Migration stimulating factor (MSF) is a truncated oncofetal fibronectin isoform expressed by fetal and tumor-associated cells. MSF mRNA is distinguished from other fibronectin isoforms by its size (2.1 kb) and the inclusion of a specific intronic sequence at its 3′ end. Initial Northern blot analysis with a MSF-specific probe indicated the presence of this 2.1-kb transcript and an additional unexpected 5.9-kb RNA present in both MSF-secreting (fetal) and nonsecreting (adult) fibroblasts. Our investigations into the nature of these transcripts and their relationship to MSF protein secretion revealed that the 5.9-kb mRNA is a second MSF-encoding transcript. Both these mRNAs have identical coding sequence and differ only in the length of their intron-derived 3′-untranslated region (UTR). The 5.9-kb MSF mRNA is retained in the nucleus whereas the 2.1-kb mRNA is not. MSF-secreting fetal fibroblasts have significantly lower nuclear levels of the 5.9-kb mRNA and correspondingly higher cytoplasmic levels of the 2.1-kb transcript than their nonsecreting adult counterparts. Adult fibroblasts induced to secrete MSF by treatment with transforming growth factor-β1 displayed similar changes in their respective levels of MSF mRNA, but not those of a control gene. When cloned downstream of a reporter gene, only the longer 3′-UTR retained coding sequence within the nucleus. We conclude that expression of MSF protein is regulated by 3′-UTR truncation of the 5.9-kb nuclear-sequestered “precursor” MSF mRNA and nuclear export of mature 2.1-kb message. Inducible 3′-UTR processing represents a novel regulatory mechanism involved in cancer pathogenesis that may open new avenues for therapeutic gene delivery.
Migration stimulating factor (MSF) is a truncated oncofetal isoform of fibronectin (1). The 2.1-kb MSF mRNA is transcribed from the Fn1 gene and is generated by read-through of the intron separating the 12th and 13th exons of this gene (exons III-1a and III-1b). As a consequence, MSF mRNA contains (a) a 1,941-bp sequence identical to the 5′ coding sequence of fibronectin, extending to and including exon III-1a, and (b) a contiguous intron-derived 30-bp coding sequence and 165-bp 3′-untranslated region (UTR) containing five in-frame stop codons and terminating in a poly(A) tail. The deduced 657-amino-acid sequence comprises the 70-kDa NH2 terminus of fibronectin and a unique (intron-derived) 10-amino-acid COOH terminus.
Immunohistochemical studies (using antibodies raised against the MSF-specific COOH-terminal decamer) indicate that MSF is expressed by keratinocytes, fibroblasts, and vascular endothelial cells in fetal, but not adult, skin (1). MSF is not significantly expressed by cells in healthy breast tissue but may be exuberantly expressed by mammary carcinoma cells as well as by tumor-associated stromal fibroblasts and endothelial cells. Recombinant MSF protein displays a number of potent bioactivities both in vitro and in vivo, including stimulation of cell migration, hyaluronan synthesis, and angiogenesis (2–6). MSF-specific bioactivity has been detected in the serum of >90% of breast cancer patients compared with ∼10% of healthy age- and sex-matched controls (7) and initial ex vivo observations suggest that elevated MSF expression in breast tumors at presentation is associated with poorer prognosis.1
Perrier et al., in preparation.
Using a MSF-specific probe (based on the 175 bp of intron-derived sequence at the 3′ end of the MSF mRNA transcript), Northern blot analysis of RNA extracted from MSF-expressing (fetal) and nonexpressing (adult) fibroblasts indicated the presence of both the 2.1-kb MSF transcript and a 5.9-kb RNA in the two cell types. In view of the unexpected nature of these findings, the objectives of this study have been to (a) characterize the 5.9-kb RNA, (b) understand its relationship to the previously cloned 2.1-kb MSF transcript, and (c) elucidate the mechanistic basis of the apparently discordant relationship between the expression of these two transcripts and the production of MSF protein.
Materials and Methods
Cells and reagents. The fetal (n = 5) and adult (n = 5) fibroblast lines used in this study were established and maintained as previously described (1).
Recombinant human transforming growth factor (TGF)-β1 was purchased from Sigma (Poole, Dorset, England, United Kingdom). The monoclonal anti–green fluorescent protein (GFP; JL-8) and polyclonal anti-fibronectin antibodies (A0245) were purchased from BD Biosciences (Palo Alto, CA) and DAKO (Ely, Cambridgeshire, England, United Kingdom), respectively. The two anti-MSF monoclonal antibodies were produced in-house. The first (mabVSI7.1) was raised against the MSF-unique 3′ decamer, VSIPPRNLGY, and was used for immunodetection. The second (mabPepQ5.1) was a function-neutralizing antibody and was raised against the bioactive IGD motif and flanking amino acids, TNEGVMYRIGDQWDKQHDMGH, located in the ninth type I module of MSF; mabPepQ5.1 effectively abrogated MSF mitogenic activity (1).
RNA extraction and Northern blots. RNA was extracted from the cytoplasm and nucleus separately and treated with DNase I, as instructed by the manufacturer (RNeasy kit: Qiagen, Crawley, West Sussex, England, United Kingdom). Northern blots were done nonisotopically using the Strip-EZ RNA, BrightStar-Psoralen-Biotin, NorthernMax-Gly, and BrightStar BioDetect-Biotin kits (Ambion, Austin, TX). Blots detecting GFP mRNA were done on polyadenylated [poly(A)] mRNA isolated from total RNAs using the MicroPoly(A)Purist kit (Ambion). Northern blot probes were generated by PCR and cloned into the pGEM-T Easy vector (Promega, Madison, WI), which was subsequently modified to include a CU Minus T3 promoter sequence (Ambion) to aid full-length probe manufacture.
Reverse transcription and PCR. PCR primer sequences for all reactions are provided in Table 1. These include Northern blot probe generation (probes U, GFP, and 1-4), mapping of the 5.9-kb MSF 3′-UTR [primer sets PCR 1-10 and poly(A)], and quantitative real-time PCR [primer sets MSF21, MSF59, GFP, human glyceraldehyde-3-phosphate dehydrogenase (GAPDH), and mouse GAPDH]. Reverse transcription was done at 42°C with a final concentration of 5 mmol/L MgCl2 using the ImProm II system (Promega).
|Primer set .||Oligomer (sense 5′-3′) .||Oligomer (antisense 5′-3′) .|
|Primer set .||Oligomer (sense 5′-3′) .||Oligomer (antisense 5′-3′) .|
The PCR used to identify the location of the 5.9-kb MSF mRNA poly(A) tail employed a sense primer in combination with two antisense primers (a and b, Table 1). The primers were used at an a/b ratio of 1:9. A total of 10 pmol of sense and antisense primers were used in this reaction and, apart from three initial cycles at annealing temperature 42°C, the remainder of the amplification was a standard touchdown PCR. Standard touchdown PCR using Platinum Taq polymerase (Invitrogen, Paisley, Scotland, United Kingdom) was used for all the other amplification reactions, apart from the quantitative PCR which was previously described (1).
PCR strategy for distinguishing 2.1-kb and 5.9-kb MSF cDNAs. The 2.1-kb MSF mRNA was distinguished from the 5.9-kb transcript on the basis of size by using an oligo-dT primer with a 5′ M13 sequence modification [5′-GTAAAACGACGGCCAGTG(T)18-3′] for reverse transcription followed by amplification using a coding sequence–specific primer in combination with an M13 sequence–specific primer (primer set: MSF21). The 5.9-kb MSF cDNA was amplified using primers (primer set: MSF59) amplifying intronic sequence <100 bp upstream of its poly(A) tail. All primer sequences are shown in Table 1. Independent amplification of the RNA used for the reverse transcription reactions confirmed the absence of contaminating genomic DNA.
Green fluorescent protein vector manufacture and transfection. The pEGFP-C1 vector (Clontech, Mountain View, CA) was modified using the upstream BsrGI site in combination with either the EcoRI site for inserting the 2.1-kb MSF 3′-UTR or the PstI site for the 5.9-kb MSF 3′-UTR. The appropriate sequences (terminal 7 bp of GFP and in-frame stop codon) were added by mutational PCR. The 5.9-kb MSF 3′-UTR was assembled by cloning three smaller PCR fragments. All constructs were sequenced in both directions at all stages of manufacture to ensure sequence fidelity. NIH 3T3 murine fibroblasts were transfected using FuGENE 6 as instructed (Roche, Lewes, East Sussex, England, United Kingdom). Experiments were initially done with uncloned lines. These were later repeated with individual subclones established from these lines. Individual clones performed in exactly the same manner as the parental line from which they were established.
Statistical analyses. Overall differences were assessed by the Kruskal-Wallis test and comparisons between individual groups were made using the two-tailed Mann-Whitney U test.
Identification and molecular characterization of a second MSF mRNA. The previously cloned MSF cDNA is a 2.1-kb transcript (Fig. 1A), the upstream 1,941 bases of coding sequence being derived from the first 12 exons of Fn1 and the following downstream 175 bases from the contiguous intronic sequence separating the 12th and 13th exons (1). The truncated size of the MSF message and the presence of the intron-derived 3′ sequence distinguish MSF from all other “full-length” Fn1 transcripts.
RNAs extracted from fetal (MSF-secreting) and adult (MSF nonsecreting) fibroblast lines were subjected to Northern blot analysis using the downstream intron-derived 175 bases of the 2.1-kb MSF sequence as a MSF-specific probe (probe U, Fig. 1B). Unexpectedly, this probe hybridized to RNA from both the MSF-secreting and the nonsecreting cell lines, picking out, in each case, two transcripts: one consistent with the previously identified 2.1-kb transcript and a more abundant larger one.
These RNAs were analyzed using a further series of probes: probe 1 (derived from coding sequence upstream of U probe) hybridized to fibronectin mRNA and 2.1-kb MSF mRNA (as predicted), as well as the new transcript. Two other probe sets were used to characterize the downstream sequence that accounts for the increased size of the new transcript. The first set, derived from downstream Fn1 coding sequence (probe 2), only hybridized to fibronectin mRNAs. A second set of probes (3 and 4) were based on the intron-derived sequence downstream of the 3′ 175 bases retained in the 2.1-kb MSF mRNA. Both probes hybridized solely to the newly identified mRNA (Fig. 1B). PCR (Fig. 1C and D) and sequencing revealed that the upstream 2.4 kb of this transcript included part of the Fn1 5′-UTR, spliced sequence from the first 12 Fn1 exons and a significantly larger segment of the retained intron than that found in the 2.1-kb MSF mRNA.
The full extent of the retained intronic sequence was mapped using a series of 10 overlapping PCR reactions (Fig. 2A), which were subsequently cloned and sequenced. PCR reaction 1 (Fig. 2B) using cDNA templates obtained from two representative fibroblast lines (one adult and one fetal) confirmed that exons I-9 and III-1a were spliced together upstream of the contiguous intron-derived sequence. This reaction confirmed that the retained intronic sequence extended further downstream than that found in the 2.1-kb MSF mRNA. The next eight reactions (space restrictions only allow reactions 2 and 9 to be shown; Fig. 2B) all gave the same result: the cDNAs, but not the RNAs, from both MSF-secreting and nonsecreting fibroblasts were amplified successfully, as was the genomic DNA control. These results confirm that the intronic sequence detected is genuinely part of the mRNA and not due to contaminant genomic DNA in the RNA preparation. PCR reaction 10 worked only on the genomic DNA control, indicating that the larger MSF mRNA had terminated before this point. A final PCR reaction using a modified oligo-dT primer identified the position of the poly(A) tail (Fig. 2C), which was the same for MSF-secreting and nonsecreting fibroblast mRNA (Fig. 2C and D) and was consistent with that predicted from the location of the canonical poly(A) signal sequence (10) located at this region of the intron (Fig. 2D). Sequence analysis of the entire cDNA (European Molecular Biology Laboratory accession no. AJ849445) revealed that (a) it is ∼5.9 kb long from its translational start codon to the beginning of its poly(A) tail, and (b) it encodes the same protein as the 2.1-kb MSF mRNA, differing from it only in the amount of intron-derived sequence within its 3′-UTR. Accordingly, this transcript is henceforth referred to as 5.9-kb MSF mRNA.
The subcellular distribution of 2.1-kb and 5.9-kb MSF mRNAs is associated with the expression of migration stimulating factor protein. In an attempt to elucidate the functional significance of both MSF mRNAs, we quantified the subcellular distribution of these transcripts in both MSF-secreting (fetal) and nonsecreting (adult) fibroblast lines (Fig. 3). Quantitative PCR of cDNA derived from nuclear and cytoplasmic RNA revealed that essentially all of the 5.9-kb MSF mRNA (over 97%) was confined to the cell nucleus, where it was ∼10-fold more abundant than its 2.1-kb counterpart. The 2.1-kb MSF transcript was also detected in cytoplasmic RNA at a level between 10% and 50% of that present in the nucleus. Significantly, the nuclear levels of the 5.9-kb MSF transcript were lower and the cytoplasmic levels of the 2.1-kb MSF transcript were higher (P = 0.0159, in each case) in the MSF-secreting fibroblasts compared with their nonsecreting counterparts. In contrast, the subcellular distribution of a control mRNA, GAPDH, did not significantly differ between MSF secretors and nonsecretors. These data suggest that MSF mRNAs levels vary in accordance with MSF protein expression status in fetal and adult cells.
We have previously shown that treatment of adult fibroblasts with TGF-β1 induces the expression of MSF.2
S.L. Schor et al., in preparation.
Processing of the 5.9-kb MSF mRNA 3′-untranslated region regulates protein secretion. These observations show that significant changes in the nuclear level of the 5.9-kb MSF mRNA and the cytoplasmic level of the 2.1-kb MSF mRNA are associated with the switch from a nonsecreting to a MSF-secreting phenotype. To ascertain the possible regulatory function of the 3′-UTR and the functional relationship between the two MSF mRNAs and protein expression, we fused both 3′-UTRs to a GFP reporter gene. The GFP-expressing vector pEGFP-C1 was modified so that either the 2.1-kb or 5.9-kb MSF 3′-UTR was introduced immediately downstream of the GFP coding sequence. A stop codon was introduced between the coding and noncoding sequences. These constructs along with the unaltered pEGFP-C1 vector were transfected into the murine fibroblast line NIH 3T3. Permanently transfected lines, selected by G418, were incubated with TGF-β1 or control medium. Lines transfected with either the unaltered vector (data not shown) or 2.1-kb MSF 3′-UTR (Fig. 5A) fluoresced regardless of TGF-β1 treatment. In contrast, 5.9-kb MSF 3′-UTR–transfected cells fluoresced very poorly until treated with TGF-β1. The heterogeneous fluorescence pattern exhibited by these cells persisted in subclones derived from these parental lines for at least 4 months in culture.
Western blot analysis (Fig. 5C) revealed that the GFP protein expressed by the 2.1-kb MSF 3′-UTR– and 5.9-kb MSF 3′-UTR–transfected lines was, as expected, 28 kDa and seemed to be unaltered by TGF-β1 treatment. The GFP protein expressed by vector-transfected cells is larger (32 kDa) because the cloning site is in-frame and adds 26 amino acids to the encoded GFP. This protein was unaltered by TGF-β1 treatment. Quantitative PCR analyses (Fig. 5B) revealed that the GFP mRNA concentration was the same in the nucleus and cytoplasm of vector-transfected cell lines. In 2.1-kb MSF 3′-UTR–transfected lines, the cytoplasmic GFP mRNA concentration was ∼75% of that found in the nucleus. The ratio of cytoplasmic to nuclear GFP mRNA levels in vector- or 2.1-kb MSF 3′-UTR–transfected lines was not significantly affected by TGF-β1 treatment. In 5.9-kb MSF 3′-UTR–transfected lines, <20% of the nuclear levels of mRNA were found in the cytoplasm, indicating substantial trapping of GFP mRNA in the cell nucleus. After TGF-β1 treatment, the cytoplasmic mRNA concentration more than doubled (P = 0.0286) to >50% of the level of nuclear mRNA. Relative levels of a control mRNA, encoding GAPDH, were the same in all these transfected lines and were unaffected by TGF-β1 treatment. Northern blot analyses revealed that no change in GFP transcript size occurred between the nucleus and cytoplasm or between TGF-β1–treated or untreated cells in either the vector- or 2.1-kb MSF 3′-UTR–transfected lines. However, evidence of mRNA processing (shortening of the nuclear 6-kb GFP transcript) could be seen in the cytoplasmic GFP mRNAs obtained from the 5.9-kb MSF 3′-UTR–transfected lines (Fig. 5D).
Fibroblasts cultured in vitro contain 2.1-kb and 5.9-kb species of MSF mRNA as detected by Northern blot analysis. Both species share an identical coding region consisting of the first 12 exons of the fibronectin gene (up to and including exon III-1a) and the immediately contiguous 175 bp of the intron separating exons III-1a and III-1b. They differ solely with respect to the length of the remaining intron-derived 3′-UTR, which plays a unique role in the regulation of MSF expression. We suggest that the expression of MSF protein is dependent on (a) initial transcription of the 5.9-kb mRNA and its sequestration within the nucleus, (b) cleavage of 5.9-kb mRNA 3′-UTR to generate the smaller 2.1-kb transcript, and (c) nuclear export of the 2.1-kb mRNA to the cytoplasm, where it may be translated (Fig. 6). Generation of the 2.1-kb MSF mRNA seems to be the critical step in regulating the nuclear export of the MSF coding sequence and its subsequent availability for translation into protein. This is in marked contrast to the direct relationship between mRNA transcription and protein production that has been described for full-length fibronectin isoforms (11, 12).
Significantly, our results indicate that all cultured fibroblasts examined transcribe MSF mRNA irrespective of whether these cells express MSF protein. This observation stands in marked contrast to ex vivo data, indicating that the majority of normal skin and mammary fibroblasts do not express MSF mRNA (as assessed by in situ hybridization) or protein (as assessed by immunolocalization; ref. 1). This apparent discrepancy is consistent with identified phenotypic differences between cells in vivo and their in vitro cultured counterparts in terms of the expression of various oncofetal stress response molecules, such as the extracellular domain B (EDB) isoform of full-length fibronectin (11). Such phenotypic differences are commonly ascribed to a “partial activation” of tissue cells by the in vitro culture environment, such as the presence of serum (a wound healing stimulus) and the artificial nature of the substratum (13, 14). Our observations accordingly suggest that the control of MSF expression (in vivo as well as in vitro) is regulated by a multistep process of cell activation characterized, at least in part, by the initiation of 5.9-kb MSF mRNA transcription and its subsequent truncation to the 2.1-kb species.
The 3′-UTR regulates mRNA stability by a number of well-characterized mechanisms. One such mechanism is mediated by cis-acting sequences such as the ARE instability sequence (15–17). In this respect, we have already reported that a canonical ARE instability sequence is present within the 2.1-kb MSF mRNA 3′-UTR. MSF mRNA is consequently over 30-fold more unstable than other Fn1 transcripts that do not contain this element (1). Further, mutational analysis has confirmed that this canonical sequence is indeed functional and that the 2.1-kb MSF mRNA 3′-UTR confers mRNA instability not only to MSF but also to other nonrelated coding sequences when attached immediately downstream of the stop codon.3
Kay et al., in preparation.
The 3′-UTR may have a profound effect on protein translation as a consequence of its effect on transcript stability, mediated through either cis-acting sequences (15–17) or structural mRNA elements providing cleavage site protection (18, 19). The 3′-UTR may also regulate transcript availability in either a positive or negative fashion (20–22) by several other mechanisms, including the regulation of nuclear export by 3′-UTR cis-acting sequences (23, 24) and mRNA sequestration into cytoplasmic mRNA ribonucleoprotein complexes during development (25). Significantly, all these reported facets of 3′-UTR functionality depend on alterations in trans-activating RNA binding proteins rather than on primary changes in the 3′-UTR itself (15–25). Reduced protein expression resulting from changes in 3′-UTR length has been described in cases of β thalassaemia (26) and myotonic dystrophy (27) but, to our knowledge, MSF represents the first instance of a physiologic mechanism whereby nuclear mRNA transport is critically regulated by inducible changes in the 3′-UTR. The cleavage event and its functional sequelae accordingly represent a novel physiologically relevant mechanism for controlling gene expression. The ability of the 5.9-kb MSF mRNA 3′-UTR to control the expression of a non-MSF coding sequence (GFP) suggests that this regulatory mechanism may occur in other gene systems and, indeed, may be clinically exploited in the development of novel targeted therapeutic delivery systems. In this regard, our data indicate that the 3′-UTR–mediated release mechanism seems to operate in a promoter/enhancer–independent fashion, thereby potentially affording more subtle control over the magnitude of therapeutic protein delivery than systems that rely on tissue-specific promoter control.
As is the case for previously characterized full-length oncofetal isoforms of fibronectin (e.g., those containing the extracellular domain B module), the transcription and translation of MSF message are regulated by both developmental and physiologic signals, such as TGF-β1 (28–30). Our data indicate that TGF-β1 increases 3′-UTR processing of the nuclear 5.9-kb MSF mRNA. However, TGF-β1 may also increase the generation of 5.9-kb MSF mRNA as its nuclear level decreases less (23.7%) than the corresponding increase of the cytoplasmic level of 2.1-kb MSF mRNA (63.6%) after TGF-β1 treatment. As the levels of the two MSF mRNA forms are predicted to be altered at an equimolar rate, this implies that some of the decrease in the nuclear mRNA reservoir may be offset by an increase in its production.
TGF-β1 can act both as a tumor suppressor and as a significant stimulator of tumor progression, invasion, and metastasis (31, 32). At early stages of tumor development, it acts directly to suppress cancer cell outgrowth but, as the disease progresses, its activities on both the cancer and surrounding stromal cells stimulate tumor progression (31, 33). Indeed, an investigation of TGF-β1 gene polymorphisms in breast cancer suggests that polymorphisms in the signal peptide, which increase TGF-β1 secretion, result in a moderate but significant increase in the risk of developing invasive disease (34). The ability of TGF-β1 to modulate MSF protein secretion may be one mechanism by which it contributes to the severity of tumor progression.
Although used as an exemplar in this study, TGF-β1 is not alone in its ability to stimulate MSF secretion. Other studies have shown that transfection of adult fibroblasts with constructs expressing dominant-negative p53 mutants also induces changes in MSF mRNA and MSF protein secretion,4
Jones et al., in preparation.
In summary, this article highlights a novel nuclear mRNA sequestration mechanism which uniquely regulates the secretion of the oncofetal cytokine MSF. It is anticipated that this information will contribute to our understanding of the mechanisms underpinning the reexpression of MSF during cancer progression and point to novel clinical intervention strategies for reversing this process.
Grant support: Engineering and Physical Sciences Research Council, NIH (USA), The Breast Cancer Campaign, Joint Research into Aging and Diabetes UK studentship, Tayside Area Oncology Fund, and Dundee University Anonymous Trust.
We thank Rene Burnett for her invaluable technical assistance.