological processes involved in CRC development and progression will help generate new diagnostic and prognostic approaches. Previous studies have identified altered metabolism as a common feature in carcinogenesis, and quantitative measurement of this altered activity (metabonomics/metabolomics) has the potential to generate novel metabolite-based biomarkers for CRC diagnosis, staging and prognostication. In the present study we applied high-resolution magic angle spinning nuclear magnetic resonance (HR-MAS NMR) spectroscopy to analyze metabolites in intact tumor samples (n = 83) and samples of adjacent mucosa (n = 87) obtained from 26 patients undergoing surgical resection for CRC. Orthogonal partial least-squares discriminant analysis (OPLS-DA) of metabolic profiles identified marked biochemical differences between cancer tissue and adjacent mucosa (R2 = 0.72; Q2 = 0.45; AUC = 0.91). Taurine, isoglutamine, choline, lactate, phenylalanine, tyrosine (increased concentrations in tumor tissue) together with lipids and triglycerides (decreased concentrations in tumor tissue) were the most discriminant metabolites between the two groups in the model. In addition, tumor tissue metabolic profiles were able to distinguish between tumors of different T- and N-stages according to TNM classification. Moreover, we found that tumoradjacent mucosa (10 cm from the tumor margin) harbors unique metabolic field changes that distinguish tumors according to Tand N-stage with higher predictive capability than tumor tissue itself and are accurately predictive of 5-year survival (AUC = 0.88), offering a highly novel means of tumor classification and prognostication in CRC. KEYWORDS: colorectal cancer, high resolution magic angle spinning nuclear magnetic resonance, tissue biopsy, metabonomics



INTRODUCTION

CRC through the development of novel tools for early detection and personalized therapeutics. It is now acknowledged that CRC arises through a stepwise accumulation of genetic and epigenetic alterations, and extensive “bottom up” systems biology research has identified several distinct CRC phenotypes arising through specific molecular pathways (chromosomal instability pathway, mismatch repair

Colorectal cancer (CRC) is the third leading cause of cancerrelated death in developed countries, with an estimated 1 million new cases diagnosed every year worldwide.1,2 Primary surgery represents the mainstay of treatment in CRC, but despite advances in screening programs3 and chemoradiotherapy regimens,4,5 the 5-year survival remains below 60% in most European countries.6 A more complete understanding of the molecular mechanisms responsible for CRC development and progression is critical to improving outcomes for patients with © 2012 American Chemical Society

Received: October 25, 2012 Published: December 16, 2012 959

dx.doi.org/10.1021/pr3010106 | J. Proteome Res. 2013, 12, 959−968

Journal of Proteome Research

Article

pathway, serrated pathway).2 Metabolic phenotyping and metabonomic approaches7 provide a holistic “top-down” viewpoint from which to characterize cancers. Metabonomics represents a systems approach for studying in vivo, the dynamic changes in hundreds of low molecular-weight metabolites in intact tissue or body fluid samples in relation to disease processes or other stimuli.8 Recent studies have confirmed the potential for clinically translatable metabonomic approaches in the diagnosis and management of CRC.9−11 Biofluid analysis by nuclear magnetic resonance (NMR) and mass spectrometry (MS) can distinguish patients with localized12 and metastatic disease.13 Pharmacometabonomic approaches14 enable predictions on likely outcomes of interventions to be made based on preinterventional trainingtest set models and have a role in the predose prediction of toxicity and chemo-resistance in patients undergoing chemotherapy for CRC.15 The application of high-resolution magic angle spinning (HR-MAS) NMR spectroscopy16 to intact tissue biopsies has been shown to differentiate between benign and malignant tissue with high sensitivity and specificity.17 We have previously used this approach to define the metabolic “field effects” associated with other malignancies in the upper gastrointestinal tract, suggesting that the localized metabolic modification of “normal tissue” induced by neighbouring tumor is significant and consistent.18 Tumor field effects can be defined as biochemical and physiological abnormalities in morphologically normal tissue induced by adjacent tumor activity. The NMR-based analytical approach has significant translatable capacity as it requires minimal sample preparation, it is rapid, automated and high throughput, and it can deliver clinically relevant data in near real-time.19 This analytical strategy will form part of the wider systems biology approach to patient stratification envisaged by the ‘Precision Medicine’ initiative.20 However, this approach has yet to demonstrate its full prognostic capacity, as prior to this study the approach has not been applied to a patient cohort with long-term oncological follow-up data. In this prospective study we hypothesized that HR-MAS NMR profiling of CRC and adjacent macroscopically normal (or “offtumor”) mucosa taken from the same resection specimen could provide accurate molecular information with which to characterize tumors and build metabolite-based prognostic models.



Table 1. Description of Patients and Sample Groups tumor samples

off-tumor samples

83 (8 unclassified)

87 (6 unclassified)

10 (38) 16 (62)

33 42

35 46

20 (77) 6 (23)

62 13

73 8

26 (100)

75

81

1 (4) 20 (77) 1 (4) 4 (15)

3 72 No sample 8

2 76 3 6

1 (4) 3 (12) 10 (38) 12 (46)

8 20 47

3 8 24 46

2 (8) 10 (38) 14 (54) 0 (0)

3 23 49

9 25 47

12 (46) 14 (54)

26 49

34 47

3 (12) 3 (12) 3 (12) 17 (64)

3 4 13 55

1 5 10 65

2 (8) 13 (50) 11 (42)

9

12

69

69

5 (19) 21 (81)

17 58

24 57

19 (73) 4 (15) 3 (12)

58 14 3

58 20 3

patients

MATERIALS AND METHODS

Patient Recruitment and Sample Collection

This study was approved by the institutional review board at Imperial College Healthcare NHS Trust (REC reference number 07/H0712/112). Tissue specimens and related clinical data were collected with written consent from 26 consecutive patients undergoing elective CRC resection at St Mary’s Hospital (London, UK). All patients were followed up using the local colorectal cancer Multidisciplinary Team (MDT) protocol over a five-year period. The clinico-pathological characteristics of these patients are summarized in Table 1. During surgery extracted cancer specimens were taken on ice to the pathology department where they were opened by a pathologist and sampled. Fresh samples were retrieved from tumor (n = 22) and adjacent macroscopically normal colorectal mucosa 5 to 10 cm from the tumor margin (n = 23). These samples were snap frozen immediately and transferred to a freezer at −80 °C prior to processing.

Total number, n

26

Age, median (range) Follow-up (months), median (range) Gender, n (%) M F Location of primary tumor, n, (%) Colon Rectum Pathology, n, (%) Adenocarcinoma Differentiation, n, (%) Well differentiated Moderately differentiated Poorly differentiated Not recorded T Stage, n, (%) T1 T2 T3 T4 UICC Stage, n, (%) I II III IV Lymph node status, n, (%) N0 N1/N2 Surgery, n, (%) Right hemicolectomy Left hemicolectomy Subtotal colectomy Anterior resection 5-FU based chemotherapy, n, (%) Neo-adjuvant Adjuvant No Preoperative EBRT1, n, (%) Yes No Relapse Freeat 5 years, n, (%) Yes No (metastasis) No (other causes)

72 (26−87) 49 (5−64)

1

EBRT = External Beam Radio Therapy.

Sample Preparation and HR-MAS NMR Experiments

Tissue samples were kept on ice at all times during the preparation process. Three samples were prepared for HR-MAS NMR from each original tissue sample where the harvested volume permitted it. These samples were processed in triplicate as a data validation step and to compensate for anticipated tissue heterogeneity. An average of 10 mg of tissue was packed into disposable 30 μL Teflon inserts. D2O saline solution (0.9% in 960

dx.doi.org/10.1021/pr3010106 | J. Proteome Res. 2013, 12, 959−968

Journal of Proteome Research

Article

Figure 1. Metabolic differences between CRC tumor tissue and adjacent off-tumor tissue. (A) Scatter plot for the OPLS model of tumor (red ●) and offtumor colorectal mucosa (■). (B) ROC curve for the OPLS model. (C) Coefficient plot for the OPLS model where positive and negative signals denote metabolites whose concentrations were increased and decreased respectively in tumor tissue compared to off-tumor tissue. The color code provides visual description of correlation coefficient values for each signal. The higher the value, the more contribution this signal has in the discrimination between the two groups. Labels correspond to the signals of the metabolites that are discriminant for the model.

The matrix contained information from the region −1 to 10 ppm and the resolution used was 0.00055 ppm. Spectral regions containing water, ethanol and polyethylene glycol (PEG) signals, were excluded from the analysis. Spectra were normalized to tissue mass and unit variance (UV) scaling. Unsupervised (PCA, Principal Component Analysis, data not shown) and supervised (OPLS, Orthogonal Projections to Latent Structures) multivariate statistical analysis were performed using SIMCA-P+ (v. 12.0.1, Umetrics AB, Sweeden) and in-house software built on Matlab. Two types of cross validation were used to obtain the Q2 statistical parameter for each modelthe typical 7-fold cross validation, and the “leaving one patient out” (Supporting Information, Supplementary Table) to account for the statistical paradigm of having several samples coming from the same patient. Matlab scripts were mainly used to identify the most discriminant signals between the two groups in comparison. Receiver operating characteristic (ROC) curves were also determined using the cross-validated predicted Y-values of the HR-MAS NMR OPLS-DA data in the Matlab environment, and the area under the curve (AUC) was also calculated for each model. Univariate statistical analysis was performed in Microsoft Office Excel 2007 and Welch correction was applied to the t test values. Fold changes were calculated for the discriminant metabolite signals for each of the models. Additionally, the t test values were multiplied by the number of discriminant signals

mass) was added to the insert to complete the volume avoiding the formation of air bubbles. The inserts were introduced into ZrO2 rotors and then into a refrigerated tray at −20 °C in preparation for experimental acquisitions. Samples remained in the tray for 0−24 h during experimental processing. No metabolic profiling differences were observed due to length of time in the refrigerated tray prior to measurement (data not shown), indicating the biochemical stability of the samples was adequate over this period. Relaxation edited experiments (1D-Carr−Purcell−Meiboom−Gill (CPMG) spin−echo) were acquired on a 400 MHz Bruker Avance III spectrometer equipped with an automated sampling robot and a TBI HR-MAS probe. Twohundred fifty-six free induction decays (FID) were accumulated for each experiment in 64 K points using a 20 ppm window centered on the water signal. The relaxation delay was set at 4 s and a water presaturation pulse was applied during this period to cancel the water signal. The total spin−spin relaxation delay was 75.6 ms. Bruker software packages were used to process the spectra (TopSpin 2.2 and Amix 3.9.9) and the methyl signal of alanine was used for calibration (1.47 ppm). Additional J-res experiments were acquired to help with metabolite identification. Metabolites were identified based on literature17,21 and using the STOCSY approaches for data driven structural assignment.22,23 Statistical Analysis of Spectroscopic Data

The CPMG spectra (n = 170) were converted into statistical matrices using in-house tools developed in MATLAB (v. 2011a). 961

dx.doi.org/10.1021/pr3010106 | J. Proteome Res. 2013, 12, 959−968

Journal of Proteome Research

Article

Table 2. Discriminant Metabolites for the Different OPLS Models Based on Correlation Values Higher than 0.2a T classification T = 4 vs T = 1−3 metabolite Lipids/ Triglycerides

Valine Iso-butyrate Alanine Leucine Acetate Iso-glutamine Creatine Tyrosine

Phenylalanine

Choline Phosphocholine Taurine

Scyllo-inositol Lactate β-Glucose Formate

chem. group -CH3

δ ppm (multiplicity) 0.90 (m)

-CH2C−CH2 CH-CH3 -CH3 -CH3 -CH3 β-CH2, γCH -CH3 β-CH2

1.29 (m) 2.00 (m) 5.33 (m) 1.0 (d) 1.02 (d) 1.0 (d) 1.47 (d) 1.72 (m)

-CH3 -CH2β-CH2 α-CH H3,H5 H2,H6 β-CH2 H2,H6

3.02 (s) 3.93 (s) 3.06 (dd) 3.98 (dd) 6.89 (m) 7.17 (m) 3.12 (dd) 7.33 (m)

H4 H3,H5 N-(CH3)3 O−CH2 N-(CH3)3 O−CH2 -CH2−NH -CH2SO3

7.38 (m) 7.43 (m) 3.20 (s) 4.26 (m) 3.22 (s) 4.19 (t) 3.26 (t) 3.42 (t)

-O−H -CHH1 -CH

3.35 (s) 4.15 (q) 4.64 (d) 8.45(s)

1.92 (s) 2.34 (m)

tumor vs offtumor

tumor

off-tumor

N stage N = 1,2 vs N = 0 tumor

−5

−1.3 (0.9)

−5

−1.7 (0.3)

−3.2* (4 × 10 ) −3.9* (2 × 10 ) −1.7* (0.006) −2.5* (0.0002)

off-tumor

survival deceased vs alive off-tumor

+1.3 (0.8) +1.54* (0.03) +1.4 (0.6) + 1.3 (0.3) +1.4* (0.011) +1.8 (0.14) +1.7* (1.0 × 10−7) +1.3 (0.09) +1.3 (0.3) +1.7* (2 × 10−6) +1.6* (5 × 10−6) +1.4* (0.04) +1.5* (0.006)

+1.3 (1.0) +1.3 (1.4) +1.3 (0.12) +1.2 (1.2)

+1.5* (0.002) +1.7* (0.0007) +1.8* (0.003) +

+1.3 (0.8) + +1.5* (0.00014) +1.5* (4 × 10−5) +1.5* (0.0004) +1.5* (5 × 10−5) +1.8* (7 × 10−9) +2.2* (1.2 × 10−11)

+1.2 (0.9) +1.12 (1.3)

+1.3* (0.03) +1.3 (0.08) −1.4 (0.2)

+1.5* (3 × 10−5) −1.8 (0.06) −2.8* (0.009)

a

Fold changes in concentration show the increment or decrease of the metabolites in one set of tissue compared to the other one, as stated in the table. Numbers in brackets correspond to the Bonferroni corrected Welch t test values. Statistically significant values (p < 0.05) and are labeled with an asterisk.

Information (Supporting Information, Supplementary Figure 1). Metabolic profiles from tumor tissue (n = 83) and from offtumor tissue (n = 87) were analyzed using chemometric methods in order to evaluate their predictive power to clinically classify CRC tumors (Table 1). As a proof of principle, we evaluated the differences between the metabolic profiles of tumor tissue and off-tumor tissue (Figure 1). The results obtained matched those previously observed by our group17 with the main differences due to the exclusion of exometabolites such as ethanol or PEG. The OPLS model shows good separation between the sample groups and explains a large percentage of variability with good predictability (R2Y = 0.72 and Q2Ycum = 0.45, Figure 1A). The ROC curve was also calculated for the model (Figure 1B) and the value obtained (AUC = 0.91) ratifies the strength of the model to discriminate between tissue from malignant tumor or its surroundings. Colorectal cancer tumor tissue presented higher concentrations of isoglutamine, choline, phosphocholine, taurine, lactate, tyrosine and phenylalanine, while lipids and triglycerides appear to be lower in concentration when compared

in each model to account for multiple comparisons in the same group (Bonferroni correction).



RESULTS

Proof of Principle: Differences between CRC Tumor Tissue and CR Mucosa from the Same Patients

Biopsies from 26 patients undergoing elective surgery for CRC were collected from center of tumor and from 5−10 cm away from the tumor margin (off-tumor samples). Tumors were primary colonic in 20 patients (77%) and primary rectal in the remaining 6 (23%). There were 19 matched paired samples, 3 additional tumor samples and 4 additional off-tumor samples. For some of the patients, two different biopsies were collected from cancerous and/or off-tumor tissue and they were treated as individual samples. From each biopsy 1 to 3 replicates were prepared and analyzed by HR-MAS NMR spectroscopy. An example of the heterogeneity observed for three replicates analyzed from the same biopsy is shown in Supporting 962

dx.doi.org/10.1021/pr3010106 | J. Proteome Res. 2013, 12, 959−968

Journal of Proteome Research

Article

Figure 2. CRC T-stage classification using HR-MAS metabolic profiles of tumor tissue and adjacent off-tumor tissue. Scatter plots for the OPLS models built using (A) tumor and (B) off-tumor tissue: the model shows the predictability for tumors classified as T = 4 (red ●) compared to those classified otherwise, T = 1−3 (■). (C and D) ROC curves for the above OPLS models. (E) Corresponding hematoxylin and eosin (H&E) stained tissue sections (8 μm thickness) taken from the center of tumors graded T1−T4 (all N0 tumors).

should be biochemically different, and so they should present specific metabolic profiles that could be characterized by HRMAS NMR. Furthermore, this should be true not only for tumor tissue, but also for adjacent normal colonic mucosa, which is likely to harbor distinct metabolic field changes of possible prognostic relevance. To examine this, metabolic profiles acquired from tumor tissue were modeled according to Tstage. Similarly, off-tumor tissue (OTT) profiles were also modeled on the basis of the histological result of the matching tumor tissue. In both cases, samples were divided into two

to off-tumor tissue (Table 2). The contribution of each metabolite to the model was evaluated using the OPLS loading plot (Figure 1C), which describes the correlation coefficients for each NMR signal. Signals belonging to the discriminant metabolites have been labeled in the plot. Classification of Tumor and Off-tumor Tissue According to Clinical T-classification

In the present study we hypothesized that the histopathological features used to classify CRC according to TNM staging criteria 963

dx.doi.org/10.1021/pr3010106 | J. Proteome Res. 2013, 12, 959−968

Journal of Proteome Research

Article

Figure 3. Prediction of lymph node status in CRC using HR-MAS metabolic profiles of tumor tissue and off-tumor tissue. Scatter plots for the OPLS models of (A) tumor tissue and (B) off-tumor tissue for patients lymph node negative disease, N0 (■) and patients with lymph node positive disease, N1/2 (red ●). (C and D) Corresponding ROC curves for the previous OPLS models. (E) Corresponding H&E stained tissue sections (8 μm thickness) taken from the center of tumors graded N0−N2 (all T3 tumors).

0.23, Figure 2B and AUC = 0.80, Figure 2D). Based on these models, higher concentrations of taurine account for the main local metabolite-based differences between T4 tumors and T1−3 tumors (Table 2). In the OTT, the concentration of valine, alanine, phenylalanine and tyrosine are increased for T4 tumors compared to T1−3. Conversely metabolites such as glucose and formate are less abundant in OTT for T4 tumors.

classification groups because of limited patient numbers in the early tumor cohort: T1−3 tumors and T4 tumors. The tumor tissue plot explains the majority of the variability, even though the predictability is low (R2Y = 0.83 and Q2Ycum = 0.23, Figure 2A), as reflected by the ROC curve plot (AUC = 0.75, Figure 2C). The OTT model shows a similar predictive value for the Tstage, although the ROC curve data for this model is improved compared to the tumor tissue one (R2Y = 0.68 and Q2Ycum = 964

dx.doi.org/10.1021/pr3010106 | J. Proteome Res. 2013, 12, 959−968

Journal of Proteome Research

Article

Classification of Tumor and Off-tumor Tissue According to Lymph Node Status of CRC

The capacity of the tumor and OTT models to predict lymph node status was also assessed. The metabolic differences between N0 (lymph node negative, LN −ve) and N1/2 (lymph node positive, LN +ve) tumors were compared for each of the two sets of samples. The score plots for the OPLS models of tumor tissue demonstrated good class separation (R2Y = 0.93 and Q2Ycum = 0.35, Figure 3A) and predictability power (AUC = 0.88, Figure 3B). However, the OTT produced a stronger predictive model of lymph node status (R2Y = 0.92 and Q2Ycum = 0.39, Figure 3B) with clear separation of LN −ve and LN +ve cases (AUC = 0.92, Figure 3D). A lower intensity of triglycerides signals in tumor tissue samples from patients with LN +ve disease, compared with those with LN −ve disease was observed. There was also an increase in the concentration of creatine and scyllo-inositol in tumor tissue from LN +ve patients. Signals from leucine and phenylalanine were stronger in the OTT samples of LN +ve patients compared to those with LN −ve disease (Table 2). Corresponding hematoxylin and eosin (H&E) stained sections for tumors presented in the statistical models are shown in Figures 2E and 3E. Lymph node involvement and metastatic spread was assessed using standard H&E staining of resected nodes to provide a complete pTNM stage (images not shown). Finally, multivariate statistical models were created to predict cancer-specific survival based on metabolite profiles from 5-year follow up data. Samples from patients who died of causes other than cancer were excluded from this classification (Table 1). The OPLS model comparing tumor tissue from patients alive at 5 years and those who died owing to local or distant cancer relapse found no predictive value (data not shown). However, the statistical model for the OTT samples from patients alive or deceased at 5 years explains the metabolic variability between samples and offers predictive capacity (R2Y = 0.75 and Q2Y (7 fold) = 0.29, Figure 4A). The ROC curve confirmed the capacity of the model for survival prediction based on OTT samples (AUC = 0.88, Figure 4B). Higher concentrations of isobutyrate, acetic acid and choline in OTT samples defined patients who experienced disease relapse compared to those who remained relapse free at 5 years. The Q2 values obtained using a “leave one patient out” cross validation were lower than those calculated using 7-fold cross validation, showing some correlation between samples coming from the same patient. However Q2 values remained positive in all cases and some models still revealed very high predictability. Table 2 provides a univariate analysis of each discriminant metabolite for the different clinical outcome models (tumor tissue, T stage, nodal status and survival). These metabolites were identified by the OPLS models (Figures 1−4) and thus they covary between groups. The Bonferroni correction was therefore applied to account for this (see Materials and Methods). This analysis confirmed the significance of the individual metabolite previously identified by the multivariate analysis. However it also demonstrated two important issues: (i) The multivariate statistical model has a greater discriminant value than a single statistically significant metabolite that changes in concentration according to clinical outcome measures; (ii) The larger the number of statistically significant metabolites changing in concentration between two groups in an OPLS model, the lower the statistical power of the individual metabolites when modified by the Bonferroni correction.

Figure 4. Survival prediction power for CRC using HR-MAS metabolic fingerprints of tumor and off-tumor tissue samples. (A) Scatter plot of the OPLS models of tumor and off-tumor tissue samples for patients alive (■) and deceased at 5 years after surgery (red ●). (B) ROC curve for the previous model.



DISCUSSION In current clinical practice, the biochemical information residing in resected cancer specimens does not form a routine part of the histopathological appraisal process. Tumor tissue when histologically analyzed at a cellular level is prone to subjective interpretation and is dependent on the expertise of the examining pathologist. The findings of the present study support the hypothesis that systematic analysis of metabolic data from tissue by platforms such as HR-MAS NMR provides a powerful means with which to generate objective, quantifiable and oncologically relevant diagnostic and prognostic information. Previous metabonomics-based studies using this technology have proven the utility of this approach for the analysis of different tumors including brain,24 colon17 and breast25 with high predictive value for disease diagnosis.13 However, until now, there has been little analysis of the biochemistry of the tissue adjacent to cancers. It is evident that a tumor affects its surroundings not only because of the energy demands of growing cancers with high levels of cell turn over, but also because of the biochemical factors that are secreted locally by tumors themselves. This effect is termed “field cancerization”26 and has previously been shown to be detectable 965

dx.doi.org/10.1021/pr3010106 | J. Proteome Res. 2013, 12, 959−968

Journal of Proteome Research

Article

by HR-MAS NMR.18 Our results show for the first time that these observable field changes contain highly relevant diagnostic and prognostic information. The metabolic data seen in the tumor specimens are consistent with those previously published by our group.17 Specifically a decrease in the concentration of lipids, especially triglycerides in the tumor tissue compared with OTT was seen. Also levels of choline, phosphocholine, lactate, isoglutamine and especially taurine were found to be higher in tumor samples. The lower concentration of lipids and triglycerides in the tumor tissue has been attributed to the effect of rapid cellular regeneration and the high cancer cell lipid demand for cell membrane biosynthesis.17 We also observed a decrease in concentration of triglycerides in tumor tissue from LN +ve vs LN −ve patients (Table 2). The increased potential for dissemination in LN +ve cancers is likely to place additional demands on triglyceride consumption and may explain this observation. The increase in concentration of choline and its derivatives can also be related to rapid cellular regeneration in tumors, as these compounds form important cell membrane constituents.17 Taurine concentrations correlate not only with the presence of tumor, but also with the extent of local tumor advancement (Tstage). Further and more quantitative studies are necessary to analytically determine the range of taurine concentration in CRC. Taurine is a sulfur amino acid, which in the brain acts as a neurotransmitter, and it has been found in other tumor tissues including gliomas,27 squamous-cell carcinoma,28 prostate cancer and liver metastasis.29 Taurine is also used to build protein blocks, acts as a cell-membrane stabilizer, and is a key facilitator in ion-transport mechanisms.30 These functions might explain its role in the tumor tissue, helping to stabilize cancer cells and regulate their osmotic potentials. The observation of elevated isoglutamine in tumor tissue is novel and may be related to the degradation product of muramyl dipeptide (N-acetyl-muramyl-L-alanyl-D-isoglutamine, MDP), which is itself a degradation product of gut microbial cell walls from epithelial adherent bacteria. MDP is known to mediate interactions with the innate immunity protein NOD2, which is an important regulator of commensal gut microbial activity.31 NOD2 is required for the expression of antimicrobial peptides known as cryptdins32 which are believed to play an important role in innate protection against colitis and colitis-associated colorectal cancer.33 The high levels of iso-glutamine in tumor tissue biopsies compared to the off-tumor ones might indicate that the degradation of gut microbial cell walls is higher in a cancerous environment compared with OTT. MDP has also been proved to induce an anorectic effect due to satiety or sickness behavior when administrated to mice by peritoneal injection, while it is quickly excreted through the urine when administered orally.34 MDP is metabolized in the bowel, and it can be the accumulation of subproducts, such as isoglutamine, that leads to the profound anorexia so frequently seen in the setting of cancer. Clearly the possible role of isoglutamine as a marker molecule for tumor microbial interactions via MDP warrants further investigation. Amino acids such as valine, tyrosine and phenylalanine, and other metabolites such as glucose and formate were found to be discriminant in the OTT samples of patients with T1−3 tumors compared to those with T4 tumors. The decrease in glucose levels in the OTT of T4 cases likely reflects the increased energy demands of these locally advanced tumors from the neighboring gut luminal environment. Chan et al. also observed a decrease in glucose concentration of tumor tissue compared to OTT, which

can be explained in the same way with the higher consumption of energy by cancerous tissue. Increased phenylalanine concentrations have previously been reported in tumor tissue compared to OTT,17 and interestingly in the present study we see an increase in levels of this amino acid in the OTT from T4 cancers. Amino acids, such as valine, phenylalanine and tyrosine might indicate a higher protein degradation rate in the OTT samples from patients with more locally aggressive tumors. Aromatic amino acids, such as phenylalanine and tyrosine are related with gut microbial activity and the differences in concentration of these metabolites for the OTT samples suggests that gut microfloral activity is altered by neighboring cancerous tissue. Formate is one of the main metabolites involved in the onecarbon metabolism pathway of cells,35 and several genes encoding enzymes related with one-carbon metabolism have been found to be up-regulated in colorectal cancer.36,37 Finally, we observed lower concentrations of iso-butyrate, acetate and choline in the off-tumor samples of patients who were alive and disease-free at 5-years postsurgery compared to those who had died from disease relapse. Isobutyrate and acetate are products of fiber fermentation by gut microbiota.21 Increased OTT concentrations of these molecules in patients who succumbed to disease relapse supports the notion that the gut microbiome may play an important role in CRC outcome. High concentrations of choline found in the OTT of patients with inferior survival may reflect higher cell proliferative activity in these tumors with more pronounced local spread to the surroundings. As local or distant disease relapse after curativeintent surgery occurs in up to 40% of patients undergoing CRC resection,38 novel means of predicting disease-free survival are required in order to develop individually tailor therapy. Our findings indicate that HR-MAS NMR metabolic profiling may offer a means of addressing this critical issue. Building on previous studies,17 this article shows the importance of assessing the functional metabolic state of cancer specimens. We have shown for the first time that the biochemical features observed in tissue adjacent to tumors can be of prognostic significance. Our results indicate that there is value in analyzing not only tumor tissue, but also the tissue surrounding the cancerous area in terms of tumor classification, lymph node staging and survival prediction. HR-MAS NMR spectroscopy is a powerful analytical platform in this respect and highly translatable into clinical practice: sample preparation is quick and straightforward, the sample is not destroyed (so it can be used for additional assessments), the spectra are stable, objective and highly reproducible and the metabolic profile of the tissue can be obtained within around 20 min of sample acquisition, allowing rapid and accurate tissue assessment. Metabonomic analysis therefore has the capacity to augment current clinical diagnostic tools and may serve as a translatable tool for more accurate staging of patients with colorectal cancer. The metabolic phenotype of colorectal cancer provided by in vivo analysis has a significant potential in the development of personalized surgical and therapeutic strategies in these patients.



ASSOCIATED CONTENT

S Supporting Information *

Supplemental figure and table. This material is available free of charge via the Internet at http://pubs.acs.org. 966

dx.doi.org/10.1021/pr3010106 | J. Proteome Res. 2013, 12, 959−968

Journal of Proteome Research



Article

J. R.; Nicholson, J. K. Pharmaco-metabonomic phenotyping and personalized drug treatment. Nature 2006, 440, 1073−7. (15) Backshall, A.; Sharma, R.; Clarke, S. J.; Keun, H. C. Pharmacometabonomic profiling as a predictor of toxicity in patients with inoperable colorectal cancer treated with capecitabine. Clin. Cancer Res. 2011, 17, 3019−28. (16) Beckonert, O.; Coen, M.; Keun, H. C.; Wang, Y.; Ebbels, T. M.; Holmes, E.; Lindon, J. C.; Nicholson, J. K. High-resolution magic-anglespinning NMR spectroscopy for metabolic profiling of intact tissues. Nat. Protoc. 2010, 5, 1019−32. (17) Chan, E. C.; Koh, P. K.; Mal, M.; Cheah, P. Y.; Eu, K. W.; Backshall, A.; Cavill, R.; Nicholson, J. K.; Keun, H. C. Metabolic profiling of human colorectal cancer using high-resolution magic angle spinning nuclear magnetic resonance (HR-MAS NMR) spectroscopy and gas chromatography mass spectrometry (GC/MS). J. Proteome Res. 2009, 8, 352−61. (18) Yakoub, D.; Keun, H. C.; Goldin, R.; Hanna, G. B. Metabolic profiling detects field effects in nondysplastic tissue from esophageal cancer patients. Cancer Res. 2010, 70, 9129−36. (19) Kinross, J. M.; Holmes, E.; Darzi, A. W.; Nicholson, J. K. Metabolic phenotyping for monitoring surgical patients. Lancet 2011, 377, 1817−9. (20) Mirnezami, R.; Nicholson, J.; Darzi, A. Preparing for precision medicine. N. Engl. J. Med. 2012, 366, 489−91. (21) Wang, Y.; Holmes, E.; Comelli, E. M.; Fotopoulos, G.; Dorta, G.; Tang, H.; Rantalainen, M. J.; Lindon, J. C.; Corthésy-Theulaz, I. E.; Fay, L. B.; Kochhar, S.; Nicholson, J. K. Topographical variation in metabolic signatures of human gastrointestinal biopsies revealed by highresolution magic-angle spinning 1H NMR spectroscopy. J. Proteome Res. 2007, 6, 3944−51. (22) Cloarec, O.; Dumas, M. E.; Craig, A.; Barton, R. H.; Trygg, J.; Hudson, J.; Blancher, C.; Gauguier, D.; Lindon, J. C.; Holmes, E.; Nicholson, J. Statistical total correlation spectroscopy: an exploratory approach for latent biomarker identification from metabolic 1H NMR data sets. Anal. Chem. 2005, 77, 1282−9. (23) Cloarec, O.; Dumas, M. E.; Trygg, J.; Craig, A.; Barton, R. H.; Lindon, J. C.; Nicholson, J. K.; Holmes, E. Evaluation of the orthogonal projection on latent structure model limitations caused by chemical shift variability and improved visualization of biomarker changes in 1H NMR spectroscopic metabonomic studies. Anal. Chem. 2005, 77, 517−26. (24) Martinez-Bisbal, M. C.; Marti-Bonmati, L.; Piquer, J.; Revert, A.; Ferrer, P.; Llacer, J. L.; Piotto, M.; Assemat, O.; Celda, B. 1H and 13C HR-MAS spectroscopy of intact biopsy samples ex vivo and in vivo 1H MRS study of human high grade gliomas. NMR Biomed. 2004, 17, 191− 205. (25) Sitter, B.; Bathen, T. F.; Singstad, T. E.; Fjosne, H. E.; Lundgren, S.; Halgunset, J.; Gribbestad, I. S. Quantification of metabolites in breast cancer patients with different clinical prognosis using HR MAS MR spectroscopy. NMR Biomed. 2010, 23, 424−31. (26) Slaughter, D. P.; Southwick, H. W.; Smejkal, W. Field cancerization in oral stratified squamous epithelium; clinical implications of multicentric origin. Cancer 1953, 6, 963−8. (27) Lehtimaki, K. K.; Valonen, P. K.; Griffin, J. L.; Vaisanen, T. H.; Grohn, O. H.; Kettunen, M. I.; Vepsäläinen, J.; Ylä-Herttuala, S.; Nicholson, J.; Kauppinen, R. A. Metabolite changes in BT4C rat gliomas undergoing ganciclovir-thymidine kinase gene therapy-induced programmed cell death as studied by 1H NMR spectroscopy in vivo, ex vivo, and in vitro. J. Biol. Chem. 2003, 278, 45915−23. (28) El-Sayed, S.; Bezabeh, T.; Odlum, O.; Patel, R.; Ahing, S.; MacDonald, K.; Somorjai, R. L.; Smith, I. C. An ex vivo study exploring the diagnostic potential of 1H magnetic resonance spectroscopy in squamous cell carcinoma of the head and neck region. Head Neck 2002, 24, 766−72. (29) Moreno, A.; Lopez, L. A.; Fabra, A.; Arus, C. 1H MRS markers of tumour growth in intrasplenic tumours and liver metastasis induced by injection of HT-29 cells in nude mice spleen. NMR Biomed. 1998, 11, 93−106. (30) Maher, S. G.; Condron, C. E.; Bouchier-Hayes, D. J.; Toomey, D. M. Taurine attenuates CD3/interleukin-2-induced T cell apoptosis in an

AUTHOR INFORMATION

Corresponding Author

*Tel: +44 (0)20 7594 3194. Fax: +44 (0)20 7594 3066. E-mail: j. [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS Biomedical Research Council/NIHR funding (UK), (awarded to A.D. and J.K.N.). We thank Mr. Andrew Smith for collecting the tissue samples and Dr. Kirill A. Veselkov for statistical consultation.



REFERENCES

(1) Hewish, M.; Lord, C. J.; Martin, S. A.; Cunningham, D.; Ashworth, A. Mismatch repair deficient colorectal cancer in the era of personalized treatment. Nat. Rev. Clin. Oncol. 2010, 7, 197−208. (2) Cunningham, D.; Atkin, W.; Lenz, H. J.; Lynch, H. T.; Minsky, B.; Nordlinger, B.; Starling, N. Colorectal cancer. Lancet 2010, 375, 1030− 47. (3) Lee, D.; Muston, D.; Sweet, A.; Cunningham, C.; Slater, A.; Lock, K. Cost effectiveness of CT colonography for UK NHS colorectal cancer screening of asymptomatic adults aged 60−69 years. Appl. Health Econ. Health Policy 2010, 8, 141−54. (4) Bosset, J. F.; Calais, G.; Mineur, L.; Maingon, P.; Radosevic-Jelic, L.; Daban, A.; Bardet, E.; Beny, A.; Briffaux, A.; Collette, L. Enhanced tumorocidal effect of chemotherapy with preoperative radiotherapy for rectal cancer: preliminary results–EORTC 22921. J. Clin. Oncol. 2005, 23 (5), 620−7. (5) Bosset, J. F.; Collette, L.; Calais, G.; Mineur, L.; Maingon, P.; Radosevic-Jelic, L.; Daban, A.; Bardet, E.; Beny, A.; Ollier, J. C. EORTC Radiotherapy Group Trial 22921. Chemotherapy with preoperative radiotherapy in rectal cancer. N. Engl. J. Med. 2007, 355, 1114−23. (6) Verdecchia, A.; Francisci, S.; Brenner, H.; Gatta, G.; Micheli, A.; Mangone, L.; Kunkler, I. EUROCARE-4 Working Group. Recent cancer survival in Europe: a 2000−02 period analysis of EUROCARE-4 data. Lancet Oncol. 2007, 8, 784−96. (7) Nicholson, J. K.; Lindon, J. C. Systems biology: Metabonomics. Nature 2008, 455, 1054−6. (8) Nicholson, J. K.; Lindon, J. C.; Holmes, E. ‘Metabonomics’: understanding the metabolic responses of living systems to pathophysiological stimuli via multivariate statistical analysis of biological NMR spectroscopic data. Xenobiotica 1999, 29, 1181−9. (9) Ikeda, A.; Nishiumi, S.; Shinohara, M.; Yoshie, T.; Hatano, N.; Okuno, T.; Bamba, T.; Fukusaki, E.; Takenawa, T.; Azuma, T.; Yoshida, M. Serum metabolomics as a novel diagnostic approach for gastrointestinal cancer. Biomed. Chromatogr. 2012, 26, 548−58. (10) Backshall, A.; Alferez, D.; Teichert, F.; Wilson, I. D.; Wilkinson, R. W.; Goodlad, R. A.; Bamba, T.; Fukusaki, E.; Takenawa, T.; Azuma, T.; Yoshida, M. Detection of metabolic alterations in non-tumor gastrointestinal tissue of the Apc(Min/+) mouse by 1H MAS NMR spectroscopy. J. Proteome Res. 2009, 8, 1423−30. (11) Qiu, Y.; Cai, G.; Su, M.; Chen, T.; Liu, Y.; Xu, Y.; Ni, Y.; Zhao, A.; Cai, S.; Xu, L. X.; Jia, W. Urinary metabonomic study on colorectal cancer. J. Proteome Res. 2010, 9, 1627−34. (12) Hirayama, A.; Kami, K.; Sugimoto, M.; Sugawara, M.; Toki, N.; Onozuka, H.; Kinoshita, T.; Saito, N.; Ochiai, A.; Tomita, M.; Esumi, H.; Soga, T. Quantitative metabolome profiling of colon and stomach cancer microenvironment by capillary electrophoresis time-of-flight mass spectrometry. Cancer Res. 2009, 69, 4918−25. (13) Bertini, I.; Cacciatore, S.; Jensen, B. V.; Schou, J. V.; Johansen, J. S.; Kruhoffer, M.; Luchinat, C.; Nielsen, D. L.; Turano, P. Metabolomic NMR fingerprinting to identify and predict survival of patients with metastatic colorectal cancer. Cancer Res. 2012, 72, 356−64. (14) Clayton, T. A.; Lindon, J. C.; Cloarec, O.; Antti, H.; Charuel, C.; Hanton, G.; Provost, J. P.; Le Net, J. L.; Baker, D.; Walley, R. J.; Everett, 967

dx.doi.org/10.1021/pr3010106 | J. Proteome Res. 2013, 12, 959−968

Journal of Proteome Research

Article

in vitro model of activation-induced cell death (AICD). Clin. Exp. Immunol. 2005, 139, 279−86. (31) Petnicki-Ocwieja, T.; Hrncir, T.; Liu, Y. J.; Biswas, A.; Hudcovic, T.; Tlaskalova-Hogenova, H.; Kobayashi, K. S. Nod2 is required for the regulation of commensal microbiota in the intestine. Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 15813−8. (32) Kobayashi, K. S.; Chamaillard, M.; Ogura, Y.; Henegariu, O.; Inohara, N.; Nunez, G.; Flavell, R. A. Nod2-dependent regulation of innate and adaptive immunity in the intestinal tract. Science 2005, 307, 731−4. (33) Saleh, M.; Trinchieri, G. Innate immune mechanisms of colitis and colitis-associated colorectal cancer. Nat. Rev. Immunol. 2011, 11, 9− 20. (34) Fosset, S.; Fromentin, G.; Rampin, O.; Lang, V.; Mathieu, F.; Tome, D. Pharmacokinetics and feeding responses to muramyl dipeptide in rats. Physiol. Behav. 2003, 79, 173−82. (35) Holmes, E.; Loo, R. L.; Stamler, J.; Bictash, M.; Yap, I. K.; Chan, Q.; Ebbels, T.; De Iorio, M.; Brown, I. J.; Veselkov, K. A.; Daviglus, M. L.; Kesteloot, H.; Ueshima, H.; Zhao, L.; Nicholson, J. K.; Elliott, P. Human metabolic phenotype diversity and its association with diet and blood pressure. Nature 2008, 453, 396−400. (36) Selvakumar, P.; Lakshmikuttyamma, A.; Lawman, Z.; Bonham, K.; Dimmock, J. R.; Sharma, R. K. Expression of methionine aminopeptidase 2, N-myristoyltransferase, and N-myristoyltransferase inhibitor protein 71 in HT29. Biochem. Biophys. Res. Commun. 2004, 322, 1012−7. (37) Knock, E.; Deng, L.; Krupenko, N.; Mohan, R. D.; Wu, Q.; Leclerc, D.; Gupta, S.; Elmore, C. L.; Kruger, W.; Tini, M.; Rozen, R. Susceptibility to intestinal tumorigenesis in folate-deficient mice may be influenced by variation in one-carbon metabolism and DNA repair. J. Nutr. Biochem. 2011, 22, 1022−9. (38) Andre, N.; Schmiegel, W. Chemoradiotherapy for colorectal cancer. Gut 2005, 54, 1194−202.

968

dx.doi.org/10.1021/pr3010106 | J. Proteome Res. 2013, 12, 959−968