Anal Bioanal Chem DOI 10.1007/s00216-015-8662-x

RESEARCH PAPER

Characterizing metabolic changes in human colorectal cancer Michael D. Williams 1 & Xing Zhang 1 & Jeong-Jin Park 3 & William F. Siems 1 & David R. Gang 3 & Linda M. S. Resar 4,5,6 & Raymond Reeves 2 & Herbert H. Hill Jr 1

Received: 16 October 2014 / Revised: 13 March 2015 / Accepted: 24 March 2015 # Springer-Verlag Berlin Heidelberg 2015

Abstract Colorectal cancer (CRC) remains a leading cause of cancer death worldwide, despite the fact that it is a curable disease when diagnosed early. The development of new screening methods to aid in early diagnosis or identify precursor lesions at risk for progressing to CRC will be vital to improving the survival rate of individuals predisposed to CRC. Metabolomics is an advancing area that has recently seen numerous applications to the field of cancer research. Altered metabolism has been studied for many years as a means to understand and characterize cancer. However, further work is required to establish standard procedures and Electronic supplementary material The online version of this article (doi:10.1007/s00216-015-8662-x) contains supplementary material, which is available to authorized users. * Raymond Reeves [email protected] * Herbert H. Hill, Jr [email protected] Linda M. S. Resar [email protected] 1

Department of Chemistry, Washington State University, Pullman, WA 99164, USA

improve our ability to identify distinct metabolomic profiles that can be used to diagnose CRC or predict disease progression. The present study demonstrates the use of direct infusion traveling wave ion mobility mass spectrometry to distinguish metabolic profiles from CRC samples and matched nonneoplastic epithelium as well as metastatic and primary tumors at different stages of disease (T1–T4). By directly infusing our samples, the analysis time was reduced significantly, thus increasing the speed and efficiency of this method compared to traditional metabolomics platforms. Partial least squares discriminant analysis was used to visualize differences between the metabolic profiles of sample types and to identify the specific m/z features that led to this differentiation. Identification of the distinct m/z features was made using the human metabolome database. We discovered alterations in fatty acid biosynthesis and oxidative, glycolytic, and polyamine pathways that distinguish tumors from non-malignant colonic epithelium as well as various stages of CRC. Although further studies are needed, our results indicate that colonic epithelial cells undergo metabolic reprogramming during their evolution to CRC, and the distinct metabolites could serve as diagnostic tools or potential targets in therapy or primary prevention. Keywords Bioanalytical methods . Ion mobility mass spectrometry . Metabolomics . Colorectal cancer . Cancer biomarkers . Mass spectrometry ICP-MS

2

School of Molecular Biosciences, Washington State University, Pullman, WA 99164, USA

3

Institute of Biological Chemistry, Washington State University, Pullman, WA 99164, USA

4

Department of Medicine, The Johns Hopkins University School of Medicine, Baltimore, MD 21205, USA

Introduction

5

Department of Oncology, The Johns Hopkins University School of Medicine, Baltimore, MD 21205, USA

6

Institute for Cellular Engineering, The Johns Hopkins University School of Medicine, Baltimore, MD 21205, USA

Colorectal cancer (CRC) remains a primary cause of cancer related mortality in the USA today. It is also the third most commonly diagnosed cancer among men and women in this country [1]. In fact, current estimates predict that 5 % of the

M.D. Williams et al.

population will be diagnosed with CRC sometime during their lifetime [1]. Moreover, almost 50,000 men and women are expected to die from this disease in 2015 in the USA alone. Early detection is key to improving survival and outcomes for CRC. Unfortunately, only 40 % of individuals diagnosed with CRC present at early, localized stages of the disease. These individuals have a 90 % five year survival rate (FYSR), while those diagnosed at late stages have only a 13 % FYSR [2]. Thus, further research is needed to uncover novel diagnostic and prognostic tools. Reprogramming energy metabolism has emerged as a hallmark of cancer [3]. Cancer cells rely on altered cellular metabolism to meet their anabolic demands that result from deregulated cellular growth and aberrant differentiation. The best-studied alteration in tumor metabolism is the dependence upon aerobic glycolysis or the Warburg effect. This change results in excess pyruvate and lactate production due to a switch from oxidative phosphorylation to glycolysis, which appears to provide advantages to tumor growth. These and other alterations result in a unique metabolic profile, characteristic of cancer cells, and provide the opportunity to distinguish cancer cells from normal cells. Current investigations are aimed at exploiting these differences as biomarkers and therapeutic targets [3–6]. While ongoing studies of metabolic changes that occur during the development and progression of CRC may reveal new phenotypic markers for the disease, significant progress has been made in understanding genomic and proteomic alterations with disease progression. For example, the High Mobility Group A1 (HMGA1) gene is up-regulated with tumor progression in diverse tumors, including CRC [7–10]. This gene encodes two isoforms of the HMGA1 protein, HMGA1a and HMGA1b, which result from alternatively spliced mRNA [11–14]. HMGA1 proteins are chromatin remodeling proteins that alter chromosome and nucleosome structure and recruit transcription factors complexes to DNA, thus modulating gene expression. HMGA1 is among the most abundant nonhistone, chromatin binding proteins found in cancer cells [7–9]. The HMGA1 gene is highly expressed during embryonic development with low or absent expression in adult differentiated tissues [9, 10]. Embryonic stem cells [15], adult stem cells [9], and all poorly differentiated cancers studied to date [14, 16–21] are also enriched in HMGA1 proteins. In fact, poor outcomes correlate with high expression of HMGA1 in diverse tumors types [17–21]. HMGA1 was identified among the genes most enriched in primary CRC, as compared to adjacent, non-neoplastic colonic epithelium [10]. In CRC cells, HMGA1 regulates genes involved in an epithelialmesenchymal transition, suggesting that it drives tumor progression by inducing stem cell transcriptional networks [14]. HMGA1 is also required for cancer stem cell properties in CRC cells [14, 20, 22, 23]. In fact, silencing HMGA1 in CRC cells depletes tumor-initiator or cancer stem cells [14].

Transgenic mice that overexpress Hmga1 in intestinal epithelium develop hyper-proliferation, crypt fusion, and polyposis, which are also key features of premalignant human polyps [14]. Together, these findings suggest that HMGA1 is a key regulator in tumor progression in CRC. While screening colonoscopy is an effective approach to detect and diagnose CRC, it is limited by cost, low sensitivity, and involves a difficult and time-consuming procedure. Additionally, in developing countries, the necessary screening tools are not widely available. Even in areas where colonoscopy is readily accessible, many individuals fail to undergo appropriate screening. Development of new and faster screening methods is therefore urgently needed to improve the overall FYSR, provide earlier detection, reduce cost, and increase sensitivity. Because most tumor cells harness alternative metabolic pathways, advances in global metabolomics have the potential to identify individuals who may be at risk for progression to CRC from premalignant lesions and inform the ongoing care of CRC patients. Recently, the metabolomes of CRCs and adjacent, nonmalignant tissue have been examined using several techniques, including the following: magic angle spinning–nuclear magnetic resonance (MAS-NMR), proton NMR, gas chromatography–mass spectrometry (GC-MS), ultra-performance liquid chromatography–quadrupole time-of-flight mass spectrometry (UPLC-QToF-MS), and capillary electrophoresis– MS (CE-MS) [24–33]. Metabolic changes in urine and serum from CRC patients have also been investigated using platforms that include GC-MS, UPLC-QToF-MS(MS/MS), Fourier transform ion-cyclotron-MS (FTICR-MS), and proton NMR (1H-NMR) [34–41]. Evidence from these studies suggests that enhanced glycolysis and reduced oxidative phosphorylation occur in CRC, consistent with the Warburg effect [27, 28, 42, 43]. Several studies also demonstrated the accumulation of several common amino acids in CRC [27, 28, 37–41]. Thus, increasing evidence indicates that metabolite signatures have the potential to be used as new biomarkers for CRC or disease progression. A better understanding of the molecular underpinnings of altered CRC metabolism could also uncover novel agents to treat, or even prevent, CRC. Despite the body of literature on metabolic aberrations associated with CRC [42], the identification metabolic profiles that confirm the presence of CRC or metastatic disease has remained elusive. For an analytical method to be clinically valuable for diagnostic and prognostic applications, speed, efficiency, and accuracy are paramount. Thus, there is a dire need to develop better technologies to meet these requirements. Over the last decade, ion mobility–mass spectrometry (IMMS) has emerged as an effective tool for the evaluation and metabolic profiling of complex biological samples [44–48]. Ion mobility spectrometry (IMS) is highly sensitive, separating ions based upon their cross-sectional size to charge

Metabolic changes in human colorectal cancer

ratio as they travel through an electrical gradient on a millisecond time scale. Mass spectrometry (MS) provides both sensitivity and high resolution, detecting analytes based on their mass to charge (m/z) ratios. IMMS brings together IMS and MS. The Waters Synapt G2-S HDMS (G2-S) system couples a high-resolution ToF MS with a unique traveling wave IMS (TWIMS) [49] and is easily combined with various chromatographic techniques (i.e., LC, GC) prior to TWIMS–MS (TWIMMS) analysis via the unique sample introduction interface. Moreover, rapid analysis of complex metabolic samples can be achieved through direct infusion of samples into the instrument. TWIMS differs from traditional drift tube IMS (DT-IMS) in its mode of ion separation. In DT-IMS, ions are separated as they travel down a chemical gradient, through a constant electric field against the flow of a drift gas, usually nitrogen. DT-IMS is also typically performed at atmospheric pressure. TWIMS, however, is carried out at low pressures (~3 mbar) and with a time-varying electric field. In fact, it is this time-varying electrical field which carries ions through the ion mobility cell of the instrument. An electrical potential wave is created by applying, sequentially, voltage to a series of stacked rings so that the voltage moves, with time, down the drift cell. Ions Bsurf^ along this potential wave, with the larger and heavier ions rolling back as the wave passes and lighter, smaller ions being pushed forward. Thus, the ions are separated according to their mobility, by their size to charge ratio. Due to the variation of the electric field within a TWIMS separation collision cross section (CCS) cannot be directly measured using this technology, but other metrics, such as drift time, can be determined. Variance in drift time using TWIMS analysis is often sufficient to separate isomeric species as well. The primary advantage of using TWIMS in a commercial instrument is the ability to achieve meaningful ion separation at lower pressures than can be used in traditional DT-IMS, enabling more efficient ion transfer from the mobility cell to the MS detector. TWIMMS coupled to UPLC (UPLC-TWIMMS) has been previously applied for metabolic profiling of human saliva and the analysis of lipids in human plasma [50, 51]. The present work seeks to identify metabolic profiles that could be used to diagnose CRC and predict disease progression. We hypothesized that TWIMMS technology can be applied for discovery of metabolic alterations in human CRC which differentiate cancer from adjacent, non-malignant tissues (NMT), primary tumors from metastatic disease, and stages 1–4 from one another. We hypothesized further that this method would improve the speed, efficiency, and accuracy of diagnoses, making it amenable to clinical applications. To test these hypotheses, global analysis of metabolite extracts were obtained from human colonic epithelium after surgical removal from CRC patients and included Bmatched^ non-malignant samples from the same individual. These tissues were previously investigated and found to overexpress HMGA1 protein.

We used partial least squares discriminant analysis (PLS-DA) for global comparisons of our results to establish a biochemical signature that indicates the presence of CRC and disease progression. Although further validation studies are needed, our results suggest specific biomarkers and metabolic profiles could be used for diagnosis, prognosis, and staging of CRC. We also report a marked improvement in efficiency and speed of analysis with our approach as compared to traditional metabolomics platforms.

Materials and methods Human tissue samples Surgical tissue samples were obtained from the Providence Regional Cancer Center, Sacred Heart Hospital, Spokane, WA, 99204. The samples were immediately frozen on dry ice following their removal from patients and stored in a −80 °C freezer at the Spokane campus of Washington State University until their subsequent transport (on dry ice) to the WSU, Pullman, campus where experimental analyses were performed. Histologic diagnoses of all of the tissue samples were performed by experienced pathologists at the Providence Regional Cancer Center. All research and data collection involving these tissue specimens were in accordance with the National Institutes of Health Guidelines for the Protection of Human Subjects and all of the experimental protocols were approved by the Institutional Review Board (IRB) from both Sacred Heart Hospital and Washington State University. Samples and metabolite extraction The primary samples of interest were comprised of nine matched sets of human colonic tissues, which included CRC and adjacent NMT from the same individual, with the exception of a single set that consisted of both primary and metastatic CRC tissue specimens paired with a NMT sample. These sets were collected during routine colonoscopy for the removal of CRC tumors. We also included tumor stages T1–T4, which refer to tumor size and invasion. Table 1 displays the pathologic features of each sample. A small, approximately pea-sized sample was dissected from colonic biopsies, placed into a 2.5-mL glass microcentrifuge vials and weighed. Sonication of samples in 1 mL/mg of cold extraction buffer was used to extract metabolites from the tissue. Sonication was repeated, with the samples stored on ice between sonication steps. After final sonication, this mixture was centrifuged at 14,000×g for 30 min to remove cell debris (repeated as necessary) and supernatant was collected for use as metabolite extraction into a clean microcentrifuge tube. All samples were frozen at −80 °C until time of analysis.

M.D. Williams et al. Table 1 Tissue sample list including and histopathological designations. Each designated set was collected from a single individual. Type, stage, and metastatic/primary designation was determined histologically by a clinical pathologist at time of harvest and is reported here. The NMT and CRC abbreviations were adopted by the authors for ease of reference and uniformity. Histopathologically, stages T1–T4 have the following characteristics: T1—superficial tumor confined to the mucosal layer within the colon; T2—invading the submucosa; T3—invading the muscular layer (muscularis propria); T4—invading adjacent organs or perforating the peritoneum Set

Type

Stage

1

NMT CRC CRC NMT CRC NMT CRC

Normal T3 T2 Normal T3 Normal T4

NMT CRC NMT CRC NMT CRC NMT CRC NMT CRC NMT CRC

Normal T1 Normal T2 Normal T3 Normal T3 Normal T3 Normal T2

2 3 4 5 6 7 8 9

Metastatic/primary

Primary Primary Metastatic

Instrumentation A Waters ACQUITY UPLC (Waters Corporation, Manchester, UK) was used as the sample introduction tool. However, only the auto-injector and solvent manager of this unit were used in the analysis, to achieve direct injection. A stainless steel union (Valco Instruments Co. Inc., Houston, TX) took the place of the UPLC column and was inserted so that the line from the auto sampler was coupled directly to the ESI source of a Waters Synapt G2-S TWIMMS (Waters Corporation, Manchester, UK). Data acquisition, instrument setup, and configuration and user interface were provided via the MassLynx 4.1 (Waters Corporation, Manchester, UK) software package. UPLC and TWIMMS instruments were operated via separate PC computers, acting in concert, both equipped with the MassLynx software.

Metastatic

IMMS conditions Primary Primary Metastatic Metastatic Primary Primary

For negative mode, the ESI conditions were as follows: capillary voltage 2.5 kV; source temperature 90 °C; cone voltage 40 V; desolvation temperature 150 °C; desolvation gas flow 500 L/h (N2 gas); scan rate 0.1 s/scan with 0.015 s inter-scan delay. Optimized traveling wave height and speed were 29.0 V and 650 m/s, respectively. For positive mode, the ESI conditions were as follows: capillary voltage 3.65 kV; source temperature 90 °C; cone voltage 50 V; desolvation temperature 200 °C; desolvation gas flow 500 L/h (N2 gas); scan rate 0.1 s/scan with 0.015 s inter-scan delay. Optimized traveling wave height and speed were 40.0 V and 652 m/s, respectively.

Experimental approach

Data acquisition

A flow diagram of the experimental approach used in collecting and analyzing data is shown in Fig. 1. Employing these methods, we initially investigated metabolic variation between localized, primary CRCs, and adjacent NMT to determine if global differences could be detected between these samples. Primary, localized CRCs were then compared with local invasive and distal metastatic tissues as a first step in characterizing disease progression. Finally, the variance between CRC stages: T1, T2, T3, and T4, respectively, was analyzed.

Sodium formate solution was used to mass calibrate the G2-S system. Metabolite extractions were diluted 10-fold in 1:1 ACN and water, 0.1 % formic acid, then analyzed by direct infusion TWIMMS (DI-TWIMMS), using both positive and negative electrospray ionization (ESI). Analysis was streamlined by the use of the UPLC system as an introduction tool, which allowed the analysis of several samples at a time using the MassLynx sample list feature. Samples were first detected in the traveling wave IMS and then passed into the ToF mass analyzer. The instrument parameters used were optimized prior to analysis of mouse or human samples. The injection volume was 10 μL. Running buffers were as follows: (A) 0.1 % formic acid and (B) 0.1 % formic acid in ACN. Buffers A and B were initially set to 50/50 and the flow rate was set to 60 μL/ min (0–0.45 min) and then to 10 μL/min (0.5–5.0 min), for optimized ESI and reduced charge suppression and matrix effects. Mass spectrometry data were collected from 0 to 3.5 min. A timing off-set (G2-S, 3.5 min; UPLC, 5.0 min) in the two instruments was necessary to allow the large data files

Chemicals and reagents Water (LC/MS, Optima® grade), acetonitrile (ACN, LC/ MS, Optima® grade), and formic acid were obtained from Thermo Fisher Scientific (Waltham, MA). Leucine-enkephalin (Leu-enk) was obtained from Waters Corporation (Manchester, UK).

Metabolic changes in human colorectal cancer

Fig. 1 Method illustration: a colon tissue samples were collected from individuals during surgery and sent to WSU. Metabolites were extracted via sonication in cold 1:1 ACN/water. b Two-dimensional data were collected via IMS in tandem with MS (IMMS). c IMMS data were interpreted statistically via PLS-DA. Scores plots provided a visualization of statistical separation and groupings of sample types. Loading plots

allowed identification of influential ion features. d Lists of these important features were exported and analyzed for specific differences. e Direct comparisons of critical ion features lead to biomarker candidate identification and comparative analyses. f These differences were then expressed visually in charts (as shown) and tables

enough time to be saved and for the MS and UPLC to start in concert. Leu-enk was sampled, via lock-mass spray, for the first 30 s of each run at a flow rate of 5 μL/min, as an external standard to account for mass accuracy and sensitivity. Instrument cleaning was performed between each run and each sample was run randomly and in triplicate.

corrected, utilizing the Accurate Mass Measure tool within the MassLynx software, to correct for instrument calibration drift during data acquisition. MassLynx software was used to create lists of data files for each analysis along with descriptors for each file, indicating the type of sample, identifying numbers, and grouping information. MarkerLynx XS (Water Corporation, Manchester, UK) software package was then used to process the data files listed and to produce a statistical analysis of the differences within samples based on the IMS and MS data according to specified methods and parameters.

Data analysis Statistical data analyses were conducted using MassLynx and MarkerLynx software packages (Water Corporation, Manchester, UK; v 4.1). Additional analyses were performed using Microsoft® Excel (Microsoft, One Microsoft Way, Redmond, WA, USA) with add-in Merge-Tables Wizard (AbleBits.com, Add-in Express, Ltd., Homel, Belarus, BY). MassLynx software was used for manually viewing MS and IMS spectra for individual data files. Additionally, data were centered with the MS Spectrum viewer and then Lock-Mass

MarkerLynx analysis of Di-TWIMMS data While MarkerLynx software is commonly used as an analysis tool for LC/MS data, this is the first application of this software package for the investigation of IMMS data that is not paired with LC. We, therefore, provide a brief explanation of the procedure for this use of MarkerLynx. The method

M.D. Williams et al.

parameters were set up using the method editor, located in the MarkerLynx tab of the MassLynx primary interface. The method was set to process data from Function 3 (IMS). Settings for initial and final retention time were left at the default setting of 0.00, which instructs the method to use the minimum and maximum of the measured data. Low and high mass parameters were set to 50.00 and 1,200.00 (Da), respectively. Full, remaining parameters can be viewed in Fig. S1 in the Electronic Supplementary Material (ESM). Since ion mobility data must be collected continuously, as profile data, it must be converted to centroid data prior to analysis. Centroid data was generated using the Accurate Mass Measure tool within MassLynx as described above. The MarkerLynx method was then used to process centroid data files from the lists created (see above). MarkerLynx XS data viewer can then be used for reviewing drift time (reported in the BRet. Time^ column) and m/z data as well as to open EZinfo statistical package. Multivariate statistical analysis, including principal component analysis (PCA), PLS-DA, and orthogonal projections to latent structures discriminate analysis (OPLS-DA) were carried out to discriminate metabolite peaks. All data were mean centered and standard deviation scaled prior to statistical analysis. By default, MarkerLynx EZinfo software performed an initial PCA analysis of the data. Statistical parameters, including R2 and Hotelling’s T2 values, can be found in Figs. S2 through S10 in the ESM. PCA is an unsupervised multivariate statistical analysis that can be used to find variations within a data and to visualize those differences in multi-dimensional space. While PCA can be a powerful

tool for establishing sample variation, due to the limited size of our sample set and the heterogeneity of tumor tissues, we found it necessary to use a supervised statistical model, PLS-DA, to clearly visualize global differences between sample types and to identify m/z values which correspond to influential metabolites. OPLS-DA, which identifies differences between two clearly defined groups of data, was then used, primarily to provide an S-plot which was helpful in determining the metabolites that had the highest degree of influence in the observed variation. Lists of the most influential metabolites, consisting of m/z, drift time (td), and weighted average values (according to sample type), were exported to Excel spreadsheets. The weighted average values were used to calculate relative changes in each metabolite. For example, the average values in NMT samples was subtracted from the average values in CRC samples to give a comparison of the relative expression of those metabolites in CRC tissues as compared to NMTs. Metabolite identifications In order to understand the metabolites and the metabolic pathways that were affected during tumor development, we used the Human Metabolome Database (HMDB, http://www. hmdb.ca/) mass search feature. Using exact mass data (50 % change from NMT to CRC are listed

M.D. Williams et al.

values, in NMT tissues. Conversely, a value of −100 indicates the opposite to be true. Namely, there was little to absent detection in CRC with significant detection in NMT. Values exceeding +100 indicate that the metabolite was detected in CRC samples with an ORI more than double that observed in the NMT samples, but the ORI was non-zero in both sample types. As detailed in this chart, several metabolites involved in energy metabolism and fatty acid biosynthesis were observed to have strong influence on the observed separation. Additional classes of metabolites that changed significantly include amino acids (AAs), nucleotides (NTs), nucleosides and derivatives, cofactors, tricarboxylic acid (TCA) cycle and pentose phosphate pathway (PPP) intermediates, bile acids, and oxidative metabolites.

regulated metabolites in the metastatic tissues represented by positive values. Interestingly, as shown in Fig. 5, several fatty acids (FAs) and acyl-carnitines increased in tumors that had progressed from localized carcinomas to metastatic disease. Significant increases in several metabolites involved in the glycolytic pathway, such as glucose-1-phosphate (G1P) and galactinol, were also observed. In addition, there were increases in the levels of various dipeptides as well as in putrescine, a polyamine. Polyamines are important growth factors required for cellular proliferation and are known to be present in high concentrations in CRC tissues [52–57]. It is, therefore, reasonable to expect that putrescine might be up-regulated in metastatic CRC.

Metastatic vs. primary CRC

Comparing CRC stages

After successfully establishing that it was possible to distinguish metabolic differences between CRC and matched, nonmalignant colonic epithelium tissues, we sought to determine whether we could identify differences between various stages of CRC. Figure 4 shows the separation of metastatic and primary CRCs using PLS-DA. As before, OPLS-DA was used to reveal the most significant metabolic features that differentiated metastatic from primary CRCs. These differentiating features were compared in the same manner as the NMT vs. CRC comparison, with the modification that in this analysis the primary values were subtracted from the metastatic values. Figure 5 shows a delta chart, similar to that in Fig. 3, which compares metastatic to primary CRC tissues, with up-

Next, we compared metabolites from CRC samples at different stages of tumor progression. In the negative mode, the various stages of CRC were separated using PCA, an unsupervised statistical technique. As shown in Fig. 6, each stage segregated well in PCA, with the exception of a single group of outliers in the T2 stage. It should be noted that this outlying group was the result of three replicate analyses of a single tissue sample. The reason for its apparently anomalous positioning is unknown but could simply reflect the fact that the CRC specimen was highly invasive and about to extend through the muscularis propria (the outer muscular layer; T3 tumors are distinguished from T2 tumors by extension through the muscularis propria; T2 tumors have invaded the

Fig. 4 PLS-DA of metastatic (M) compared to primary (P) CRC. a Metabolomes of metastatic and primary CRCs, detected in the negative mode, separated via PLS-Da. b Metabolomes of metastatic and primary CRCs, detected in the positive mode, separated via PLS-Da

Metabolic changes in human colorectal cancer

Fig. 5 Delta charts comparing primary and metastatic CRC. The chart lists possible identifications of important, differentiating metabolic features, detected in the positive (a) and negative (b) modes, along with

the percent change of that feature in metastatic CRC tissue as primary CRC tissue

submucosa and muscularis propria without extending through it). Further, a comparison of different stages was made in order to determine whether there are observable metabolite differences between one stage and the next. The groupings were as follows: 1.—T1 vs. T2; 2.—T2 vs. T3; 3.—T3 vs. T4. As shown in Fig. 7, in both the positive and negative mode, we found that the more advanced stages could be separated from the preceding stage using PLS-DA. Furthermore, as shown in the heat maps in Figs. 8 and 9, it was possible to identify metabolic features that distinguish various stages of CRC.

We also identified potential biomarkers that indicate the presence of CRC as well as possible markers of disease progression. Since we are using direct infusion in this analysis, it is important to note that the quality of the data and the information obtained is more than likely impacted by matrix affects, such as ionization suppression and in-source fragmentation. Efforts were made to control for this eventuality and limit its effect on our results by limiting the flow rate to 10 μL per minute (the lower limit of the UPLC pump being utilized and dilution of metabolite extract by 10-fold). Nonetheless, it should be noted that matrix effects cannot be controlled for perfectly and likely to occur in the present data set. Keeping this in mind, we sought to normalize our measured intensity to limit the effect of ionization suppression on the actual results of our analysis and the determination of key differentiating metabolites. When comparing the collective CRC samples to the collective NMT samples, we discovered that nucleotides,

Discussion The current study demonstrates the ability of TWIMMS analysis to differentiate between the metabolic profiles of NMT and CRC samples, as well as between the various stages of CRC progression and, importantly, between the profiles of primary and metastatic tumors from this small sample size.

M.D. Williams et al. Fig. 6 Negative mode PCA comparison of CRC stages. Separation of the various stages (T1–T4) using this unsupervised method supported the use of the current method for identification of possible biomarkers for CRC stage development

Fig. 7 Positive mode PLS-DA comparison of CRC stages. a All stages are included in this PLS-DA scores plot, demonstrating the ability to separate the different stages. b Stages T1 and T2 were compared directly;

c stages T2 and T3 were compared directly and separated; d stages T3 and T4 were also compared directly

Metabolic changes in human colorectal cancer Fig. 8 Heat map showing the relative average values for key ion features detected in NMT (N), CRC tumor (T), primary (P), and metastatic (M) sample types. Drift time (td) and m/z values are listed in the columns on the far left. For each feature, a box is color coded to represent the relative value of that feature in the N, T, P, and metastatic (M) sample types. The CRC tumor group of samples consisted of a combination of primary and metastatic tumors, resulting in differentiation between healthy tissue and tumor tissue. However, by separating the primary and metastatic tumors, we can discern the state most responsible for the overall change in CRC tumors. Values are in overall relative intensity units (i.u.)

nucleosides, bile acids, and oxidative metabolites accounted for the separation of the NMT and CRC metabolomes. For example, hypoxanthine, an important breakdown product of adenosine monophosphate, was elevated in CRC samples as compared to NMT. Hypoxanthine production is accelerated under conditions of hypoxia in normal, healthy cells [58, 59]. The elevation of hypoxanthine in CRC is consistent with increases in oxidative stress. It could also reflect enhanced purine metabolism or decreases in the enzyme xanthine oxidase reductase (XOR), which has been reported in CRC. Indeed, elevated hypoxanthine was identified as a urinary marker for non-Hodgkin’s lymphoma and may also represent a good marker for CRC as well [58]. Other metabolites in this same metabolic pathway, such as inosine, xanthine adenosine diphosphate (ADP), and adenosine monophosphate (AMP), were also elevated in CRC. AMP can be converted to adenosine, then inosine, to eventually produce hypoxanthine. Hypoxanthine can then be broken down into xanthine, subsequently, to uric acid. Fluctuations were seen in all of these metabolites in CRC vs. NMT specimens. We also found significant increases in FAs, particularly those associated with membrane fluidity and stability [33, 40, 59–64], between non-malignant tissues and CRC samples. These findings are consistent with our results in the Hmga1 transgenic model of polyposis [14, 43] and prior reports of metabolomes observed in CRCs and other cancers [33, 40, 59–64]. Many metabolites involved in FA biosynthesis were also increased, including carnitine, acyl-carnitines, and ascorbic acid. Additionally, an

increase in glutathione (GSH), an important biomolecule for cellular protection against oxidative stress and detoxification of xenobiotics by deactivation, was detected. Increased levels of GSH in tumors have been previously reported to protect tumor cells against various anticancer drugs by disrupting their cytotoxic action [66–70]. Interestingly, the metabolic differences observed between primary and metastatic tumors involved fluctuations in many of the same metabolic pathways that distinguish CRC from NMT tissues. Many FA compounds showed significant increases from primary to metastatic CRCs, including several phosphatidylcholines (PCs), Lyso-PCs, and TGs. Increases in carnitine, acyl-carnitines, bile acids, nucleotides, nucleosides, amino acids and oxidative compounds (i.e., glutathione), as well as in the polyamine, putrescine, were also observed. It is important to note that some of the metabolite identifications made in this work are preliminary assignments. While several of the identified metabolites could be readily assigned using exact mass data (these are marked with a ‡, in Figs. 3 and 5), there were others for which isomers exist. This is particularly true for some of the phospholipid identifications, as various FA tails may be present. Nevertheless, the isomeric species are typically of the same class of compound and usually have similar physiological consequences. Wherever possible, the biological relevance of the possible assignments was taken into account when selecting which isomeric species to display in the figures. Tables with the additional isomeric identities are given in the ESM (Tables S1–S4). One key advantage of using

M.D. Williams et al. Fig. 9 Heat map comparison of CRC stages. Drift time and m/z of significant differentiating ion features are accompanied by possible metabolite identifications. For each metabolite, a box is color coded to represent the relative value of that feature in each stage, T1–T4

ion mobility coupled with mass spectrometry is the ability to separate isomeric species. For example, in our analysis, we observed species of the same m/z but with different drift time measurements. Through further work with standards and collision cross section measurements, we expect to apply this method to determine which metabolite corresponds with each drift time. The use of drift time data for isomer identification will enable us to observe differences to which MS alone is blind. As shown in Fig. 3, significant up-regulation of AMP in CRC samples was observed in both positive and negative ionization mode analyses. Because AMP activates phophofructokinase-1 (PFK1) activity, this is consistent with enhanced aerobic glycolysis as a primary energy source for tumor cells. Additionally, bile acids were increased in both ionization modes, and bile acids are important in the transport of lipids in the body. One interesting, and somewhat unexpected, finding was the significant elevation of N5-formyltetrahydofolate (THF), which is the active form of folic acid, a B vitamin, in the body. While many early studies suggest that folate supplementation could reduce the risk of

developing CRC, more recent work suggests that excessive intake of folate also increases the risk of developing CRC, particularly once a premalignant lesion has formed [71]. Because folate is an important metabolite for maintaining DNA biosynthesis, repair, and methylation, and is also involved in the production of purine and pyrimidine nucleotide bases, high levels may contribute to sustaining proliferation in CRC once an abnormal lesion forms [72]. Our results, together with prior studies, suggest that tightly regulated folate is important in tissue homeostasis in the gut, and aberrations in folate metabolism can occur with enhanced proliferation and carcinogenesis. The heat maps shown in Figs. 8 and 9 are instructive in viewing the changes in possible biomarkers that distinguish primary and metastatic tumors as well as different stages of CRC. The heat map in Fig. 8 shows the ORI values of biomarkers that were found to be important in differentiating both CRC from NMT and primary from metastatic lesions. This contrasts from the initial comparison of CRC and NMT samples, which included both metastatic and primary lesions within the CRC group. This figure shows the differences observed

Metabolic changes in human colorectal cancer

in comparing tumor vs. NMT and also includes metabolites that distinguish primary or metastatic CRCs. For example, three bile acids (12-ketodeoxycholic acid, 7-Hydroxy-3oxocholanoic acid, 7-ketodeoxycholic acid) are up-regulated in CRC. Further, the heat map clearly shows that the overall increase observed in CRC tissues was primarily due to very significant increases in these bile acids in metastatic CRCs compared with little to no change in primary CRCs. These results suggest that these bile acids could be biomarkers for more advanced disease and metastatic progression, as opposed to early stage CRCs. Also note that two of these bile acids, identified as 12-ketodeoxycholic acid and 7-hydroxy-3oxocholic acid, are isomeric species, having the same m/z, but different drift time measurements. Further work with metabolite standards of these compounds may allow us to distinguish the isomer corresponding to each drift time. The heat map in Fig. 9 shows the ORI values of potential biomarkers that differentiate the four stages of CRC (T1 to T4). Importantly, this heat map demonstrates that while a given biomarker may differentiate stage T1 from T2 or T3, it does not necessarily differentiate any one of these stages from T4. Nevertheless, we found elevated levels of carnitine, certain FAs, and membrane components at all stages of the disease. Previously, we found elevations in the same FAs and membrane components in our Hmga1 transgenic mouse model of polyposis [43], which recapitulates some of the early changes observed in precancerous polyps (stage 0, T0) in humans. However, as shown in Fig. 9, several FAs, acyl-carnitines, and some membrane components were unexpectedly found to be decreased in the later stage CRCs as compared to the earlier stages. Interestingly, a prior study showed that proliferative rates may fall once tumors begin to metastasize, and these changes could reflect a decrease in cell proliferation as the cancer cells undergo changes needed to invade and spread to distant sites. FAs and other membrane components likely correlate with increased proliferative activity [72]. Indeed, in the current analysis of various CRC stages, significant fluctuations of FAs between stages was observed, with several FAs increasing in early and mid-stages (T2 and T3) while decreasing at a later stage (T4), consistent with this observation on proliferative changes with metastatic progression. Of note, the majority of the metastatic CRC samples analyzed in this study were at stage T3 (n=3) with only a single sample at stage T4. Because the number of samples for all tumors and stages was limited, further work with larger samples sizes are needed to validate our findings. It is possible that a subset of the metabolites could correspond to genetic variation in metabolism between individual patients as opposed to tumor progression. Nonetheless, these results demonstrate the power of the TWIMMS technology to identify metabolomics differences between non-malignant colonic epithelium and CRC and possibly between localized lesions and invasive or metastatic disease [73].

Conclusions Here, we demonstrate the use of TWIMMS technology as a fast and efficient method for metabolomic profiling of CRC. The unique direct infusion setup of the instrument allowed reduction of analysis time by more than fivefold from that of traditional techniques (UPLC-MS, GC-MS, and NMR). Moreover, this approach resulted in successful differentiation of CRC from NMT tissues as well as possible segregation of different stages of CRC. Further, we identified specific metabolites and metabolic pathways that distinguish CRC from NMT. Although our sample size was limited, our results also suggest that this approach could be used to distinguish different stages in tumor invasion. Further work is needed to establish this technique as a standard for metabolic profiling of CRC and other diseased states with larger numbers of samples and more validation of identifications using MS/MS and standard analysis.

Acknowledgments Support for this work came, in part, from the National Institutes of Health grant # 5R33RR020046 (H. Hill), #NIHR03 CA182679-01 and NIH R03 CA1646677 (L. Resar), the Maryland Stem Cell Research Fund (L. Resar), and from WSU Cancer Research Development Fund #17A-2412-0165 (R. Reeves). We would also like to sincerely thank Dr. Anders Merg, M.D., and Dr. Shane McNevin, M.D., both of Sacred Heart Medical Center, Providence Regional Cancer Center, Spokane, WA, for the surgical collection of tissue samples. Without their assistance, this work would not have been possible. Thanks also go to Ms. Joan Militon, RN, and Ms. Laura Nittolo, RN, also of Sacred Heart Medical Center, for coordinating the collection and documentation of tissue samples and obtaining appropriate pathology reports. Finally, we would like to thank Mr. Gary Johnson, of the WSU Spokane campus, for collecting and storing the tissue samples immediately after their collection.

References 1.

Siegel R, Ma J, Zou Z, Jemal A (2014) Cancer statistics, 2014. CA: A Cancer J Clin 64:9–29. doi:10.3322/caac.21208 2. American Cancer Society (2014) Colorectal cancer facts & figures 2014–2016. American Cancer Society, Atlanta 3. Hanahan D, Weinberg RA (2011) Hallmarks of cancer: the next generation. Cell 144:646–674. doi:10.1016/j.cell.2011.02.013 4. Warburg O, Wind F, Negelein E (1927) The metabolism of tumors in the body. J Gen Physiol 8:519–530 5. Vander Heiden MG, Cantley LC, Thompson CB (2009) Understanding the Warburg effect: the metabolic requirements of cell proliferation. Science 324:1029–1033. doi:10.1126/science. 1160809 6. Kroemer G, Pouyssegur J (2008) Tumor cell metabolism: cancer’s Achilles’ heel. Cancer Cell 13:472–482. doi:10.1016/j.ccr.2008.05. 005 7. Reeves R, Beckerbauer L (2001) HMGI/Y proteins: flexible regulators of transcription and chromatin structure. Biochim Biophys Acta 1519:13–29 8. Fusco A, Fedele M (2007) Roles of HMGA proteins in cancer. Nat Rev Cancer 7:899–910. doi:10.1038/nrc2271

M.D. Williams et al. 9.

10.

11.

12. 13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

Resar LMS (2010) The high mobility group A1 gene: transforming inflammatory signals into cancer? Cancer Res 70:436–439. doi:10. 1158/0008-5472.CAN-09-1212 Grade M, Hörmann P, Becker S, Hummon AB, Wangsa D, Varma S et al (2007) Gene expression profiling reveals a massive, aneuploidy-dependent transcriptional deregulation and distinct differences between lymph node-negative and lymph node-positive colon carcinomas. Cancer Res 67:41–56. doi:10.1158/0008-5472. CAN-06-1514 Johnson KR, Lehn DA, Reeves R (1989) Alternative processing of mRNAs encoding mammalian chromosomal High-Mobility-Group Proteins HMG-I and HMG-Y. Mol Cell Biol 9:2114–2123 Reeves R (2001) Molecular biology of HMGA proteins: hubs of nuclear function. Gene 277:63–81 Pedulla ML, Treff NR, Resar LM, Reeves R (2001) Sequence and analysis of the murine Hmgiy (Hmga1) gene locus. Gene 271:51– 58 Belton A, Gabrovsky A, Bae YK, Reeves R, Iacobuzio-Donahue C, Huso DL et al (2012) HMGA1 induces intestinal polyposis in transgenic mice and drives tumor progression and stem cell properties in colon cancer cells. PLoS One 7:e30034. doi:10.1371/journal.pone. 0030034 Chiappetta G, Avantaggiato V, Visconti R, Fedele M, Battista S, Trapasso F et al (1996) High level expression of the HMGI (Y) gene during embryonic development. Oncogene 13:2439–2446 Shah SN, Kerr C, Cope L, Zambidis E, Liu C, Hillion J et al (2012) HMGA1 reprograms somatic cells into pluripotent stem cells by inducing stem cell transcriptional networks. PLoS One 7:e48533. doi:10.1371/journal.pone.0048533 Ben-Porath I, Thomson MW, Carey VJ, Ge R, Bell GW, Regev A et al (2008) An embryonic stem cell-like gene expression signature in poorly differentiated aggressive human tumors. Nat Genet 40: 499–507. doi:10.1038/ng.127 Pomeroy SL, Tamayo P, Gaasenbeek M, Sturla LM, Angelo M, McLaughlin ME et al (2002) Prediction of central nervous system embryonal tumour outcome based on gene expression. Nature 415: 436–442. doi:10.1038/415436a Hristov AC, Cope L, Di Cello F, Reyes MD, Singh M, Hillion JA et al (2010) HMGA1 correlates with advanced tumor grade and decreased survival in pancreatic ductal adenocarcinoma. Modern Pathol 23:98–104. doi:10.1038/modpathol.2009.139 Shah SN, Cope L, Poh W, Belton A, Roy S, Talbot CC et al (2013) HMGA1: a master regulator of tumor progression in triple-negative breast cancer cells. PLoS One 8:e63419. doi:10.1371/journal.pone. 0063419 Roy S, Di Cello F, Kowalski J, Hristov AC, Tsai H-L, Bhojwani D et al (2013) HMGA1 overexpression correlates with relapse in childhood B-lineage acute lymphoblastic leukemia. Leuk Lymphoma 54:2565–2567. doi:10.3109/10428194.2013.782610 Yanagisawa BL, Resar LM (2014) Hitting the bull’s eye: targeting HMGA1 in cancer stem cells. Expert Rev Anticancer Ther 14:23– 30. doi:10.1586/14737140.2013.859988 Huso TH, Resar LM (2014) The high mobility group A1 molecular switch: turning on cancer—can we turn it off? Expert Opin Ther Targets 18:541–553. doi:10.1517/14728222.2014.900045 Hirayama A, Kami K, Sugimoto M, Sugawara M, Toki N, Onozuka H et al (2009) Quantitative metabolome profiling of colon and stomach cancer microenvironment by capillary electrophoresis time-of-flight mass spectrometry. Cancer Res 69:4918–4925. doi: 10.1158/0008-5472.CAN-08-4806 Chan ECY, Koh PK, Mal M, Cheah PY, Eu KW, Backshall A et al (2009) Metabolic profiling of human colorectal cancer using highresolution magic angle spinning nuclear magnetic resonance (HRMAS NMR) spectroscopy and gas chromatography mass spectrometry (GC/MS). J Proteome Res 8:352–361. doi:10.1021/pr8006232

26.

Tessem M-B, Selnaes KM, Sjursen W, Trano G, Giskeodegard GF, Bathem TF et al (2010) Discrimination of patients with microsatellite instability colon cancer using 1H HR MAS MR spectroscopy and chemometric analysis. J Proteome Res 9:3664–3670. doi:10. 1021/pr100176g 27. Chae Y-K, Kang W-Y, Kim S-H, Joo J-E, Han J-K, Hong B-W (2010) Combining information of common metabolites reveals global differences between colorectal cancerous and normal tissues. Bull Korean Chem Soc 31:379–383. doi:10.5012/bkcs.2010.31.02. 379 28. Piotto M, Moussallieh F-M, Dillmann B, Imperiale A, Neuville A, Brigand C et al (2008) Metabolic characterization of primary human colorectal cancers using high resolution magic angle spinning 1H magnetic resonance spectroscopy. Metabolomics 5:292–301. doi:10.1007/s11306-008-0151-1 29. Mal M, Koh PK, Cheah PY, Chan ECY (2009) Development and validation of a gas chromatography/mass spectrometry method for the metabolic profiling of human colon tissue. Rapid Commun Mass Spectrom 23:487–494. doi:10.1002/rcm.3898 30. Jordan KW, Nordenstam J, Lauwers GY, Rothenberger DA, Alavi K, Garwood M et al (2009) Metabolomic characterization of human rectal adenocarcinoma with intact tissue magnetic resonance spectroscopy. Dis Colon Rectum 52:520–525. doi:10.1007/DCR. 0b013e31819c9a2c 31. Moreno A, Rey M, Montane JM, Alonso J, Arús C (1993) 1H NMR spectroscopy of colon tumors and normal mucosal biopsies; elevated taurine levels and reduced polyethyleneglycol absorption in tumors may have diagnostic significance. NMR Biomed 6:111– 118 32. Moreno A, Arús C (1996) Quantitative and qualitative characterization of 1H NMR spectra of colon tumors, normal mucosa and their perchloric acid extracts: decreased levels of myo-inositol in tumours can be detected in intact biopsies. NMR Biomed 9:33–45. doi:10.1002/(SICI)1099-1492(199602)9:13. 0.CO;2-G 33. Nakagami K, Uchida T, Ohwada S, Koibuchi Y, Suda Y, Sekine T et al (1999) Increased choline kinase activity and elevated phosphocholine levels in human colon cancer. Jpn J Cancer Res 90:419–424 34. Bertini I, Cacciatore S, Jensen BV, Schou JV, Johansen JS, Kruhoffer M et al (2011) Metabolomic NMR fingerprinting to identify and predict survival of patients with metastatic colorectal cancer. Cancer Res 72:356–364. doi:10.1158/0008-5472.CAN-111543 35. Cheng Y, Xie G, Chen T, Qiu Y, Zou X, Zheng M et al (2012) Distinct urinary metabolic profile of human colorectal cancer. J Proteome Res 11:1354–1363. doi:10.1021/pr201001a 36. Ritchie SA, Ahiahonu PWK, Jayasinghe D, Heath D, Liu J, Lu Y et al (2010) Reduced levels of hydroxylated, polyunsaturated ultra long-chain fatty acids in the serum of colorectal cancer patients: implications for early screening and detection. BMC Med 8:13. doi:10.1186/1741-7015-8-13 37. Qiu Y, Cai G, Su M, Chen T, Liu Y, Xu Y et al (2010) Urinary metabonomic study on colorectal cancer. J Proteome Res 9:1627– 1634. doi:10.1021/pr901081y 38. Qiu Y, Cai G, Su M, Chen T, Zheng X, Xu Y et al (2009) Serum metabolite profiling of human colorectal cancer using GC−TOFMS and UPLC−QTOFMS. J Proteome Res 8:4844–4850. doi:10.1021/ pr9004162 39. Ludwig C, Ward DG, Martin A, Viant MR, Ismail T, Johnson PJ et al (2009) Fast targeted multidimensional NMR metabolomics of colorectal cancer. Magn Reson Chem 47:S68–S73. doi:10.1002/ mrc.2519 40. Li S, Zhao X, Wu Z, Li Y, Zhu L, Cui B et al (2013) Polymorphisms in arachidonic acid metabolism-related genes and the risk and

Metabolic changes in human colorectal cancer prognosis of colorectal cancer. Fam Cancer 12:755–765. doi:10. 1007/s10689-013-9659-2 41. Nishiumi S, Kobayashi T, Ikeda A, Yoshie T, Kibi M, Izumi Y et al (2012) A novel serum metabolomics-based diagnostic approach for colorectal cancer. PLoS One 7:e40459. doi:10.1371/journal.pone. 0040459 42. Williams MD, Reeves R, Resar LS, Hill HH (2013) Metabolomics of colorectal cancer: past and current analytical platforms. Anal Bioanal Chem 405:5013–5030. doi:10.1007/s00216-013-6777-5 43. Williams MD, Zhang X, Belton AS, Xian L, Huso T, Park J-J et al (2015) HMGA1 drives metabolic reprogramming of intestinal epithelium during hyperproliferation, polyposis, and colorectal carcinogenesis. J Proteome Res 14:1420–1431. doi:10.1021/pr501084s 44. Dwivedi P, Schultz AJ Jr, HHH (2010) Metabolic profiling of human blood by high-resolution ion mobility mass spectrometry (IMMS). Int J Mass Spectrom 298:78–90. doi:10.1016/j.ijms.2010.02. 007 45. Dwivedi P, Hill HH (2008) A rapid analytical method for hair analysis using ambient pressure ion mobility mass spectrometry with electrospray ionization (ESI-IMMS). Int J Ion Mobil Spectrom 11: 61–69. doi:10.1007/s12127-008-0006-5 46. Kaplan K, Dwivedi P, Davidson S, Yang Q, Tso P, Siems W et al (2009) Monitoring dynamic changes in lymph metabolome of fasting and fed rats by electrospray ionization-ion mobility mass spectrometry (ESI-IMMS). Anal Chem 81:7944–7953. doi:10. 1021/ac901030k 47. Dwivedi P, Puzon G, Tam M, Langlais D, Jackson S, Kaplan K et al (2010) Metabolic profiling of Escherichia coli by ion mobilitymass spectrometry with MALDI ion source. J Mass Spectrom 45: 1383–1393. doi:10.1002/jms.1850 48. May JC (2012) Encyclopedia of drug metabolism and interactions, Chapter 9, Gas-phase ion mobility-mass spectrometry (IM-MS) and tandem IM-MS/MS strategies for metabolism studies and metabolomics. vol. 5. 1st ed. Wiley, Hoboken 49. Giles K, Pringle SD, Worthington KR, Little D, Wildgoose JL, Bateman RH (2004) Applications of a travelling wave-based radio-frequency-only stacked ring ion guide. Rapid Commun Mass Sperom 18:2401–2414. doi:10.1002/rcm.1641 50. Malkar A, Devenport NA, Martin HJ, Patel P, Turner MA, Watson P et al (2013) Metabolic profiling of human saliva before and after induced physiological stress by ultra-high performance liquid chromatography–ion mobility–mass spectrometry. Metabolomics 9: 1192–1201. doi:10.1007/s11306-013-0541-x 51. Shah V, Castro-Perez JM, McLaren DG, Herath KB, Previs SF, Roddy TP (2013) Enhanced data-independent analysis of lipids using ion mobility-TOFMS E to unravel quantitative and qualitative information in human plasma: enhanced data-independent analysis of lipids. Rapid Commun Mass Spectrom 27:2195–2200. doi:10. 1002/rcm.6675 52. Ignatenko N, Gerner E, Besselsen D (2011) Defining the role of polyamines in colon carcinogenesis using mouse models. J Carcinog 10:10. doi:10.4103/1477-3163.79673 53. Berdinskikh NK, Ignatenko NA, Zaletok SP, Ganina KP, Chorniy VA (1991) Ornithine decarboxylase activity and polyamine content in adenocarcinomas of human stomach and large intestine. Int J Cancer 47:496–498 54. Elitsur Y, Moshier JA, Murthy R, Barbish A, Luk GD (1992) Polyamine levels, ornithine decarboxylase (ODC) activity, and ODC-mRNA expression in normal and cancerous human colonocytes. Life Sci 50:1417–1424 55. Yerushalmi HF, Besselsen DG, Ignatenko NA, Blohm-Mangone KA, Padilla-Torres JL, Stringer DE et al (2006) Role of polyamines in arginine-dependent colon carcinogenesis inApcMin/+ mice. Mol Carcinog 45:764–773. doi:10.1002/mc.20246

56.

Bridoux MC, Ingalls AE (2010) Structural identification of longchain polyamines associated with diatom biosilica in a Southern Ocean sediment core. Geochim Cosmochim Acta 74:4044–4057. doi:10.1016/j.gca.2010.04.010 57. Linsalata M, Orlando A, Russo F (2014) Pharmacological and dietary agents for colorectal cancer chemoprevention: effects on polyamine metabolism (review). Int J Oncol. doi:10.3892/ijo.2014.2597 58. Yoo B, Kong S-Y, Jang S-G, Kim K-H, Ahn S-A, Park W-S et al (2010) Identification of hypoxanthine as a urine marker for nonHodgkin lymphoma by low-mass-ion profiling. BMC Cancer 10: 55. doi:10.1186/1471-2407-10-55 59. Baenke F, Peck B, Miess H, Schulze A (2013) Hooked on fat: the role of lipid synthesis in cancer metabolism and tumour development. Dis Model Mech 6:1353–1363. doi:10.1242/dmm.011338 60. Hopperton KE, Duncan RE, Bazinet RP, Archer MC (2014) Fatty acid synthase plays a role in cancer metabolism beyond providing fatty acids for phospholipid synthesis or sustaining elevations in glycolytic activity. Exp Cell Res 320:302–310. doi:10.1016/j. yexcr.2013.10.016 61. Zaidi N, Swinnen JV, Smans K (2012) ATP-citrate lyase: a key player in cancer metabolism. Cancer Res 72:3709–3714. doi:10. 1158/0008-5472.CAN-11-4112 62. Ide Y, Waki M, Hayasaka T, Nishio T, Morita Y, Tanaka H et al (2013) Human breast cancer tissues contain abundant phosphatidylcholine(36∶1) with high stearoyl-CoA desaturase-1 expression. PLoS One 8:e61204. doi:10.1371/journal.pone.0061204 63. Kurabe N, Hayasaka T, Ogawa M, Masaki N, Ide Y, Waki M et al (2013) Accumulated phosphatidylcholine (16:0/16:1) in human colorectal cancer; possible involvement of LPCAT4. Cancer Sci 104: 1295–1302. doi:10.1111/cas.12221 64. Ishikawa S, Tateya I, Hayasaka T, Masaki N, Takizawa Y, Ohno S et al (2012) Increased expression of phosphatidylcholine (16:0/18: 1) and (16:0/18:2) in thyroid papillary cancer. PLoS One 7:e48873. doi:10.1371/journal.pone.0048873 65. Kingsnorth AN, Lumsden AB, Wallace HM (1984) Polyamines in colorectal cancer. Br J Surg 71:791–794 66. Balendiran GK, Dabur R, Fraser D (2004) The role of glutathione in cancer. Cell Biochem Funct 22:343–352. doi:10.1002/cbf.1149 67. Traverso N, Ricciarelli R, Nitti M, Marengo B, Furfaro AL, Pronzato MA et al (2013) Role of glutathione in cancer progression and chemoresistance. Oxidative Med Cell Longev 2013:1–10. doi: 10.1155/2013/972913 68. Hussain SP, Hofseth LJ, Harris CC (2003) Radical causes of cancer. Nat Rev Cancer 3:276–285. doi:10.1038/nrc1046 69. Kondo T, Iida T (1997) gamma-GCS and glutathione—new molecular targets in cancer treatment. Gan To Kagaku Ryoho 24:2219– 2225 70. Carretero J, Obrador E, Anasagasti MJ, Martin JJ, VidalVanaclocha F, Estrela JM (1999) Growth-associated changes in glutathione content correlate with liver metastatic activity of B16 melanoma cells. Clin Exp Metastasis 17:567–574 71. Duthie SJ (2011) Folate and cancer: how DNA damage, repair and methylation impact on colon carcinogenesis. J Inherit Metab Dis 34:101–109. doi:10.1007/s10545-010-9128-0 72. Jerby L, Wolf L, Denkert C, Stein GY, Hilvo M, Oresic M et al (2012) Metabolic associations of reduced proliferation and oxidative stress in advanced breast cancer. Cancer Res 72:5712–5720. doi:10.1158/0008-5472.CAN-12-2215 73. Williams MD (2014) Discovering biochemical signatures of disease: metabolomics of colorectal cancer via ion mobility mass spectrometry. Doctoral Dissertation, Washington State University, Department of Chemistry