The neuron-restrictive silencer factor [NRSF (RE-1 silencing transcription factor/X box repressor)] is a transcriptional silencer,which we have previously implicated in deregulation of the vasopressin promoter in small cell lung cancer (SCLC). Here we describe a novel splice variant of the NRSF transcript, which is highly expressed in SCLCs. The variant was detected in both established cell lines and primary SCLC cultures as well as in some primitive neuroectodermal tumor biopsies. It was present at very low levels in human brain tissue, non-SCLC tumors, and normal bronchial epithelium. This human splice variant, which is massively overexpressed in SCLCs, incorporates a 50-bp insert between exons 5 and 6, introducing a stop codon and predicting translation of a truncated NRSF isoform. We propose that the encoded isoform may antagonize repression of the vasopressin promoter and other “neuronal” genes with neuron- restrictive silencer elements in SCLCs. Thus, up-regulated expression of this NRSF isoform may be a key early factor in defining the neuroendocrine phenotype of these tumors. The NRSF splice variant represents a specific clinical marker that could prove useful in detection of the majority of SCLCs.
SCLCs3express markers of both neuronal and epithelial cells and are characterized by the production of a variety of neuropeptides. These neuronal peptides, such as arginine vasopressin, are not normally expressed in bronchial epithelial cells or in nonneuroendocrine tumors(1), including the majority of NSCLCs.4However, a large proportion of SCLCs secrete vasopressin(2), which can cause dilutional hyponatremia syndrome in patients (3) and stimulate tumor proliferation via an autocrine loop (4, 5, 6). The hypothesis that SCLCs originate from normal neuroendocrine cells in the lung is now being superseded by the idea that these tumors originate from epithelial cells and that the acquisition of neuroendocrine features is key to the process of oncogenesis. Strong evidence for an epithelial origin of SCLCs has recently been derived from cDNA arrays (7), but the mechanism by which transcription of neuroendocrine genes is activated remains an important question.
We have previously characterized transcription factors that interact with the proximal vasopressin promoter, which contains the major elements that determine the expression of vasopressin in lung tumor cell lines (8, 9, 10). We recently identified a motif with homology to the NRSE around the transcriptional start site of vasopressin, suggesting a role for NRSF (REST/X box repressor 1) in the normal restriction of expression to neuronal cells (11). The NRSF is a large repressor protein (12, 13) that can silence the expression of neuronal genes in nonneuronal tissues via NRSE motifs (14). The complex secondary structure of NRSF includes nine zinc fingers, eight of which form a DNA binding domain,and there are two repressor domains at the NH2-and COOH-terminals (illustrated in Fig. 1). The COOH-terminal domain ninth zinc finger was recently reported to interact with a corepressor (15), whereas the NH2-terminal repressor domain is now thought to be involved in histone deacetylation (16). We have shown that overexpression of NRSF could switch off the vasopressin promoter in some SCLCs (11). However, rather than observing a loss in binding of NRSF to the vasopressin promoter NRSE motif, we noted that several SCLC complexes were present in electrophoretic mobility shift analysis. We postulated that these multiple complexes could be modified forms of NRSF, such as the splice variants recently reported by others in a rat model (17). We now report that a novel splice variant of NRSF is very highly expressed in SCLCs relative to NSCLCs or normal lung tissue. We propose that the encoded isoform antagonizes the function of normal NRSF, thus derepressing genes such as vasopressin. This isoform may represent a key factor early in establishment of neuroendocrine SCLCs and can be exploited as a specific clinical marker for detecting the majority of these tumors.
Materials and Methods
The Lu-165 SCLC line was a gift from Dr. T. Terasaki (National Cancer Center, Tokyo Japan), the NSCLC line COR-L23 and the SCLC cell lines COR-L24 and COR-L88 were gifts from Dr. P. Twentyman (established at the Medical Research Council Clinical Oncology Research Unit,Cambridge, United Kingdom). NCI-H711 (SCLC) was a gift from Dr. B. Johnson (National Cancer Institute, Bethesda, MD), U2020 (SCLC) was obtained from Dr. P. Rabbitts (Laboratory of Molecular Biology,Cambridge, United Kingdom), and GLC-19 (SCLC) was obtained from the University of Groningen (Groningen, the Netherlands). Other SCLC(NCI-H69 and NCI-H345), NSCLC (NCI-H460), and HeLa cell lines originated from the American Type Culture Collection. These cell lines were routinely maintained in RPMI 1640/10% bovine calf serum at 5%CO2/37°C. Primary normal human bronchial epithelial cells from a single donor (4653; Clonetics) and SV40-transformed human bronchial epithelial cells (HBEs; a gift from Dr. W. Franklin; University of Colorado Cancer Center, Denver,CO) were maintained in fully supplemented Bronchial Epithelial Growth Medium (Clonetics).
Primary lung tumor cells were PW3 (SCLC pretreatment, passage 4), PW5(SCLC posttreatment, passage 5), 011 (SCLC, pleural fluid, passage 6),and 002 (NSCLC, lymph node, passage 6). These were derived from lymph node and pleural aspirates by culture in selective media(HITESA = RPMI 1640 with l-glutamine,supplemented with 10 nm hydrocortisone, 5 μg/ml insulin,10 μg/ml transferrin, 10 nm estradiol, 30 nmselenite, and 0.25% BSA). Normal brain biopsies were collected at Nottingham City Hospital from fresh postmortem material, these tissues were snap-frozen in liquid nitrogen. PNET cDNA was a gift from Dr. P. Scotting (Queen’s Medical Center, Nottingham, United Kingdom).
Isolation of mRNA from Cultured Cells and Tissues.
Total cellular RNA was prepared from 2 × 107 cultured cells using a Purescript RNA kit(Gentra Systems, Inc.). Total RNA was prepared from human tissue using Trizol (Life Technologies, Inc.). The concentration and integrity of all RNAs were determined spectrophotometrically and electrophoretically.
Total cellular RNA (1 μg for cell lines or 2 μg for primary tissue)was used for reverse transcription (Promega, Southampton, United Kingdom), and the cDNA was denatured and diluted to 100 μl; 2 μl of this were used for PCR. Dynazyme II polymerase was used in magnesium-free buffer (Flowgen) supplemented to a final concentration of 1.5 mm MgCl2. As described previously, GAPDH primers (8) and β-actin primers that produce a band of 626 bp (18) were used as controls for semiquantification. Amplification of 28 (GAPDH) or 35 (β-actin)cycles was used for cell lines, and amplification of 33 or 37 cycles,respectively, was used for primary samples. NRSF primer sequences NRSFfor (5′-GAATCTGAAGAACAGTTTGTGCAT-3′) and NRSFrev(5′-TTTGAAG-TTGCTTCTATCTGCTGT3′), which amplify 554-1180(19) of the human NRSF/REST cDNA (HSU22314), were used for 35 cycles (cell lines) or for up to 45 cycles (primary material). A SCLC-specific NRSF primer, sNRSFrev (5′-ATCACACTCTAGTAAATATTACC-3′),was designed against the insert sequence and used in seminested RT-PCR to detect the splice variant in tissues where it was not seen in the standard PCR. The first round of amplification (25 cycles) was performed with NRSFfor/NRSFrev; 2 μl were then used as the template in the second round with NRSFfor/sNRSFrev (25 cycles, 521 bp). All PCR products were electrophoretically separated on 2% agarose gels, and AMRESCO 3:1 agarose (Anachem, Bedfordshire, United Kingdom) was used for high resolution (Fig. 1 b). Products were excised and purified using QIAquick columns (Qiagen, Crawley, United Kingdom) for direct automated sequencing or cloned into PCR2.1 (InVitrogen) before sequencing on an ABI Prism 377 (DNA Sequencing Facility,Queen’s Medical Center).
Results and Discussion
An NRSF Splice Variant Is Present in SCLC Cell Lines and Tumors.
We have previously described multiple NRSE-binding complexes in lung cancer cell lines, some of which were specific to SCLCs(11). RT-PCR primers were used under semiquantitative conditions to amplify the region corresponding to a cluster of zinc fingers in the DNA binding domain of NRSF (see Fig. 1,a). The overall level of the NRSF amplified product was lower in the SCLC cell lines than in the control cell lines, although control GAPDH andβ-actin amplification products were constant across the panel (data not shown). However, when the NRSF PCR reactions were further resolved,it was clear that two major products were amplified in at least seven of eight SCLC cell lines (Fig. 1 a). Sequencing of products amplified from the Lu-165 cell line confirmed that the faster migrating band, which was also amplified in all of the NSCLC cell lines and normal lung samples, was identical to wild-type NRSF. Sequencing of the lower mobility product revealed a 50-bp insert located between the boundaries of exon 5 and exon 6. We confirmed that an identical insert was present within the larger PCR product from the four other SCLC cell lines sequenced to date. Therefore, although the SCLC cell lines retain some expression of normal NRSF, the majority also express comparable levels of a splice variant. HeLa, another example of an epithelial tumor cell line, also only expressed the normal NRSF transcript (data not shown).
Interestingly, the one SCLC cell line (COR-L88) in which the splice variant was not evident (Fig. 1,a) is a morphological variant cell line that we have found to express low levels of vasopressin by nested RT-PCR (data not shown). Repeated PCR and increased resolution of the NRSF amplification products by the use of AMRESCO 3:1 agarose in fact reproducibly demonstrated that some sNRSF isoform was expressed in this cell line (Fig. 1 b). However, the normal NRSF product was much more abundant than the isoform, in contrast to the other SCLC cell lines, in which the two forms were expressed at similar levels. This finding in a variant cell line has implications in terms of both neuroendocrine profiles and clinical exploitation, as discussed below.
To confirm that the SCLC splice variant (sNRSF) was not an artifact of established tumor cell lines cultured in vitro, we have also examined primary tumors for the splice variant. Four low-passage primary cultures of lung tumors established from lymph node or pleural aspirates were screened by RT-PCR (Fig. 2). Whereas sNRSF and the wild-type repressor were both seen in 011(SCLC, passage 6), only the latter was evident in the NSCLC isolate(002, passage 6). We also studied a pair of cultures (PW3, passage 4 and PW5, passage 5) established from the same patient at the early(pretreatment) and late (posttreatment) stages of disease. sNRSF was detected in the two samples, and equivalent amounts of normal NRSF and sNRSF were seen in PW5; however, in the PW3 sample, sNRSF was the major product. These data imply that high-level expression of the aberrant splice variant is a feature of SCLC tumors but not of NSCLC tumors and that this expression is seen pretreatment and may therefore represent an early change in the development of SCLC. It has been suggested that variant SCLCs represent recurrent or drug-resistant tumors, which have reduced neuroendocrine characteristics. The reduction in the ratio of sNRSF:NRSF posttreatment (PW5) compared to pretreatment (PW3) and in a variant cell line (COR-L88) compared to other SCLC lines is consistent with this hypothesis.
The SCLC Splice Variant Is Seen in Neuronal Tumors but Is Rare in Normal Brain.
To determine whether the sNRSF splice variant was a common feature of other neuroendocrine tumors, we also amplified cDNA derived from clinical PNET samples. In contrast to SCLCs, PNET is a neuronally derived cancer of the central nervous system. We identified a variant RT-PCR product in two of four of these samples (Fig. 3,a). Sequencing confirmed that this also contained a 50-bp insert at the same position found in the SCLC variant. sNRSF may therefore represent a human splice variant encoding a neural-specific form of NRSF, which is abnormally expressed in SCLCs and other neuronal tumors. Alternatively it may be an exclusively tumor-specific form of NRSF. To address this question, we prepared RNA from several regions of human brain and investigated NRSF expression by RT-PCR (Fig. 3,b). In standard PCR reactions, only the product corresponding to normal NRSF could be visualized (Fig. 3 b, top), implying that if sNRSF is generated in normal brain, it is present at a much lower abundance than in the SCLC tumors.
These products were then used as templates for a second round of amplification, with a reverse primer specific to the novel exon of sNRSF. Whereas this amplified abundant sNRSF product from the Lu-165 SCLC cell line, a product of the same size was only seen at low levels in one of the brain samples that was isolated from an area of the cerebral hemisphere (Fig. 3 b, middle). A smaller, low abundance product was detected in the hypothalamus, whereas this and three other products of different sizes were faintly seen in the cortex sample. We conclude that sNRSF and other variants may be present at very low levels in some normal neuronal tissue, but they are uncommon. Thus, sNRSF appeared to be a predominantly tumor-specific splice variant.
Interestingly, nested PCR with the sNRSF-specific primer amplified two products from the NSCLC cell line NCI-H460. Sequencing of these products revealed that one corresponded to amplification of sNRSF, and the other corresponded to the normal product of the first-round amplification, which must have been highly abundant. These data imply that although the sNRSF splice variant was present at greatly elevated levels in SCLCs, it could also be detected, albeit at substantially lower levels, in NSCLCs. This lends support to the theory of a common epithelial origin for SCLC and NSCLC lung tumor types.
Structure and Functions of NRSF and Its Isoforms.
The SCLC splice variant, sNRSF, incorporates an additional 50-bp exon at the boundary of exons 5 and 6, which introduces a stop codon resulting in a truncated isoform of NRSF with a novel 13-amino acid sequence not present in wild-type NRSF. We have submitted the nucleotide sequence to GenBank (accession number AF228045). Sequence comparison (Fig. 4) shows that this is distinct from but related to the neural-specific NRSF isoforms (rREST4 and rREST5) previously described in rat brain(17). The 3′ portion of the sNRSF insert (16 bp) has homology to the common region of the rat neural variants, with only one mismatched base. However, sNRSF also has an additional 34-bp at sequence 5′ to this, which has not been described in rat. The presence of this splice variant in SCLCs, which is generally believed to be an epithelial-derived tumor, is of great interest because rREST4 and rREST5 have been detected exclusively in neuronal tissue. We failed to detect sNRSF in most brain tissue studied, and where the corresponding PCR product was amplified, this was of very low abundance. These data concur with the reported expression of rREST4 and rREST5, which were present at low levels in rat neuronal tissue(17).
During the preparation of the manuscript, the sequence of several mouse and human neural-specific NRSF transcripts expressed in neuronal-derived neuroblastoma tumors has been published(20). Whereas variants that correspond to rREST4 and rREST5 were described in the mouse, the major human splice variant has an insertion of 62 bp between exons 5 and 6. This human neuroblastoma transcript (20) differs from the human SCLC transcript described here by an additional 12 bp at the 3′ end; this is the same sequence that distinguishes rREST5 from rREST4. Although the two human splice variants translate as the same isoform, they are likely to be derived by different mechanisms.
A Potential Role for sNRSF in Neuropeptide Expression.
The rat neural-specific splice variants were originally described as retaining repressor activity (17). However, a recent publication (21) describes an antisilencer mechanism for murine REST4 that prevents NRSF silencing of the cholinergic gene locus through a protein-protein interaction, implying a role in neuronal transcription. This recent description of an antagonistic role for at least one NRSF isoform (21) and our discovery of a human NRSF splice variant in SCLCs provide strong support for our earlier hypothesis that abnormal forms of NRSF lead to derepression of vasopressin expression in these tumors (11). We propose that sNRSF antagonizes the function of normal NRSF and that this isoform could potentially derepress a number of neuroendocrine genes contributing to the neuroendocrine pathology of SCLC and the expression of a number of autocrine growth factors. Because it has been suggested that relative concentrations of NRSF and REST4 control expression levels (21), those tissues with very low levels of sNRSF,for example, NSCLC (Fig. 3,b), would be unlikely to express sufficient variant for antisilencer activity. We have also shown that ratios of sNRSF:wild-type NRSF vary between classical and variant (Fig. 1,b) or pre- and posttreatment SCLCs (Fig. 2), implying that this isoform may be less important in later stage or drug-resistant SCLC tumors, which often show reduced neuroendocrine properties.
Application of sNRSF in SCLC Diagnosis.
This human SCLC NRSF splice variant was not detected by standard RT-PCR in normal lung or in other nonneuronal tissue. However, we have detected high relative levels of sNRSF in most primary SCLC tumors. From the data presented here, this represents a largely tumor-specific clinical marker that is highly expressed in the majority of SCLCs. We propose to use sNRSF in the detection of these tumors(22). Our data imply that high-level sNRSF expression would be a useful clinical marker in early detection of SCLC, whereas the loss of sNRSF in a tumor may be an indicator of recurrent or resistant disease.
We thank Julie Stanley for excellent technical assistance and Drs. Samreen Ahmed, Paul Scotting, and Colin Clelland for provision of samples.
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.
Supported by the Association for International Cancer Research, St Andrews, Scotland.
The abbreviations used are: SCLC, small cell lung cancer; NSCLC, non-SCLC; NRSE, neuron-restrictive silencer element; NRSF, neuron-restrictive silencer factor; REST, RE-1 silencing transcription factor; PNET, primitive neuroectodermal tumor; GAPDH,glyceraldehyde-3-phosphate dehydrogenase; RT-PCR, reverse transcription-PCR.
M. Garcia-Ocejo, S. Ahmed, J. M. Coulson,and P. J. Woll. Mitogenic neuropeptides and receptors in lung cancer: potential autocrine circuits and early markers, submitted for publication.