EXPERIMENTAL

CELL

RESEARCH

193,

331-338 (19%)

Generation of Full-Length cDNA Recombinant Vectors for the Transient Expression of Human Fibronectin in Mammalian Cell Lines SYLVIEDUFOUR,*ALEJANDROGUTMAN,~FLORENCEBOIS,* NEDLAMB,$ JEANPAULTHIERY,*'~ AND ALBERTO R. KORNBLIHTT~ *Laboratoire

de Physiopathologie du Dtkeloppement, CNRS URA 1337, Ecole Normale Sup&ieure, 46, rue d’lJlm, 75230 Paris Cedex 05, France; $Centre de Recherche de Biochimie Macromol~culaire, CNRS-INSERM, 34033 Montpellier, France; and PINCEBZ-CONICET, Vuelta de Obligado 2490, 1428 Buenos Aires, Argentina

tides as a consequence of a complex pattern of alternative splicing of the precursor mRNA. Three regions of alternative splicing have been identified. From the carboxy to the amino termini, these regions are the type III connecting segment (IIICS), also called V, which maps between the penultimate and the last, type III repeats, it is encoded by a single exon which is subdivided to yield up to five mRNA variants (referred to as IIICS120, IIICS95, IIICS89, IIICSO, and IIICS64 for the amino acid length of the polypeptides that they encode); the ED1 or extra domain I, also called EIIIA or EDA, involving a 270-nucleotide exon encoding exactly one type III unit which may be alternatively included or excluded from mature FN mRNA; and the ED11 or extra domain II, also called EIIIB or EDB, involving a 273-nucleotide exon with properties similar to those of EDI. The alternative splicing of FN mRNA is cell-specific: ED1 and ED11 are always excluded from plasma FNs which are synthesized in hepatocytes, whereas the proportion of EDI+ and EDII+ FN mRNAs may vary in different cell types according to physiological and pathological conditions [5-81. The precise functions of ED1 and ED11 segments are still unknown. On the contrary, accumulated evidence confers specific cell-binding activities to the variable IIICS region [4, 9, 111. As a first step to study the function of the different binding sites, structural domains, and alternatively spliced segments of human FN without dissecting them from the whole molecule, we have developed a series of full-length human cDNA clones, suitable for the transient expression of FN by both transfection and microinjection techniques. Due to the large size of FN mRNAs (-8.0 kb) no full-length cDNA clones have been previously isolated. Instead, a series of overlapping short cDNA clones had to be assembled in order to obtain cDNA inserts encoding the different splicing variants of human preprofibronectin.

In order to study the roles of the different alternatively spliced variants of human fibronectin (FN) as well as of its binding sites and structural domains in the processes of extracellular matrix assembly, cell adhesion, and cell migration, we constructed expression vectors coding for different human full-length FN polypeptides and deleted versions of these constructs. We expressed them transiently in mammalian cells by calcium phosphate transfection and microinjection techniques. While the deleted recombinants were expressed by both transfection and microinjection, the full-length recombinants could be only expressed by microinjection. Calcium phosphate transfection leads to the accumulation of recombinant FN in cytoplasmic vesicles of both matrix-forming and nonforming cells. On the contrary, microinjection leads to the accumulation of recombinant FN in cytoplasmic vesicles in cells that do not form a matrix, but to the rapid incorporation into the extracellular matrix of matrix-forming cells in addition to a cytoplasmic localization. Identical results were obtained when the FN signal and propeptides were c 1991 Academic Press, Inc. replaced by those of E-cadherin.

INTRODUCTION Fibronectins (FNs) are high molecular weight glycoproteins that play important roles in cell contact processes by binding macromolecules such as collagen, fibrin, glycosaminoglycans, and themselves, as well as specific receptors (integrins) (for reviews see [l-4]). Each FN polypeptide of about 250,000 Da shows three types of internal sequence repeats, the so-called homology types I, II, and III, with average lengths of 40, 60, and 90 amino acid residues, respectively. In humans, a single FN gene gives rise to 20 polypep-

MATERIALS AND METHODS

’ Present address: Institut de Chimie Biologique, Facultb de MQdecine, Strasbourg, 67035 Cedex, France. * To whom correspondence and reprint requests should be addressed.

Numbering of human FN mRNA sequences. The longest human FN mRNA species (EDII+, EDI+, IIICS120) has 8392 nucleotides 331 All

0014.4827191 $3.00 Copyright 0 1991 by Academic Press, Inc. rights of reproduction in any form reserved.

DUFOUR

332 pXHFPst

1.0

pFH6

PUCI 34

pFH54

pFH60

PAT 75

c !ncs09

’

pFHlO0

/-

5665bp

TAA

lEDII:EDI;IIICS891

FIG. 1. Strategy for the generation of pFH100. All starting clones have been described in the references quoted in the text, except for pUC134 which is the result of subcloning the insert of pFHI34 [13] into pUC18. Abbreviations for restriction enzyme sites: A, Asp7181; B, BstXI; Ba, BarnHI; Bg, RglII; C, ClaI; E, EcoRI; Ea, EagI; EC, EcoRV; H, HindIII; N, NcoI; P, PstI; S, SalI; Sa, SacI; SC, ScaI; Sm, SmaI; St, StyI; T, TuqI; X, XhoI; Xb, XbaI; Xm, XmaI.

from the cap site (nucleotide number 1) to the poly(A) addition site. Within this, relevant sequences are located as follows: ATG, 268-270; EDII, 4063-4335; RGDS (cell-binding site), 5110-5121; EDI, 54315700; IIICS120, 6514-6873; and TAA, 769997701. For practical reasons, all restriction enzyme sites used in the successive subclonings were numbered according to this 8392.nucleotide-long mRNA, even in those cases in which alternatively spliced segments are absent from the clones. All the cDNA clones used thoughout this work code for the IIICS89 variant, and in consequence, they all lack the sequence between positions 6782 and 6873. Plasmid constructions: Full-length clones. We first constructed a full-length cDNA clone, pFH100, in the cloning vector pUC18, using a series of previously characterized cDNA clones: pFH60 [12], pFH54, pUC134, and pFH6 [13], as well as part of the genomic clone pXHF Pstl.0 [14], kindly provided by Dr. S. Bourgeois, The Salk Institute. The strategy to generate pFHlO0 (EDII-, EDI-, IIICS89) is depicted in Fig. 1. pFHlO0 contains an extra dodecanucleotide plasmid sequence between positions 346 and 347. This sequence introduces four extra amino acids (ISFR) within the propeptide between the residue 27 (serine) and the residue 28 (lysine) (151 without altering the rest of the reading frame and is particularly useful for the distinction between endogenous and transiently expressed hFN mRNAs by Sl nuclease protection (see below). Figure 2 shows the strategy used to generate full-length clones representing other splicing variants by fragment replacement: pFHlO1 (EDII-, EDI+, IIICS89), pFH102 (EDIIt, EDII, IIICS89), and pFH103 (EDII+, EDI+, IIICS89). Expres,sion uectors. The full-length FN cDNAs were put under the control of the SV40 early promoter and enhancer plus the cytomegalovirus enhancer. For this, a plasmid expressing the acid fibroblast

ET AL. growth factor (FGF), ~267.3 [16], was modified as shown in Fig. 3. Modifications included: (i) elimination of the SmaI site of ~267.3 by cutting and ligation of EcoRI linkers; (ii) replacement of the XhoIl BglII acid FGF cDNA insert by a double-stranded synthetic oligonucleotide of FN sequences 264-288, flanked by XhoI and BglII sites; and (iii) cloning of the SmaI (288)lScaI (8157) fragment from any of the FN full-length clones into the SmaI contained in the synthetic oligonucleotide. In this way, the full-length clones pFH100, pFH101, pFH102, and pFH103 gave rise to the expression clones pCEF100, pCEF101, pCEF102, and pCEF103, respectively. A deleted version of pCEF103, pCEFA3, was created by excision of a 3211-bp fragment between the XbaI (916) and Asp7181 (4161) sites, blunt-ended in by Klenow and religation. pCEFA3 codes for the signal and propeptides, the first 3: type I repeats, and the carboxy terminal half of hFN from the middle of the ED11 segment, without interruption of the translation reading frame. All the preceding expression clones contain only five nucleotides of the FN mRNA 5’ noncoding region. A variant of pCEF53 in which the 5’ noncoding region, pre- and propeptides, and first 3: type I repeats of FN were replaced by a segment containing 69 nucleotides of the 5’ noncoding region and the sequences coding for the pre- and propeptides (27 + 129 aa) and 16 amino acids of the mature mouse E-cadherin was obtained by subcloning the 582bp blunt-ended EcoRI fragment of pEM1 [17], kindly supplied by Dr. M. Takeichi, into the blunt-ended HindIIIIAsp7181 sites of pCEFA3. This clone is named pCE E-cadha3. The correct reading frame of all the sites affected by subclonings within coding regions was checked by Maxam and Gilbert sequencing. Cell cultures. Mouse NIH 3T3 fibroblasts and sarcoma S180, simian COST, and human HeLa cells were cultured in DMEM supplemented with 10% (v/v) fetal calf serum. Transfections. Transfections by the calcium phosphate method were done according to Gorman [18]. MicroinJection. Cells growing on glass coverslips were microinjetted with plasmid DNA essentially as described before [27]. After

l::sOg

pFHlO0

--

pFHlO21EOIl*, EDI-, IlIC89)

EDI 2 530 bp

pFHlO1

pFHl03lEDl~:

IEDU-

EDI’.llICS891

EDI*,IIICS89)

FIG. 2. Construction of clones pFH101, pFH102 and pFH103 from pFH100. The 2530-bp EcoRI fragment containing the ED1 sequence was derived from pFHll1 1191. The 476.bp SacIIBstXI fragment containing the ED11 sequence was derived from a human cDNA clone kindly provided by Dr. L. Zardi. Abbreviations are the same as in Fig. 1.

X

Bg L--S ATG

j

aFGF

Ba

-Ba +

OLlGONUCLEOTlDE

~267.3 (Sm+El

mesas

7881 bp

from

pFH103

Ba EQrh

pCEF103 Modifications FIG. 3. are the same as in Fig.

of’ the expression

clone

~267.3

and construction

of’the

full-length

hFN

expression

vector

pCEF103.

Abbreviations

1. CELL 1

CELL 3

CELL 2

IIICS 89

pCEF103

CELL 1

CELL 3

CELL 2

pCEFA3 IIICS 89

EDI

t

112 EDII

E-Cadherin ,,,a

pCE-ECadh-A3

CELL 3

CELL 1 J\ high affinity +

I

I fibrin H

FGDs

HHHm

w2 t

CELL 2

EDI

IIICS 89

112 EDII FIG. repeats:

FN polypeptides 4. ovals, type II repeats:

encoded by the expression wide rectangles, type III

vectors pCEF103, pCEF1\3, and pCE E-cadhA3. Symbols: narrow rectangles, type repeats. Cells 1, 2, and 3 indicate different cell binding sites along the FN polypeptide. 3.13

I

334

I)IIFOIJR

ET AI,. and by running a ‘7.5% polyacrylamidelNa Dod SO4 gel electrophoresis as previously described 1201. Immurzofluorc~sce~cr. Cells were fixed in 3.7% (v/v) formaldehyde in PRS for 1 h. In the case of intracellular staining, cells were permeabilized by treatment with 0.5% (v/v) Triton X-100 in 50 mM MES, pH 6.1, for 4 min. Indirect immunofluorescence was performed using as first antibodies either Ist9, a mouse monoclonal against human ED1 epitopes [21], or 137, a rabbit polyclonal against the 160.kDa fragment of human FN that also recognizes other vertebrate FNs.

RESULTS

-56n

INPUT PROBE

il 188n 32P-!Jp

PROBE PROTECTED BY: EXOGENOUS hFN mRNA

n

56n 32PU

ENDOGENOUS hFN mRNA

FIG. 5. Sl nuclease analysis of RNA-DNA hybrids. RNA was extracted from HeLa cells (lanes 1, 2, 5, and 6) and from Cos cells (lanes 3 and 4) nontransfected (lanes 1,3, and 5) or transfected with pCEFA3 (lanes 2 and 4) or pCEF103 (lane 6). HeLa cells were cotransfected with pMK16 SV40, a plasmid expressing the large T antigen of SV40. The 56.base protected band reflects the endogenous human or simian FN mRNA. The 18%base protected band reflects the exogenous FN mRNA.

Figure 4 shows the three different FN polypeptides encoded by the expression vectors pCEF103, pCEFA3, and pCE E-cadhA3. The first two contain hFN while the third contains mouse E-cadherin pre and pro sequences. When COST, HeLa, NIH 3T3, or S180 cells were transfected with either pCEFlO1 or pCEF103 by the calcium phosphate method, no immunofluorescence-reactive cells were observed. Immunofluorescence was carried out with the anti-human ED1 monoclonal antibody, Ist9, which showed negligible immunoreactivity with the nontransfected cells. Cells transfected with pCEF103 showed small amounts of exogenous FN mRNA as detected by Sl nuclease protection (Fig. 5, lanes 5 and 6). This assay distinguishes between endogenous and exogenous FN mRNAs. However, it does not determine whether the exogenous FN mRNA is full size or fragmented. Full-size exogenous FN mRNA would have a length similar to that of the endogenous FN mRNA and it would not be specifically detected by Northern blotting. By in uitro transcription of the insert of pCEFlO3 in a SP6 polymerase system and subsequent translation in a rabbit reticulocyte system (not

230 kDa--+

two steps of purification on cesium chloride gradients, plasmids were ethanol-precipitated twice and resuspended in injection buffer containing 50 mM Hepes, pH 7.25, 10 mM NaCl, supplemented with 0.5 mg/ml purified rabbit immunoglobulins from nonimmune serum, to act subsequently in the identification of microinjected cells. Cells were microinjected with between 0.1 and 0.5 mg/ml plasmid DNA. Three to five hours after microinjection cells were fixed and processed for immunofluorescence. RNA extraction and SI n&ease protections. Total RNA was extracted from confluent monolayers by the method of Favaloro et al. [26]. A 18%bp SmaIIStyI fragment of pFHlO0 was subcloned into the SmaIlXbaI sites of pSVL (Pharmacia). From this clone, a l-kb XmaIlEcoRV fragment was isolated, labeled by filling in with the Klenow enzyme and [otBBP]dCTP, and used as a probe to detect FN mRNA by Sl nuclease mapping. Hybridization in formamide (57°C) and subsequent steps were as described [19]. Cell cultures were Radioactive labeling of cells and detection of FN. labeled for 4 h with [%]methionine (25 &i/6 cm dish). The presence of newly synthesized FN was assessed in the conditioned medium and in the cell extract by immunoprecipitation with anti-FN antibodies

c

200kDa

e

116kDa

a--

92.5 kDa

144kDa--,

1

2

3

4

FIG. 6. Na Dod SO,/polyacrylamide gel electrophoresis analysis of the products labeled in uiuo with [?S]methionine from COST cells nontransfected (lanes 3 and 4) or transfected with pCEFA3 (lanes 1 and 2). Cell extracts were immunoprecipitated with an unrelated rabbit polyclonal antibody (lanes 2 and 4) or with the rabbit polyclonal anti-hFN antibody 137 (lanes 1 and 3). The 144.kDa band in lane 1 has the expected size of the product of pCEFA3.

VECTORS

FOR TRANSIENT

EXPRESSION

FIG. 7. Immunofluorescence staining of COST cells transfected with pCEFL3 (a) or pCE E-cadhA (h). Previously to the reactions with the first antibody, Ist9, cells were fixed and permeabilized with detergent (see Materials and Methods). The arrow indicates a cell in which matrix deposition of recombinant FN was observed.

shown), we demonstrated that our full-length constructs are able to synthesize full-size FN polypeptides, immunoprecipitable by both Ist9 and 137 antibodies. In consequence, the absence of transient expression of the protein in transfections with pCEF103 could be due to a poor transfection efficiency caused by the large size of the plasmid and/or by some difficulty in the in viva translation of the pCEF103 transcript. In order to overcome this problem, we constructed a deleted version of pCEF103, called pCEFA3, which is 3211 bp smaller than pCEF103 (see Figs. 3 and 4). This construct could be expressed in COST cells both at the RNA level (Fig. 5, lanes 3 and 4) and at the protein level (Fig. 6). Furthermore, staining with Ist9 of pCEFA3-transfected cells showed the accumulation of exogenous, deleted FN in cytoplasmic vesicles (Fig. 7a). Little or none of the deleted FN was incorporated into the extracellular matrix (ECM). Similar results were obtained when a construct in which the dodecanucleotide plasmid se-

OF FIBRONECTIN

335

quence (see Materials and Methods) was eliminated (not shown). This means that the lack of incorporation into the matrix of the product of pCEFA3 is not due to the inclusion of four extra amino acids within the pro sequence. To test whether the FN pre and pro sequences were determinant in the intracellular accumulat,ion of deleted FN, we transfected COST cells with pCE-EcadhA3, in which the pre and pro sequences of FN were replaced by those of mouse E-cadherin. In this experiment, the number of cells producing deleted FN and the level of deleted FN synthesized by each cell were higher than with pCEFA3. Furthermore, although most of the exogenous FN was located in intracellular vesicles (Fig. 7b), some cells incorporated deleted FN into their matrices (Fig. 7b, arrow). HeLa cells transfected with pCEFA3 or pCE E-cadhA showed only vesicular accumulation of exogenous FN (not shown). In this case, FN expression vectors were cotransfected with a plasmid expressing SV40 large T antigen in order to allow the replication of the pCE constructs, mimicking in this way the situation of COST cells. Cos cells synthesize large amounts of endogenous FN. This is shown at the mRNA level by the strong 56-nucleotide band observed in the Sl nuclease protection assays (Fig. 5, lanes 3 and 4). Despite this, their FN matrix is scarce (Fig. 7b). In order to assess the expression of our recombinants in a matrix-producing cell line, we transfected NIH 3T3 cells with pCE-EcadhA3. The results of this kind of experiment are in Fig. 8 and show that the monoclonal Ist9 does not stain NIH 3T3 cells (Fig. Sa), while the polyclonal 137 recognizes their matrix (Fig. Sb). pCE E-cadhA3-transfected cells express deleted FN in cytoplasmic vesicles and a small proportion of it is incorporated into the preexisting matrix (Figs. 8d and Se, see arrows). Since in the above experiments the efficiency of transient expression was low and most of exogenous FN did not reach its normal localization in the matrix, we decided to try other methods of introducing DNA into cells. DEAE-dextran-mediated transfection and lipofection [23] gave exactly the same results as calcium phosphate transfections. However, microinjection of our recombinants gave substantially different results. When pCE E-cadhA was microinjected in NIH 3T3 cells, some cells showed the already observed intravesicular staining, while others showed good incorporation of deleted FN into the matrix (not shown). Unexpectedly, microinjection of full-length constructs like pCEF103 produced this time expression of full-length human FN in NIH 3T3 cells and full incorporation into the ECM (Fig. 9a). IstS-positive fibrils appear as soon as 4 h after cytoplasmic microinjection of the plasmid pCEF103. Microinjection of pCEFlO3 in mouse sarcoma S180

336

DUFOUR

ET AI,.

FIG. 8. Double immunofluorescence staining and phase-contrast of NIH 3T3 cells nontransfected (a-c) or transfected with pCE Em cadhA (d-f). Cells were stained with Ist9 (a, d) and 137 (b, e) antibodies after fixation and permeabilization with Triton X-100. The arrows indicate FN fibrils containing recombinant hBN, recognized by both lst9 and 137.

adherent cells, that synthesize little or no FN and do not form FN-ECM, led to the intracellular accumulation of full-length FN with high efficiency. Cytoplasmic localization of FN was observed both when the plasmid was injected in the nucleus or in the cytoplasm (Figs. 9b and 9c).

DISCUSSION We have constructed several recombinant vectors for the transient expression of full-length human FN in mammalian cells. These constructs include all the possible combinations of absence and presence of EDI and

VECTORS

FOR TRANSIENT

EXPRESSION

OF FIBRONECTIN

337

the exogenous FN was carried out by indirect immunofluorescence with Ist9, a monoclonal that recognizes human ED1 sequences coded for by our recombinants,

but that shows negligible immunoreactivity

FIG. 9. Microinjection of matrix-producing (NIH 3T3) and nonproducing (S180) cells with recombinant plasmids expressing hFN. (a) NIH 3T3 cells were microinjected in the nucleus with pCEF103 and stained with Ist9, after detergent permeabilization. (b, c) S180 sarcoma cells were microinjected with a mixture ofpCEF103 and normal rabbit IgG into the nucleus (arrow) or into the cytoplasm. Four hours later the cells were fixed, permeabilized, and double-stained with a Huorescein-labeled anti-rabbit IgG antibody (b) and with Ist9 and a rhodamine-labeled anti-mouse antibody (c). Staining in b allows identification of the microinjected cells and the site of microinjection, as revealed by the nuclear staining ofthe cell marked with an arrow in b.

ED11 exons, which are alternatively spliced in uivo in a cell-specific manner, but whose functions are still unknown. Others have successfully produced full-length rat FN cDNA clones by a strategy involving in uivo cDNA synthesis [24]. We had to assemble a series of short human cDNA segments due to the lack of longer clones produced by primary cDNA synthesis. Three transfection methods currently used for the transient expression in mammalian cells, such as calcium phosphate precipitation, DEAE-dextran, and lipofection proved to be ineffective for the transient, expression of our full-length human FN cDNA clones. However, microinjection allowed us to get expression of exogenous full-length hFN both in matrix-producing (NIH 3T3) and nonproducing (Y180) cells. Detection of

with endoge-

nous FNs from HeLa, COST, NIH 3T3, or S180 cells. Deleted versions of hFN, similar to the rat deminectins expressed by Schwarzbauer et al. [22, 251, could be expressed in our hands by both transfection and microinjection methods. Recombinant deleted hFN coded for by pCEFA3 or pCE E-cadhA tended to accumulate in cytoplasmic vesicles both in matrix-forming (NIH 3T3) and nonforming (SlSO) cells. In NIH 3T3 cells we also observed a minor incorporation of recombinant deleted hFN into the preexisting matrix. Accumulation of recombinant FN intravesicularly implies that endoplasmic reticulum translocation has occurred. This abnormal localization could be due to an incomplete processing of pre- and/or propeptides. Alternatively, if matrix formation is very sensitive to the rate of FN synthesis, and this rate is too high in each transfected or microinjected cell, the system might be overwhelmed, accumulating FN in transport vesicles or perhaps diverting it to wrong vesicles, with little incorporation into the matrix. Whatever the case, the abnormal localization observed does not depend on the nature of the FN pre and pro sequences since similar results were obtained with recombinant FNs linked to the E-cadherin pre and pro sequences. The role of the FN pro sequence is still unknown. Our results involving its replacement by a different pro sequence (E-cadherin) or its modification by the insertion of four extra amino acids (see Plasmid constructions) would indicate that it does not play a relevant role in determining the final localization of FN. Recently, Guan et al. [24] have reported the expression of full-length recombinant rat FNs by stable transfectants of NIH 3T3 cells and lymphoid WEHI 231 cells. The latter do not synthesize endogenous FN. Their stable transfectants secrete recombinant FN but do not assemble their own matrix. Similar results were obtained in this paper when microinjecting pCEF103 into adherent sarcoma S180 cells. In these cells there is little or no endogenous FN synthesis. Transient expression of pCEF103 in S180 cells must lead to secretion of recombinant FN, since no matrix formation is observed. In the same microinjection conditions, pCEF103 leads to the incorporation of recombinant hFN into the preexisting matrix of NIH 3T3 cells. The microinjection transient expression system described here provides an important tool for studying the molecular mechanisms of matrix formation, as well as its kinetics. Microinjection of plasmids coding for partially deleted or mutated FN variants, together with vectors expressing integrins, will allow us to assess the

338

DUFOUR

role of the different biologically active sequences of FN in its secretion and matrix formation. A good example of the kind of kinetic experiments that can be performed with a transient expression systern is illustrated by the result in Fig. 9a, where we determine that the time needed for transcription, transport of the mRNA to the cytoplasm, translation, membrane translocation, secretion, and deposition of FN in the ECM is less than 4 h. Results reported here will open to us and others a new way to investigate the multiple biological functions of fibronectins., We thank Drs. Suzanne Bourgeois, Masatoshi Takeichi, and Michael Jaye for providing the plasmid pXHF Pstl.0, pEM1, and ~267.3, respectively. We also thank Dr. Luciano Zardi for the gift of Ist9 and the unpublished cDNA clone carrying the human EDII, and Dr. Kenneth Yamada for the synthetic oligonucleotides used in this paper. This work was supported by the CNRS and the “Fundacion Antor chas” and was part of the French/Argentine scientific cooperation agreement between the CNRS and the CONICET.

REFERENCES 1. 2.