Genes upregulated in a metastasizing human colon carcinoma cell line

Abstract

Differential gene expression between the metastatic human colon cancer cell line HT29p and its nonmetastatic counterpart HT29-MTX was revealed by suppression subtractive hybridization. Fifty-eight individual genes showed increased mRNA levels in HT29p cells. Only 15 of these genes had been related to transformation in previous studies; the majority of genes are new candidates encoding proteins relevant for the metastatic process. Cancer profiling arrays as well as in situ hybridization study revealed that at least some of the genes obtained in the SSH screen are also differentially expressed in human tumors. © 2004 Wiley-Liss, Inc.

Cancer is the result of an accumulation of genetic changes. This is best exemplified for colon cancer for which about 5 to 7 genetic hits trigger the stepwise progression from normal epithelium via polyp formation to invasive tumor cells and finally metastatic, life threatening, malignant tumors.1 Identification of the genes affected during tumor progression was facilitated by the study of inherited diseases with high cancer risk, e.g., familial adenomatous polyposis (FAP) that allowed the identification of the APC gene, a component of the Wnt pathway,2 or hereditary nonpolyposis colon cancer (HNPCC) that identified the Mut S gene and defective mismatch repair to be important for colon cancer development.3 Genetic changes in cancer cells occur frequently in genes that encode components involved in signal transduction, e.g., APC or Ras, the central regulator of the MAP kinase pathway that is frequently mutated in all different kinds of tumors. As a result of such mutations, the genetic programs in cancer cells are changed resulting in up or down regulation of genes as compared to normal cells or when different stages of transformation are compared.

Of particular interest are those genes whose expression is changed in a metastatic cancer cell. Unfortunately, the majority of these genes is still unknown. For 2 reasons this gap should be filled: i) knowing the genes that are activated or repressed during tumor progression would allow a mechanistic understanding of metastasis-relevant processes and ii) relevant genes can lead to new therapeutic targets, an important goal since future cancer therapy will be based on more than 1 target as mono-therapies have shown major limitations.

Identification of genes upregulated in metastasizing tumor cell lines relies on comparisons of expression profiles. There are several ways to compare patterns of gene expression, including differential hybridization screening, representational difference analysis (RDA), differential display, serial analysis of gene expression (SAGE), suppression subtractive hybridization (SSH) and cDNA microarray technology. Each method has advantages. We made use of SSH that allows detection of low abundancy transcripts and compared 2 closely related colon carcinoma cell lines differing in their ability to metastasize. The identification of low abundancy transcripts is important since regulatory genes, e.g., metastasis promoting or suppressing genes, may belong to this class.

We have identified 100 differentially expressed cDNA clones that correspond to 58 individual genes. These genes encode proteins with different functions, e.g., structural proteins, proteins involved in signal transduction or transcription factors. Several of these genes were tested in human tumor material and showed statistically significant up regulation in primary tumors.

Material and methods

Cells

The human HT29p (parental cell line) colon carcinoma cell line and a methotrexate-resistant (10⁻⁵ M) variant (HT29-MTX) of this cell line was obtained from Dr. A. Zweibaum (INSERM Unit 178, Paris, France). The characteristics of these cell lines have been previously described.4, 5 Cells were routinely grown in Dulbecco's modified Eagle's medium (DMEM, Invitrogen, Karlsruhe, Germany) supplemented with 10% heat-inactivated fetal calf serum (PAA Laboratories, Cölbe, Germany).

Metastasis assay

Scid-beige mice (CB-17/IcrHsd-scid-bg) were purchased from Harlan, Borchen, Germany. Ten mice each were injected subcutaneously with 1×10⁶ of either HT29p or HT29-MTX cells. When the tumors at the site of injection had reached a diameter of 2 cm, the animals were sacrificed, and the primary tumor, draining lymph nodes and major organs were collected in 10% neutral buffered formalin. The tissues were embedded in paraffin and 6 μm sections were cut and stained with hematoxylin and eosin for microscopic analysis.

Construction of the SSH library and screening

The construction of the subtractive library was essentially based on the method described previously6 using the PCR-Select cDNA subtraction kit (Clontech, Heidelberg, Germany).

The resulting subtracted cDNA library was amplified and cloned in the PCRII.1 TA vector (Invitrogen, Karlsruhe, Germany); 2,200 clones were picked and amplified by colony PCR. Equal amounts of the PCR reactions were loaded on 2 identical high-density gels. The gels were transferred onto a nitrocellulose membrane and probed with radioactively labeled cDNAs prepared from the tester and the driver cell lines. This procedure is referred to as reverse Northern blots.

Computational analysis

The SSH sequences were analyzed using a BLASTN 2.1.3 sequence similarity search.7 The clones were compared to the sequences contained in public human databases, including all the nonredundant GenBank and EMBL sequences, and the GenBank EST sequences. Prior to the BLASTN analysis, we removed the flanking nested primers, keeping only the inserted sequence in order to avoid false BLASTN hits. These sequences were then compared to all others to identify and remove the redundant clones. The remaining 72 clones were compared to the public databases for the identification of the genes. Some of these sequences originated from different parts of the same gene reducing the number of individual genes finally to 58.

Northern blot analysis

Poly A+ RNAs were prepared according to classical methods from the tester (HT29p) and the driver cell line (HT29-MTX). Aliquots of 2 μg polyA⁺ RNA were loaded on a 1% formaldehyde-agarose gel and blotted overnight in 10×SSC onto hybond N⁺ membrane (Amersham, Freiburg, Germany). The filters were crosslinked using the U.V. stratalinker 2400 (Stratagene Europe, Amsterdam, Netherland). Two different methods were used for Northern blot hybridizations. In the classical radioactive method, cDNA probes were labeled with ³²P dCTP and hybridization was detected on X-ray films. In the nonradioactive procedure, antisense RNA probes were produced and labeled in an in vitro transcription assay using digoxigenin-dUTP. Subsequent handling followed the protocol recommended by Roche Diagnostics (Mannheim, Germany).

Cancer c-DNA microarrays

Cancer profiling arrays carrying 241 paired cDNA samples from individual patients were purchased from Clontech, Heidelberg, Germany). The membranes were pre-hybridized in ExpressHyb hybridization solution (Clontech). The cDNA probes were labeled with ³²P using the Rediprime Random Labeling system (Amersham, Freiburg, Germany). The probes were hybrized to the membrane in ExpressHyb solution at 68°C for 1 hr. After washing at 50°C, the positive signals were detected by exposure to X-ray film or by phosphoimager for quantification.

In situ hybridization

The BST2 cDNA was purchased from RZPD (Deutsches Resourcenzentrum für Genomforschung, Berlin, Germany), cloned in a bidirectional vector and transcribed in vitro using SP6 or T7 polymerase. The slides containing normal and cancer tissues were purchased from BioCat (Heidelberg, Germany). The in situ hybridization was performed according to standard hybridization protocols (Roche Diagnostics, Mannheim, Germany). Briefly, de-paraffinized sections were rinsed with TBS (50 mM Tris-HCl, pH 7.5, and 150 mM NaCl) and treated with 200 mM HCl followed by washings with TBS. Thereafter, the slides were incubated in 0.5% acetic-anhydride solution in 100 mM Tris-HCl, pH 8, for 10 min. After several washes in TBS, the sections were treated with proteinase K in TBS containing 2 mM CaCl₂ at 37°C for 20 min. The reaction was stopped with TBS at 4°C for 5 min. The samples were then dehydrated in graded ethanol and then briefly washed with chloroform. RNA probes were produced using the linearized vector and digoxigenin labeled UTP in an in vitro transcription assay. The probes were denatured and hybridized to the slides at 65°C overnight. After several washes, the slides were subjected to a colorimetric assay using an anti-digoxigenin antibody.

Results

HT29 cells with different metastatic potential

The HT29 colon carcinoma cell line has been extensively used to study proliferation and differentiation of intestinal cells. The so-called parental cell line (HT29p) corresponds to an undifferentiated non polarized population of cells growing in multiple layers.4 From this cell line several descendants were derived that are more or less differentiated. The HT29-MTX cells were selected from HT29p for their resistance to lethal doses of methotrexate.5 They form a homogenous population of polarized cells that are highly differentiated and produce gastric mucins.

To determine their metastatic potential, HT29p and HT29-MTX cells were injected subcutaneously into scid-beige mice. When tumors at the site of injection had reached 2 cm in diameter the animals were euthanized. The primary tumor, draining lymph nodes and major organs were collected for histological evaluation. The lymph nodes from animals injected with HT29p cells had large metastases (8 out of 8 animals, an inguinal lymph node is shown in Fig. 1a,b) involving in most cases at least half the node. Metastases were not detected in other organs. Lymph node metastases were not present in animals injected with HT29-MTX cells (Fig. 1c,d was taken from an axillary lymph node). Thus, the undifferentiated HT29p cells metastasize readily through the lymph nodes while HT29-MTX cells do not metastasize. The metastatic potential of the HT29p cells has also been proven in a different assay using immunosuppressed newborn rats (Dr. L. Remy, INSERM U45, Lyon, France, unpublished results).

Figure 1. Metastatic potential of HT29 cells. Scid-beige mice were injected with 1x10⁶ HT29p or HT29-MTX cells. When the tumors at the site of injection had reached a diameter of 2 cm the animals were killed and the draining lymph nodes were prepared for histological examination. A large metastasis (identified by the white outline) is present in the lymph node of a mouse injected with HT29p cells (a, ×4; b, ×10). There were no metastases in lymph nodes of mice injected with HT29-MTX cells (c, ×4; d, ×10).

A suppression subtractive hybridization (SSH) library of HT29p cells

To establish an SSH-library of cDNAs over-expressed specifically in the metastatic tumor cell line HT29p, cDNAs of the HT29-MTX cells (driver cells) were subtracted from the cDNAs of the HT29p cells (tester cells). The protocol of subtractive suppression hybridization and the isolation of differentially expressed clones has been described.6 Two thousand two hundred clones were picked from which 100 were truly differentially expressed in the tester cell line as determined by reverse Northern blotting. Since some of these clones were redundant or originated from different parts of the same gene, these 100 clones corresponded in fact to 58 individual genes as deduced by sequence analysis. Table I lists the different genes and classifies them according to the function of the corresponding proteins. The majority of proteins is involved in the maintenance of cell structure and cell interactions. Another interesting group comprises proteins involved in signal transduction and regulation of gene expression. Other proteins regulate proliferation and metabolism. Interestingly, 15 of these genes have already been implicated in transformation and/or metastasis formation, as indicated. Six genes (keratin-8, connexin 26, HMGB1, phospholipase A2, ribosomal protein S7 and annexin A1) were also identified as metastasis-specific in an SSH screen of pancreatic or mammary metastatic tumor cells versus their nonmetastatic counterparts.8

Table I. Metastasis-Specific Genes1

Functional group	Name	Gene symbol	Number of Hits	Notes
1 All genes identified by the SSH screen made on HT29p metastatic cells vs. HT29-MTX nonmetastatic cells are listed according to their functions. There exists an internet link for the gene symbol and the references cited in the note section.
Ribosomal proteins	Ribosomal protein S20	RPS20	2
	Ribosomal protein S26	RPS26	1
	Ribosomal protein S9	RPS9	1
	Ribosomal protein S7	RPS7	1	Associated with metastatic phenotype Nestl et al., 2001
	Ribosomal protein L13	RPL13	1	Expression in human breast tumors Adams et al., 1992
	Ribosomal protein L14	RPL14	1
	Ribosomal protein L39	RPL39	1
Metabolism	Glucuronidase β, (MPS7)	GUSB	1
	Phospho-glycerate-kinase (PGK-A)	PGK1	1
	Lactate dehydrogenase (LDH1)	LDHA	1
	Methylene tetrahydrofolate dehydrogenase	MTHFD2	1
	2′-5′oligoadenylate synthetase (p100)	OAS3	1
Cell cycle/ replication	Protein kinase, regulatory unit 1B (CDC28)	CKS1B	1
	Origin recognition complex, subunit 6-like	ORC6L	1
	Ribonukleotid reductase	RRM1	1	Tumor suppressive in lung tumors Gautam and Bepler, 2003
	Microtubule-associated protein, RP/EB family member 1 (EB1)	MAPRE1	1	Associates with APC Su et al., 1995
Signalling	Phosphotyrosine phosphatase, non transmembrane (PTP-PEST)	PTPN12	3	Aberrant transcripts in colon cancer Takekawa et al., 1994
	Neurotrophic tyrosine kinase, receptor, type 1	NTRK1	1	Overexpressed in ovarian carcinoma Davidson et al., 2001
	Phospholipase A2 (14-3-3 ζ, □□□□-1)	YWHAZ	7	Involved in spreading Zhu et al., 2003 associated with metastatic phenotype Nestl et al., 2001
	Inositol polyphosphate-5-phosphatase F (SAC2)	INPP5F	1
	Acid cluster protein 33, masparatin	ACP33	1
Structure/ cell-cell-interaction/ matrix interaction	Keratin 8	KRT8	1	Associated with metastatic phenotype Nestl et al., 2001
	GAP junction protein β2, connexin 26	GJB2	1	Associated with metastatic phenotype Nestl et al, 2001Bertucci et al., 2004
	Annexin A1, lipocortin I	ANXA1	1	Associated with metastatic phenotype Nestl et al., 2001
	FilaminA, actin binding protein 280	FLNA		Localized in focal adhesions and substrate of PKC Tigges et al., 2003
	FilaminB, actin binding protein 278	FLNB	1	Integrin binding van der Flier et al., 2002
	Elastin microfibril interfacer 3, multimerin 2, emilin like protein, EndoGlyx-1	MMRN2	1
	Syndecan binding protein (syntenin)	SDCBP	1	Promotes cell migration in metastatic breast tumors Koo et al., 2002
	Calcium and integrin binding 1 (calmyrin)	CIB1	1	Involved in platelet spreading Naik et al., 2003
	Glucosaminyl (N-acetyl) transferase 3, mucin type	GCNT3	2
	CDw92 antigen, choline transporter-like protein CTL1	CDW92	1
	β-2 microglobulin	B2M	1
	Amyloid beta (A4) precursor-like protein 2	APLP2	1
	Bone marrow stromal cell antigen 2	BST2	1	Overexpressed in myeloma Ohtomo T et al., 1999
	Retinoic acid receptor responder (tazarotene induced) 1	RARRES1	1
Transcription/ splicing	Polymerase (RNA) II (DNA directed) polypeptide F	POLR2F	1	Subunit of RNA polymerase II Acker, J et al., 1994
	TATA box binding protein (TBP)-associated factor, RNA polymerase II, H, 30kD	TAF10	1	Required for cell cycle progression Metzger D et al., 1999
	Ets homologous factor, (ESE-3)	EHF	3
	Transcription factor 7-like 2 (T-cell specific, HMG-box) TCF-4	TCF7L2	1	Associated with primary transformation in colorectal cancer Marchenko et al., 2002
	High-mobility group box 1, amphoterin	HMGB1	1	Abundantly expressed in breast carcinoma Brezniceanu et al., 2003, expressed in gastrointestinal tumors Choi et al., 2003, expressed in melanoma, Poser et al., 2003 associated with metastatic phenotype Nestl et al., 2001
	Splicing factor, arginine/serine-rich 3, (SRP20)	SFRS3	2
	Heterogeneous nuclear ribonucleoprotein R	HNRPR	1
	RNF6: RING finger protein 6	RNF6	1
Degradation/ folding	Proteasome (prosome, macropain) 26S subunit, ATPase, 5, MSUG1 protein	PSMC5	3
	Proteasome (prosome, macropain) subunit, alpha type, 7	PSMA7	1
	Protein disulfide isomerase, thioredoxin domain containing 7 (P5)	TXNDC7	1
Transport/ reorganization	Lysosomal associated transmembrane protein 4 beta, (LC27)	LAPTM4B	1	Upregulated in hepatocellular carcinoma Shao et al., 2003
	Golgin 67	GOLGIN-67	2
Function unclear	Interferon-induced, hepatitis C-associated microtubular aggregate protein, (MTAP44)	IFI44	2
	Interferon inducible protein 27	IFI27	1	Expressed in breast carcinoma Rasmussen et al., 1993
	Interferon alpha-inducible protein (clone IFI-6-16), interferon-induced protein 56	G1P3	12
	Interferon-induced protein with tetratricopeptide repeats 1	IFIT1	3
	Small acidic protein	SMAP	2
	Chromosome 1 open reading frame 29	C1orf29	3
	Similar to hypothetical protein MGC17347	LOC159090	2
	HN1 like, (C16orf34)	HN1L	1
	HSPC014	C13orf12	2	Overexpressed in promoted hepatocellular carcinomas Shibutani et al., 2002
	IMAGE:5302006	Hs.351680	1

Northern blot analysis

Thirteen genes were selected for a validation of enhanced expression in HT29p cells as compared to HT29-MTX cells by Northern blot analysis (Fig. 2). We chose all genes of unknown function (shown are only IFI44, IFI27, IFIT1, G1P3 or C13orf12) and compared them with others that had already been described to be related to metastatic processes (YWHAZ, PTPN12, KRT8, BST2 and LAPTM4B). PolyA⁺ RNA from HT29p and HT29-MTX cells were hybridized with the respective probes (Fig. 2). The RNA abundance of all sequences tested was elevated in the metastatic HT29p cell line in comparison to HT29-MTX cells confirming the results of the SSH-data and the reliability of the method. We can therefore extrapolate to the other genes and assume that they are also truly over-expressed in the HT29p metastatic cell line.

Figure 2. Northern blot analysis of HT29 mRNAs. Probes corresponding to 13 genes that showed upregulation of mRNA transcripts in the HT29p cells as compared to the HT29-MTX cells were used for Northern blot analysis (Material and methods). The blots were rehybridized with an actin or a GAPDH probe and revealed equal loading of the 2 cell lines RNAs (not shown).

Cancer array analysis

To prove the relevance of our library for human cancer, we tested some of the genes on cancer profiling arrays. These arrays are composed of cDNAs from human tumor material and from the corresponding normal tissue of the same patient. 241 matched cDNA pairs were spotted side by side on a nylon membrane. The cDNAs originated from 10 different cancer types and the respective normal tissues. As an example, the hybridization of cDNAs from colon cancer and normal colon tissue with 2 of the differentially expressed genes is shown (Fig. 3). The data of such hybridizations were converted into numbers by densitometric evaluation and were normalized to tubulin or actin expression (Table II). The numbers of patients with enhanced expression in their tumors (≥ 2-fold over that in normal tissue) are listed. The percentage relates these numbers to all cases examined. As Table II shows, genes upregulated in the HT29p cells were also upregulated in the majority of samples of primary human cancers of the digestive track (colon and rectum) of the genitary system and of the lung and breast. The genes tested were not upregulated in kidney cancer. For other cancers, we cannot derive a conclusion since the number of samples was too small.

Figure 3. Hybridization of cancer profiling arrays. Cancer profiling arrays [matched pairs of cDNA obtained from tumor material (T) and normal tissue (N) of the same patient] were hybridized with various probes that were upregulated in a metastasis-specific manner. An example of colon cancer (and normal tissue) from 19 different patients hybridized with 2 different probes is shown.

Table II. Evaluation of the Cancer Profiling Array Hybridization1

Tumor gene (no. of samples)	IFIT1	LAPTM4B	BST2	C13orf12	IFI27	G1P3	IFI44
1 Probes obtained from 7 genes that showed upregulation in the metastatic HT29p cells as compared to the HT29-MTX cells were hybridized to cancer profiling arrays (Material and methods); an example is shown in Figure 3. The hybridization was quantified using a phosphoimager and related to comparable hybridization between the matched pairs (tumor vs. normal tissue) to either a tubulin or an actin probe. When this normalization revealed an intensity more than 2-fold in the tumor sample the gene was considered upregulated. The numbers indicate the number of patients where this upregulation occurred. The percentage refers to the total number of patients examined for a particular tumor.
Breast (53)	8 (15%)	7 (13%)	16 (30%)	1 (2%)	14 (26%)	15 (28%)	4 (8%)
Uterus (44)	6 (14%)	26 (59%)	24 (55%)	3 (7%)	24 (55%)	17 (38%)	13 (30%)
Colon (39)	12 (31%)	22 (56%)	15 (38%)	10 (26%)	8 (21%)	19 (39%)	21 (54%)
Stomach (27)	6 (22%)	6 (22%)	11 (41%)	3 (11%)	1 (4%)	7 (26%)	6 (22%)
Ovary (16)	6 (38%)	9 (56%)	7 (44%)	2 (13%)	12 (75%)	11 (69%)	5 (31%)
Lung (21)	3 (14%)	11 (52%)	3 (14%)	1 (5%)	3 (14%)	5 (24%)	6 (29%)
Kidney (20)	0 (0%)	0 (0%)	2 (10%)	1 (5%)	5 (25%)	2 (10%)	2 (10%)
Rectum (19)	6 (32%)	7 (37%)	11 (58%)	7 (37%)	9 (47%)	12 (63%)	9 (47%)
Thyroid (6)	0	0	1	0	0	0	0
Prostate (4)	3	1	2	2	1	1	0
Cervix (1)	0	1	1	0	0	0	1
Intestine (2)	0	0	1	1	1	1	1
Pancreas (1)	1	0	1	0	1	0	1

In situ hybridization

For the BST2 gene that is a surface protein that so far has only been related to B cell differentiation and their transformation into myeloma9 and whose expression was now clearly upregulated in Northern blotting and cancer array analysis (Figs. 2, 3), we performed in situ hybridizations with frozen sections of human tumors derived from stomach, esophagus, breast, lung and liver. All tumor cells in sections from the digestive tract, breast and lung cancers were stained positively. This result correlates with the data obtained with the tumor array analysis. The liver tumor, however, did not stain with the BST2 probe.

The examples tested by array analysis and in situ hybridization allows to conclude that several of the genes upregulated in HT29p cells may be relevant for human cancer in general.

Discussion

Using SSH to identify differentially expressed genes between the metastasizing HT29p cells and the nonmetastasizing HT29-MTX cells, we found 58 unique RNA sequences that were more abundant in the metastatic cell line. These sequences are either related to loss of differentiation (the nonmetastatic differentiated HT29-MTX cell line was derived from the nondifferentiated HT29p cells) or to the tumor progression phenotype of HT29p cells. The specific expression pattern of a subset of these 58 gene sequences was confirmed by Northern analysis. Also a subset was examined for expression in primary human tumors and found upregulated in cancer of the digestive and reproductive tract, of lung and breast. Among the 58 genes there were 10 of unknown function.

From a putative functional context, the expression of several genes in metastatic cancer makes sense: several proteins encoded by the isolates have been associated with tumor progression previously (see references in Table I); the functions of others could be related to migration or proliferation. The majority of sequences, however, has not previously been isolated from metastasizing tumor cells.

The SSH protocol permitted to identify transcripts of transcription factors presumably expressed at low rate, e.g., EHF, HMGB1 and TCF7L2. These transcription factors fit well into a program of tumor progression. EHF is an epithelial cell-specific member of the Ets family.10 Ets members, yet not EHF, have been found in cancer-specific chromosomal translocations. One particularly interesting target of EHF is the c-Met promoter. The latter contains 3 high affinity EHF binding sites.10 c-Met is a growth factor receptor which together with its ligand HGF promotes invasive growth.11

HMGB1 appears to be a multifunctional protein. It is involved in the transcriptional regulation of p53,12 protects mammalian cells from apoptosis13 and might trigger metastasis formation by binding to the receptor tyrosine kinase RAGE when released from necrotic cells.14 Elevated expression of HMGB1 occurred in human primary breast carcinoma,13, in gastrointestinal cancers15 and in melanoma.16.

TCF7L2, also called TCF-4, is the central regulator of the Wnt signaling pathway. Complexed with β-catenin TCF-4 turns on a program of proliferation in epithelial cells.17 Stabilization of β-catenin upon inactivation of the tumor suppressor gene APC initiates 1 of the pathways to colon cancer.18

TAF 10, ribosomal proteins and other housekeeping proteins need to be expressed at elevated rate in tumor cells. It is not clear whether they exert a more specific role in the potential to metastasize. Interestingly upregulation of the ribosomal protein S7 has also been observed in a metastasizing mammary carcinoma cell line.8

One of the signaling proteins, PTPN12, a cytoplasmic protein tyrosine phosphatase, has been described as a potential tumor suppressor gene, e.g., dephosphorylating the Abl onco-protein.19 Aberrant transcripts were found in colon cancer.20 Our isolate may be an aberrant product, but we have not yet determined the exact transcript structure.

The neurotrophic receptor tyrosine kinase type 1 (NTRK1) was originally isolated from neuronal tissues but is also expressed in many other normal and malignant human tissues.21 The receptor was found on the cell surface of advanced stages of ovarian cancer22 and pancreatic carcinoma.23 From the same group of proteins, YWHAZ interacts with the cytoplasmic tail of ADAM,22 a transmembrane protease with a potential role in cell adhesion.24

For most other proteins, it is futile to speculate on a putative role in tumor progression. Of course, migrating cells require transient cell-cell and cell-matrix contacts. Several isolates may fall into this category. Keratin-8 had been identified as a metastasis-specific gene in another screen using a pair of rat mammary carcinoma cells of differing metastatic potential.8 It binds the intermediate filaments and is essential for migration.25 Other genes not yet reported as linked to cancer progression include filamin A that cross links actin filaments to membrane glycoproteines, emilin or filamin B which mediate cell-cell contact or migration.

The interferon-inducible protein 27 and HSPC014 have been found overexpressed in mammary and hepatocellular carcinoma respectively in the past.26, 27 but their function is still unclear.

Our study, together with other screens, forms an entry point into exploring gene functions in human metastatic cancer and into a search for new targets for intervention. 4

Figure 4. In situ hybridization of tumor tissues. Tissue array slides (BioCat) were hybridized with an antisense probe (upper panel) or a sense probe (lower panel) of the BST2 gene. Hybridization of the digoxigenin-labeled RNA was detected by an anti-digoxigenin antibody using the colorimetric assay provided by Roche Diagnostics. Brown staining indicates positively stained cells. The blue staining corresponds to the hematoxylin detection that enables to visualize the cells.

Acknowledgements

We thank N. Howells and D. Weih for expert advice in metastasis assays. V. Orian-Rousseau received a scholarship from “La Fondation pour la Recherche Medicale”, France.