The Disparate Twins: A Comparative Study of CXCR4 and CXCR7 in SDF-1α–Induced Gene Expression, Invasion and Chemosensitivity of Colon Cancer

This article is featured in Highlights of This Issue, p. 523

The members of the human chemokine superfamily build up a network contributing to cancer metastasis, survival, and proliferation. During the last years, scientific interest has contributed to the discovery of a unique interaction of stromal cell-derived factor-1 (SDF-1; CXCL12) with CXC chemokine receptor 4 (CXCR4), because this ligand–receptor interaction accomplishes pivotal roles for signaling in normal tissue, but also in cancer cells (1, 2). CXCR4 is the most commonly found chemokine receptor on cells of various tumors (1, 3), including colorectal cancers (4). High expression of CXCR4 was observed in primary tumors of patients with colorectal cancer (stage IV) and correlated with a significant reduction of median overall survival due to liver metastasis (5, 6). A gradient of SDF-1 established by secretion from cells of lymph nodes, liver, and lung into the blood flow serves as an attractant for CXCR4-expressing metastatic colorectal cancer cells that have detached from the primary tumor (1, 7, 8).

Recently, it has been shown that SDF-1 also binds to CXCR7, another member of the CXC chemokine receptor family (9, 10). This interaction was shown to induce cellular adhesion, proliferation, and survival in vitro and in vivo (10, 11). In contrast, CXCR7 signaling induced by binding of the ligand I-TAC (interferon-inducible T-cell alpha chemoattractant; CXCL11) resulted in ambiguous observations concerning its influence on migration (9, 10). In humans, CXCR7 is expressed in embryonic and adult tissue as well as on tumor cells (10, 12, 13). Elevated levels of CXCR7 were observed in aggressive colon (14) and colorectal cancer brain metastases (15).

Activation of CXCR4 via SDF-1α was shown to trigger intercellular pathways, for example, phospholipase C-β and phosphoinositide 3-kinase (16). The CXCR4/SDF-1 axis was described to regulate various signaling pathways controlling tumor cell proliferation and survival, angiogenesis, or metastasis in a multitude of leukemias and solid tumor entities (16). Thus, inhibition of this ligand–receptor interaction has emerged as a promising target in tumor treatment. Analysis of the signaling responses after binding of SDF-1α to CXCR7 revealed an induction of transendothelial migration and adhesion of human tumor and nonmalignant cells (11, 17).

However, very little is known about the consequence of the binding of SDF-1α to either CXCR4 or CXCR7 on gene expression, and about the resulting different function of cells toward invasion or therapy response. An elucidation of differential mechanisms and particular molecular targets following CXCR4 or CXCR7 stimulation could help to improve existing, or suggest new, regimens for the treatment of colorectal tumors. Thus, we performed microarray expression analyses followed by mechanistic analyses to shed light on common or opposed signaling cascades initiated by CXCR4 and CXCR7 downstream of SDF-1α.

Tumor and corresponding normal tissue samples of untreated patients with colorectal cancer were obtained from the tumor bank of the Mannheim Medical Faculty, University of Heidelberg. All patients had given informed consent before surgical intervention. The ethics committee of the Medical Faculty Mannheim of the University of Heidelberg approved the sampling. Sample sections were prepared by pathologists before snap freezing and subsequent storage in liquid nitrogen. Tissue sections were cut into 1 to 5 μm sections in a cryotome precooled to −25°C.

The colon cancer cell lines SW480 and HT-29, the human fibrosarcoma cell line HT-1080, the human embryonic kidney cell line 293T, and the mouse FBMD-1 feeder cells were cultured as previously described (18).

Methods are described in Supplementary Material S1.

Methods are described in Supplementary Material S2.

All SW480 cell lines were transfected with siRNAs either against AXL or AKR1C3, or scrambled control (Life Technologies) using Lipofectamine 2000 (Life Technologies). Transfection efficiency reached >70%. Knockdown was confirmed by quantitative real-time PCR (qPCR). Functional assays were performed 48-hour posttransfection.

Total RNA or microRNA (miRNA) was extracted and purified from snap frozen colorectal cancer tissues and cell lines (2 × 10⁶ cells/entity) with the RNeasy or the miRNeasy Mini Kit (Qiagen), respectively, according to the manufacturer's instructions. RNA concentration was determined with a Nanodrop ND-1000 Spectrophotometer (Peqlab). RNA quality was ascertained with the Experion electrophoresis station (Bio-Rad).

Reverse transcription of total RNA was carried out with M-MuLV Reverse Transcription Reagents (Fermentas). Reverse transcription of miRNA was performed with the miScript reverse transcription Kit (Qiagen).

For determination of gene expression, qPCR was performed with 2.5 ng cDNA or 5 ng reversely transcribed miRNA (rt-miRNA) per sample. TATA box binding protein (TBP) was used as reference for cDNA normalization. RNU6B, miR-346 and miR-491 were not regulated in the analyzed cell entities and used as references for rt-miRNA normalization. All qPCR reactions were performed with the QuantiTect or miScript Primer Assays and the SYBR Green PCR Kit (Qiagen) and carried out in triplicates using a Light Cycler 480 instrument (Roche) with the following thermocycling conditions: 95°C for 15 minutes, 40 cycles of 15 seconds at 95°C, 30 seconds at 55°C and 30 seconds at 72°C (cDNA) or 70°C (rt-miRNA).

Cells (10⁴) were seeded in growth medium supplemented with SDF-1α, I-TAC (100 ng/mL each) or with 6 μmol/L of the AKR1C3 inhibitor [AI, 3-(4-(trifluoromethyl)phenylamino)benzoic acid; Calbiochem] in 96-well plates. On the next day, medium was replaced by medium containing the chemotherapeutic drug, 5-fluorouracil (5-FU; Sigma-Aldrich), in increasing concentrations (25–16,000 μg/mL) ± 100 ng/mL SDF-1α or I-TAC or increasing concentrations (100–4,000 ng/mL) of the chemokines and 6 μmol/L AI. Following incubation for 48 hours at 37°C, cytotoxicity was analyzed by MTT assay (18). Half-maximal inhibitory concentration (IC₅₀) values were determined by the CalcuSyn Software (Biosoft) as already described (18).

A 96-well plate was coated with PBS/0.1% gelatine solution, and a feeder layer of FBMD-1 cells was established as described (18). SW480 cells were incubated for 1 hour at 4°C in RPMI medium. Subsequently, cell suspensions were added to the pre-established feeder layer and allowed to migrate for 24 hours at 37°C. Thereafter, a mixture of RPMI and FBMD1 medium ±5-FU in increasing concentrations (4–16 mg/mL) was added to the respective wells with PBS as negative control. After further 48 hours of incubation, all cells were harvested and analyzed by FACS. SW480 cells were stained with anti-human-MIC A/B (1:10; eBioscience) for 30 minutes at 4°C to discriminate from murine FBMD-1 cells. Dead cells were excluded using propidium iodide (1 mg/mL; Sigma-Aldrich).

Cell invasion of plerixafor-pretreated, untreated or transfected SW480 cells toward increasing concentrations of human recombinant SDF-1α or I-TAC (25–400 ng/mL; PeproTech GmbH) was examined using Matrigel (BD Biosciences) as described before (19).

SW480wildtype, SW480CXCR4, SW480CXCR7, and SW480CXCR4/CXCR7 cells were exposed to 100 ng/mL SDF-1α (conditions as described in 20, 21) for 3 and 12 hours. After RNA extraction, microarray analyses on Illumina Human Sentrix-8 BeadChip arrays (Illumina, Inc.) were performed at the Genomics and Proteomics Core Facility at DKFZ, Heidelberg.

Transcriptional profiling was done using Illumina Chip SentrixV3. Microarray data were reported to GEO (http://www.ncbi.nlm.nih.gov/geo/), accession number GSE40017. Microarrays were annotated using an Entrezgene based customer CDF-file and raw signals were quantile normalized. The mixed model ANOVA for differential gene expression, correlation, and unsupervised hierarchical cluster analysis was performed with the JMP Genomics (SAS) software package (version 4.0).

Genes were only considered significant if they were differentially regulated during SDF-1α treatment (>2-fold compared with unstimulated SW480wildtype cells), but not if SW480wildtype cells were differentially regulated during SDF-1α treatment (<1.1-fold).

Gene Set Enrichment Analysis (GSEA; version 3.0) was performed to discover possibly modified biologic pathways and miRNAs in treated cells. In brief, an mRNA set (named after a gene ontology term or a miRNA) is an a priori defined set of mRNAs, which are involved in a gene ontology pathway or considered to be regulated by a particular miRNA. All mRNAs of an mRNA set were ranked to a list according to differential gene expression, with the most upregulated genes on the top. A normalized enrichment score (NES) and the statistical significance (nominal P-value) were calculated (22). As a cutoff for each cell line, mRNA sets significantly deregulated in at least one time point as well as inversely regulated in both cell lines were chosen as miRNA targets. We have selected gene ontology terms based on following criteria: P > 0.5 in stimulated and unstimulated SW480wildtype cells, and P < 0.01 in all time points of, that is, SW480CXCR4.

Preparation of protein lysates, electrophoretic separation and blotting was previously explained (23). The membranes were probed with primary antibodies against EGFR and GAPDH overnight at 4°C, incubated with the appropriate horseradish peroxidase conjugated secondary antibody (α-rabbit and α-mouse) and visualized by chemiluminescence (ECL; Amersham Biosciences).

Formalin-fixed paraffin-embedded 3 μm sections of tumor tissue of five patients with advanced disease (stage III/IV, pT3/4, pN0/2, M0/1, and G2/3), who gave informed consent, were immunostained using a biotin–streptavidin–peroxidase method (VECTASTAIN Elite ABC Standard Kit; Vector Laboratories). The sections underwent deparaffinization by xylene, rehydration, and were then immersed in Target Retrieval Solution (DAKO), and boiled for 30 minutes. The sections were incubated for 5 minutes in 3% hydrogen peroxide in distilled water, washed in PBS and incubated with 10% goat serum and 1% bovine serum albumin in PBS for 30 minutes. Following avidin/biotin blocking for 15 minutes, respectively, the sections were incubated overnight at 4°C with 10 μg/mL primary anti-CXCR4 (MAB172, clone 44716, R&D Systems), anti-CXCR7 (MAB4227, clone 358426, R&D Systems), and anti-CXCL12/SDF-1 (MAB350, clone 79018, R&D Systems) antibodies. The sections were then incubated with 25 μg/mL of the appropriate biotin-labeled secondary antibody (BAF007, R&D Systems) for 60 minutes at room temperature and with ABC reagent for 30 minutes. The slides were stained for 5 minutes with Nova Red (Vector Laboratories) and then counterstained with Mayer's Hemalaun (Sigma-Aldrich), dehydrated and mounted in Eukitt.

The expression of CXCR4, CXCR7, or CXCL12 was blindly evaluated and semi-quantitatively assessed as an immunostaining score (0 negative; 1 weak; 2 moderate; 3 strong) in tumor and corresponding stroma cell populations, by an experienced pathologist (MJT), in parallel to a scientist of our group.

Quantitative real-time PCR data were compared using the two-sided unpaired Welsh t test for log-10 transformed values. Statistical analyses of transwell and chemosensitivity assays data were described previously (18, 19).

Gene expression of CXCR4, CXCR7, SDF-1α, I-TAC, AKR1C3, AXL, IGFBP7, RIN2, EFNB2, DUSP5, miR-217, and miR-218 was measured in tumor and normal tissues of patients. The association between gene expression and different pathophysiologic parameters (G, pN, and M) was examined using Kendall τ rank correlation analysis (KC; SAS version 9.2, SAS Institute Inc.). Correlation of gene expressions (gene expression versus gene expression), and gene expression versus pT was analyzed using the Pearson method (PC; SAS version 9.2; SAS Institute Inc.). A P ≤ 0.05 was considered statistically significant.

In previous studies, high expression of CXCR4 had been observed in primary tumors of stage III and IV in patients with colorectal cancer and correlated with reduced survival due to liver metastasis (5, 6). In contrast, few data are available elucidating the relevance of CXCR7 expression for colorectal cancer progression. Thus, we compared the expression of CXCR4, CXCR7, SDF-1α, and I-TAC in tumor samples with corresponding normal tissues isolated from 52 patients with colorectal cancer, who were not treated before surgery, either with colon (n = 36) or rectum (n = 16) carcinomas (Table 1).

CXCR4 and CXCR7 were expressed in all tumor samples with consistently higher absolute expression of CXCR4 (Fig. 1). Expression of both genes was elevated in tumors with stages III and IV (CXCR4: PC: 0.28; P = 0.046 and CXCR7: PC: 0.35; P = 0.010), increased with tumor size (pT; CXCR4: PC: 0.29, P = 0.039; CXCR7: PC: 0.26, P = 0.058), and lymph node involvement (pN; CXCR4: KC: 0.23, P = 0.037; CXCR7: KC: 0.28, P = 0.011). Additionally, expression of CXCR7 was significantly elevated in poorly differentiated tumor tissues (G; KC: 0.33, P = 0.006). Expression of I-TAC decreased with increasing tumors stages (PC: −0.27, P = 0,011). In samples of metastatic colorectal cancer, expression of I-TAC was reduced (PC: 0.3, P < 0,001) and negatively correlated with expression of CXCR7 (PC: −0.23, P = 0.001). In contrast, SDF-1α was expressed similar as CXCR4 (PC: 0.36, P < 0.009).

Immunohistochemistry was performed on representative tumor tissue samples of patients with advanced colorectal cancer, showing strong expression of CXCR4, CXCR7, and SDF-1α in tumor and moderate expression in some stroma cell populations (Fig. 2). CXCR4 showed a predominantly nuclear distribution. CXCR7 and SDF-1α staining was observed in the cytoplasm. A very interesting observation was that staining of SDF-1α could only be detected in tumor cells but not in stroma cells, hypothesizing that the ligand might be receptor-bound predominantly in tumor cells.

These results pose the question whether the interaction of the ligand SDF-1α with either CXCR4 or CXCR7 or the ligand I-TAC with CXCR7 results in differences, or similarities, of receptor-specific gene regulation, and thus in guidance of different, or similar, signaling pathways.

To answer this question, three stable chemokine receptor overexpressing SW480 cell lines (SW480CXCR4, SW480CXCR7, SW480CXCR4/CXCR7; the lentiviral generation of the cell lines is described in detail in S2) as well as untransduced SW480 cells were compared in microarray analysis following stimulation with the ligand SDF-1α for different periods (0, 3, and 12 hours). SW480 cells do neither express SDF-1α nor I-TAC as determined by qPCR (data not shown). Hierarchical clustering analysis of gene expression patterns resulted in two diverging groups: expression of some genes of SW480wildtype and SW480CXCR7 cells was either inverse as compared with SW480CXCR4 and SW480CXCR4/CXCR7 cells, or not different (Fig. 3A).

A total of 101 genes were differentially expressed in transduced cells compared with untransduced cells after stimulation with SDF-1α (P < 0.05). From this list, 38 (SW480CXCR4) and 34 (SW480CXCR4/CXCR7) genes, partly overlapping (>52%), had been differentially expressed in transduced cells compared with untransduced cells, already without any SDF-1α stimulus (0 hour). In SW480CXCR4 cells, 89 of the 101 genes were differentially expressed after stimulus with SDF-1α: 20 genes were downregulated and 69 genes were upregulated as compared with untreated SW480wildtype cells. In contrast, in SW480CXCR7 cells, 8 genes only were significantly differentially expressed after exposure to SDF-1α, compared with untreated SW480wildtype cells. Overexpression of both receptors simultaneously (SW480CXCR4/CXCR7 cells) and treatment with SDF-1α revealed that three genes were significantly less and 45 were stronger expressed than in untreated untransduced SW480wildtype cells: 41 of these 48 genes also showed a significant differential expression in SW480CXCR4 cells, but not in SW480CXCR7 cells.

To examine the potential biologic relevance of both chemokine receptors, differentially expressed genes were classified according to their biologic process, cellular component, and molecular function defined in gene ontology. In SW480CXCR4 cells, significantly deregulated genes were assigned to the following biologic processes: positive regulation of cell migration, and proliferation. In SW480CXCR7 cells, GSEA revealed “fatty acid, lipid, and cellular lipid metabolic processes” as the most significant biologic processes that were deregulated.

Seven candidates from the 101 genes significantly deregulated by SDF-1α stimulus in SW480CXCR4 or SW480CXCR7 cells and affiliated to the gene ontology terms described above were chosen for further validation by quantitative real-time PCR (qPCR): AXL (AXL receptor tyrosine kinase; Fig. 3B), C5 (complement component 5; Fig. 3C), RRAS [related RAS viral (r-ras) oncogene homolog; Fig. 3D], AKR1C3 (aldo-keto reductase family 1, member C3; Fig. 3E), TNNC1 (troponin C type 1; Fig. 3F), IGFBP7 (insulin-like growth factor binding protein 7; Fig. 3G), and IL24 (interleukin 24, Fig. 3H).

AXL, assigned to the biologic processes “positive regulation of cell migration and proliferation,” IGFBP7, and IL24 were highly expressed in SW480CXCR4 as well as in SW480CXCR4/CXCR7 cells. Transcription of AXL, C5, IGFBP7, and IL24 showed a significant upregulation in response to SDF-1α in both cell lines (Fig. 3B, G, H).

Expression of epidermal growth factor receptor (EGFR; Supplementary Fig. S3A), another candidate gene of the biologic processes “positive regulation of cell migration and proliferation” and of major interest in the context of colorectal cancer, was stimulated by SDF-1α in CXCR4-expressing cell lines (Supplementary Fig. S3A and S3B).

After exposure to SDF-1α, the only marginal upregulation of AXL, C5, and EGFR in SW480CXCR7 cells was similar to that observed in untransduced control cells (Fig. 3B, C, S3A). Similarly in all 4 cell lines, expression of RRAS was increased after exposure to SDF-1α (Fig. 3D).

AKR1C3 was chosen as a candidate gene for “fatty acid, lipid and cellular lipid metabolic processes.” The high expression and strong upregulation of AKR1C3 after stimulation with SDF-1α for 3 hours in SW480CXCR7 cells could be confirmed (Fig. 3E).

TNNC1 was oppositely expressed in SW480CXCR4 and SW480CXCR7 cells (Fig. 3F). In contrast to SW480CXCR4 cells where TNNC1 was continuously less expressed than in the control cells, stimulation with SDF-1α resulted in a strong increase of expression in SW480CXCR7 cells. TNNC1 and RRAS were the only genes with similar expression kinetics in SW480CXCR7 and SW480CXCR4/CXCR7 cells.

In conclusion, we were able to confirm the microarray results in SW480 cells for several selected genes by qPCR. Concomitantly, the same genes also showed higher transcription in HT-29 cells with high endogenous CXCR4 expression as compared with HT-29 cells with low CXCR4 expression (for cell line description see ref. 18, data not shown).

Because several differentially expressed genes (AXL and EGFR) were assigned to the biologic processes “positive regulation of cell migration and proliferation,” and are known to regulate cell migration and invasion, the chemotactic behavior of the transduced SW480 cell lines toward SDF-α was analyzed in transwell systems coated with Matrigel including the CXCR4 antagonist plerixafor (Fig. 4A). Although SDF-α was not expressed endogenously, treatment with SDF-1α significantly stimulated Matrigel invasion of SW480CXCR4, SW480CXCR4/CXCR7, and SW480CXCR7 cells (vs. the respective unstimulated cells; P < 0.05) in a dose-dependent manner (Supplementary Fig. S4A). However, the proportion of invading SW480CXCR7 cells remained at a level even below unstimulated untransduced cells. In SW480CXCR4/CXCR7 cells, signaling via CXCR4 obviously dominated the invasiveness toward SDF-1α. The addition of plerixafor reduced SDF-1α–induced invasiveness only in CXCR4-overexpressing cells (P = 0.003 for SW480CXCR4 and P < 0.001 for SW480CXCR4/CXCR7 vs. control cells), but had no effect on CXCR7-overexpressing cells.

Thus, SDF-1α stimulated invasion of CXCR4-expressing cells, whereas plerixafor had an anti-invasive effect. In contrast with CXCR4, CXCR7 expression might play only a minor role in tumor cell invasion.

Addressing the contradicting findings on I-TAC-mediated invasion, we analyzed the invasive capacity of the CXCR7-expressing cells in Matrigel assay. SW480 cells do not express I-TAC endogenously. I-TAC dose-dependently (Supplementary Fig. S4B) increased the invasive capacity of CXCR7-expressing cells (P < 0.03). As additionally observed for SDF-1α treatment, invasion toward I-TAC remained at a level even below unstimulated untransduced cells. No stimulation was seen in CXCR4/CXCR7-co-expressing cells, SW480CXCR4 cells or SW480wildtype cells.

These findings led to the conclusion that I-TAC exclusively stimulated invasion of CXCR7 cells to a similar exiguous amplitude as observed for the action of SDF-1α.

In concordance with results in thyroid cancer cells (24), our microarray expression analyses revealed a pronounced upregulation of AXL expression via SDF-1α/CXCR4 (Fig. 3B). We have previously shown that knockdown of CXCR4 reduced the invasive potential of SW480CXCR4 cells as well as expression of AXL (data not shown; ref. 19). In invasion assays, we showed that siRNA-mediated knockdown of AXL expression reduced the number of invaded cells in SDF-1α-treated CXCR4-expressing cells significantly, compared with scrambled control (P < 0.001; Fig. 4A).

In CXCR7-expressing cells, SDF-1α increased invasion and induced a transient upregulation of AKR1C3 expression (Fig. 3E). Using siRNA or a specific inhibitor of AKR1C3, AI, we observed that both resulted in reduced numbers of invaded cells in SDF-1α-treated CXCR4-expressing cells significantly, compared with scrambled control (P < 0.001; Fig. 4A). In contrast to AKR1C3 inhibition, knockdown of AKR1C3 expression successfully reduced the invasive potential of SW480CXCR7 cells (P = 0.016).

Taken together, we observed functional consequences for the gene regulation downstream of SDF-1α/CXCR4 and SDF-1α/CXCR7. AXL and AKR1C3 might be suggested as potential candidates in the invasion pathway.

To expand our observations in colon cancer cell lines that expression of AXL, AKR1C3, and IGFBP7 is regulated downstream of SDF-1α/CXCR4 or SDF-1α/CXCR7 (Fig. 3B, E, and G), we performed expression analyses to tumor and normal tissue of patients with colorectal cancer (Supplementary Fig. S5). Expression of AXL was reduced in tumors classified pT3/4 (PC: −0.18; P < 0.005). Expression of IGFBP7 increased with metastatic disease [N and M: (KC: 0.23; P < 0.001)] and with expression of CXCR4 (PC: 0.41; P < 0.001) and CXCR7 (PC: 0.33; P < 0.001) in tumor tissue. Expression of AKR1C3 highly correlated with expression of SDF-1α (PC: 0.83; P = 0.025) in tumors.

These results insinuate that expression of AKR1C3 in patients with colorectal cancer might be regulated via SDF-1α/CXCR7. Moreover, expression of AXL and IGFBP7 was observed in colorectal cancer tissue and regulation via SDF-1α/CXCR4 might be a negative prognostic factor for disease progression.

We have already shown that CXCR4-overexpression boosts chemosensitivity of colon carcinoma cells (18). Thus, the antidromic expression of genes in SW480CXCR4 and SW480CXCR7 cells according to cell survival (i.e., AKR1C3, C5, IL24, and IGFBP7) raised the question if these cells differ in their response to chemotherapy. We analyzed cytotoxicity in a stroma-dependent cell kill assay after treatment with 5-FU, which is a standard chemotherapeutic drug in colorectal cancer treatment. The stromal environment was mimicked by using the murine SDF-1α-producing stromal FBMD-1 cells in a coculture with SW480CXCR4, SW480CXCR7, SW480CXCR4/CXCR7, SW480wildtype or SW480EGFP (enhanced green fluorescent protein) cells (as a transduction control cell line). As expected, treatment with 5-FU (4, 8, and 16 mg/mL) increased cytotoxicity compared with untreated cells in all cell lines (Fig. 4B). Both control cell lines (SW480wildtype and SW480EGFP cells) showed similar chemosensitivity. However, cell toxicity was significantly increased in SW480CXCR4 cells compared with these control cells after exposure to 5-FU (SW480CXCR4 vs. SW480wildtype 8 mg/mL P = 0.029, 16 mg/mL P = 0.004; SW480CXCR4 vs. SW480EGFP 4 mg/mL P = 0.050, 8 mg/mL P = 0.030, 16 mg/mL P = 0.003). Chemosensitivity of SW480CXCR4/CXCR7 cells was less but also significantly increased (SW480CXCR4/CXCR7 vs. SW480wildtype 8 mg/mL P = 0.002, 16 mg/mL P = 0.007; SW480CXCR4/CXCR7 vs. SW480EGFP 4 mg/mL P = 0.030, 8 mg/mL P = 0.040, 16 mg/mL P = 0.047). In contrast, SW480CXCR7 cells showed increased chemoresistance under the same conditions (SW480CXCR7 vs. SW480wildtype 8 mg/mL P = 0.022, 16 mg/mL P = 0.014; SW480CXCR7 vs. SW480EGFP 8 mg/mL P = 0.034, 16 mg/mL P = 0.016) as compared with controls.

These results conclude that CXCR4-expressing tumor cells are more chemosensitive than CXCR7-expressing tumor cells in a SDF-1α-secreting environment.

Following the question if both, SDF-1α and I-TAC, are capable of hampering chemotherapeutic responses in colon cancer cells, we performed MTT assays (Table 2). Dose-response curves using increasing concentrations (100, 400, and 4,000 ng/mL) of SDF-1α showed an increase of chemosensitivity in CXCR4- and CXCR7-expressing cells toward 5-FU (SW480CXCR4 cells all concentrations, SW480CXCR4/CXCR7, SW480CXCR7, and SW480wildtype cells 100 and 400 ng/mL: P < 0.05) compared with untreated cells. In contrast, no changes in response to 5-FU in all established cell lines were observed in dose-response curves using increasing concentrations (100, 400, and 4,000 ng/mL) of I-TAC.

We conclude that SDF-1α contributes to the chemosensitivity of chemokine receptor-expressing colon cancer cells, whereas I-TAC has no effect.

After we have shown SDF-1α-dependent regulation of expression of AXL downstream of CXCR4, the chemoresponsiveness after siRNA-mediated knockdown of the enzyme was analyzed in cytotoxicity assays (Table 2). Compared with scrambled control cells, combination treatment with SDF-1α and 5-FU resulted in a significant increase in sensitivity of SW480CXCR4 cells lacking AXL (P < 0.001). This effect could not be observed in AXL-siRNA-transfected CXCR4/CXCR7 or wildtype cells.

Summarizing these results, AXL expressed via SDF-1α/CXCR4, participates in invasion processes as well as in tumor cell survival cascades.

Gene expression is partly regulated posttranscriptionally by miRNAs targeting the 3′UTR (untranslated region) of mRNAs, resulting in inhibition of translation or mRNA degradation (25). Our microarray data were applied to an in silico GSEA predicting that miR-217 and miR-218 might be involved in CXCR4 or CXCR7 signaling. In silico, 7 nt sequences in the mRNA of genes were used as target structures for prediction of specific miRNA binding (26). In GSEA, based on one recognition sequence, a set of mRNAs was identified putatively targeted by one specific miRNA.

GSEA predicted 95 mRNAs as targets for miR-217 and 339 mRNAs for miR-218. Both mRNA sets were less expressed in SW480CXCR4 as well as in SW480CXCR4/CXCR7 cells and highly expressed in SW480CXCR7 cells. In accordance with this, qPCR showed opposed expression for both miRNAs: The SDF-1α stimulus resulted in increased expression of miR-217 and miR-218 in SW480CXCR4 cells and in decreased expression in SW480CXCR4/CXCR7 cells. In SW480CXCR7 cells, SDF-1α signaling downregulated expression of both miRNAs after 3 hours, but stimulated expression after 12 hours (Supplementary Fig. S6A and S6B).

Expression of three genes, candidates from those in silico mRNA sets putatively targeted by miR-217 and miR-218, reflecting the opposing microarray expression profiles compared with their respective miRNAs, were also analyzed by qPCR and confirmed the expression determined by the microarray: RIN2 (Ras and Rab interactor 2; Supplementary Fig. S6C) as a predicted target of miR-217 and DUSP5 (dual specificity phosphatase 5; Supplementary Fig. S6D) and EFNB2 (ephrin-B2; Supplementary Fig. S6E) as putative targets of miR-218 were significantly less expressed in unstimulated SW480CXCR4 and SW480CXCR4/CXCR7 cells compared with SW480wildtype and SW480CXCR7 cells (Supplementary Fig. S6C–S6E). An SDF-1α stimulus had a marginal effect on RIN2 expression in transduced cells but upregulated its expression in SW480wildtype cells after 3 hours. The extreme downregulation of RIN2 expression in the transduced cells after 12 hours is perfectly consistent with the enhanced expression of miR-217 at the same time point (Supplementary Fig. S6C). Expression of EFNB2 and DUSP5 was enhanced in all 4 cell lines 3 hours after simulation with SDF-1α; however, this was less pronounced in CXCR4-overexpressing cells (Supplementary Fig. S6D and S6E). These gene expression kinetics in SW480CXCR4 and SW480CXCR7 cells were well correlated with the opposite kinetics for miR-218 expression.

Thus, signaling by SDF-1α via CXCR4 and CXCR7 inversely regulated expression of miR-217 and miR-218, and their putative targets in colon cancer cells.

Expression of miR-217 and miR-218 and their hypothesized targets RIN2, EFNB2, and DUSP5 were also analyzed in tumor and corresponding normal tissues of the 52 patients with colorectal cancer (Supplementary Fig. S7). In primary tumor tissues from patients suffering from metastatic disease, expression of miR-217 was significantly reduced (KC: −0.25; P = 0.046), whereas expression of miR-217 and miR-218 showed a trend to increase with pT (PC: 0.24; P < 0.114). We observed elevated expression of RIN2, EFNB2, and DUSP5 in colorectal tumor tissues of stage III and IV samples compared with lower stages (RIN: PC: 0.27; P = 0.054; EFNB2: PC: 0.24; P = 0.081; DUSP5: PC: 0.37; P = 0.008). Additionally, miRNA target expression was significantly correlated with pathologic features characterizing tumor progression [RIN2: N: (KC: 0.31; P = 0.005), M: (KC: 0.21; P = 0.070); EFNB2: N: (KC: 0.20; P = 0.072); DUSP5: N: (KC: 0.25; P = 0.027)].

High RIN2 expression was observed in rectal tumors invading through the muscularis propria into the subserosa (T3) or directly into other organs (T4; PC: 0.44; P = 0.092), as also seen in T3 and T4 colorectal cancer tumors. Moreover, RIN2 in trend was inversely expressed compared with miR-217 expression in rectal tumor tissue (PC: −0.41; P = 0.1322). DUSP5 expression correlates to expression of CXCR4 (PC: 0.27; P = 0.103) and CXCR7 (PC: 0.33; P = 0.003).

These data are certainly preliminary, but nevertheless, our findings in cell lines and correlations with aggressive stages of tumor tissue indicate that miR-217 and miR-218 and their targets should be followed as interesting putative CXCR4 or CXCR7 regulated players in further mechanistic studies.

Colorectal cancer is the most common malignancy of the gastrointestinal tract. The prognosis significantly declines for patients with tumors in high stages spreading to lymph nodes or distant sites. Elevated expression of the 2 chemokine receptors CXCR4 and CXCR7 in renal and cervical cancer was shown to predict reduced survival (27, 28). Our current analysis of tumor tissue from 52 patients with colorectal cancer also revealed increased transcription of CXCR4 and of CXCR7 in colorectal tumors from stages III and IV. Furthermore immunohistochemistry with tumor samples of five representative patients with advanced disease (stage III/IV, pT3/4, pN0/2, M0/1, and G2/3) revealed that CXCR4 was localized predominantly in the nucleus of tumor tissue, whereas CXCR7 was observed in the tumor cell cytoplasm. The nuclear localization of CXCR4 is in agreement with previously published results (29), probably enabled by a nuclear localization motif in the polypeptide (30). We hypothesize that the nuclear appearance of CXCR4 serves as a concealment to survive chemotherapeutic regimens. In our findings, where expression of CXCR4 and SDF-1α increased with disease progression, and SDF-1α staining was restricted to the tumor cells and not localized in the surrounding stroma, might suggest that only in tumor cells, SDF1 signaling can occur and initiate metastasis. Therefore, we suggest CXCR4 and CXCR7 as putative markers for poor prognosis of patients with colorectal cancer to investigate in further independent studies.

Because both receptors possess SDF-1α as a common ligand, we were interested whether the interaction of SDF-1α with either CXCR4 or CXCR7 might result in activation of similar or different signaling pathways and therefore gene expression in colorectal cancer cells.

For a detailed microarray expression analysis of differentially expressed genes in response to signaling via SDF-1α/CXCR4 or SDF-1α/CXCR7, we used lentiviral vectors to establish pools of SW480 colon cancer cells overexpressing CXCR4 and CXCR7, individually or combined.

Already without stimulus by SDF-1α, various genes were differentially expressed in SW480CXCR4 cells. The lentiviral transduction per se as a cause for this effect is highly unlikely, since then these expression changes should have also occurred in all of the other transduced cell lines. More convincing as an explanation is a ligand-independent signaling following multimerization of CXCR4 receptors that were highly expressed in SW480CXCR4 cells (19). Such a ligand-independent signaling had been described for multimerization of CXCR4 (31).

SDF-1α exposure of SW480CXCR4 cells resulted in upregulation of most genes that were differentially expressed. Compared with this, the set of differentially expressed genes after stimulation of SW480CXCR7 and SW480wildtype cells with SDF-1α was completely different and partly even inverse. Differential expression in SW480CXCR4/CXCR7 cells was similar to SW480CXCR4 cells, indicating that CXCR4 signaling dominates CXCR7 signaling. Only very few genes (e.g., TNNC1, RRAS) showed an SDF-1α-dependent upregulation simultaneously in SW480CXCR7 and SW480CXCR4/CXCR7 cells, with a contrary weak expression of TNNC1 or a similar increased expression of RRAS in SW480CXCR4 cells.

GSEA supposed that miR-217 and miR-218 might be functional for control of mRNA translation after SDF-1α exposure. The mRNAs of RIN2 or EFNB2 and DUSP5 were identified as targets for miR-217 or miR-218, respectively. It was shown previously that both miRNAs were expressed in solid tumors affecting tumorigenesis (32, 33). In agreement with findings in pancreatic carcinoma tissues (33), we showed that expression of miR-217 in primary colorectal tumors of patients with metastatic disease was decreased. Furthermore, expression analyses in the 16 rectal tumor samples confirmed an inverse expression of miR-217 and its putative target RIN2. As a player in vesicular trafficking, RIN2 might activate RAB5B by guanine nucleotide exchange and contribute to invasiveness. However, its function and this interaction have still to be emerged.

GSEA of all differentially expressed genes revealed a strong SDF-1α dependent upregulation of genes involved in cell migration, invasion and metastasis (i.e., AXL, C5, and EGFR) in CXCR4-overexpressing SW480 cells, and a significant upregulation of genes involved in lipid and fatty acid metabolism (i.e., AKR1C3) in SW480CXCR7 cells. We showed that lentiviral knockdown of CXCR4 overexpression (19) resulted in reduced expression of AXL and EGFR even compared with unstimulated untransduced SW480 cells, and that knockdown of both CXCR4 and AXL independently reduced invasion in colon cancer cells. In agreement with our results in colon cancer cells, AXL was identified as a transcriptional target of SDF-1 in CXCR4-expressing thyroid cancer cells (24), and has been associated with solid cancer metastasis (23, 34, 35). A synergistic effect of EGFR and CXCR4 has been observed in metastatic disease (36), whereas C5 triggered cell migration (37).

AKR1C3 was shown to be required for the synthesis of steroid hormones and prostaglandins as well as for detoxification of lipid peroxidation products (38). Because lipid metabolism plays a role in chemosensitivity of various cancer cell lines, we assumed that AKR1C3 might be a potential survival factor upon SDF-1α stimulus in SW480CXCR7 cells (39). In our microarray, expression of AKR1C3 was upregulated after SDF-1α treatment in SW480CXCR7 and SW480CXCR4/CXCR7 cells. Thus it was hypothesized, that the enzyme might modify 5-FU metabolism (40), for example, by activating of protein kinase C, and alter control of DNA repair and apoptosis, or by detoxifying intracellular reactive oxygen species produced by certain anticancer drugs. Overexpression of AKR1C1, sharing more than 80 % amino acid sequence similarity with AKR1C3, was detected in chemoresistant solid tumors (40). However, knockdown of AKR1C3 expression or chemical inhibition failed to hamper chemosensitivity in response to SDF-1α in CXCR7-expressing cells. Thus, AKR1C3 might not contribute to survival of colon cancer cells toward chemotherapy.

Furthermore, the increased chemosensitivity of CXCR4-expressing cells against 5-FU (18) was not caused by upregulation of AXL gene expression after stimulation with SDF-1α, because an siRNA-mediated knockdown of AXL additively increased chemosensitivity of SW480CXCR4 cells. Thus, these results are in agreement with observations in lung cancer (41), showing that AXL confers chemoresistance. We conclude that due to the dominating CXCR4 signaling, colon cancer cells expressing CXCR4 and CXCR7 might be sensitive against 5-FU.

Several in vivo models have independently shown that CXCR4 is a key player in colorectal cancer metastasis (42–44). In contrast, the attempts confirming this pathologic function also for the CXCR7 receptor remain contradictory (9, 10, 45, 46). Binding of either SDF-1α or I-TAC to CXCR7 might result in differing signaling cascades. In our experiments, the addition of SDF-1α or I-TAC to the medium resulted in minimal but significantly enhanced invasive capacity of SW480CXCR7 cells. Nevertheless, these cells were only weakly invasive compared with moderately invasive SW480CXCR4/CXCR7 cells and highly invasive SW480CXCR4 cells.

Simultaneous expression of both receptors was shown to result in homo- and in heterodimers (47). Therefore, in SW480CXCR4/CXCR7 cells, SDF-1α could be scavenged by CXCR7 (Supplementary Fig. S8) due to its higher affinity to this receptor, in consequence, CXCR4 signaling would be weakened. Thus, the intermediate invasive behavior of SW480CXCR4/CXCR7 cells compared with cells expressing either CXCR4 or CXCR7 can be explained.

In contrast to both cell entities overexpressing CXCR4 (19), there was no inhibitory effect of the CXCR4-inhibitor plerixafor on SDF-1α-mediated invasiveness of SW480CXCR7 cells. This can be explained by the inability of plerixafor to block the SDF-1α binding site at the CXCR7 chemokine receptor (Supplementary Fig. S9), as demonstrated for the binding site at CXCR4 in vitro (19, 48) and in vivo (7). In conclusion, CXCR7 probably contributes to progression and chemoresistance to 5-FU but only marginally to the invasive capacity of colorectal cancer cells.

In patients with colorectal cancer, we observed a decrease in expression of I-TAC with increased disease status. Our findings were in line with integrative analyses on gene expression signatures in colorectal cancer performed by the Cancer Genome Atlas Network proposing I-TAC as a marker for less-aggressive disease (49). Moreover, expressions of I-TAC and CXCR7 were negatively correlated in our set of patients with colorectal cancer.

In summary, our results obtained in CXCR4- and CXCR7-overexpressing cells showed which cellular pathways might be activated by each receptor. Our results with SW480CXCR4/CXCR7 cells (where CXCR4 signaling dominates CXCR7 signaling) are probably mostly relevant for translation to an in vivo situation. The disparate behavior of colorectal cancer cells after either CXCR4 or CXCR7 signaling demands further mechanistic studies, and an unexpected attention in the clinical application of chemokine receptor antagonists.

Jonathan Sleeman is a group leader in KIT Karlsruhe. No potential conflicts of interest were disclosed by the other authors.

Conception and design: D. Heckmann, P. Maier, S. Laufs, S. Fruehauf, H. Allgayer

Development of methodology: D. Heckmann, P. Maier, J. Sleeman, S. Fruehauf, H. Allgayer

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Heckmann, J. Sleeman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Heckmann, P. Maier, S. Laufs, L. Li, F. Wenz, H. Allgayer

Writing, review, and/or revision of the manuscript: D. Heckmann, P. Maier, L. Li, F. Wenz, S. Fruehauf, H. Allgayer

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. Heckmann, M.J. Trunk, F. Wenz, W. Jens Zeller, H. Allgayer

Study supervision: D. Heckmann, J. Sleeman, F. Wenz, S. Fruehauf, H. Allgayer

The authors thank Hans-Jürgen Engel, Annette Gruber, and Philipp Münzer for technical assistance.

This work was supported in part by grant M39.1 of the H.W. & J. Hector Foundation, grant 10-2089-FI 1 of Deutsche Krebshilfe/Dr. Mildred-Scheel-Stiftung, and grant 0315-452-C of the Federal Ministry of Education and Research (BMBF; to S. Fruehauf, H. Allgayer, S. Laufs). H. Allgayer, in addition, was supported by the Alfried Krupp von Bohlen und Halbach Foundation (Award for Young Full Professors; Essen, Germany), Hella Bühler Foundation (Heidelberg, Germany), Dr. Ingrid zu Solms Foundation (Frankfurt/Main, Germany), the FRONTIER Excellence Initiative of the University of Heidelberg, the Walter Schulz Foundation (Munich, Germany), Deutsche Krebshilfe (Bonn, Germany), the German–Israeli Project Cooperation DKFZ-MOST (Ministry of Science and Technology), the Wilhelm-Sander Foundation (Munich, Germany), and two grants of the HIPO/POP-Initiative, DKFZ Heidelberg.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.