Surgical specimens and corresponding normal tissue from the same samples were collected from patients with CRA (n = 30), CRC of different tumor categories (n = 48) and synchronous or metachronous CRLM (n = 16) who underwent surgical resection at our department between January 2003 and october 2006. Informed consent for tissue procurement was obtained from all patients. The study was approved by the ethics commission of the Ärztekammer of the Saarland, Germany. The clinical variables presented in Table 1 and 2 were obtained from the clinical and pathological records and are in accordance with the UICC/TNM classification 33.

Tissue samples were collected immediately after resection, snap frozen in liquid nitrogen and then stored at -80°C until they were processed under nucleic acid sterile conditions for RNA and protein extraction. Tumor samples were taken from vital areas of histopathologically confirmed (R.M.B. and M.W.) adenocarcinomas and liver metastases, respectively. As corresponding normal tissue we used adjacent unaffected mucosa, 2–3 cm distal to the resection margin from the same resected adenocarcinoma or liver specimen, respectively. All tissues obtained were reviewed by a minimum of two experienced pathologists (M.W. et al.) and examined for the presence of tumor cells. As minimum criteria for usefulness for our studies we only chose tumor tissues in which tumor cells occupied a major component (> 80%) of the tumor sample.

Total RNA was isolated using RNeasy columns from Qiagen (Qiagen, Hilden, Germany) according to the manufacturer's instructions. RNA integrity was confirmed spectrophotometrically and by electrophoresis on 1% agarose gels. For cDNA synthesis 5 μg of each patient total RNA sample were reverse-transcribed in a final reaction volume of 50 μl containing 1× TaqMan RT buffer, 2.5 μM/l random hexamers, 500 μM/l each dNTP, 5.5 mM/l MgCl₂, 0.4 U/μl RNase inhibitor, and 1.25 U/μl Multiscribe RT. All RT-PCR reagents were purchased from Applied Biosystems (Applied Biosystems, Foster City, CA). The reaction conditions were 10 min at 25°C, 30 min at 48°C, and 5 min at 95°C.

Expression profiles of CXCL1, CXCL5 and CXCL6 were monitored in CRA and CRC tissue specimens and corresponding CRLM by Q-RT-PCR. Adjacent, disease and tumor free colorectal epithelium and liver samples served as control groups, respectively. All Q-RT PCR assays containing the primer and probe mix were purchased from Applied Biosystems and utilized according to the manufacturer's instructions. PCR reactions were carried out using 10 μl 2× Taqman PCR Universal Master Mix No AmpErase^® UNG (Applied Biosystems) and 1 μL gene assay, 8 μL RNase-free water and 1 μl cDNA template (50 mg/l). The theoretical basis of the Q-RT assays is described in detail elsewhere 34. All reactions were carried out in duplicate along with no template controls and an additional reaction in which reverse transcriptase was omitted to assure absence of genomic DNA contamination in each RNA sample. For signal detection, the ABI Prism 7900 sequence detector (Applied Biosystems), was programmed to an initial step of 10 min at 95°C, followed by 40 thermal cycles of 15 s at 95°C and 10 min at 60°C and the log-linear phase of amplification was monitored to obtain C_T values for each RNA sample.

Gene expression of all target genes was analyzed in relation to the levels of the slope matched housekeeping genes phosphomannomutase (PMM1) and cyclophilin C (CycC) 35. Since reporting of data obtained from raw C_T values falsely represent the variations, we converted the individual C_T values to the linear form as follows:

Fold difference = 2 − ( mean C T  pathological tissue − mean C T  calibrator ) = 2 − delta C T MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xI8qiVKYPFjYdHaVhbbf9v8qqaqFr0xc9vqFj0dXdbba91qpepeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=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@8BD9@

Consequently, the normal tissue becomes the 1 × sample, and all other quantities are expressed as an n-fold difference relative to this tissue.

Protein lysates from frozen tissues were extracted with a radioimmunoprecipitation (RIPA) cell lysis and extraction buffer (Pierce, Rockford, USA). Total protein content was assessed by using the Pierce BCA protein assay reagent kit (Pierce).

Tissue concentrations of CXCL1, CXCL5 and CXCL6 in the lysates were determined by sandwich-type ELISA according to the manufacturer's protocol: R&D systems (DuoSet, R&D Systems Inc. Minneapolis, Minnesota, USA). The absorbance was read at 450 nm.

Operative specimens were routinely fixed in formalin and subsequently embedded in paraffin. Before staining, 4-μm thick paraffin-embedded tissue sections were mounted on Superfrost Plus slides, deparaffinized and rehydrated in graded ethanol to deionized water. The sections were microwaved with an antigen retrieval solution (Target Retrieval, Dakocytomation, Carpinteria, CA, USA) and after blocking of endogenous peroxidase activity with 3% hydrogen peroxide, the sections were further blocked for 30 minutes with normal rabbit serum. Overnight incubation at 4°C with primary goat polyclonal anti-human CXCL5 antibody (15 μg/ml, AF254, R&D Systems, Minneapolis, Minn., USA) or primary goat anti-human CXCL1 antibody (3.5 μg/ml, sc-1374, Santa Cruz Biotechnology, Santa Cruz, Calif., USA) was followed by incubation of secondary biotinylated biotinylated rabbit anti-goat IgG antibody and the avidin-biotin-peroxidase reaction (Vectastain ABC ELITE Kit, Vector Laboratories, Burlingame, CA, USA). After colour reaction with aminoethylcarbazol solution (Merck, Darmstadt, Germany), tissues were counterstained with haematoxylin. Negative controls were performed in all cases omitting primary antibody. All pictures were taken using a digital camera (Olympus DP71, Olympus Optical Co, Ltd, Tokyo, Japan).

Laser microbeam microdissection (LMM) was employed for obtaining pure tumor cell and pure normal cell samples for subsequent genetic analysis. LMM was performed on three samples for each tissue type for CXCL1, CXCL5 and CXCL6. Histochemical staining was used on cryo sections before microdissection. Specimen preparation, microdissection and catapulting were performed following a laser pressure catapulting protocol according to the manufacturer's instructions (P.A.L.M. Microlaser Technologies, Bernried, Germany). RNA was extracted using the P.A.L.M. RNA extraction kit and for reverse transcription the invitrogen reverse transcription kit (Invitrogen Life Technologies, Karlsruhe, Germany) was applied. Subsequently quantitative PCR analysis was performed.

Expression profiles of CXCL1, CXCL5 and CXCL6 in the different groups are shown as mean and standard error of the mean (SEM). Statistical calculations were done with the MedCalc software package (MedCalc software, Mariakerke, Belgium) 36. Where appropriate, either the Student's t-test or the Wilcoxon's rank sum test was applied to test for group differences of continuous variables. A P value of 0.05 or less was deemed significant. To measure how variables are related we performed a bivariate correlation analysis using the spearman ρ correlation coefficient.

CXCL1 and CXCL5 mRNA expressions were significantly up-regulated in all CRC tissue specimens in relation to the matched tumor neighbor tissues (P < 0.001, respectively; Fig. 1). In addition, CXCL1 also revealed significant up-regulation in the CRA tissues (P < 0.05). In contrast, no significant up-regulation of CXCL6 mRNA expression was observed in either tissue type (Fig. 1). In CRA tissues also CXCL5 showed no significant up-regulation, indicating that CXCL5 is not overexpressed in preliminary stages of CRC. However, CXCL5 overexpression in CRC tissues revealed a much more pronounced 80-fold up-regulation in comparison to CXCL1, which only showed a 5-fold overexpression in CRC patients. Furthermore, we detected a significant clinicopathological association between CXCL5 expression and early tumor categories of CRC (P < 0.001), which did not occur for CXCL1 or CXCL6 (Fig. 2). In metastatic samples, CXCL1 and CXCL5 expression was also significantly elevated relative to corresponding normal liver tissues (P < 0.05, respectively, Fig. 3). Again, CXCL5 expression was much more pronounced. Between metastases and corresponding primary tumors we observed no significant difference in mRNA expression (Fig. 3). Notably, CXCL5 expression was significantly lower in the corresponding primary tumors (n = 14) of the CRLM presented in Fig. 3 compared to the entire CRC patient cohort (n = 48), which comprised all tumor categories 1–4 as shown in Fig. 1. Showing that lower T- categories (T1–2) express almost 100-fold significantly more CXCL5 than higher T-categories (T3–4) (Fig. 2B), led us to review the T- categories of the corresponding primary tumors of the CRLM. It turned out that 11 out of 14 tumors were classified as T3 and T4 tumors (Table 1), thus explaining the results presented in Fig. 2B.

To verify that up-regulation of CXCL1 and CXCL5 is generated exclusively by tumor cells rather than inflammatory cells or other non-malignant cells in the tissue blocks, we routinely analysed Q-RT-PCR expressions of several sections of microdissected tumor and normal cells from three specimens of all tissue types under investigation. Moreover, we generally determined the differences between gene expressions from matched normal/cancer samples constituting the basis for all our calculations.

CXCL1 and CXCL5 protein expressions, as assessed by ELISA, showed a significant up-regulation in the CRC tissue specimens in comparison to the unaffected corresponding mucosa tissues, respectively (P < 0.001) as demonstrated in Figures 4A and 4B. In consistence with the results obtained at the mRNA level, CXCL5 protein overexpression was much more pronounced corresponding to a 60-fold expression increase compared to only a 3.5-fold expression increase for CXCL1. Although significant up-regulation of CXCL1 proteins was also detected in CRA tissues, the level of CXCL1 protein expression in the CRC tissues was significantly 3.5-fold higher than in the CRA tissues (P < 0.05, Fig. 4A). Similarly, CXCL5 showed significant protein up-regulation in the CRC patient tissues with respect to the CRA tissues (Fig. 4B). Again, this overexpression was much more pronounced in comparison to CXCL1, demonstrating a 20-fold up-regulation for CXCL5 versus a 3.5-fold up-regulation for CXCL1, thus indicating a prominent progressive protein expression increase in the transition from the premalignant condition to the development of colorectal malignancies. In contrast, CXCL6 protein expressions showed no significant up-regulation in either tissue type (Fig. 4C).

Comparative analysis of the CXCL1/CXCL5 absolute protein quantities in CRC patients revealed a significantly higher CXCL1 expression (P < 0.05) of almost 14000 pg/ml compared to approximately 4000 pg/ml CXCL5 (Fig. 5A). On the other hand, we also observed a significantly higher CXCL1 expression level in the CRA tissues, where 4000 pg/ml CXCL1 compared to 200 pg/ml CXCL5, indicating that the basic expression level of CXCL1 is generally higher in both disease entities in comparison to CXCL5 (Fig. 5B).

Detection of CXCL1 and CXCL5 expression was assessed by immunohistochemical staining in CRA, CRC and CRLM specimens and corresponding normal tissues (Fig 6). Immunostaining with CXCL1 -specific antibodies showed weak apical epithelial signals and weak staining of goblet cells in the corresponding normal areas of CRA and CRC tissues (Fig. 6A and 6C). Also in CRA tissues CXCL1 revealed only apical weak to medium staining of epithelial cells (Fig. 6B). In contrast, intense apical and basal CXCL1 staining signals were found in epithelial cells of CRC tissues and to a lesser extent also positive signals were observed in mesenchymal cells of the CRC tissues (Fig. 6D). In tumor neighbor tissues of CRLM we observed medium CXCL1 staining intensities of fine granular texture in hepatocytes (Fig. 6E) while Glisson's triangles were strictly negative (left upper corner of Fig. 6E). In contrast, CRLM specimens displayed intense cytoplasmic CXCL1 staining signals in ubiquitous distribution (Fig 6F). Interestingly, both colorectal tumor tissues and hepatic metastatic tumor tissues showed apical and basal distribution of CXCL1 signals, whereas CRA tissues and unaffected corresponding neighbor tissues of CRC and CRA displayed only apical CXCL1 signals. Also for CXCL5 we detected only very weak staining intensities in the unaffected corresponding neighbor tissues of CRC and CRA restricted to some insular cytoplasmic signals in basal crypt cells (Fig.7A and 7C). Likewise, no substantial immunostaining was detected within CRA tissue specimens for CXCL5 with the exception of very insular positive signals in mesenchymal or goblet cells (Fig. 7B). In contrast, we observed intense CXCL5 staining within CRC specimens mainly concentrated in epithelial cells and only randomly interspersed positive signals in mesenchymal cells (Fig. 7D). These findings indicate that the high CXCL5 expression detected in our mRNA and protein analysis is attributed to positive signals of tumor cells rather than infiltrating immune cells.

In tumor neighbor tissues of CRLM we detected no substantial CXCL5 reactivity. Only insular dotlike, cytoplasmic signals were randomly interspersed in some hepatocytes (Fig. 7E). In contrast, we observed intensive staining signals in epithelial and mesenchymal cells within tumor areas of CRLM specimens (Fig. 7F). Well confined from the positive signals in the tumor areas we also observed unspecific staining of bile pigments (left upper image border in Fig. 7F).

While we observed no significant up- or down regulation of CXCR2 mRNA expression in CRA and CRLM tissue specimens, we found significant CXCR2 up-regulation in CRC tissues (P < 0.05) (Fig. 8).