Background: Eph receptors and their ligands, the ephrins, represent a large class of cell–cell communication molecules with well-defined developmental functions. Their role in healthy adult tissues and in human disease is still largely unknown, although diverse roles in carcinogenesis have been postulated.
Methods: We established a set of fluorescent PCR probes and primers for the definition of individual gene expression profiles of 12 different Eph receptors and 8 ephrins in 13 different healthy tissues. The mRNA expression profiles were studied in human lung, colorectal, kidney, liver, and brain cancers.
Results: The family of Eph receptors/ephrins was widely expressed in adult tissues with organ-site-specific patterns: EphB6 was highest in the thymus, compatible with an involvement in T-cell maturation. Brain and testis shared a unique pattern with EphA6, EphA8, and EphB1 being the most prominent. EphA7 had a high abundance in the kidney vasculature. Ephrin-A3 was up-regulated 26-fold in lung cancer, and EphB2 was up-regulated 9-fold in hepatocellular carcinoma. EphA8 was down-regulated in colon cancer, and EphA1/EphA8 was down-regulated in glioblastomas.
Conclusion: Eph/Ephrin genes are widely expressed in all adult organs with certain organ-site-specific patterns. Because their function in adult tissues remains unknown, further analysis of their role in disease may disclose new insights beyond their well-defined meaning in development.
The Eph receptors represent the largest family of receptor protein tyrosine kinases and interact with ligands called ephrins. The Eph receptors and ephrins are divided into the two subclasses, A and B, on the basis of their sequence homologies, structures, and binding affinities (1). Altogether, nine EphA, five EphB, five ephrin-A, and three ephrin-B members are currently known in humans. Ephrin-A ligands are tethered to the outer leaflet of the plasma membrane with a glycosyl-phosphatidyl-inositol anchor. In contrast, ephrin-B ligands have a transmembrane domain and a short cytoplasmic tail. Because both receptors and ligands are membrane-bound, a direct cell–cell contact is necessary for ligand binding and activation of the signaling cascades. As a unique feature, bidirectional signaling is initialized in both the receptor and the ligand-bearing cell (2). This bidirectional Eph/ephrin signaling between cells is fundamentally involved in developmental processes that require organized patterning and movement of cells, such as axonal guidance, patterning of hindbrain rhombomeres, and maintenance of cellular boundaries in the development of the central nervous system (3) or in the remodeling of blood vessels(4). Eph/ephrin signaling is also essential for the correct formation of crypts and villi in the intestinal epithelium (5).
More recently, the genes for Eph receptors and ephrins have been recognized to be differentially expressed in various human tumors, e.g., malignant melanoma, neuroblastoma, prostate cancer, breast cancer, small cell lung cancer, endometrial cancer, esophageal cancer, gastric cancer, and colorectal cancer (6)(7)(8)(9)(10)(11). Profound distortion of expression patterns could be correlated with altered tumor behavior, e.g., increased invasiveness, increased metastatic potential, and prominent vascularization, and consequently with poor patient outcome. Stimulation of EphA receptors was shown to cause morphologic changes in tumor cells, such as rounding and detachment (12); on the other hand, ligand signaling enhanced integrin-dependent attachment and migration potential (13). Hence, altered contact guidance and differential attachment may determine the expansion and metastatic spread of tumor cells. Because Eph receptors and ephrins are also major determinants of tumor angiogenesis in concert with vascular endothelial growth factor and angiopoietins (14), this field of the signalome attracts more and more attention in the context of cancer research. Furthermore, an important role in tissue maintenance and morphogenesis in regenerative processes can be assumed, because most of the Eph receptors and ephrins are also present in healthy adult tissues.
The medical interest in this family is further enhanced by recent reports on successful peptide, antibody, and antisense targeting of Eph receptors (15)(16). Novel structural data on Eph/ephrin complexes have opened the gate for computational structure-based screens of small molecule databases to find Eph agonists and antagonists (17). Although the importance of Eph signaling in many diseases, particularly in cancer, is apparent and some tools for therapeutic invention already exist, researchers and clinicians are hampered by the lack of complete data sets describing the tissue-specific expression of this large family of genes.
We therefore sought to systematically establish a gene expression survey of the individual members of this large family in various healthy human tissues. A TaqMan® probe/primer set was designed for real-time reverse transcription-PCR (RT-PCR) for the detection and quantification of gene expression of all currently known human members of the Eph receptor and ephrin family except for EphA5 and EphA10. For EphA7, EphB2, EphB4, EphB6, ephrin-B1, ephrin-B2, and ephrin-B3, immunohistochemistry was established to visualize some conclusive patterns of tissue-specific expression. Thirteen different healthy human tissues were analyzed. We also investigated squamous cell lung carcinoma, hepatocellular carcinoma, colorectal carcinoma, renal cell carcinoma, and glioblastoma and identified several new candidates that are differentially expressed in those tumors compared with their corresponding healthy tissues. These differentially expressed family members add new potential diagnostic and prognostic markers to our current knowledge and set the groundwork for further Eph/ephrin-targeted molecular cancer therapy.
Material and Methods
RNA was collected from 13 different healthy human tissues and 4 different human tumor entities. For each tissue, RNA was pooled from as many donors as possible to equalize potential interindividual differences. The numbers of individuals ranged from 2 (liver) to 47 (prostate) for healthy tissue samples and from 2 (lung and kidney) to 8 (glioblastoma) for tumors. Detailed sample information, including histopathologic classification of the tumors, is shown in Table 1⇓ . Healthy human tissue RNA was obtained from Clontech (BD Biosciences Clontech) except for colon and bladder RNA. Additionally, RNA from healthy lung, liver, kidney, and bladder tissue was obtained from Stratagene. Colon RNA was harvested from biopsies of healthy human individuals. For bladder RNA, healthy tissue was isolated from a cystectomy specimen.
Tumor RNA from lung, liver, colon, and kidney tumors was obtained from Ambion. The RNA of three additional colon carcinomas was harvested from tumors resected in our local Department of Surgery (University of Regensburg). Likewise, the RNA from six glioblastomas and two anaplastic astrocytomas was obtained from our local Department of Neurology (University of Regensburg). The RNA from three glioblastoma cell lines (LN 229, G 109, and 722) was kindly provided by Jürgen Schlegel (Institute of Pathology, LMU, Munich, Germany), the RNA of nine glioblastoma cell lines (243-28, U87-P41, U373-P253, HTZ17-P42, U118-P543, 411-12, HTZ402, HTZ408, and HTZ 410) was obtained from the Department of Neurology, University of Regensburg.
In brief, the specimens from healthy tissues and tumors were cut in small pieces immediately after resection and submersed in RLT buffer (Qiagen). After disruption with a mortar and pestle, samples were shredded in a Qiagen shredder column for homogenization, and RNA was isolated according to the protocol of the RNeasy Mini Kit (Qiagen). RNA quality and quantity were assessed with use of the Agilent 2100 bioanalyzer (Agilent Technologies); the RNA was then stored at −80 °C.
Human Reference RNA (Stratagene) was used to generate the calibration curves. For the calibration curves for ephrin-A2 and EphA8, Human Reference RNA was pooled 1:1 with RNA from fetal human brain (Clontech).
First-strand cDNA was synthesized with the Reverse Transcription Kit from Promega (Madison) according to the manufacturer’s protocol. To quantify the transcribed gene-specific RNA, we performed TaqMan real-time PCR (PE Applied Biosystems) on an ABI Prism 7900 HT Sequence Detection System as described elsewhere (18). The measured fluorescence signal of a given reaction showed a strong correlation with the number of copies of the amplified products. VIC-labeled eukaryotic 18S rRNA TaqMan predeveloped assays served as an endogenous control (Applied Biosystems). Probes and primers for TaqMan analysis were designed on the basis of gene-specific nonhomologous DNA sequences of the corresponding members. For four members, predeveloped commercial assays could be used. Details on the primers and probes are shown in Table 2⇓ .
relative quantification of gene expression
The calibration curve method was used to determine the relative amounts of gene expression products as described elsewhere (18)(19). In brief, serial dilutions of a stock cDNA were prepared to give concentrations of 50, 25, 12.5, 6.25, and 3.125 mg/L of RNA starting material. The calibration curves were constructed, and the amounts of input mRNA of the investigated genes were calculated by the TaqMan software. The reactions were performed in triplicate. The calculated amount of input mRNA was reported in nanograms. The values were normalized by dividing the amount of the investigated RNA by the amount of 18S rRNA. The normalized values therefore are unitless and can be used to compare the relative amount of target mRNA in different samples. When we compared a series of samples, one was designated as calibrator, and each normalized value was divided by this calibrator, giving values representing the x-fold amount of mRNA compared with this calibrator. This means that in healthy tissues, the tissue displaying the lowest expression of the investigated gene was set as the calibrator (Fig. 1⇓ ). For tumor samples, the corresponding healthy tissue served as the calibrator (Fig. 2⇓ ).
Immunohistochemistry was performed for EphA7, EphB2, EphB4, EphB6, ephrin-B1, ephrin-B2, and ephrin-B3. Sections (1.5 μm thick) of paraffin-embedded tissues were deparaffinized and rehydrated. The sections were incubated in citrate buffer (pH 6) at 90 °C for 40 min. After the slides were washed with H2O and phosphate-buffered saline, they were blocked with 20 mL/L H2O2 in CH3OH at 4 °C for 30 min. The slides were again washed with H2O and incubated with horse serum for 20 min to suppress nonspecific binding. The samples were then incubated overnight with the primary antibody dilution at 4 °C [EphB2, 1:75; EphB4, 1:75; EphB6, 1:150; EphA7, 1:100; ephrin-B2, 1:75 (Santa Cruz Biotechnology); ephrin-B1, 1:15; ephrin-B3, 1:25 (Zymed)]. After the slides were washed with H2O, they were incubated with a 1:50 dilution of biotinylated anti-goat secondary antibody for 30 min at room temperature. For staining, ABC Peroxidase complex (Linearis) and 3,3′-diaminobenzidine (Dako) were added. The reaction was stopped after 5 min with H2O. Slides for which the primary antibody was left out served as negative controls. The sections were evaluated by expert pathologists (T.V. and F.B.).
eph receptor and ephrin gene expression profile in healthy human tissue
Gene expression analysis of 12 Eph receptors and 8 ephrins was performed in 13 different human tissues, including brain, lung, liver, spleen, colon, small intestine, kidney, bladder, prostate, testis, uterus, thymus, and bone marrow. All currently known human Eph receptors and ephrins were investigated except EphA5 (assay design failed repeatedly) and EphA10 (sequence only recently reported) (20). TaqMan real-time RT-PCR technology was used to obtain a quantitative gene expression profile. The amount of gene-specific mRNA is displayed as the x-fold gene expression of each tissue compared with the calibrator tissue (i.e., the tissue with the lowest expression). To give a better idea of the relative expression, we classified the fold changes as 20th percentiles and translated them to a linear scale from 1 to 5 (0–20% = 1; 21–40% = 2, 41–60% = 3; 61–80% = 4; 81–100% = 5) and indicate them as dots in Table 3⇓ and Fig. 1B⇑ .
Negative mRNA expression was assumed if no fluorescence signal was detected after 45 cycles of PCR. Table 1⇑ in the Data Supplement that accompanies the online version of this article (http://www.clinchem.org/content/vol50/issue3/) gives the mean threshold (Ct) values of all investigated Eph/ephrin members and tissues, which enable estimation and comparison of the expression ranges for the different Eph/ephrin family members in any given tissue. Interestingly, except for EphA8 and ephrin-A2, all members of the family were expressed in all investigated healthy tissues, but the relative amounts of the transcripts varied considerably. Although some Eph receptors were rather widely expressed, we observed some remarkable restrictions: the EphA6, EphB1, and EphA8 genes were preferentially highly expressed in brain and testis. These tissues have very little in common, but they are both immunoprivileged organs. This raises the possibility that those receptors retain a certain antigenic potential. Interestingly, EphA6 gene expression decreases dramatically in colorectal cancer (see below). Another restriction phenomenon is the preferred expression of the EphB6 gene in brain and thymus. This confirms previous findings that EphB6 plays a pivotal role in T-cell maturation (21)(22) as well as T-cell signaling (23). Immunohistochemistry confirmed that the EphB6 gene is highly expressed in the adult human thymus, particularly in the cells at the margin of the follicles and in the Hasall bodies (Fig. 3F⇓ ). On the other hand, EphB2 seems to be the leading Eph receptor in the intestine and is produced primarily in the epithelium as shown by immunohistochemistry (Fig. 3C⇓ ). Recently in EphB2 −/− knockout mice, EphB2 was shown to be relevant for the maintenance of crypts and villus organization (5). High amounts of EphB1 were produced only in the colon, brain, and testis. EphA3 was highest in the uterus, bladder, and prostate. In contrast to the abovementioned restrictions, the EphB3 and EphB4, EphA1, EphA2, and EphA7 genes are widely expressed. As a rule, bone marrow showed the lowest Eph receptor and ephrin expression. The kidney had the highest amounts of EphA7 mRNA of the investigated tissues. Interestingly, immunohistochemical staining for EphA7 showed preferential production in the walls of blood vessels in the kidney and the liver (Fig. 3⇓ , D and E).
In analogy to the receptors the genes encoding the ligands fell into two categories, those that were widely expressed and those whose expression was more restricted. Ephrin-A3 and -B3 were preferentially expressed in the brain, whereas ephrin-A2 and -A4 expression was highest in the colon. In contrast, ephrin-A1, -A5, -B1, and -B2 were the most widely expressed family members. The high ephrin-B2 expression of the lung tissue was localized, particularly in the bronchial epithelium (Fig. 3B⇑ ). Similarly, in the colon ephrin-B2 (Fig. 3A⇑ ) and ephrin-B3 were preferentially expressed in the epithelial cells rather than in the stroma. The expression profiles of all Eph receptor and ephrin gene family members are shown in Table 3⇑ .
differential expression of eph receptors and ephrins in human cancers
Because Eph receptors and ephrins have been reported to be involved in the carcinogenesis of various human malignancies, we investigated five of the more frequently occurring tumor entities (lung, liver, colon, kidney, and brain) for differential expression of Eph/ephrin family members to identify new candidates that might serve as diagnostic and prognostic markers or as therapeutic targets. RNA from the tumors was pooled to minimize the interindividual variance (range, 2–8; see Table 1⇑ ). The amounts of the various family members expressed in the corresponding healthy tissues served as calibrators, and the amount of RNA in the tumors was calculated as a positive or negative fold change relative to the calibrator (Fig. 2⇑ and Table 4⇓ ). Fold changes >5 were set as the threshold for significance.
Ephrin-A3 mRNA expression was up-regulated 26-fold in squamous cell lung carcinoma. It is also interesting that ephrin-A2 mRNA was expressed in squamous cell lung carcinoma, but not in healthy lung tissue; therefore, no fold change could be calculated. EphA7 was up-regulated 25-fold in hepatocellular carcinomas compared with healthy liver tissue, and EphB2 was up-regulated 9-fold in this carcinoma. Immunohistochemistry for EphB2 confirmed the up-regulation of this molecule on the protein level for well-differentiated hepatocellular carcinomas (Fig. 3G⇑ ), whereas in poorly differentiated carcinomas (sometimes with adipose degeneration), EphB2 staining was not stronger than in the surrounding benign liver tissue. In the colon carcinomas, EphA6, -A7, and -B1 were down-regulated; EphA6 expression was 94-fold lower than in the healthy colon tissue. Interestingly, EphA8 mRNA expression was detected only in the colon tumors, and not in the healthy colon tissue. Thus this receptor, which is physiologically restricted to brain and testis, fulfills the criteria of a cancer/testis antigen gene and therefore deserves attention as a possible novel therapeutic target in colorectal cancer. Similar to colorectal carcinoma, in renal cell carcinomas EphA6 was down-regulated sixfold. Furthermore, investigation of the pooled RNA of eight glioblastoma biopsies revealed six candidates, which were more than fivefold differentially expressed compared with healthy brain tissue: EphA2, ephrin-A4, and EphB4 were up-regulated, whereas EphA1, EphA8, and ephrin-A5 were down-regulated (Fig. 4A⇓ ). To confirm our findings, we investigated the quantitative expression of these six members in 12 different glioblastoma cell lines (primary cell lines and established cell lines) and compared the results with expression in healthy brain tissue. As shown in Fig. 4B⇓ , differential expression of the six Eph/ephrin members in the pooled biopsies and the cell lines seemed to be fairly congruent. Interestingly, EphA1 and -A8 gene expression was completely lost in almost all investigated cell lines. Immunohistochemistry for EphB4 in glioblastoma displayed positively staining cells partly surrounding blood vessels (Fig. 3H⇑ ).
Despite the increasing knowledge about the functions of Eph receptors and ephrins regarding essential issues such as embryonic development (24), angiogenesis(25), immunology (23), cell motility(26), hemostasis (27), and cancer(28), to our knowledge a systematic expression profile of the whole family for various benign human tissues has not been published. Because TaqMan real-time RT-PCR is a precise method for quantitative measurement of mRNA with high sensitivity (only 50 ng of total RNA/well is necessary), high specificity (gene-specific probes), reproducibility, and the possibility of high-throughput analysis by use of 384-well assays (29), it is a method of choice for surveys of mRNA expression of whole gene families. We therefore used this method to determine the Eph receptor and ephrin ligand expression profiles of various benign human tissues and a selection of human tumors.
The most important finding of our study is that the family displays very ubiquitous expression in almost all investigated tissues, suggesting an essential role of the Eph receptors and ephrins not only in developmental processes, but also in the maintenance of organized adult tissues. Nevertheless, each tissue seems to have a specific inventory of Eph receptors and ephrins. Some members even show rather restrictive expression: e.g., EphB6 expression was highest in the thymus (111-fold compared with the calibrator tissue). This finding is compatible with previous studies that identified EphB6 as critically involved in T-cell signaling and embryonic thymus development in mice (22)(23)(30). Another peculiar profile was found in brain and testis, in which EphA6, EphA8, and EphB1 are relatively highly expressed. This seems to be quite unique among healthy human tissues. One possible but speculative interpretation is that the distribution of members in a certain organ is determined during differentiation of the main cellular components. Assuming large homolog functions of A- or B-class receptors, this could just be a random event, but considering the necessity of expressing this large variety in higher vertebrates vs few family members in, e.g., the nematode Caenorhabditis elegans (31), chances are that the individual family members are functionally diverse. In this context, our analysis can be helpful in disclosing more of the functional diversity of Eph/ephrin molecules on the basis of certain organ-specific expression patterns.
However, investigation of pooled tissues has some limitations. As our data show, expression is often compartmentalized in a certain cell type, which is not reflected by the isolation of RNA from whole organ samples. Furthermore, mRNA expression may not correlate with protein concentrations and signaling function because of posttranscriptional regulation mechanisms. To deal with this problem, immunohistochemistry can give further insights. For example, EphA7 mRNA was up-regulated 26-fold in hepatocellular carcinoma, but immunohistochemistry showed only weak expression of this protein in parenchymal liver cells in both the healthy tissue and the tumor. On the other hand, staining for EphA7 was typically strong and intense in blood vessels for both the liver and the kidney. Therefore, the “up-regulation” of EphA7 RNA in hepatocellular carcinomas may be attributable to higher vascularization in the investigated tumor. Expression might also depend on the histologic subtype of the cancer. Correlations with the histologic tumor type or with the outcome data of patients need to be further analyzed on the basis of large sample numbers, e.g., by use of tissue array technology (32).
Because the Eph/ephrin family is attracting more attention in cancer research, we used our TaqMan RT-PCR high-throughput assay to identify new differentially expressed members in five different tumor entities. To our knowledge, this is the first investigation of the Eph/ephrin expression profile in squamous cell lung carcinoma, hepatocellular carcinoma (except for EphA1, which was initially identified in a hepatocellular carcinoma cell line as the first receptor of this family), and renal cell carcinoma. A set of Eph/ephrin members for which expression was up- or down-regulated more than fivefold in carcinomas compared with healthy tissue were identified in this study (e.g., ephrin-A3 in lung carcinoma, EphB2 in hepatocellular carcinoma, and EphA6 in renal cell carcinoma). For colon carcinoma, only some members of the EphB/ephrin-B family have been investigated (33). Our data support the notion of up-regulated EphB2 expression in colorectal cancer, but in our study this was quite subtle (1.3-fold). Expression in glioblastomas has been reported only for EphA5 (34). We identified six members of the family that appear to be differentially expressed more than fivefold in this cancer entity.
Previously, mainly the up-regulation of selected members was described as typical for certain cancers (11)(28)(35). There are few reports describing down-regulation of Eph/ephrins in cancer (36). We identified several members that were strongly down-regulated in the tumors (e.g., EphA6 was 94-fold down-regulated in colorectal cancer). Furthermore, we have recently found a loss of EphB6 expression in malignant melanomas (37), whereas other receptors, such as EphA2, and ligands, such as ephrin-B2, are up-regulated (6)(7). This again raises the question of whether different receptors or ligands are functionally diverse although they are structural homologs and show no stringent binding specificity. Taken together, the concept that Eph receptors are “oncogenes” and therefore are usually up-regulated (28) requires a new look on the basis of our data. However, it must be considered that down-regulation of Eph/ephrin mRNA in tumors may also be attributable to a different distribution of cell types in the tumor tissue compared with the corresponding benign tissue. In that case, the down-regulation of mRNA would represent an epiphenomenon without any functional relevance regarding oncogenesis.
Differential expression of receptor tyrosine kinases and their ligands is of high interest in oncology because these molecules are assumed to be of functional importance in oncogenesis, mediating the enhanced proliferation, migration, and metastatic potential of tumor cells as well as tumor angiogenesis (6)(7)(12)(14)(38)(39). Some findings also indicate a role of Eph receptors in tumor immunology. In particular, the loss of expression of a certain receptor could be an indicator of the presence of T-cell clones that suppress tumor cell clones with an abnormal expression pattern (40). Consequently, the appearance of EphA8, which is usually restricted to the immunoprivileged organs brain and testis, in colorectal carcinoma might alert the immune system. It will be interesting to see whether the prognosis for patients with EphA8-positive tumors is different from the prognosis for those with EphA8-negative tumors.
Apart from their significance as prognostic markers (9), overexpressed Eph receptors such as EphB2 in hepatocellular carcinoma or ephrin-A2/A3 in squamous cell lung carcinoma may serve as therapeutic targets (15). Antagonistic ephrin mimetic peptides that bind selectively to Eph receptors can be used to antagonize Eph receptor-mediated pathways in overexpressing cancers. These peptides can also be used for selective drug delivery to Eph-(over)expressing cells, which is of great interest for cancer therapy. On the other hand, applications for benign processes such as support of nerve growth or epithelial regeneration can be envisioned because these molecules take part in tissue maintenance and (re)organization (16).
TaqMan-based surveys may be helpful to select potential candidates for such therapies and to predict possible effects on other tissues that express the targeted molecules. Taken together, our results provide the first systematic survey of the Eph receptor tyrosine kinase family and their ephrin ligands for various benign human tissues. High-throughput analysis using TaqMan technology is a suitable method for performing such surveys. Our results confirm recent findings that members of this family are frequently differentially expressed in cancer and add several new candidates that might represent both prognostic markers and therapeutic targets.
We thank Bettina Federhofer, Doris Gaag, Lydia Künzel, Sabine Troppmann, and Nadine Wandtke for excellent technical support. We also thank Christoph Moehle, Mario Probst, Jürgen Schlegel, and Matthias Woenckhaus for technical consultation, provision of cell lines, and histopathologic diagnostic advice. This work was supported in part by the Deutsche Forschungsgemeinschaft (SFB 585/A8) and by the Dr. H. Maurer-Grant.
- © 2004 The American Association for Clinical Chemistry