Human preimplantation embryonic cells are similar in phenotype to cancer cells. Both types of cell undergo deprogramming to a proliferative stem cell state and become potentially immortal and invasive. To investigate the hypothesis that embryonic genes are re-expressed in cancer cells, we prepare amplified cDNA from human individual preimplantation embryos and isolate embryo-specific sequences. We show that three novel embryonic genes, and also the known gene, OCT4, are expressed in human tumours but not expressed in normal somatic tissues. Genes specific to this unique phase of the human life cycle and not expressed in somatic cells may have greater potential for targeting in cancer treatment.
During very early development, there is a wave of de-programming of the differentiated oocyte and sperm genetic complements which is accompanied by genome-wide demethylation (Monk et al., 1987; Monk, 1990). This demethylation is marked in the inner cell mass cells of the blastocyst and continues in the primordial germ cells which delineate soon after implantation (McMahon et al., 1983). We have hypothesized that the erasure of gametic genetic programming returns the embryonic cells to the ground state of developmental totipotency (i.e., capable of giving rise to any cell type, including germ cells, in chimaeric mice), and that the archetypal stem cell is the primordial germ cell (Monk, 2001). In line with this, cells of the inner cell mass of the preimplantation blastocyst and the primordial germ cells of the post-implantation embryo are the only potentially immortal cells in the mammalian life cycle. Once removed from the constraints of development in vivo, they can be cultured indefinitely in vitro (Evans and Kaufman, 1981; Matsui et al., 1992) remaining undifferentiated and totipotent. Following transplantation into adult mice they give rise to tumours. Cells from the early embryo are also invasive – trophectoderm cells of the blastocyst invade the uterine stroma after implantation and primordial germ cells are migratory as they find their way to the developing gonad of the foetus (Fujimoto et al., 1977).
Broadly speaking, cancer cells are also immortal, undifferentiated and invasive. Therefore, it might be expected that cancer cells will express genes in common with these very early embryonic cells, especially genes specifically associated with deprogramming and return to the undifferentiated and proliferative stem cell state, and the maintenance of that state. These genes would not, by definition, be expressed by an individual's somatic cells which are committed to differentiation and senescence and, as such, they could be excellent candidates to target in the treatment of cancer and/or for the development of a DNA cancer vaccine (Gilboa, 1999).
In order to test the hypothesis that embryonic genes may be re-expressed in cancer cells, we set out to isolate human embryo-specific expressed genes and test for their expression in a panel of human cancers.
Isolation of embryo-specific expressed gene sequences from human oocytes and embryos
We devised techniques of sufficient sensitivity to prepare amplified cDNA from human single preimplantation embryos using a strategy of making cDNAs from a number of individual human embryos at each of the 4-cell, 8-cell and blastocyst stages rather than from pooled samples (Holding et al., 2000). Primordial germ cells from a male and a female 10 week human foetus were pooled in batches of 200 to 300 cells for preparation of cDNA (Goto et al., 1999). To isolate embryo-specific expressed sequences we chose differential display of our embryonic cDNAs in comparison with foetal somatic cell cDNAs. Although microarrays are now considered to be the most powerful method of identification of differentially expressed genes, they are not suitable for our analysis at this stage. Microarrays prepared from expressed sequences in somatic cells cannot, by definition, contain embryo-specific sequences. Those prepared from expressed sequences from the entire genome, or from cancer cells, may contain some embryonic genes, depending on the abundance of the transcript; however, they would not currently be recognized as such.
Using differential display, the patterns for gene expression of the embryos and primordial germ cells were compared with those of 10-week foetal somatic tissues, brain, muscle and gut (Figure 1). This procedure identifies embryonic genes, A, B, C, D and E, expressed in the embryos and not in the somatic tissues. The three somatic cDNAs were chosen to represent the early ectoderm, mesoderm and endoderm lineages, respectively, although it is realized that there may be specific somatic tissue genes that are not represented. A 10-week foetus was used in order to focus on gene sequences specifically expressed in the early embryo and to exclude foetal-specific genes (also identified by these procedures (Holding and Monk, unpublished)). Note that the patterns of gene expression in the preimplantation embryos are sparse and variable and that the differential display bands A, B, C, D and E do not appear in all embryo cDNAs at all stages. The variable patterns of expressed genes is an inherent property of preimplantation embryos resulting from degradation of maternally-inherited mRNAs and the gradual onset of embryonic gene expression (Holding et al., 2000; Salpekar et al., 2001).
Confirmation of embryo-specificity of expression of the five differential display sequences
The embryo-specific bands, A, B, C, D and E, in Figure 1 were excised from the gel, cloned and sequenced. Primers were designed within each differential display sequence and used for PCR amplification of the sequence within the embryo, primordial germ cell and foetal somatic cDNAs. Figure 2 shows confirmation of embryo-specific expression of these sequences. Expression is absent or very much reduced in the somatic tissues. Expression of the housekeeping gene, GAPDH, was used as a control for cDNA quantity and quality.
Expression of the five new embryonic sequences in cancer cells
Figure 3 shows the expression of the five embryo-specific sequences on a panel of tumour and normal tissue cDNAs (Multiple Tissue cDNAs (MTCs) supplied by BD Clontech, UK). All sequences are expressed in one or more of the tumour samples. Sequences B and D are also expressed in the normal somatic tissue cDNAs and were not further analysed. However, expression of A, C, and E is relatively low or absent in all the non-tumour cell controls except placenta and testis. The expression in placenta and testis further confirms the embryo- and germ cell-specific nature of these three genes.
Database analysis and extended sequences of embryo/cancer genes A, C and E
The three new embryo/cancer sequences, A, C and E, were subjected to BLAST database analysis (Altschul et al., 1997; Karlin and Altschul, 1990, 1993) to extend these human gene sequences. The general approach was first to identify, in the EMBL database, ESTs which matched the embryo-specific differential display sequences. The EST sequences found were then used to search the Ensembl genome Golden Path database which revealed chromosome locations and intron/exon boundaries.
Figure 4 shows the 3′ differential display sequence (lower case in Figure 4) together with further upstream sequences determined by database analysis for each of the embryo/cancer gene sequences, A, C and E. For embryo/cancer gene A (Figure 4a), the differential display sequence matches the 3′ end of a 944 bp sequence derived from six exons of an uncharacterized gene at chromosome 3q13.13. It also matches the 3′ ends of two probable pseudo-gene sequences on chromosomes 1 and 3 (an observation consistent with expression and transposition in the germ line), and shows homology with fragments of this sequence at three other regions. For embryo/cancer gene sequence C (Figure 4b) the differential display sequence matches the 3′ end of a predicted transcript of 8 exons, totalling 1062 bp, of a gene of unknown function located at chromosome 3p25.2. The differential display sequence of embryo/cancer gene E (Figure 4c) consists mainly of long terminal repeat sequence with two short segments localizing the sequence specifically to chromosome 8p23.1 It matches the 3′ end of an EST which extends the sequence for three exons upstream into an uncharacterized gene at this location giving a sequence of 589 bp. The LTR sequence incorporated at the 3′ end of this gene provides a polyadenylation signal. This acquisition and use of retroviral insertions as regulatory 3′ sequences has been previously observed for two other novel genes not associated with cancer (Mager et al., 1999). Primers specific for each upstream exon thus identified for each of the three genes were used as 5′ primers together with the most 3′ differential display sequence primer, to confirm, by PCR, specificity of expression of the extended sequences to embryo cDNAs.
Embryo- and tumour-specific expression of extended embryo/cancer gene sequences, and of OCT4
The extended sequences of A, C and E were screened for specific expression in embryo and tumour cDNAs using primers at the most 5′ and 3′ ends (see Figure 4: note that primers for extended sequence E were designed to avoid repetitive motifs). Figure 5 shows expression of the extended sequences of A, C and E in three embryonic cDNAs and three tumour cDNAs, and the absence of expression of these gene sequences in three normal tissue cDNAs. As in Figure 3, embryo/cancer gene A is expressed in only one colon sample and the abundance of embryo/cancer gene C is low in tumours.
Thus encouraged by the confirmation that three out of the five embryo-specific expressed sequences tested were re-expressed in cancer cells, we tested for the re-expression of the known embryonic gene, OCT4, in the panel of tumours. Figure 5 shows that OCT4 (gene F in Figure 5) is also expressed at a high level in blastocyst and cancer cell cDNAs but is expressed at much lower level in normal tissue cDNA. We also tested a range of cancer cell lines for expression of our embryo/cancer genes C and E and of OCT4. All three gene sequences were expressed in various cancer cell lines but were not expressed in immortalized fibroblasts (data not shown).
Tissue specificity of expression of embryo/cancer genes confirmed by Northern blot
The PCR products of the extended gene sequences were used as probes to determine tissue specificity of expression by Northern blot. Figure 6 shows mRNA transcripts in testis for embryo/cancer genes A, C and E of approximately 3.5, 3 and 2 kb respectively. Embryo/cancer gene sequence C is also detected in ovary in the Northern blot. However, the expression of this gene is markedly higher in ovarian cancer as determined by PCR analysis (data not shown).
Processes occurring during tumorigenesis may be similar to processes occurring in early development. They have, in common, double strand breaks and damage to the DNA – the aftermath of meiosis in development, or exposure to radiation or radiomimetic chemicals in somatic cells. They also share, as a common trigger for deprogramming, the cessation of informational signalling – the follicle cells withdraw their processes from the maturing oocyte, the somatic cells may suffer injury or blockage to information appropriate to a differentiated cell. The exact nature of the initiating event is unknown but, in both processes, development and tumorigenesis, the result is genome-wide demethylation (Monk et al., 1987; Monk, 1990; Gama-Sosa et al., 1983; Goelz et al., 1985; Feinberg et al., 1988). At the same time, promoter hypermethylation and silencing of other genes, including tumour suppressor genes is observed (Jurgens et al., 1996; Jones and Laird, 1999). Increased methylation may be a default state following cessation of gene activity – in other words, ‘use it or lose it’ – and it is possible that the hypermethylation observed for specific genes in cancer cells and cultured cells is a consequence of their decreased transcription with the change in cell phenotype to the proliferative stem cell state. It is curious that, although genome-wide demethylation is observed in embryos (Monk et al., 1987) and cancer (Gama-Sosa et al., 1983; Goelz et al., 1985; Feinberg et al., 1988), the level of the methylating enzyme, DNA methyl-transferase, is actually very high in embryos (Monk et al., 1991) and tumours (Kautiainen and Jones, 1986; El-Deiry et al., 1991).
Another common feature of embryos and tumour cells is an increase in expression and transposition of retrotransposons, presumably as a result of global demethylation. Endogenous retroviruses (Piko et al., 1984; Kuff and Leuders, 1988; Hsieh et al., 1987) and long interspersed nucleotide elements (L1 LINE; Packer et al., 1993; Branciforte and Martin, 1994; Jurgens et al., 1996) are normally silenced by methylation but they are activated in embryos (Piko et al., 1984; Packer et al., 1993) and cancer cells (Kuff, 1990; Hsieh et al., 1987). Links between repair of DNA damage and demethylation, activity of retrotransposons and cancer have been reported in the literature but are not fully understood (Lengauer et al., 1997). It is not clear whether the derepression of these repetitive elements is a cause or a consequence of tumour progression. The known association of many viruses with cancers makes it seem highly likely that they will have a causative role (reviewed in Urnovitz and Murphy, 1996).
The hypothesis implicit in the comparison between embryos and cancer cells is that a somatic cell with damage to its DNA, or disruption of normal informational signalling, may undergo the same deprogramming, genome-wide demethylation and transition to an undifferentiated proliferative stem cell path as the early embryo. Like the embryo, the tumour cell is triggered to ‘start again’. We do not propose that the deprogramming of a somatic cell would return the cell to the totipotent state of the primordial germ cell but rather towards the undifferentiated proliferative stem cell state appropriate to its tissue of origin. However, it should be noted that adult somatic cells may deprogramme towards a stem cell state associated with many diverse tissues when provided with the appropriate culture medium, supporting feeder cells and ‘cocktail’ of growth factors (reviewed in Blau et al., 2001), and the nucleus of an adult somatic cell may deprogramme to the totipotent state when transferred to an enucleate oocyte (reviewed in Solter, 2000).
Although we do not know as yet the function of the new embryo/cancer genes isolated in our work, it is possible that they function in the OCT4 pathway. As with our new embryonic genes, expression of the gene coding for the octamer-binding-protein, OCT4 (also known as OCT3) is specific to oocytes, the pluripotent cells of the embryo and the germ line in mouse and human (Scholer et al., 1989; Abdel-Rahman et al., 1995; Goto et al., 1999). Previously, an investigation of POU homeobox gene expression in breast cancer revealed expression of at least four POU genes, and showed expression of the OCT3/OCT4 embryonic transcription factor in the breast cancers tested (Jin et al., 1999).
Our aim is to find a ‘universal’ cancer antigen from amongst these novel embryonic genes whose function may be associated with the initiation of deprogramming (demethylation) and a return to the proliferative stem cell state associated with the immortality and/or invasiveness of cancer cells. Early embryogenesis is a unique phase of the human life cycle and we may expect that there are genes expressed at this time that will not be expressed in the normal somatic cells of the adult. Thus, the re-expression of these embryonic genes in cancer cells may have a high potential for cancer therapy in the future. We are continuing with our work to identify further embryo/cancer genes and to begin functional analyses to determine their role in the cancer cell phenotype.
Materials and methods
Procedures for preparation of the embryonic cDNAs, differential display and the recovery, cloning and sequencing of embryo-specific differential display bands are described in Holding et al. (2000). Briefly, mRNAs from lysed embryos were isolated by binding of their polyA tails to polyT attached to magnetic beads, copied by reverse transcriptase into cDNA, and the cDNA amplified by PCR using a Clontech SMART kit (BD Clontech, UK). Differential display PCR was performed as described by Liang and Pardee (1997). Embryo-specific bands were excised from the gel and cloned into pGemTEasy and sequenced. Primers within each of the obtained sequences (see Figure 4 for location and sequence of primers in the differential display sequences) were used for confirmation of embryo-specificity. The cycling parameters are given below.
Screening of tumour and normal cDNAs for expression of embryo-specific sequences were carried out using panels of multiple tumour cDNAs (MTC Panel I, Panel II and Human Tumor MTC Panel) supplied by BD Clontech (UK). Although supplied as ‘normalized’ for housekeeping genes, we nevertheless checked quality and quantity using the three housekeeping genes, GAPDH, beta-actin and HPRT (not shown), in our PCR system and adjusted loading amounts of the different cDNAs accordingly. We used at least 3 μl of each cDNA per 25 μl PCR reaction mixture comprising Perkin Elmer Amplitaq Buffer I, 1.25 U Amplitaq, 200 μM dNTPs, and 25 pmoles each primer. The primer sequences are shown in Figure 4, except for the OCT4 primers which are OCT4a: 5′ GACAACAATGAAAATCTTCAGGAGA 3′ and OCT4b: 5′ TTCTGGCGCCGGTTACAGAACCA 3′. The PCR conditions were 30 cycles of each: denaturation 95°C for 1 min, annealing temperature (see below) for 1 min, and strand elongation 72°C for 1.5 min. Annealing temperatures were – Sequence A: 60°C; Extended Sequence A: 55°C; Sequence B: 55°C; Sequence C: 55°C; Extended Sequence C: 55°C; Sequence D: 60°C; Sequence E: 60°C; Extended Sequence E: 60°C; OCT4 60°C.
Northern analysis was carried out according to manufacturer's instructions. A Multiple Tissue Northern Blot (MTN II) from BD Clontech UK was hybridized with labelled PCR products from Extended Sequences A, C and E. These were excised from the gels, purified using Qiagen Gel Purification kit and labelled directly using Amersham Ready-To-Go DNA Labelling Beads (−dCTP) with 32PdCTP (ICN: 3000 Ci/mmol). Hybridization was carried out in ExpressHyb Hybridization Solution (Clontech) at 68°C for 1 h. Blots were washed twice for 15 min in 2×SSC, 1% SDS at room temperature, then twice for 15 min in 1×SSC, 1% SDS at 50°C, and exposed to Kodak BioMax film at −70°C for up to 3 days.
This work is supported by the BBSRC. We thank Ashreena Salpekar for her assistance with quantitating the cDNAs from Clontech and David Faulkes and Gary Williams (HGMP) for their assistance with the database analysis.