Distinct expression pattern of the full set of secreted phospholipases A2 in human colorectal adenocarcinomas: sPLA2-III as a biomarker candidate

Recent studies suggest that secreted phospholipases A2 (sPLA2s) represent attractive potential tumour biomarkers and therapeutic targets for various cancers. As a first step to address this issue in human colorectal cancer, we examined the expression of the full set of sPLA2s in sporadic adenocarcinomas and normal matched mucosa from 21 patients by quantitative PCR and immunohistochemistry. In normal colon, PLA2G2A and PLA2G12A were expressed at high levels, PLA2G2D, PLA2G5, PLA2G10 and PLA2G12B at moderate levels, and PLA2G1B, PLA2G2F and PLA2G3 at low levels. In adenocarcinomas from left and right colon, the expression of PLA2G3 was increased by up to 40-fold, while that of PLA2G2D and PLA2G5 was decreased by up to 23- and 14-fold. The variations of expression for sPLA2-IID, sPLA2-III and sPLA2-V were confirmed at the protein level. The expression pattern of these sPLA2s appeared to be linked respectively to the overexpression of interleukin-8, defensin α6, survivin and matrilysin, and downregulation of SFRP-1 and RLPA-1, all these genes being associated to colon cancer. This original sPLA2 profile observed in adenocarcinomas highlights the potential role of certain sPLA2s in colon cancer and suggests that sPLA2-III might be a good candidate as a novel biomarker for both left and right colon cancers.

Sporadic colorectal cancer is a major human health concern in industrialised countries and is the third leading cause of mortality by cancer. The accumulation of genetic alterations in several key genes (APC, K-ras, DCC and p53) has been associated with tumour development (Fearon and Vogelstein, 1990). In particular, the activation of the Wnt/b-catenin/Tcf-4 pathway is known to play a key role in colon tumorigenesis, leading to overexpression of a large number of target genes, including c-myc, cyclin D1, MMP-7, ITF-2, Il-8, gastrin, uPAR, cryptdin/defensin and EPR-1 genes (Kolligs et al, 2002;Levy et al, 2002). Moreover, the participation of tumour modifier genes may also contribute to pathogenesis, providing a basis to individual genetic predispositions, in addition to environmental and diet influences (Wiesner et al, 2003;Yan et al, 2004;Rakoff-Nahoum and Medzhitov, 2007). A comprehensive characterisation of such tumour modifier genes must take into account the stage of cancer and the location of the tumour within the large intestine, that is, 'right' or 'left' colon. This information may lead to a better understanding of the molecular events involved in colorectal cancer, and may provide new therapeutic targets to improve patient treatment outcomes (Williams et al, 2003;Barrier et al, 2005). It is also of interest to investigate the expression of such genes to identify new biomarkers of human colorectal cancers for early diagnosis and prognosis (Williams et al, 2003;Barrier et al, 2005).
In humans, 10 secreted phospholipases A 2 (sPLA 2 s) have been described in the last decade, and the analysis of their functions in physiological and pathophysiological conditions is under intense investigation. Human sPLA 2 s have been classified into groups IB, IIA, IID, IIE, IIF, III, V, X, XIIA and XIIB according to their structural properties (Laye and Gill, 2003;Rouault et al, 2003;Murakami et al, 2005;Cummings, 2007). Several of these proteins have recently been proposed either as biomarkers of pathologies (Smith et al, 2003;Mallat et al, 2005;Wootton et al, 2007) or as therapeutic targets (Laye and Gill, 2003;Cummings, 2007;Henderson et al, 2007). These enzymes have the capacity to generate biologically active lipid mediators such as lysophosphatidic acid (LPA) and arachidonic acid (AA), which can be further converted into prostaglandin E 2 (PGE 2 ). These lipid mediators are known to be involved in cell proliferation, cell migration, angiogenesis, and are likely to play a role in the initiation and/or progression of colorectal cancer (Morioka et al, 2000;Krause and DuBois, 2001;Laye and Gill, 2003;Mills and Moolenaar, 2003;Murakami et al, 2005). AA can also participate in apoptosis by activating sphingomyelinases and ceramide production (Dong et al, 2005;Ilsley et al, 2005).
Secreted phospholipases A 2 have been recently proposed as targets for anticancer drugs (Laye and Gill, 2003;Cummings, 2007), and there is increasing evidence for their involvement in various human cancers. Indeed, the expression levels of sPLA 2 -V and sPLA 2 -X are modified in human lung cancer (Masuda et al, 2005b), and those of sPLA 2 -V and sPLA 2 -XIIA in human ovarian cancer (Gorovetz et al, 2006). In particular, the overexpression of human sPLA 2 -IIA in gastric adenocarcinoma was proposed to be related to prolonged survival and less frequent metastasis (Leung et al, 2002;Aggarwal et al, 2006). By contrast, sPLA 2 -IIA overexpression has been associated to oncogenic effects in prostate cancer (Sved et al, 2004) and is related to poor prognosis (Graff et al, 2001). The polymorphism of the sPLA 2 -IIA gene, PLA2G2A, is also associated to some phenotypic features of patients with familial adenomatous polyposis (Yanaru-Fujisawa et al, 2007). More recently, human sPLA 2 -XIIB was shown to be upregulated in 50% of patients with hepatitis C virus-associated hepatocellular carcinoma (Smith et al, 2003). Finally, a protective role for sPLA 2 -IIA against Apc Min -induced intestinal tumours has been established in mice (MacPhee et al, 1995;Cormier et al, 1997). However, data on sPLA 2 and human cancers are still sparse, and often rely on only one or a few sPLA 2 s.
In this study, we have analysed the expression pattern of the full set of sPLA 2 s in tumour vs normal matched mucosa from patients with adenocarcinomas located in the left and right colon. Variation in the expression of a number of other genes associated with colon tumorigenesis and/or inflammation was also examined to establish possible gene coregulations. Our data provide the first comprehensive analysis of the expression pattern of all sPLA 2 s in normal human colon mucosa as well as adenocarcinomas. These data further support the fact that several sPLA 2 s may contribute to the initiation, progression or modulation of colon tumorigenesis, and may provide new potential tumour markers for this disease.

Human tissues and RNA isolation
Fourteen colon adenocarcinoma specimens from the left colon (descending part of the colon) and seven colon adenocarcinoma specimens from the right colon (ascending part of the colon) were obtained from surgical resections according to the French and American institutional guidelines. No specimen was from rectum. The patients did not receive chemotherapy or radiotherapy prior to surgery. The specimens were from 42-to 85-year-old patients (median age: 71 years), 14 men and 7 women. Samples were well (3 out of 21), moderately (14 out of 21) or poorly (4 out of 21) differentiated sporadic adenocarcinomas, with pTNM classification ranging pT2-4 N0-2 M0-1. Neither adenoma nor Duke'D adenocarcinoma was examined in this study. All samples had an expression of hMLH1 and hMSH2 unchanged as evaluated by immunohistochemistry and quantitative PCR (qPCR) (data not shown). Tissues from the non-necrotic part of the tumour and from distant normal mucosa were snap frozen in liquid nitrogen and stored at À801C. Each tissue sample (20-100 mg) was mixed with 700 ml of lysis buffer plus b-mercaptoethanol (nucleospin RNA II kit; Macherey-Nagel, Hoerdt, France) in a green-cap tube containing Lysing Matrix D (Q-BIOgene, Illkrich, France), and tissue disruption was achieved with the fast prep instrument (FP220A; Q-BIOgene). Total RNA was then isolated using the nucleospin RNA II kit, including DNase treatment. RNA concentration was determined by OD 260 and RNA quality was evaluated by analysis on an Agilent Bioanalyzer (Agilent Technologies, Les Ulis, France).

Quantitative PCR
First-strand cDNA was synthesised from 5 mg of total RNA using 100 U of MMLV reverse transcriptase (#M170A; Promega, Charbonnières-les bains, France) in a final volume of 50 ml with 500 ng of random primers (#C118A; Promega). Quantitative PCR was carried out in 96-well ABgene plates using the GENEAMP 5700 sequence Detection System apparatus (Applied Biosystems, Courtaboeuf, France) with the qPCR Master Mix Plus for SYBR s Green I (Eurogentec, Angers, France). All reactions were performed in a total volume of 16 ml and contained 50 ng of reverse transcribed RNA (based on the initial RNA concentration) and 250 nM of each primer set. The primer sets were designed using the Primer Express program from Applied Biosystems for the following human genes: PLA2G2A (NM_000300), sPLA 2 -IID gene (PLA2G2D) (NM_012400), sPLA 2 -IIE gene (PLA2G2E) (NM_0145891), sPLA 2 -IIF gene (PLA2G2F) (NM_022819), Most primer sets were designed to span an intron in order to avoid amplification from potential traces of genomic DNA in the total RNA preparations. Only the primer sets for MLH1, uPAR, EPR-1, MMP-7 and bcl-2 genes were not spanning an intron. For primer sets spanning an intron, we checked that no amplification signal was obtained using human genomic DNA as template in the qPCR (data not shown). The sequences of the designed primer sets are available on request. We used Qiagen commercial primer sets for cPLA 2 -IVA gene (PLA2G4A) (ref. QT00085813) and sPLA 2 -IB gene (PLA2G1B) (ref. QT00000637). We used published primer sets for RLPA-1 and RLPA-2 genes (Shida et al, 2004) and HD-6 gene (Andreu et al, 2005). The efficiency and specificity of each primer sets were validated using either serial dilutions of cloned human sPLA 2 cDNAs or mixed human tissue cDNA for the other genes. Moreover, when enough total RNA was collected, negative controls without added reverse transcriptase were performed. Thermal cycling was performed at 951C for 10 min, followed by 40 cycles comprising each a denaturation step at 951C for 15 s, and an annealing/extension step at 601C for 1 min. Amplification of the appropriate product was verified by analysing the dissociation curve that was obtained after PCR with the following steps: 15 s at 951C, 20 s at 601C, and then a slow ramp of 20 min from 60 to 951C. The abundance of the mRNA target was calculated relative to the expression of the reference gene TOP1 and is expressed as 2 DCt , where DC t ¼ C t (gene of interest)ÀC t (TOP1). The choice of TOP1 as a reference gene was determined using a Human GeNorm kit (PrimerDesign Ltd, Southampton, UK), allowing the determination of the best reference gene among 12 widely used reference genes. The data were also validated using GAPDH as a reference gene (data not shown). When the relative level of expression was plotted for normal mucosa (n) (Figure 1), we used the formula 2 20ÀDCtðnÞ , with DC t ¼ C t (n)geneÀC t (n)TOP1. The C t (n) values for TOP1 were typically around 20. When the expression of each gene in normal mucosa vs tumours (Figures 2 -4) was plotted, the comparative C t method (detailed in the ABI Prism 7700 Sequence Detection System User Bulletin no. 2) was used to determine the relative quantities of mRNA, and we changed the sign in order to get the lowest level of expression at the bottom and the highest level of expression at the top of the y axis, that is, ÀDC t ¼ À(C t gene of interestÀC t TOP1). When the expression of each gene was compared between the tumour (t) and the matched normal mucosa (n), the decrease or increase factor f in the tumour vs the normal mucosa ( Figure 5A and B) was calculated with the formula f¼ 2 IDCtðnÞÀDCtðtÞI À1, , and I the absolute value of the term DC t (n)ÀDC t (t).

Immunohistochemistry
All experiments were performed using paraffin-embedded tissues from the same patients as those used for the qPCR experiments.
The expression of sPLA 2 -IID, sPLA 2 -III, sPLA 2 -V and COX-2 proteins was analysed in tumours and normal matched tissues from three to four patients. Consecutive 4-mm tissue sections were deparaffinised in xylene and rehydrated in graded alcohol dilutions. Immunolabelling was performed using avidin -biotinperoxidase technique (Vectastain ABC kit; Vector, Burlingame, CA, USA). Before immunostaining, endogenous peroxidase activity was inhibited with 0.1% hydrogen peroxide in methanol for 30 min. Colour development was achieved with 3-amino-9-ethylcarbazole, and sections were finally counterstained with haematoxylin. Specific rabbit polyclonal antibodies against recombinant human sPLA 2 -IID, sPLA 2 -III and sPLA 2 -V were produced as described (Haas et al, 2005), and used at working dilutions of 1/300, 1/250 and 1/100, respectively. All anti-sPLA 2 antibodies were tested for specificity towards the various sPLA 2 s and were shown to be highly specific for each human sPLA 2 (Haas et al, 2005). Negative controls were performed by omission of the primary antibody. To further check for the specificity of labelling with sPLA 2 -IID-, sPLA 2 -III-and sPLA 2 -V-specific antibodies, competition experiments were performed in which the antibody was preincubated with the corresponding purified recombinant sPLA 2 used for immunisation, prior to covering the tissue slides. The competition was performed by preincubating the relevant antibody solution with 200 nM of human recombinant sPLA 2 -III or 100 nM of human recombinant sPLA 2 -IID and sPLA 2 -V for 1 h at room temperature. Purified recombinant proteins were obtained as described (Rouault et al, 2007). Human COX-2 antibodies were from Cayman (Montigny-le Bretonneux, France) (ref. 160112, working dilution 1/500), human MLH1 antibodies were from BD Pharmingen (Le Pont-de-Claix, France) (clone G168-178, working dilution 1/100), and human MSH2 antibodies were from Calbiochem (Darmstadt, Germany) (clone FE-11, working dilution 1/125).

Statistics
We determined that the sample data from left and right colons followed a Gaussian distribution using the D'agostino -Pearson normality test. Normal mucosa and tumour paired data for each gene were analysed using the Bonferroni test, and statistical significances were represented by *Po0.05, **Po0.01 and Figure 2 Expression level of sPLA 2 genes in human adenocarcinomas and normal matched mucosa. The expression level of sPLA 2 genes and of the M-type sPLA 2 receptor (PLA2R1) was measured in adenocarcinomas and normal matched mucosa for the 14 left-sided samples. The data were obtained after normalisation with TOP1 used as reference gene and using the formula ÀDC t ¼ C t gene of interestÀC t TOP1. N, normal tissues; T, tumour tissues. ***Po0.001. We looked for correlation between all genes examined in this study by using the Mev3.1 software and hierarchical clustering to analyse different linkages to cluster genes and samples. Euclidean distance was used to calculate the distance between two genes.

Expression levels of sPLA 2 genes in normal human colon mucosa
We first looked at the mRNA levels for the different sPLA 2 genes in normal mucosa and compared their relative expression levels. Our data show dramatic differences in the relative expression of sPLA 2 . The lowest gene expression level was found for PLA2G3, and the highest gene expression level was found for PLA2G2A ( Figure 1). The level of PLA2G2A expression was particularly high. It is above that of TOP1 and almost reaches that of GAPDH, which are both used as reference genes and are known to be highly expressed in many tissues, including colon. By contrast, the expression of PLA2G3 was generally very low, in fact below the limit of detection in many samples (Figure 1). The expression levels of PLA2G2F and PLA2G1B were also low, while that of other sPLA 2 genes were moderate. The expression of PLA2G2E was not detectable (data not shown). We also compared the expression levels in normal mucosa of left and right colon. A significant difference in gene expression levels was seen for PLA2G2F, PLA2G1B, PLA2G12B and PLA2G10 (respectively 30-, 13-, 24-and 40-fold lower expression in rightsided normal mucosa than in left-sided normal mucosa). No or minor differences were observed in gene expression for PLA2G2A, PLA2G2D, PLA2G3, PLA2G5 and PLA2G12A between right-and left-sided normal mucosa (Figure 1).

Expression levels of sPLA 2 genes in left and right colon adenocarcinomas vs normal human colon mucosa
We next compared sPLA 2 gene expression in adenocarcinomas vs normal mucosa. The raw data for samples from the left colon are shown in Figure 2. The fold-increase or fold-decrease in the expression of each gene in adenocarcinomas vs normal mucosa is represented in Figure 3. Interestingly, we observed a 40-fold increase in PLA2G3 expression in the adenocarcinomas. In contrast, we found 23-and 14-fold decreases for PLA2G2D and PLA2G5, respectively. Similar data were obtained for samples from the right colon, with a 22-fold increased expression in tumour vs normal mucosa for PLA2G3, and 10-and 44-fold reduced expression for PLA2G2D and PLA2G5, respectively (Figure 4). No variation in expression between the tumours and normal mucosa was observed for PLA2G1B, PLA2G2A, PLA2G2F, PLA2G10, PLA2G12A and PLA2G12B and for the M-type sPLA 2 receptor (PLA2R1) (Figures 3 and 4). The expression of PLA2G2E PLA2G1B * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * Figure 3 Expression level of sPLA 2 , tumour-related and inflammationrelated genes in human left colon adenocarcinomas. After normalisation with the TOP1 reference gene, the expression level for each gene in the tumour was compared to that in the normal matched mucosa. The zero value indicates no variation, a positive value indicates an increased expression level in tumour vs normal tissue and a negative value indicates a decreased expression level in tumour vs normal tissue (see Materials and Methods). The incidence of variation in the different samples is indicated. PLA2G1B * * * * * * * * * * * * * * * * * * * * *

Expression levels of genes involved in inflammation and tumorigenesis in left and right colon adenocarcinomas vs normal human colon mucosa
Since sPLA 2 s may exert a coordinate action with other genes in tumorigenesis and associated inflammation, we also examined the expression of a panel of genes known to be involved in inflammation and colorectal tumorigenesis (Figures 3 and 4).
In left colon samples, we did not observe any differences in expression levels between adenocarcinomas and normal mucosa for most of the inflammation-related genes examined in this study, that is, ptgs2, ptges1, ptges2, TNF-a, Il-1a, Il-6, Il-10, PPARg, PPARd, MMP-9 and PLA2G6. Only a slight decrease in the expression of ptgs1 and PLA2G4A was observed. With the exception of ptges2 and Il-1a, we found that the inflammationrelated genes were already expressed at high levels in the normal mucosa.
We also looked at the expression of urokinase plasminogen activator (u-PA) and its receptor (u-PAR) that play a role in cell adhesion and cell migration, and are involved in late stages of tumour development, contributing to tumour cell invasion and metastatic spread (Terada et al, 2005). No changes were observed in u-PA and u-PAR gene expression levels.
RLPA-1 and RLPA-2 genes code for two distinct LPA receptors (Mills and Moolenaar, 2003). While the expression of RLPA-2 did not change, that of RLPA-1 was decreased by 20-fold in tumour vs normal matched mucosa.
IL-8, EPR-1, HD-6 and MMP-7 are target genes of the Wnt/b-cat/ tcf-4 pathway (Kolligs et al, 2002;Levy et al, 2002;Wheatley and McNeish, 2005). Interestingly, a marked increase in expression was found for Il-8, HD-6 and MMP-7 expression in the tumours (8-, 22and 70-fold, respectively). The level of EPR-1 was not changed. SFRP-1 codes for an antagonist of the Wnt pathway, which binds to the frizzled receptor and blocks frizzled -Wnt interaction (Suzuki et al, 2004). Our results show a dramatic 130-fold decrease in the expression of SFRP-1 in the tumour.
Similar data were obtained for most genes in the right-sided samples ( Figure 4). Indeed, no variation in expression was observed for most inflammatory-related genes, while a dramatic increase (280-fold) and a strong decrease (125-fold) were found for MMP-7 and SFRP-1, respectively. However, in contrast to left-sided samples, we observed no variation in HD-6 and Il-8 expression levels and a strong increase (45-fold) in EPR-1 expression.

Incidence and coregulation of studied genes
We first looked at the incidence of the above variations, that is, the number of samples showing an increase or decrease in the expression level of a particular gene within the 21 patients used in this study (Figures 3 -5). The increase in PLA2G3 expression showed a high incidence (12 out of 14 for left-sided samples and 5 out of 7 for right-sided samples), which was similar to that of MMP-7 gene overexpression (10 out of 14 for left-sided samples and 6 out of 7 for right-sided samples). The incidence for the overexpression of HD-6 and Il-8 was lower, since these two genes were not significantly upregulated in right-sided samples. The decrease in PLA2G2D and PLA2G5 expression also showed a high incidence (12 out of 14 and 5 out of 7; and 11 out of 14 and 5 out of 7, respectively), which was similar to that of RLPA-1 and SFRP-1 gene expression (13 out of 14 and 6 out of 7; and 14 out of 14 and 7 out of 7, respectively). We did not observe any difference between  Figure 5 Hierarchical clustering of genes. Heatmaps comparing the log 2 ratio between the expression level in adenocarcinomas and normal epithelium for the 14 patients with left-sided adenocarcinomas (A) and for the 7 patients with right-sided adenocarcinomas (B) are shown. The distance corresponds to an Euclidean distance calculated using the Mev3.1 software.
female and male patients used in our study for the variations of expression of PLA2G3, PLA2G2D and PLA2G5 (data not shown). Euclidean distance is commonly used to evaluate linkages in the expression of different genes. We used this method to point out linkages between the variations in expression level of the most relevant genes. In left-sided adenocarcinomas ( Figure 5A), a first cluster (cluster I) of upregulated genes was observed containing PLA2G3 and several genes related to colon cancer such as HD-6, MMP-7 and Il-8. A second cluster (cluster II) of several downregulated genes, including SFRP-1, PLA2G2D, RLPA-1, PLA2G5, PLA2G4A and ptgs1, was observed. These two clusters appeared most likely to discriminate adenocarcinomas from normal tissues. Remarkably, there are 3 sPLA 2 genes out of 11 genes in these two clusters. The other genes did not show any linkage with colon tumorigenesis, since their expression levels were up-or downregulated or not changed among the different patients. When looking at right-sided samples ( Figure 5B), similar patterns were observed even though the number of patients was lower.
Protein expression levels of sPLA 2 -IID, sPLA 2 -III and sPLA 2 -V in colon adenocarcinomas and normal matched mucosa To confirm that the variation of expression for PLA2G2D, PLA2G3 and PLA2G5 is also observed at the protein level, immunohistochemical analyses were performed on tissue sections from the same patients as those used for qPCR. For each sPLA 2 protein analysed, tissue sections from three or four patients showing a significant increase or decrease of sPLA 2 expression were selected and immunostained ( Figure 6). The labelling for sPLA 2 -IID and sPLA 2 -V was markedly decreased in tumour cells compared to the normal matched epithelium, in accordance with the qPCR data. By contrast, the labelling for sPLA 2 -III was absent or very low in the normal epithelium, with a significant increase in tumour tissue. The absence of labelling in competition experiments (data not shown) or when the primary antibody was omitted ( Figure 6) demonstrated that the signals observed for the three sPLA 2 proteins were specific. In addition, immunohistochemical analysis of COX-2 showed a more intense protein expression in tumours than in normal mucosa in a subset of patients (data not shown).

DISCUSSION
Several lines of evidence have been accumulated during the past decade to support the role of sPLA 2 s in cancer pathogenesis. Their role in human cancer, however, has not been clarified. As a first step towards addressing this issue, we have analysed the expression patterns of the full set of human sPLA 2 s in colorectal cancer tissue samples and normal matched mucosa. Our results indicate that (i) several sPLA 2 s, including sPLA 2 -IIA, sPLA 2 -X and sPLA 2 -XIIA, are highly expressed in both normal and tumour colon tissues; (ii) the expression levels of sPLA 2 -III, sPLA 2 -IID and sPLA 2 -V are dramatically altered in adenocarcinomas at both mRNA and protein levels; and (iii) their mRNA profiles are part of gene expression clusters with other genes associated with inflammation and cancer. Together, our data are further suggestive of a role of these sPLA 2 s in colorectal cancer and open the possibility that sPLA 2 s, in particular sPLA 2 -III, may provide novel cancer biomarkers.
sPLA 2 gene expression levels in colon adenocarcinomas and normal human colon Colon adenocarcinomas analysed in this study are likely cancers without high microsatellite instability. Indeed, we used immunohistochemical analysis as a reliable method for screening DNA mismatch repair defects (Lindor et al, 2002), and found that the expression of hMLH1 and hMSH2 was unchanged. Furthermore, we analysed the normal mucosa and the tumour tissues from the right and left colon separately because of their distinct gene expression patterns (Glebov et al, 2003;Birkenkamp-Demtroder et al, 2005). We observed similar levels of expression for most sPLA 2 s in left and right normal mucosa, with PLA2G2A having a high expression level, PLA2G3 being not detectable in most samples and the other sPLA 2 genes showing intermediate levels of expression ( Figure 1). Interestingly, we observed marked alterations in the expression levels of PLA2G2D, PLA2G5 and PLA2G3 in colon adenocarcinomas. In contrast, there was no difference in the expression levels of the other sPLA 2 genes (Figures 2 -4). The absence of tumour-related alterations in the gene expression of PLA2G2A and PLA2G10 in human sporadic colon cancer tissues is consistent with earlier studies (Dimberg et al, 1998;Osterstrom et al, 2002;Murakami et al, 2005). These data obtained in humans differ from those obtained in mice, which showed an increased expression of PLA2G2A and no variation in PLA2G5 expression in tumours from azoxymethane-treated mice . Secreted phospholipase A 2 immunolabelling in representative colon adenocarcinomas and normal matched mucosa. For sPLA 2 -IID and sPLA 2 -V expression, a strong staining in the surface epithelium of the normal mucosa was observed, contrasting with the weak or absent staining observed in tumour cells. For sPLA 2 -III expression, a very faint staining was observed in normal epithelial cells, contrasting with the moderate to strong staining observed in tumour cells. For each sPLA 2 , a representative immunolabelling from 1 patient out of 3 -4 patients analysed is shown. As shown in the bottom picture, no staining is observed when the primary antibody is omitted. Scale bar is 100 mm.

Expression level of genes involved in inflammation and tumorigenesis in colon adenocarcinomas and normal human colon
No significant changes between adenocarcinomas and normal colon were observed for most inflammation-related genes. It should be noted that many of these genes were expressed at high levels within the normal mucosa, in particular, MMP-9, PLA2G6 and ptgs2. A recent study by Chen et al (2004) has also shown high mRNA expression levels of several inflammation-related genes in normal-appearing colon mucosa of patients with colon cancer. Therefore, our data would suggest a possible pre-existing inflammatory condition within the normal mucosa distant from the tumours, and likely explain the absence of increased expression in adenocarcinomas vs normal mucosa for several inflammation-related genes. The slight decrease in ptgs1 (the COX-1 gene) appeared in agreement with that observed in colon adenocarcinomas of stage III (Duke's C) patients (Church et al, 2004). The COX-1 protein is considered to exert dual opposing effects in cancer, acting as either a tumour suppressor or a tumour initiator (Chulada et al, 2000). The slight decrease in PLA2G4A expression that we observed may be related to the dual opposing effects of cPLA 2 -IVA in cancer, that is, proliferative effects via metabolism of AA into eicosanoids, such as PGE 2 , and antiproliferative effects via AA-dependent ceramide production leading to apoptosis . The absence of increase in ptgs2 (the COX-2 gene) levels in our set of patients is in accordance with two recent qPCR studies also performed on patients with colon cancer (Church et al, 2004;Gustafsson et al, 2007), and may be related to the inflammatory status of the normal mucosa discussed above. The fact that we did not observe an increase of COX-2 at the mRNA level, while we detected an overexpression at the protein level by immunohistochemistry as previously reported (Wendum et al, 2003), is in line with the posttranscriptional regulation of COX-2 expression (Dixon et al, 2003). Therefore, mRNA and protein levels of COX-2 may not be closely linked, as recently proposed in colon cancer tissues (Gustafsson et al, 2007) and colon cancer cells (Dixon et al, 2003).
Downregulation of sPLA 2 -IID and sPLA 2 -V expression in human colon adenocarcinomas A marked decrease in the expression level of PLA2G2D and PLA2G5 was observed in both left-and right-sided adenocarcinomas (Figures 2-4). These variations were also observed at the protein level by immunohistochemical analyses. Indeed, sPLA 2 -IID and sPLA 2 -V were found to be present in epithelial cells of normal mucosa, and their expression was markedly reduced in tumours ( Figure 6). The decreased expression of sPLA 2 -IID and sPLA 2 -V in tumours may suggest a protective role of these sPLA 2 s, as it has been proposed for sPLA 2 -IIA in mice (MacPhee et al, 1995;Cormier et al, 1997). A similar decrease in expression of sPLA 2 -IID and sPLA 2 -V was also described in gastric tissues with signet-ring cell carcinoma (Masuda et al, 2005a). Interestingly, the PLA2G2A, PLA2G2C, PLA2G2D, PLA2G2E, PLA2G2F and PLA2G5 genes reside within the same region of human chromosome 1 at p35 -36.1 (Valentin et al, 2000), a region frequently altered in colorectal cancer (Spirio et al, 1996). The decreased expression for PLA2G2D and PLA2G5 appears to be linked (cluster II, Figure 5) to that observed for PLA2G4A, ptgs1, SFRP-1 and RLPA-1, which have been proposed as tumour suppressor genes in colon cancer (Chulada et al, 2000;Shida et al, 2004;Suzuki et al, 2004;Dong et al, 2005).

Upregulation of PLA2G3 and Wnt target genes in human colon adenocarcinomas
A marked increase in sPLA 2 -III expression level was observed in both left-and right-sided adenocarcinomas, a finding that was confirmed by immunohistochemical analysis. No or very weak sPLA 2 -III expression was observed in normal epithelial cells, but a robust expression was observed in tumours ( Figure 6). Our findings are consistent with the recent observation that sPLA 2 -III can trigger the proliferation of human colon cancer cells in vitro (Murakami et al, 2005). Parallel to the increase in PLA2G3 gene expression level, we observed an increase in the expression level of four target genes of the Wnt/b-cat/Tcf-4 pathway: Il-8 (Levy et al, 2002), HD-6 (also called defensin a6) (Kolligs et al, 2002), MMP-7 (also called matrilysin) (Nelson et al, 2000) and EPR-1 (also called survivin) (Kolligs et al, 2002), as well as a concomitant dramatic decrease in SFRP-1 gene expression level. Therefore, our data are consistent with the activation of the Wnt/b-cat/Tcf-4 pathway in the development of colon adenocarcinomas. It will be of interest to determine whether the expression of PLA2G3 is related to the activation of the Wnt/b-cat/Tcf-4 pathway, as already suggested for PLA2G2A in gastric cancer (Aggarwal et al, 2006). Our data further support the role of LPA in the pathogenesis of colon cancer. LPA exerts its effects through at least three different receptors: RLPA-1, RLPA-2 and RLPA-3 (Mills and Moolenaar, 2003). It has been demonstrated that the Wnt/b-cat/Tcf-4 pathway is involved in the proliferative effects of LPA through binding to RLPA-2 (Yang et al, 2005). We observed a decreased expression of RLPA-1 and a sustained expression of RLPA-2, which is in agreement with the predominant expression of RLPA-2 in adenocarcinomas (Shida et al, 2004). Whether sPLA 2 -III plays a role in the production of LPA, which in turn activates RLPA-2, remains to be established.
The sPLA 2 -III as a novel potential biomarker of human colon cancer Because of their upregulation during colorectal carcinogenesis, Wnt target genes and their associated products have been examined for their potential use as biomarkers. The level of Il-8 protein was found to be increased in colorectal cancer patients (Terada et al, 2005). A significant increase in the level of defensin a6 has also been found in patients with colon cancer (Nam et al, 2005). Although Il-8 and defensin a6 levels have been recently proposed as markers of human colorectal cancer (Nam et al, 2005;Terada et al, 2005), our data show that they would detect only leftsided adenocarcinomas (Figures 3 -5). In contrast, the increased expression of sPLA 2 -III was observed in both left and right colon adenocarcinomas, indicating that the analysis of sPLA 2 -III levels would detect cancers located in both left and right colon. MMP-7 is a matrix metalloprotease that has been associated with tumour invasion and metastasis (Nelson et al, 2000). We found that the expression pattern of PLA2G3 was more similar to that of MMP-7, which was also increased in both left and right colon adenocarcinomas (Figures 3 -5).
In conclusion, this work is the first comprehensive analysis of the expression pattern of the full set of sPLA 2 s in human colon cancer. The distinct expression pattern observed for sPLA 2 genes suggests that mRNA profiling of the full set of human sPLA 2 s may be useful to detect colon tumours either by analysing their expression pattern in tumours (Barrier et al, 2005), in circulating blood cells (Burczynski et al, 2005) or directly in serum on circulating mRNA, as recently proposed for other genes (Li et al, 2006). Moreover, we have observed a dramatic increase in sPLA 2 -III expression in both left-and right-sided adenocarcinomas, suggesting that sPLA 2 -III may represent a novel broad molecular biomarker of colon cancer. It will be of interest to determine whether the expression level of sPLA 2 -III also increases at earlier stages of tumorigenesis, including adenomas. It will also be useful to detect the sPLA 2 -III protein in human colon biopsies, stools or serum using the recently time-resolved fluoroimmunoassays developed for the different human sPLA 2 s (Nevalainen et al, 2005).