Skip to main content
NCBI home page
Molecular Oncology logoLink to Molecular Oncology
. 2013 Sep 14;8(1):39–49. doi: 10.1016/j.molonc.2013.08.010

Chemical mapping of the colorectal cancer microenvironment via MALDI imaging mass spectrometry (MALDI‐MSI) reveals novel cancer‐associated field effects

R Mirnezami 1,†, K Spagou 2,†, PA Vorkas 2, MR Lewis 2, J Kinross 1, E Want 2, H Shion 3, RD Goldin 4, A Darzi 1, Z Takats 2, E Holmes 2, O Cloarec 2,5,1,✉, JK Nicholson 2
PMCID: PMC5528498  PMID: 24112879

Abstract

Matrix‐assisted laser desorption ionisation imaging mass spectrometry (MALDI‐MSI) is a rapidly advancing technique for intact tissue analysis that allows simultaneous localisation and quantification of biomolecules in different histological regions of interest. This approach can potentially offer novel insights into tumour microenvironmental (TME) biochemistry. In this study we employed MALDI‐MSI to evaluate fresh frozen sections of colorectal cancer (CRC) tissue and adjacent healthy mucosa obtained from 12 consenting patients undergoing surgery for confirmed CRC. Specifically, we sought to address three objectives: (1) To identify biochemical differences between different morphological regions within the CRC TME; (2) To characterise the biochemical differences between cancerous and healthy colorectal tissue using MALDI‐MSI; (3) To determine whether MALDI‐MSI profiling of tumour‐adjacent tissue can identify novel metabolic ‘field effects’ associated with cancer. Our results demonstrate that CRC tissue harbours characteristic phospholipid signatures compared with healthy tissue and additionally, different tissue regions within the CRC TME reveal distinct biochemical profiles. Furthermore we observed biochemical differences between tumour‐adjacent and tumour‐remote healthy mucosa. We have referred to this ‘field effect’, exhibited by the tumour locale, as cancer‐adjacent metaboplasia (CAM) and this finding builds on the established concept of field cancerisation.

Keywords: MALDI-MSI, Molecular imaging, Colorectal cancer, Tumour microenvironment, Lipid profiling

Highlights

  • We investigate colorectal cancer (CRC) lipid biochemistry using MALDI imaging mass spectrometry (MALDI‐MSI).

  • We examine changes in lipid composition across different bio‐compartments within the CRC tumour microenvironment.

  • We report tumour‐adjacent lipidomic ‘field effects’ in morphologically healthy colorectal mucosa.

  • We report a novel and efficient method for sample‐to‐sample comparative analysis in MSI based profiling.

1. Introduction

1.1. Imaging mass spectrometry

The tumour microenvironment (TME) is thought to play a critical role in solid organ cancer development and progression (Hogan et al., 2012 Jul 1) and TME profiling approaches are gaining increasing attention in cancer research (Straussman et al., 2012 Jul 26; Engelhardt et al., 2012 Mar 20; Tarin and 2012 Feb 3). Imaging mass spectrometry (MSI) utilising techniques such as matrix‐assisted laser‐desorption ionisation (MALDI), (Caprioli et al., 1997 Dec 1; Schwamborn et al., 2010 Sep; Gruner et al., 2012) desorption electrospray ionization (DESI) (Gerbig et al., 2012 Jun; Eberlin et al., 2013 Jan 29; Takats et al., 2004 Oct 15) and secondary ion mass spectrometry (SIMS) (Passarelli et al., 2011 Nov; Cillero‐Pastor et al., 2012 Nov 6) have demonstrated early promise in this area, offering a means of chemically mapping tissue sections, without the requirement for specific staining/labelling agents. The major advantage of MSI as an analytical tool lies in the ability to measure biochemical activity with direct correlation to morphological features of interest, bypassing the need for time‐consuming steps such as laser‐capture micro‐dissection. The latter by its very nature results in significant disruption to the tumour niche, preventing visualisation of biomolecules within their native backdrop.

To date a number of studies have successfully used MSI approaches to characterize region‐specific protein/peptide distribution within the cancer TME (Gruner et al., 2012; Aichler et al., 2013 Apr 16; Morgan et al., 2013 Mar; Meding et al., 2012 Mar 2). More recently, the ability of MSI to assess the localisation and concentration distribution of chemotherapeutic compounds in cancer tissue has also been demonstrated, offering the opportunity to investigate anti‐cancer drug mechanisms and efficacy from an entirely novel perspective (Bouslimani et al., 2010 Feb; Rompp et al., 2011 Jul).

Of the available MSI techniques MALDI has been the most commonly employed in the recent literature (Caprioli et al., 1997 Dec 1; Schwamborn et al., 2010 Sep; Gruner et al., 2012). This technique is a ‘matrix‐assisted’ approach that involves the application of a chemical solution (referred to as a ‘matrix solution’ in this context) to the tissue surface to crystallise biomolecules prior to ionisation (Caprioli et al., 1997 Dec 1). Commonly employed solutions for this purpose include α‐cyano‐4‐hydroxycinnamic acid (CHCA) and 2,5‐dihydroxybenzoic acid (DHB). (Cohen et al., 1996 Jan 1) Where employed in this study the term ‘matrix’, refers to this experimental solution and should not be confused with the conventional biological definition of ‘matrix’ that refers to the inter‐cellular material found within connective tissue.

The MALDI technique is very well suited to the localisation of proteins, which occupy the upper end of the mass‐to‐charge (m/z) spectrum. The successful application of MALDI‐based methods to profiling biomolecules at the lower end of the m/z range (<1000 m/z = metabolites/lipids) is more challenging as matrix and/or matrix‐analyte cluster peaks (generated as a consequence of matrix deposition) can interfere with the detection of target low molecular weight compounds (Goto‐Inoue et al., 2011 Nov). A novel method for matrix‐associated peak removal has recently been proposed by our group to address this issue and facilitate effective MALDI‐based profiling of lower m/z value molecular species (Fonville et al., 2012 Feb 7).

1.2. Exploring the cancer lipidome via imaging mass spectrometry

Lipid profiling in particular represents an attractive avenue for novel cancer biomarker discovery as there is growing evidence to suggest that membrane lipids play a vital role in carcinogenesis (Sparvero et al., 2012 Jul; Meriaux et al., 2010 Apr 18). Lipids are key cell membrane constituents (Quehenberger et al., 2011 Nov 10) and serve several critical physiological roles including regulation of energy metabolism, (Quehenberger et al., 2011 Nov 10) cellular signalling (Fernandis et al., 2007 Apr) and trafficking of immune cells. (Dykstra et al., 2003) Disordered lipid metabolism at the cellular level is now recognised as a hallmark feature across a variety of cancer subtypes. (Smith et al., 2008 Jan) Imaging mass spectrometry‐based techniques offer a means of mapping lipid biochemistry in unparalleled detail within different tissue subtypes comprising the TME. Current methods for lipid detection and localisation such as staining with Nile Red, Oil Red O or osmium tetroxide can be applied to frozen sections (van Goor et al., 1986), and while these can reveal the distribution of a handful of specific lipid fractions, they lack the ability to provide a more holistic lipid profile, as can be achieved with MSI.

To date a limited number of studies have applied MSI‐based approaches to identify specific lipid signatures with respect to CRC; Shimma et al. identified region‐specific lipid profiles in CRC liver metastasis samples obtained from a single patient (Shimma et al., 2007 Aug), and more recently Thomas et al. found a panel of lipid‐based biomarkers to be up‐ and down‐regulated in CRC liver metastases (Thomas et al., 2013 Feb 8). These preliminary studies have highlighted the potential of lipid profiling in identification of disease‐relevant lipid signatures for next‐generation biomarker discovery.

1.3. Study objectives

In the current study we have applied MALDI‐MSI analysis to determine region‐specific lipid signatures in colorectal tissue samples (CRC, n = 12; healthy ‘tumour‐remote’ colorectal mucosa, n = 12; healthy ‘tumour‐adjacent’ colorectal mucosa, n = 12) obtained from patients with confirmed CRC. This study methodology is novel as until recently most studies that have employed MALDI‐MSI to evaluate primary CRC tissue have focused principally on protein expression, (Djidja et al., 2010 May; Casadonte et al., 2011 Nov) with little or no consideration to downstream lipid biochemistry.

Specifically we have sought to address three primary objectives in this study:

  • (1)

    To determine whether MALDI‐MSI can reveal regional differences in lipid biochemistry in different tissue types within the CRC microenvironment.

  • (2)

    To determine basic lipidomic differences between cancerous and healthy colorectal tissue via MALDI‐MSI analysis.

  • (3)

    To perform MALDI‐MSI – based comparison of lipid biochemistry in ‘tumour‐adjacent’ and ‘tumour‐remote’ healthy colorectal mucosa. The aim here was to identify lipid‐based ‘field effects’ in cancer‐adjacent tissue to supplement established field cancerisation principles.

2. Theory/calculation

2.1. Super‐spectral data treatment for efficient sample‐to‐sample comparison

At present the complexity of data analysis represents a significant challenge to realising the translational potential of MSI. (Alexandrov, 2012) In MSI experiments ionisation is applied directly to the tissue surface across a pre‐defined array of co‐ordinates, each of which generates its own unique, spatially localised mass spectrum. This produces vast, multidimensional data that are challenging to interpret with conventional multivariate methods. To simplify data analysis we have devised a method where all spectra from a given tissue section are combined to create a single ‘super’ mass spectrum, which can then be used to perform efficient sample‐to‐sample comparison. This method can be used to highlight molecular differences between cancerous and healthy tissue sections; the distribution of biomolecules found to be up‐ or down‐regulated in cancer can then be visualised on individual MALDI‐MSI images.

3. Materials and methods

3.1. 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 informed and written consent from 12 patients undergoing elective surgical resection for confirmed CRC at St Mary's Hospital (London, UK). The clinico‐pathological characteristics of these patients are summarized in the Supporting Information (Table S1). Resected cancer specimens were taken on ice to the pathology department immediately after surgical extraction and were evaluated by a single pathologist (RDG) prior to sampling. Fresh tissue was retrieved from tumour centre (n = 12) and macroscopically cancer‐free ‘tumour‐adjacent’ (directly adjacent to tumour; n = 12) and ‘tumour‐remote’ (10 cm from cancer margin; n = 12) colorectal mucosa. These samples were immediately transferred to a freezer at −80 °C prior to processing.

3.2. Sample preparation and MALDI‐MSI analysis

Tissue samples were cryo‐sectioned using an SME Cryostat (ThermoShandon, USA) with consecutive sections generated for MALDI‐MSI (12 μm) and histopathological (6 μm) analysis to allow co‐registration of morphology and biochemistry. Sections were thaw mounted onto plain glass microscopy slides before those obtained for histopathology were stained with haematoxylin and eosin (H&E). For MALDI‐MSI analysis, a uniform coating of α‐cyano‐4‐hydroxycinnamic acid (CHCA) matrix solution (Huwiler et al., 2003 Dec) was applied to designated tissue sections using an automated spray system (TM Sprayer, HTX Technologies, NC, USA). MALDI‐MSI analysis was carried out using an SYNAPT G2 high definition mass spectrometer (Waters Corporation, Manchester, UK) equipped with an Nd:YAG (neodymium‐doped yttrium aluminium garnet) laser for sample ionisation. (Hart et al., 2011 Jul) Spectra were collected in the mass range m/z 50 to 1000. Further description of the analytical parameters and data pre‐processing steps employed here is provided in the Supplementary Information.

3.3. Pattern recognition

Pattern recognition of spectral or imaging data is used to reduce the complexity and density of analytical data (dimensionality reduction) and to find patterns between clinical samples based on biochemical similarities. (Trygg et al., 2007 Feb) The two most commonly employed pattern recognition methods in metabolic profiling are principal component analysis (PCA) and partial least squares (PLS) regression analysis. (Trygg et al., 2007 Feb) PCA is an unsupervised method that illustrates the overall spread within a dataset by summarising sample variation as a series of ‘latent variables’ (LV), which can be used to model data. (Trygg et al., 2007 Feb) PLS methods are an extension of PCA and are linear regression techniques that, in our application, use class information (for example whether a sample has been obtained from cancerous or healthy tissue) to maximise the discrimination between groups of observations, hence the term PLS‐DA is used (DA standing for discriminant analysis).

In this study we have combined MALDI‐MSI data with a PLS‐based pattern recognition approach to determine biochemical differences between different tissue types. The PLS method takes original data points (mass spectra generated for a particular tissue location) and projects them onto a new set of axes. This creates a more intuitive overview of the data, according to weighted sums/combinations of chemical features (referred to as latent variables, LV). The predictive capacity of a PLS model is encompassed by the Q 2 statistic which indicates whether the classification (or biomarker combination) is robust. (Trygg et al., 2007 Feb) The value for Q 2 ranges from 0 to 1; a Q 2 of 0 implies that the model has no predictive capacity above random chance/no‐model estimates, while a Q 2 of 1 implies that the generated model is ‘perfect’. In the case of very poor model predictive capacity the Q 2 can even assume a negative value. Thus the higher the Q 2 value the greater the predictive capacity of a given model to correctly discriminate between classes.

4. Results

4.1. Identification of topographically discrete lipid signatures in different biological compartments within the CRC TME

In order to identify differences in lipid biochemistry according to tissue type, we first selected regions of interest (cancer, smooth muscle, stroma and fibrous tissue) from H&E stained centre of tumour tissue sections. In addition we included in this model tumour‐adjacent mucosa and tumour‐remote mucosa (sampled from adjacent to tumour margin and 10 cm from tumour margin respectively), to identify any metabolic ‘field effects’ in the immediate cancer vicinity. Next, mass spectra were collected from pixels corresponding to these tissue regions on MALDI‐MSI images. A PLS‐DA model was then fitted to differentiate MALDI‐MSI profiles according to tissue type, as ‘non‐cancerous’ or ‘cancerous’ (Figure 1). A Q 2 value of 0.79 was obtained for this model indicating robust separation of biochemical profiles according to tissue morphology. A p‐value for this model was also calculated as an additional means of assessing model predictive capacity (cancerous versus all non‐cancerous tissue types, p < 0.001). The PLS‐DA scores plot for this model is shown in Figure 1 and reveals visible differences in MALDI‐MSI derived chemical signatures according to tissue type. Different non‐cancerous tissue types have been colour coded according to morphology and these appear to cluster closely together, suggesting the presence of specific lipid signatures in each tissue region.

Figure 1.

Figure 1

Open in a new tab

PLS‐DA scores plot revealing distribution of MALDI‐MSI profiles obtained from cancerous and non‐cancerous tissue regions colour‐coded according to morphology. The x‐axis denotes grouping of samples according to tissue class and the y‐axis indicates a PLS‐derived ‘score’ based on a weighted sum of molecular features for each data point; the closer the score is to ‘1’ the more likely it is to represent a cancerous region and likewise the closer the score is to ‘−1’ the more likely it is to be healthy based on identified chemical signatures. The plot reveals clear differences between cancerous and non‐cancerous tissue, and also reveals clustering of healthy tissue profiles according to tissue morphology. Data points from tumour‐adjacent healthy tissue are seen to lie in close proximity to cancer compared to tumour‐remote.

To determine the topographical specificity of m/z values found to be more abundant in cancer we selectively projected these values back onto ‘blank canvas’ MALDI‐MSI images. The accuracy of localisation of these biomolecules in cancer regions was evaluated by co‐registering the reconstructed MALDI‐MSI images with corresponding H&E stained sections. Examples of reconstructed images are shown in Figure 2 (A–F). Selective projection of m/z 478.3 (found to be elevated in cancer relative to healthy) onto cancer‐bearing MALDI‐MSI images, revealed close correlation between the localisation of this molecule and areas of cancer confirmed on corresponding H&E stains (Figure 2A–C). By contrast projection of m/z 478.3 onto non‐tumour‐bearing tissue sections showed no classification and a marked reduction in signal intensity, as anticipated (Figure 2D–F).

Figure 2.

Figure 2

Open in a new tab

MALDI‐MSI images from sections of tumour‐bearing (A–C) and non‐tumour‐bearing tissue (D–F). Selective projection of m/z 478.3 onto MALDI‐MSI images reveals localisation of these molecular signatures within cancer‐bearing areas when comparison is made with corresponding H&E images. By comparison, projection of m/z 478.3 onto non‐tumour bearing tissue sections (D–F) reveals poor signal intensity as expected

4.2. Identifying cancer‐associated ‘field effects’ via comparison of tumour‐adjacent and tumour‐remote tissue

The differences we observed between tumour‐adjacent and tumour‐remote tissue may be of prognostic significance and we sought to further investigate these using PLS‐DA. The aim here was to determine the lipid fingerprint of tumour‐adjacent healthy mucosa in comparison to tumour‐remote tissue, and tumour itself.

Figure 3 presents the PLS‐DA scores plot for differentiating between tumour‐adjacent and tumour‐remote tissue (Q 2 = 0.76). The visible clustering on the scores plot implies the presence of chemical differences in morphologically bland colorectal tissue according to proximity to tumour. Figure 4A shows a three‐class PLS‐DA scores plot for tumour, tumour‐remote and tumour‐adjacent colorectal tissue plotted according to biochemical similarities/differences. This plot reveals visible overlap between data points acquired from tumour (blue) and tumour‐adjacent (pink) sites – suggesting that the tumour field, even when morphologically free of cancer, exhibits metabolic properties, which closely mimic those of neighbouring cancerous zones. This stepwise molecular shift is illustrated more intuitively in the vertical scatter plot (Figure 4B). The statistical capacity of this model to correctly distinguish samples as belonging to tumour, tumour‐adjacent or tumour‐remote classes is summarised in the form of Q 2 statistics as Q 2 = 0.64, Q 2 = 0.67 and Q 2 = 0.69, respectively.

Figure 3.

Figure 3

Open in a new tab

A PLS‐DA scores plot for MALDI‐MSI profiles obtained from tumour‐adjacent (pink) and tumour‐remote (blue) tissue sections. The x‐ and y‐axes here refer to the ‘latent variables’ (LV) responsible for the greatest between‐class variation that were used for model construction. B Alternative representation of PLS scores for improved visual interpretation of data spread. In this plot the x‐axis denotes grouping of samples according to tissue class (tumour‐adjacent and tumour‐remote) and the y‐axis indicates a PLS‐derived ‘score’ based on a weighted sum of molecular features for each data point; the closer the score is to ‘1’ the more closely it resembles the cancer‐associated MALDI‐MSI profile.

Figure 4.

Figure 4

Open in a new tab

A PLS‐DA scores plot summarising the molecular relationships between MALDI‐MSI profiles acquired from tumour (blue), tumour‐adjacent (pink) and tumour‐remote (green) tissue sections. The x‐ and y‐axes refer to the ‘latent variables’ (LV) responsible for greatest between‐class variation that were used for model construction. B Alternative representation of PLS scores for improved visual interpretation of data spread. The x‐axis denotes grouping of samples according to tissue class (tumour‐adjacent, tumour‐remote, tumour) and the y‐axis indicates a PLS‐derived ‘score’ based on a weighted sum of molecular features for each data point. Both plots reveal significant overlap of data points obtained from tumour‐adjacent tissue and tumour itself.

4.3. Defining chemical differences between cancerous and non‐cancerous colorectal tissue according to regional lipid distribution

For efficient group‐wise comparison (cancer samples versus healthy samples) a ‘super’ mass spectrum was generated for every tissue section from centre of tumour (n = 12) and healthy tissue 10 cm from tumour margin (n = 12) by aggregating spectral features form each image into one overall (‘super’) spectrum. A PLS‐DA model was then generated to reveal differences between cancerous and healthy tissue based on these ‘super’ spectral characteristics. A ‘leave‐one‐patient‐out’ approach was used for cross‐validation which means that each pair of slides corresponding to the same patient was removed before fitting the model.

Figure 5A shows the PLS‐DA scores plot generated for this model, as well as a selection of MALDI‐MSI images for cancerous and healthy tissue sections. These images demonstrate a visible trend towards more abundant molecular signal intensity in cancer‐bearing relative to healthy tissue sections. This information was then used to construct a ‘loadings plot’ (Figure 5B) where the specific molecular species responsible for discrimination can be determined (Cloarec et al., 2005 Jan 15). With respect to chemical assignment, the loadings plot (Figure 5B) reveals that molecular species m/z 478.3 (M1), m/z 504.3 (M2) and m/z 760.6 (M3) are found in significantly higher concentrations in cancer‐bearing samples relative to healthy tissue. These three m/z ions exhibited a similar regional distribution within tissue sections as shown in Figure 6C, appearing to be highly abundant in cancer‐cell bearing areas (Figure 6B and C).

Figure 5.

Figure 5

Open in a new tab

A PLS‐DA scores plot revealing discrimination between tissue sections obtained from cancerous (blue) and healthy tissue (10 cm from the tumour margin, red). Corresponding MALDI‐MSI images around this plot show visibly greater intensity of molecular ion signatures (indicated by increased yellow/orange colouration relative to blue) from samples obtained from centre of tumour (images have been numbered for ease of cross‐referencing with plot data points). B PLS loadings plot based on the model shown in A; this figure allows specific m/z values responsible for discrimination between groups to be identified and their relative contributions to this difference can be evaluated. The x‐axis on this plot corresponds to different m/z species, and the y‐axis indicates magnitude of difference and direction of difference in terms of chemical expression patterns. Positive coefficients (m/z species 478.3 (M1), 504.3 (M2) and 760.6 (M3)) signify increased abundance of the corresponding ion in tumour tissue relative to healthy.

Figure 6.

Figure 6

Open in a new tab

A Overlapped MALDI spectral profiles for cancerous tissue (red), non‐cancerous tissue (blue) and from regions outside the margin of the tissue section (i.e. matrix solution only; black); B Digitised image of the corresponding H&E stained section for this sample with morphological regions of interest defined by a solid line (Mc: Mucosa, T: tumour tissue, M: Muscle); C Distribution of m/z ions that were expressed in higher levels in tumour tissue is explored using targeted MALDI‐MSI image reconstruction with selective localisation of m/z values of interest (m/z 478.3, 504.3 and 760.6). This shows clear localisation of these chemical compounds in tumour‐bearing areas and relative absence of associated signal in muscle and mucosa.

Figure 6A illustrates spectra obtained from two morphologically distinct tissue section regions (tumour bearing tissue (red), smooth muscle (blue)). A black spectrum depicting chemical signatures from outside of the tissue margin is also provided. Overlapping these spectra shows that m/z ions 478.3, 504.3 and 760.6 are present in significantly higher intensity in cancer‐bearing tissue relative to muscle. In addition ion images of these m/z values were generated in order to compare the localisation of m/z 478.3 and 504.3 in cancerous and non‐cancerous tissue samples (Figure 7). Tandem mass spectrometry (MS/MS) experiments confirmed the assignment of m/z 760.6 as phosphatidyl‐choline (PC) species 16:0/18:1, m/z 478.3 as LysoPC(16:0) and m/z 504.3 as LysoPC(18:1).

Figure 7.

Figure 7

Open in a new tab

MALDI‐MSI ion images revealing the distribution of m/z 478.3 (A and B) and m/z 504.3 (C and D) in cancer‐bearing (centre of tumour; A and C) and non‐cancer‐bearing (healthy mucosa 10 cm from the tumour margin; B and D) tissue sections. These ionic species are seen to be specifically over‐expressed in cancerous regions with little expression evident in healthy tissue.

5. Discussion

In this small feasibility study we have developed and applied a novel processing strategy for MALDI‐MSI data analysis that confirms the presence of lipidomic differences between cancerous and healthy colorectal tissue and demonstrates tumour‐adjacent field effects with respect to lipid metabolism that may have translational significance. In addition this approach can provide chemical data to supplement current histological approaches for profiling the tumour microenvironment (TME).

Accurate molecular characterisation of the TME represents a major research priority in systems oncology (Swartz et al., 2012 May 15; Hanahan et al., 2011 Mar 4). The TME is comprised of a variety of cell types as well as pro‐tumorigenic factors and a myriad of signalling molecules, all of which are thought to be involved in tumour development, survival and progression. (Hanahan et al., 2011 Mar 4) Phenotyping the TME from a lipidomic/metabolic perspective has the potential to provide important biological information downstream of the cancer genome and proteome, and recent in vitro work has confirmed that the tumour niche exhibits compartmentalised metabolic activity. (Martinez‐Outschoorn et al., 2011 Aug 1) For example, tumour cells are thought to impose conditions of oxidative stress on neighbouring cells creating adjacent zones of cellular catabolism, mitochondrial dysfunction and aerobic glycolysis. (Martinez‐Outschoorn et al., 2011 Aug 1) MALDI‐MSI profiling approaches offer the opportunity to non‐destructively characterise this activity, particularly with respect to the cancer lipidome. However, several refinements are necessary in order to facilitate the widespread application of MALDI‐MSI approaches in the clinical setting. Firstly, data treatment methodologies must be developed that can reduce the ‘high‐dimensionality’ of MALDI‐MSI data and display results in a streamlined and readily interpretable way. In addition MALDI‐MSI datasets present unique challenges where the aim is to build a model for comparison across multiple tissue samples. In the present study we have developed a novel technique for integration of data from a given MALDI‐MSI image to create a ‘super’ mass spectrum summarizing all captured information efficiently. We have demonstrated that using this approach, super‐spectra generated from different tissue sections can be subjected to multivariate analysis to identify discriminant molecular features between groups (in this case cancerous versus healthy tissue).

The primary aim of this study was to develop lipid‐based profiles according to tissue type (Figure 1). We were able to generate distinct profiles for different tissue regions (tumour, stroma, fibrous tissue, smooth muscle, submucosa) after identification of features of interest on H&E sections and co‐registration of these with MALDI‐MSI images. In addition the PLS‐DA scores plots for this model highlighted some interesting differences between tumour‐remote and tumour‐adjacent healthy mucosa (1, 3, 4). Chemical profiles obtained from tumour‐adjacent tissue occupied dimensional spaces on the scores plot that were very close to, and in many cases overlapped by, cancer tissue. This is not a new observation per se and the findings of the current study support previous work related to the concept of ‘field cancerisation’. This idea was introduced by Slaughter (Slaughter et al., 1953 Sep) and Orr (Orr and 1954 Mar) in the 1950s who independently observed that tumour‐adjacent tissues exhibit significant changes in morphology prior to being invaded by cancer. In the present study the changes observed in tumour‐adjacent tissue relate principally to lipid metabolism which may indicate an important role for lipids in the priming of neighbouring tissue prior to cancer invasion. We have opted to use the term ‘cancer‐adjacent metaboplasia’ (CAM) to describe these tumour‐adjacent metabolic field effects, though this finding requires more investigation in view of the small sample size in this study. Developing a system for grading the extent of CAM/metabolic field cancerisation may offer a new avenue of prognostic molecular profiling. In addition, the ability to utilise mass spectrometric (MS) methods for identifying and delineating the tumour field, offers a potential means of developing novel technologies for in vivo application. We have recently developed and validated a ‘real‐time’ MS‐based method of tissue characterisation using a technique we have termed Rapid Evaporative Ionisation Mass Spectrometry (REIMS) (Takats et al., 2012 Feb; Sachfer et al., 2011 Mar 1; Sasi‐Szabó et al., 2013). With this technique, different tissues being dissected with electro‐cautery are seen to exhibit characteristic molecular ion patterns within the cautery‐induced vapour plume. This can be analysed within milliseconds using mass analyser technology to guide the operating surgeon (Sasi‐Szabó et al., 2013). Our finding of unique MS‐based lipid signatures in the tumour field can potentially be applied to this technology to offer the cancer surgeon a means of confirming when tumour deposits are in close proximity, and ‘skirting around’ these within the CAM field, to avoid disrupting/cutting into tumour itself. This could be particularly useful in disciplines such as hepatic surgery for example, where the solid, ‘closed‐box’ nature of the anatomical site reduces the sensitivity of tactile lesion perception.

From a biochemical perspective our study has demonstrated elevated levels of phosphatidyl‐choline PC(16:0–18:1), LysoPC(16:0) and LysoPC(18:1) in cancerous tissue regions. This is not a new finding in biochemical terms, as altered PC metabolism in cancerous tissue is well established and is being used to develop in vivo magnetic resonance spectroscopy (MRS) based approaches to tumour localisation, across a variety of cancer subtypes (Klomp et al., 2011 Dec; Wilson et al., 2009 Feb; Venkatesh et al., 2012 Mar). Our results are in agreement with previous ex vivo tissue profiling studies in breast and thyroid cancer (Williams et al., 1993 Apr; Ishikawa et al., 2012). Phosphatidyl‐choline (16:0/18:1) was found in significantly increased concentrations in cancerous compared to healthy thyroid tissue in a recent study evaluating the significance of lipid composition in papillary thyroid cancer. (Ishikawa et al., 2012) In an earlier study by Williams et al., breast cancer cells were found to contain relatively increased concentrations of PC (16:0/18:1) compared to healthy counterparts (Williams et al., 1993 Apr).

Lipids comprise a group of diverse molecules with different structures and functionalities that play important roles in cellular processes. They are associated with a number of critical cellular functions including membrane synthesis, metabolic regulation and immunity (Gschwind et al., 2002 Nov 1; Coussens et al., 2002 Dec 19–26). Increasing evidence points towards an important role for lipids in carcinogenesis, and altered cellular phospholipid composition has recently been reported across a variety of cancer sub‐types (Doria et al., 2013 Feb; Azordegan et al., 2012 Nov 23). In addition, target tissue lipid composition has been shown to be of potentially prognostic significance in determining likely response to chemotherapy (Todor et al., 2012 Jul). Taken together, these observations indicate that comprehensive characterisation of lipidomic profiles in cancer may offer an important new avenue of discovery for markers of diagnosis, prognosis and therapeutic efficacy. Although in this study we were able to localise only a small number of lipid species, future refinements in experimental workflow/parameters and the development of increasingly sensitive mass spectrometric profiling platforms promises to lead to more comprehensive mapping of the cancer lipidome.

6. Conclusion

In this study we have used a MALDI‐MSI approach to demonstrate that different tissue regions in the CRC microenvironment exhibit distinct lipid characteristics, and this finding supports emerging evidence across a variety of other cancer sub‐types. In addition we have observed lipid‐based differences between ‘tumour‐adjacent’ and ‘tumour‐remote’ healthy colorectal mucosa, and this is in keeping with the established principle of ‘field cancerisation’, whereby cancers influence the local environment prior to invasion. These field effects require further investigation and may prove to be of prognostic significance. We have coined the term ‘cancer‐adjacent metaboplasia’ (CAM) for our preliminary findings, as the changes we observed were primarily related to lipid metabolism in the tumour field. With shorter data acquisition‐time and reduced costs foreseeable in the near future, MALDI‐MSI methods could be used to develop novel biochemistry‐driven methods for cancer phenotyping to supplement current histopathology.

Conflicts of interest

None.

Supporting information

The following are the supplementary data related to this article:

Supplementary data

Click here for additional data file. (14.7KB, docx)

Supplementary data

Click here for additional data file. (12.7KB, docx)

Acknowledgements

The authors would like to acknowledge the NIHR Biomedical Research Centre for funding currently active and previous surgical metabonomics projects at Imperial College London. PAV acknowledges the Royal Society of Chemistry for financial support. EJW acknowledges Waters Corporation for funding. In addition the authors would like to acknowledge Mr Nigel Murray (Centre for Pathology, St Mary's Hospital, London, UK) for technical support in tissue sample cryo‐sectioning and staining. The authors would also like to thank Dr Emmanuelle Claude (Principal Scientist, Pharmaceuticals and Life Sciences, Waters Corporation, Manchester, UK) for assistance provided with metabolite identification in the present study.

Supplementary data 1.

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.molonc.2013.08.010.

Mirnezami R., Spagou K., Vorkas P.A., Lewis M.R., Kinross J., Want E., Shion H., Goldin R.D., Darzi A., Takats Z., Holmes E., Cloarec O. and Nicholson J.K., (2014), Chemical mapping of the colorectal cancer microenvironment via MALDI imaging mass spectrometry (MALDI‐MSI) reveals novel cancer‐associated field effects, Molecular Oncology, 8, doi: 10.1016/j.molonc.2013.08.010.

References

  1. Aichler, M. , Elsner, M. , Ludyga, N. , Feuchtinger, A. , Zangen, V. , Maier, S.K. , 2013 Apr 16. Clinical response to chemotherapy in oesophageal adenocarcinoma patients is linked to defects in mitochondria. J. Pathol.. [DOI] [PubMed] [Google Scholar]
  2. Alexandrov, T. , 2012. MALDI imaging mass spectrometry: statistical data analysis and current computational challenges. BMC Bioinfo.. 13, (Suppl 16) S11 [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Azordegan, N. , Fraser, V. , Le, K. , Hillyer, L.M. , Ma, D.W. , Fischer, G. , 2012 Nov 23. Carcinogenesis alters fatty acid profile in breast tissue. Mol. Cell Biochem.. [DOI] [PubMed] [Google Scholar]
  4. Bouslimani, A. , Bec, N. , Glueckmann, M. , Hirtz, C. , Larroque, C. , 2010 Feb. Matrix-assisted laser desorption/ionization imaging mass spectrometry of oxaliplatin derivatives in heated intraoperative chemotherapy (HIPEC)-like treated rat kidney. Rapid Commun. Mass Spectrom.. 24, (4) 415–421. [DOI] [PubMed] [Google Scholar]
  5. Caprioli, R.M. , Farmer, T.B. , Gile, J. , 1997 Dec 1. Molecular imaging of biological samples: localization of peptides and proteins using MALDI-TOF MS. Anal Chem.. 69, (23) 4751–4760. [DOI] [PubMed] [Google Scholar]
  6. Casadonte, R. , Caprioli, R.M. , 2011 Nov. Proteomic analysis of formalin-fixed paraffin-embedded tissue by MALDI imaging mass spectrometry. Nat. Protoc.. 6, (11) 1695–1709. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Cillero-Pastor, B. , Eijkel, G. , Kiss, A. , Blanco, F.J. , Heeren, R.M. , 2012 Nov 6. Time-of-flight secondary ion mass spectrometry-based molecular distribution distinguishing healthy and osteoarthritic human cartilage. Anal Chem.. 84, (21) 8909–8916. [DOI] [PubMed] [Google Scholar]
  8. Cloarec, O. , Dumas, M.E. , Trygg, J. , Craig, A. , Barton, R.H. , Lindon, J.C. , 2005 Jan 15. 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.. 77, (2) 517–526. [DOI] [PubMed] [Google Scholar]
  9. Cohen, S.L. , Chait, B.T. , 1996 Jan 1. Influence of matrix solution conditions on the MALDI-MS analysis of peptides and proteins. Anal Chem.. 68, (1) 31–37. [DOI] [PubMed] [Google Scholar]
  10. Coussens, L.M. , Werb, Z. , 2002 Dec 19-26. Inflammation and cancer. Nature. 420, (6917) 860–867. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Djidja, M.C. , Claude, E. , Snel, M.F. , Francese, S. , Scriven, P. , Carolan, V. , 2010 May. Novel molecular tumour classification using MALDI-mass spectrometry imaging of tissue micro-array. Anal. Bioanal. Chem.. 397, (2) 587–601. [DOI] [PubMed] [Google Scholar]
  12. Doria, M.L. , Cotrim, C.Z. , Simoes, C. , Macedo, B. , Domingues, P. , Domingues, M.R. , 2013 Feb. Lipidomic analysis of phospholipids from human mammary epithelial and breast cancer cell lines. J. Cell Physiol.. 228, (2) 457–468. [DOI] [PubMed] [Google Scholar]
  13. Dykstra, M. , Cherukuri, A. , Sohn, H.W. , Tzeng, S.J. , Pierce, S.K. , 2003. Location is everything: lipid rafts and immune cell signaling. Annu. Rev. Immunol.. 21, 457–481. [DOI] [PubMed] [Google Scholar]
  14. Eberlin, L.S. , Norton, I. , Orringer, D. , Dunn, I.F. , Liu, X. , Ide, J.L. , 2013 Jan 29. Ambient mass spectrometry for the intraoperative molecular diagnosis of human brain tumors. Proc. Natl. Acad. Sci. U S A. 110, (5) 1611–1616. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Engelhardt, J.J. , Boldajipour, B. , Beemiller, P. , Pandurangi, P. , Sorensen, C. , Werb, Z. , 2012 Mar 20. Marginating dendritic cells of the tumor microenvironment cross-present tumor antigens and stably engage tumor-specific T cells. Cancer Cell.. 21, (3) 402–417. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Fernandis, A.Z. , Wenk, M.R. , 2007 Apr. Membrane lipids as signaling molecules. Curr. Opin. Lipidol.. 18, (2) 121–128. [DOI] [PubMed] [Google Scholar]
  17. Fonville, J.M. , Carter, C. , Cloarec, O. , Nicholson, J.K. , Lindon, J.C. , Bunch, J. , 2012 Feb 7. Robust data processing and normalization strategy for MALDI mass spectrometric imaging. Anal Chem.. 84, (3) 1310–1319. [DOI] [PubMed] [Google Scholar]
  18. Gerbig, S. , Golf, O. , Balog, J. , Denes, J. , Baranyai, Z. , Zarand, A. , 2012 Jun. Analysis of colorectal adenocarcinoma tissue by desorption electrospray ionization mass spectrometric imaging. Anal Bioanal. Chem.. 403, (8) 2315–2325. [DOI] [PubMed] [Google Scholar]
  19. Goto-Inoue, N. , Hayasaka, T. , Zaima, N. , Setou, M. , 2011 Nov. Imaging mass spectrometry for lipidomics. Biochim. Biophys. Acta. 1811, (11) 961–969. [DOI] [PubMed] [Google Scholar]
  20. Gruner, B.M. , Hahne, H. , Mazur, P.K. , Trajkovic-Arsic, M. , Maier, S. , Esposito, I. , 2012. MALDI imaging mass spectrometry for in situ proteomic analysis of preneoplastic lesions in pancreatic cancer. PLoS One. 7, (6) e39424 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Gschwind, A. , Prenzel, N. , Ullrich, A. , 2002 Nov 1. Lysophosphatidic acid-induced squamous cell carcinoma cell proliferation and motility involves epidermal growth factor receptor signal transactivation. Cancer Res.. 62, (21) 6329–6336. [PubMed] [Google Scholar]
  22. Hanahan, D. , Weinberg, R.A. , 2011 Mar 4. Hallmarks of cancer: the next generation. Cell. 144, (5) 646–674. [DOI] [PubMed] [Google Scholar]
  23. Hart, P.J. , Francese, S. , Claude, E. , Woodroofe, M.N. , Clench, M.R. , 2011 Jul. MALDI-MS imaging of lipids in ex vivo human skin. Anal Bioanal. Chem.. 401, (1) 115–125. [DOI] [PubMed] [Google Scholar]
  24. Hogan, N.M. , Dwyer, R.M. , Joyce, M.R. , Kerin, M.J. , 2012 Jul. Mesenchymal stem cells in the colorectal tumor microenvironment: recent progress and implications. Int. J. Cancer. 131, (1) 1–7. [DOI] [PubMed] [Google Scholar]
  25. Huwiler, K.G. , Mosher, D.F. , Vestling, M.M. , 2003 Dec. Optimizing the MALDI-TOF-MS observation of peptides containing disulfide bonds. J. Biomol. Tech.. 14, (4) 289–297. [PMC free article] [PubMed] [Google Scholar]
  26. Ishikawa, S. , Tateya, I. , Hayasaka, T. , Masaki, N. , Takizawa, Y. , Ohno, S. , 2012. Increased expression of phosphatidylcholine (16:0/18:1) and (16:0/18:2) in thyroid papillary cancer. PLoS One. 7, (11) e48873 [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Klomp, D.W. , van de Bank, B.L. , Raaijmakers, A. , Korteweg, M.A. , Possanzini, C. , Boer, V.O. , 2011 Dec. 31P MRSI and 1H MRS at 7 T: initial results in human breast cancer. NMR Biomed.. 24, (10) 1337–1342. [DOI] [PubMed] [Google Scholar]
  28. Martinez-Outschoorn, U.E. , Lin, Z. , Trimmer, C. , Flomenberg, N. , Wang, C. , Pavlides, S. , 2011 Aug 1. Cancer cells metabolically "fertilize" the tumor microenvironment with hydrogen peroxide, driving the Warburg effect: implications for PET imaging of human tumors. Cell Cycle. 10, (15) 2504–2520. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Meding, S. , Nitsche, U. , Balluff, B. , Elsner, M. , Rauser, S. , Schone, C. , 2012 Mar 2. Tumor classification of six common cancer types based on proteomic profiling by MALDI imaging. J. Proteome Res.. 11, (3) 1996–2003. [DOI] [PubMed] [Google Scholar]
  30. Meriaux, C. , Franck, J. , Wisztorski, M. , Salzet, M. , Fournier, I. , 2010 Apr 18. Liquid ionic matrixes for MALDI mass spectrometry imaging of lipids. J. Proteomics. 73, (6) 1204–1218. [DOI] [PubMed] [Google Scholar]
  31. Morgan, T.M. , Seeley, E.H. , Fadare, O. , Caprioli, R.M. , Clark, P.E. , 2013 Mar. Imaging the clear cell renal cell carcinoma proteome. J. Urol.. 189, (3) 1097–1103. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Orr, J.W. , 1954 Mar. The relative importance of cell and field in cancer. East Afr. Med. J.. 31, (3) 101–106. [PubMed] [Google Scholar]
  33. Passarelli, M.K. , Winograd, N. , 2011 Nov. Lipid imaging with time-of-flight secondary ion mass spectrometry (ToF-SIMS). Biochim. Biophys. Acta. 1811, (11) 976–990. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Quehenberger, O. , Dennis, E.A. , 2011 Nov 10. The human plasma lipidome. N. Engl. J. Med.. 365, (19) 1812–1823. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Rompp, A. , Guenther, S. , Takats, Z. , Spengler, B. , 2011 Jul. Mass spectrometry imaging with high resolution in mass and space (HR(2) MSI) for reliable investigation of drug compound distributions on the cellular level. Anal Bioanal. Chem.. 401, (1) 65–73. [DOI] [PubMed] [Google Scholar]
  36. Sachfer, K.C. , Szaniszlo, T. , Gunther, S. , Balog, J. , Denes, J. , Keseru, M. , 2011 Mar 1. In situ, real-time identification of biological tissues by ultraviolet and infrared laser desorption ionization mass spectrometry. Anal Chem.. 83, (5) 1632–1640. [DOI] [PubMed] [Google Scholar]
  37. Sasi-Szabó, L. , Balog, J. , Kinross, J. , Lewis, M.R. , Muirhead, L.J. , Veselkov, K. , 2013. The ‘Intelligent surgical knife’: real-time classification of tumor Chemistry in surgery via Rapid Evaporative ionization mass spectrometry. Sci. Translational Med.. in press [DOI] [PubMed] [Google Scholar]
  38. Schwamborn, K. , Caprioli, R.M. , 2010 Sep. Molecular imaging by mass spectrometry–looking beyond classical histology. Nat. Rev. Cancer. 10, (9) 639–646. [DOI] [PubMed] [Google Scholar]
  39. Shimma, S. , Sugiura, Y. , Hayasaka, T. , Hoshikawa, Y. , Noda, T. , Setou, M. , 2007 Aug. MALDI-based imaging mass spectrometry revealed abnormal distribution of phospholipids in colon cancer liver metastasis. J. Chromatogr. B Analyt Technol. Biomed. Life Sci.. 855, (1) 98–103. [DOI] [PubMed] [Google Scholar]
  40. Slaughter, D.P. , Southwick, H.W. , Smejkal, W. , 1953 Sep. Field cancerization in oral stratified squamous epithelium; clinical implications of multicentric origin. Cancer. 6, (5) 963–968. [DOI] [PubMed] [Google Scholar]
  41. Smith, R.E. , Lespi, P. , Di Luca, M. , Bustos, C. , Marra, F.A. , de Alaniz, M.J. , 2008 Jan. A reliable biomarker derived from plasmalogens to evaluate malignancy and metastatic capacity of human cancers. Lipids. 43, (1) 79–89. [DOI] [PubMed] [Google Scholar]
  42. Sparvero, L.J. , Amoscato, A.A. , Dixon, C.E. , Long, J.B. , Kochanek, P.M. , Pitt, B.R. , 2012 Jul. Mapping of phospholipids by MALDI imaging (MALDI-MSI): realities and expectations. Chem. Phys. Lipids. 165, (5) 545–562. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Straussman, R. , Morikawa, T. , Shee, K. , Barzily-Rokni, M. , Qian, Z.R. , Du, J. , 2012 Jul 26. Tumour micro-environment elicits innate resistance to RAF inhibitors through HGF secretion. Nature. 487, (7408) 500–504. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Swartz, M.A. , Iida, N. , Roberts, E.W. , Sangaletti, S. , Wong, M.H. , Yull, F.E. , 2012 May 15. Tumor microenvironment complexity: emerging roles in cancer therapy. Cancer Res.. 72, (10) 2473–2480. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Takats, Z. , Wiseman, J.M. , Gologan, B. , Cooks, R.G. , 2004 Oct 15. Mass spectrometry sampling under ambient conditions with desorption electrospray ionization. Science. 306, (5695) 471–473. [DOI] [PubMed] [Google Scholar]
  46. Takats, Z. , Denes, J. , Kinross, J. , 2012 Feb. Identifying the margin: a new method to distinguish between cancerous and noncancerous tissue during surgery. Future Oncol.. 8, (2) 113–116. [DOI] [PubMed] [Google Scholar]
  47. Tarin, D. , 2012 Feb 3. Clinical and biological implications of the tumor microenvironment. Cancer Microenviron.. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Thomas, A. , Patterson, N.H. , Marcinkiewicz, M.M. , Lazaris, A. , Metrakos, P. , Chaurand, P. , 2013 Feb 8. Histology-driven data mining of lipid signatures from multiple imaging mass spectrometry analyses: application to human colorectal cancer liver metastasis biopsies. Anal Chem.. [DOI] [PubMed] [Google Scholar]
  49. Todor, I.N. , Lukyanova, N.Y. , Chekhun, V.F. , 2012 Jul. The lipid content of cisplatin- and doxorubicin-resistant MCF-7 human breast cancer cells. Exp. Oncol.. 34, (2) 97–100. [PubMed] [Google Scholar]
  50. Trygg, J. , Holmes, E. , Lundstedt, T. , 2007 Feb. Chemometrics in metabonomics. J. Proteome Res.. 6, (2) 469–479. [DOI] [PubMed] [Google Scholar]
  51. van Goor, H. , Gerrits, P.O. , Grond, J. , 1986. The application of lipid-soluble stains in plastic-embedded sections. Histochemistry. 85, (3) 251–253. [DOI] [PubMed] [Google Scholar]
  52. Venkatesh, H.S. , Chaumeil, M.M. , Ward, C.S. , Haas-Kogan, D.A. , James, C.D. , Ronen, S.M. , 2012 Mar. Reduced phosphocholine and hyperpolarized lactate provide magnetic resonance biomarkers of PI3K/Akt/mTOR inhibition in glioblastoma. Neuro Oncol.. 14, (3) 315–325. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Williams, C.M. , Maunder, K. , 1993 Apr. Fatty acid compositions of inositol and choline phospholipids of breast tumours and normal breast tissue. Eur. J. Clin. Nutr.. 47, (4) 260–267. [PubMed] [Google Scholar]
  54. Wilson, M. , Davies, N.P. , Grundy, R.G. , Peet, A.C. , 2009 Feb. A quantitative comparison of metabolite signals as detected by in vivo MRS with ex vivo 1H HR-MAS for childhood brain tumours. NMR Biomed.. 22, (2) 213–219. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

The following are the supplementary data related to this article:

Supplementary data

Click here for additional data file. (14.7KB, docx)

Supplementary data

Click here for additional data file. (12.7KB, docx)

Articles from Molecular Oncology are provided here courtesy of Wiley

RESOURCES