Sufficiency of the Reactive Site Loop of Maspin for Induction of Cell-Matrix Adhesion and Inhibition of Cell Invasion: Conversion of Ovalbumin to a Maspin-like Molecule*

and mg/ml Matrigel. MDA-MB-231 cells (5 x 104) were incubated in the absence or presence of 0.5 m M rHIS/FLAG maspin, ovalbumin, the maspin/ovalbumin chimeras, the R340Q and R340A maspin mutants or the RSL peptide in Leibovitzs L-15 medium containing 1% FBS, 1x MITO+ and 10 m g/ml ciprofloxacin. After 48-hr of incubation at 37oC in a humidified incubator without CO2, the cells in the bottom well were labeled with the fluorogenic vital dye calcein AM (5 m M final concentration). Fluorescence was measured using a CytoFluor (cid:228) fluorescence microplate reader (Millipore) with Ex: 480 nm and Em: 530 nm filters. Cells in duplicate wells without transwell inserts served as controls for cell proliferation and/or death during the incubation period. Invasion was calculated by dividing the relative fluorescence value of invading cells by that of total cells plated in duplicate wells without transwell inserts. Invasion of the control was set at 100%. The experiments were repeated 4-5 times with 3-5 replicates per experiment. Substitution at the putative P1 retained inhibitory activity while this activity was lost in the Ala mutant (RA). The ovalbumin maspin/ containing the RSL of maspin (OV/Masp-RSL/OV C) inhibited invasion of the MDA-MB-231 mammary gland carcinoma cells further establishing the role of the RSL. These results suggest inhibition of carcinoma cell invasion by maspin utilizes the P10-P5’ residues within the RSL. The RSL peptide was tested and the results confirmed a functional role of this region. Similar to the effect on increased cell-ECM adhesion, this peptide was sufficient to inhibit invasion of the carcinoma cells through Matrigel. maspin with and and keratinocyte


INTRODUCTION
Maspin is a 42 kDa protein synthesized by normal epithelial cells of a variety of mammalian organs such as mammary gland, prostate, skin, and cornea (1), (2). Synthesis of maspin has also been identified in the nonepithelioid cells of the corneal stroma, both in situ and in cell culture (2). The expression of maspin however, is lost after the second passage of cultured stromal cells. These later passage cells mimic the wound activated stromal fibroblasts which are much more mobile than stromal cells in the normal corneas. Expression of maspin is also lost or down regulated in many invasive carcinoma cells (3,4). Down regulation of maspin in carcinoma tissues correlates with progression and metastasis of tumors (5)(6)(7). Both the later passage corneal stromal cells and the invasive carcinoma cells respond to exogenously added maspin (2,3).
Several biological activities of maspin have been characterized, which suggest a role for maspin as a tumor suppressor and an inhibitor of angiogenesis. Addition of recombinant maspin or transfection of the maspin gene into carcinoma cells inhibits cell migration and invasion in vitro and reduces tumor growth and metastasis in vivo (3,8). Maspin also inhibits the in vitro migration and proliferation of endothelial cells, and blocks bFGF-induced neovascularization in the rat corneal pocket model (9). To date, the underlying mechanisms of maspin action on inhibition of tumor invasion and angiogenesis are not well established. Nonetheless, maspin can regulate adhesion of cultured corneal stromal cells and carcinoma cells to extracellular matrix (ECM 1 ) molecules (2,10). Pretreatment with recombinant maspin increases adhesion of cultured corneal stromal cells to several ECM molecules, including type I and type IV collagen, laminin and fibronectin (2), whereas it induces adhesion of carcinoma cells only to fibronectin and not to gelatin, laminin, type I and type IV collagens or fibrinogen (10). Stimulation of cell-ECM adhesion by maspin likely involves a mechanism by which maspin upregulates expression of integrins since the level of α5 integrin (the α component of the fibronectin receptor) on the cell surface is induced in carcinoma cells pretreated with recombinant maspin (10).
Maspin shares sequence homology with the serpins (serine protease inhibitors) of the ovalbumintype subfamily, which includes ovalbumin, plasminogen activator inhibitor-2 (PAI-2), squamous cell carcinoma antigen (SCCA) and bomapin (PI10) (11). Most serpins are inhibitors of specific proteases that react with an exposed reactive site loop (RSL) at the top of the molecule (Fig. 1). Whether maspin is a protease inhibitor is still controversial. Early in vitro studies showed that maspin does not inhibit a number of representative proteases (12). One recent study demonstrated maspin can inhibit prostate carcinoma cell surface-associated plasminogen activation by urokinase type plasminogen activator (uPA) (13), however, another study, using uPA bound to its receptor on tumor cells and tPA bound to vascular smooth muscle cells, fibrin or the prion protein, reported maspin does not inhibit either of these two plasminogen activators (14).
Although an intact RSL of maspin is required for inhibition of migration and invasion of mammary gland and prostate carcinoma cells (8), nothing is known about which region of the maspin molecule is required for induction of increased cell-ECM adhesion. Preincubation with blocking antibodies to the RSL of maspin (8), deletion from the P7' residue of the RSL to the C-terminal end (15), and cleavage of maspin at the putative P1 site Arg within the RSL (16) destroyed the ability of maspin to regulate the migration and invasion of carcinoma cells. We hypothesized that stimulation of cell-ECM adhesion by maspin also utilizes the RSL.
We investigated the structure-function relationship of maspin using maspin/ovalbumin chimeras rather than using maspin deletion mutants as previously studied, due to a concern about proper folding of the mutant molecules. A tertiary structure model of maspin based on the crystal structure of ovalbumin (PDB: 1OVA) suggests maspin adopts a common serpin structure in which the C-terminal region (Strands B4 and B5) lies inside the molecule (Fig. 1). Deletion of this region could disrupt the overall serpin conformation resulting in an inactive molecule whereas swapping equivalent regions between maspin and a homologous serpin would potentially preserve the overall structure. Expression of yeast recombinant maspin/ovalbumin chimeras as secreted proteins increases the probability the proteins are properly folded. Ovalbumin was chosen for swapping domains with maspin because the physical properties of both molecules are similar. Maspin and ovalbumin are similar in size and share about 30 % amino acid sequence identity. Neither molecule undergoes a urea-induced unfolding transition from a stressed (S) form to a heat-stable relaxed (R) form (12,17). Furthermore, the hinge region (P8 to P12) of ovalbumin and maspin differs from that of inhibitory serpins in which a multiple alanine stretch is conserved (18). Most importantly, unlike maspin, ovalbumin does not induce cell-ECM adhesion (2).
In this study, we explored the function of the RSL domain of maspin on cell-ECM adhesion and tumor invasion using a region swapping approach between maspin and ovalbumin to preserve the serpin structure. The studies reported here document that 1) the maspin RSL domain, but not the C-terminal region, is required, 2) the RSL peptide is sufficient for induction of increased cell-ECM adhesion of corneal stromal cells and carcinoma cells and inhibition of carcinoma cell invasion, 3) replacement of the RSL of ovalbumin with that of maspin can convert ovalbumin to a maspin-like molecule and 4) the RSL can compete for specific binding of maspin to carcinoma cells.

Cell Cultures
The MDA-MB-231 cells were maintained in Leibovitz's L-15 medium with 10 % FBS and 10 µg/ml ciprofloxacin at 37 o C without CO 2 . The mouse T-lymphocyte E.G7-OVA cells were grown at 37 o C with 5 % CO 2 in RPMI 1640 medium with 2 mM L-glutamine, 18 mM sodium bicarbonate, 25 mM glucose, 10 mM HEPES, 1.0 mM sodium pyruvate, 0.05 mM 2-mercaptoethanol, 0.4 mg/ml G418 and 10% FBS. The immortalized normal human corneal stromal cells (19), which mimic activated migratory wound corneal stromal cells and have the same properties as passaged human corneal stromal cells, were routinely grown in D-MEM medium containing 5% FBS, 1x MITO + and 10 µg/ml ciprofloxacin in a humidified incubator with 5% CO 2 at 34 o C.
The MDA-MB-231 human mammary carcinoma cells and immortalized corneal stromal cells do not synthesize maspin in contrast to normal mammary epithelial and corneal epithelial cells as determined by RT-PCR and Western blots using previously reported conditions for these assays (2).

Construction of Maspin/Ovalbumin-swap Mutants
The human maspin gene was previously obtained from human corneal epithelium by RT-PCR as described by Ngamkitidechakul et al. (2). Total RNA from the T lymphocyte cell line containing the ovalbumin gene was extracted using TRI-Reagent. The full-length ovalbumin gene was specifically amplified by RT-PCR using the ProSTAR™ Ultra HF RT-PCR System according to the manufacturer's procedure (Stratagene) in the Mastercycler ® personal thermocycler (Eppendorf, Westbury, NY). In the RT step, a specific ovalbumin antisense primer (Table I) was used to generate ovalbumin cDNA. PCR amplification of the cDNA was performed using two primers in which the EcoR I site was incorporated into the 5' end of the sense primer and Bgl II into the 5' end of the antisense primer. The annealing temperature for the PCR was 65 o C. The gel-purified 1.2 kb PCR product was digested with EcoR I and Bgl II, and ligated into the YEpHF expression vector (20). DNA sequencing at the MCW Protein and Nucleic Acid Facility confirmed the nucleotide sequences of both human maspin and chicken ovalbumin cDNA.
Three ovalbumin (OV)/maspin-RSL swap mutants and one chimeric maspin/OV-RSL swap mutant were constructed by the overlap extension PCR method of Ho et al. (21) using the primers listed in Table I. For each mutant construction, two sets of primers were utilized to amplify two separate overlapping PCR products that contain the mutation (or insertion) together with either upstream or down stream sequences.
These PCR products were amplified using PfuTurbo ® Hotstart DNA polymerase according to the manufacturer's instructions using the optimal annealing temperature at 60 o C. Two PCR fragments were joined together through overlapping sequences and PCR using primers from the 5' and 3' ends of the appropriate gene. An annealing temperature of 55 o C was used. The full-length PCR products were cut with EcoR I and Bgl II and cloned directly into the yeast expression vector YEpHF. The MCW Protein and Nucleic Acid Facility confirmed the reading frame and sequences of all mutants.
To construct a site-specific P1 R to Q and R to A maspin mutants, the YEpHF-Maspin template vector was directly amplified by the method of Weiner et al (22) using two primers containing the mutation sequences (Table I) and PfuTurbo ® Hotstart DNA polymerase. After 18 cycles of PCR amplification (55 o C annealing), the PCR reaction was treated with Dpn I (1 U) at 37 o C for 1 hr to remove the parental template. The purified PCR products were transformed into JM 109 E. coli cells. A colony containing the plasmid with the desired mutation was chosen and the mutated sequences confirmed by sequencing.

Expression and Purification of Yeast Recombinant HIS/FLAG-tagged Proteins
All recombinant proteins were produced as N-terminal HIS/FLAG tag fusion proteins in a yeast system. The yeast expression vector YEpHF was modified from the original vector YEpFLAG-1 by adding a six-histidine tag upstream of the FLAG peptide tag (20). Due to the presence of an α-factor leader peptide sequence, the recombinant proteins were expressed and secreted outside the S. cerevisiae yeast BJ3505 into the buffered expression medium (2% peptone, 1% yeast extract, 3% glycerol, 1% glucose and 100 mM K 2 HPO 4 , pH 6.4). After growing at 30 o C for 72 hrs, 1 liter of the culture medium was collected and concentrated to 200 ml by ultrafiltration (Amicon) under nitrogen gas at 60 psi. The concentrate was pH-adjusted to 8.0.
The yeast rHIS/FLAG-tagged protein was purified by mixing the concentrated yeast conditioned medium with 10 ml of Ni-NTA™ agarose resin overnight at 4 o C. The resin was washed with 50 mM phosphate buffer, pH 8.0 containing 300 mM NaCl and 10 mM imidazole. Subsequently, the recombinant protein was eluted with 50 mM phosphate buffer, pH 7.0 containing 300 mM NaCl and 250 mM imidazole. The eluted protein was concentrated using an Ultrafree spin concentrator (Millipore), and dialyzed in phosphate buffered saline (PBS) pH 7.4 to remove the imidazole. Identical results were obtained in the presence and absence of the N-terminal His tag and the FLAG sequence for the maspin functions measured here.

Arginine-specific Proteolysis of Recombinant Maspin and the R340Q and R340A mutants
Endoproteinase-Arg-C (clostripain) was incubated with the yeast recombinant wild-type maspin the R340Q and R340A mutants to confirm the mutagenesis. The enzyme was preactivated at room

In Vitro Tumor Cell Invasion Assay
The invasion assay was performed using the standard Matrigel method of Sternlicht et al. Competition assays of cell-maspin binding were conducted using 4 µM fluorescent-labeled maspin. The RSL peptide at 2, 10, 20-or 100-fold excess was added to cells along with BODIPY-TX-R-maspin. Non-specific binding was determined by adding a 50x excess of unlabeled maspin. Bound labeled maspin was quantified in the presence and absence of each competitor as described above. The experiments were repeated 3 times with 3-5 replicates per experiment.

Statistical Analysis
Overall differences among the treatment groups were determined using a one-way analysis-ofvariance and differences between individual treatments were determined using the Student-Newman-Keuls test by SigmaStat software (SPSS Inc., Chicago, IL).

RESULTS
To elucidate the significance of the RSL in the induction of increased cell-matrix adhesion by maspin, we constructed chimeric mutants between maspin and the homologous non-inhibitory serpin, ovalbumin, a molecule that does not alter adhesion of corneal stromal cells to ECM molecules (2). The RSL to the C-terminal end (aa 331-375), the C-terminal region downstream of the RSL (aa 346-375) and the RSL of maspin (aa 331-345) were individually replaced using the equivalent portions of ovalbumin (Fig.   1).

Effect of OV/Maspin-RSL Chimeras on Cell-ECM Adhesion
In contrast to maspin, the Maspin/OV-RSL/OV-C swap mutant containing the RSL to the Cterminal end of ovalbumin did not stimulate adhesion of either the corneal stromal cells to type I collagen ( Fig. 2A) or the carcinoma cells to fibronectin (Fig. 2B). However, when only the C-terminal domain of ovalbumin (Maspin/Masp-RSL/OV-C-Term) was swapped into the maspin molecule, adhesion was stimulated in both cell systems with all ECMs, suggesting the C-terminal portion of maspin is not directly associated with this activity. In contrast, replacement of the RSL of maspin alone with that of ovalbumin (Maspin/OV-RSL/Maspin-C-Term) abolished the effect on adhesion in both the corneal stromal cells to type I collagen ( Fig. 2A) and the carcinoma cells to fibronectin. Therefore, the RSL of maspin is required for enhancement of cell-ECM adhesion.

Effect of Mutation of the Putative P1 Residue, R340 on Cell Adhesion
To determine whether arginine at the putative P1 site within the RSL of maspin is required for maspin activity on cell-ECM adhesion, we mutated the R340 at the P1 site to a Q in one mutant and to an A in a second mutant. The R340Q mutant and wild type maspin, increased adhesion of corneal stromal cells to type I collagen (Fig. 3A) and of mammary carcinoma cells to fibronectin (Fig. 3B). The R340Q mutant was significantly more active than maspin for the carcinoma cells. In contrast, the R340A mutant lost activity. This mutant failed to induce adhesion of the corneal stromal cells (Fig. 3A) but showed intermediate activity between maspin and the control for the carcinoma cells (Fig. 3B). This loss of activity is not due to a major change in conformation because limited proteolysis using endoproteinase-Arg-C over 6 hr showed the same degradation pattern for this mutant as that for the active R340Q mutant. Thus, Gln can replace Arg at the putative P1 site but Ala cannot fully substitute for Arg at this site.

Effect of the Maspin RSL Polypeptide on Cell-ECM Adhesion
Since the RSL of maspin is required for the increase in adhesion of cells to ECM, we next examined whether the RSL is sufficient to induce this activity. A synthetic polypeptide corresponding to the 15 amino acid residues (P10-P5', aa 331-345) within the RSL of maspin was assayed to determine its effects on adhesion of cells to ECMs. The RSL peptide induced adhesion similar to maspin for both the corneal stromal cells and the carcinoma cells (Fig 4A and   B).

Conversion of Ovalbumin to a Maspin-like Molecule
To further characterize the role of the RSL of maspin, we generated and tested a chimeric ovalbumin mutant in which the RSL of maspin was exchanged for the RSL of ovalbumin. As shown in Fig. 5, the Ovalbumin/Masp-RSL/OV-C-Term was also able to stimulate adhesion of stromal cells to type I collagen and carcinoma cells to fibronectin. Thus, ovalbumin can be converted by the replacement of only the RSL into a maspin-like molecule that induces cell-ECM adhesion.

Stromal Cells to Laminin and Fibronectin
Maspin not only stimulates adhesion of human corneal stromal cells to type I collagen but also to laminin and fibronectin (2). To determine whether the RSL of maspin is critical for the stimulation of maspin adhesion to multiple extracellular matrix molecules, two mutants, the Maspin/OV-RSL/Masp-C and the Ovalbumin/Masp-RSL/OV-C were chosen for use in the adhesion assays. The ability of maspin to stimulate adhesion of the corneal stromal cells to fibronectin and laminin in addition to type I collagen was lost upon removal of the maspin RSL and its replacement with the RSL of ovalbumin (Fig. 6). In addition, the RSL of maspin was sufficient to convert ovalbumin into a molecule that can stimulate adhesion of the corneal stromal cells not only to type I collagen but also to fibronectin and laminin. Thus, the RSL of maspin is critical for the stimulation of adhesion of human corneal cells to multiple extracellular matrix molecules.

Polypeptide on Carcinoma Cell Invasion
The RSL of maspin was previously shown to be important for inhibition of carcinoma cell invasion (8,15) (16). Deletion of the P7' residue through the C-terminal end of maspin (aa 347-375) and trypsin cleavage of maspin at the P1 site Arg within the RSL resulted in loss of the activity. However, these modifications can potentially alter the overall structure of maspin leading to an inactive molecule. To better demonstrate the requirement of the RSL of maspin for inhibition of carcinoma cell invasion, we tested the maspin/ovalbumin chimeras using an in vitro tumor cell invasion assay. As predicted, the RSL of maspin is required because substitution of the RSL of maspin with that of ovalbumin (Masp/OV-RSL/Masp C) abolished the activity (Fig. 7). Substitution of Arg at the putative P1 site with Gln (RQ)

Determination of the Kinetics of Maspin Binding to Mammary Gland Carcinoma Cells and the Ability of the RSL Peptide to Inhibit Maspin Binding
The ability of the RSL peptide to mimic maspin activity led us to hypothesize that the effects induced by maspin are initiated by maspin binding to the cells through the RSL. Maspin binding to mammary gland carcinoma cells is specific and saturable ( Figure 8). The data fits best for a one binding site model with an r 2 = 0.98. The k d determined by non-linear regression was 367 ± 67 nM and the B max was 5.44 ± 0.24 pmol/10 5 cells. The calculated number of maspin binding sites per cell was 32.0 ± 2.2 x 10 6 . These values indicate that maspin binds to the carcinoma cells through a high number of low affinity sites. Using near saturating amounts of labeled maspin (4µM), the RSL peptide even at 2x (8µM) competed for specific binding of maspin to the carcinoma cells (Fig. 9). The degree of inhibition of specific maspin binding did not significantly change from 2 and 100x of the RSL peptide. These results suggest that maspin can bind to the cell surface through the RSL.

DISCUSSION
In the present study, we show that the RSL of maspin is required and sufficient to induce cell-ECM adhesion in two different cell types and for the inhibition of invasion of carcinoma cells. In contrast, mutation of the Arg 340 to an Ala residue results in the loss of these activities. This would indicate that the positive charge of Arg does not form a critical ionic bond but the side chain of this residue probably is involved in the formation of a critical hydrogen bond. Our putative maspin model predicts the side chain of Arg 340 is exposed suggesting an intermolecular interaction of this residue.
Inhibitory activity toward target proteases is lost or decreased upon mutation of a charged P1 site residue of inhibitory serpins to non-charged residues (24)(25)(26). If Arg is in fact the P1 site of maspin, the lack of an effect of the R340Q mutant on increased cell-ECM adhesion suggests inhibition of proteases such as uPA and tPA by maspin does not contribute to the induction of adhesion by maspin. A similar conclusion was drawn in a recent study, which showed that although maspin inhibits migration of vascular smooth muscle cells, it does not inhibit tPA bound to these cells nor uPA bound to tumor cells (14).
In contrast to the effect of substitution of Arg 340 with Gln on cell-ECM adhesion and invasion, the putative P1 Arg of maspin cannot be substituted with this residue for inhibition of carcinoma cell migration (9). Although both this study and ours used multiple cell types and ECMs, the results within each set of experiments are consistent. The migration of the mammary carcinoma cells through Matrigel coated membranes was inhibited by maspin but not by an R340Q mutant (9). The same results were shown for endothelial cell migration through membranes coated with gelatin and skin fibroblast and keratinocyte migration though non-coated membranes. The R340Q mutant retained the ability to stimulate adhesion of the carcinoma cells to fibronectin and the corneal stromal cells with type I collagen, fibronectin and laminin (data not shown). This mutant also inhibited invasion of mammary carcinoma cells through Matrigel, the same matrix used for the migration assay. Thus, the differences probably are due to the mode of interaction of maspin with the cells rather than the extracellular matrix molecule present. In the adhesion assay, the cells were preincubated with maspin for 18 hrs, the medium removed, the cells released with EDTA and replated in serum-and maspin-free medium on ECM coated wells for one hr (10).
In the invasion assay, the cells were mixed with maspin, placed into the Transwell inserts and allowed to migrate for 48 hrs. In the migration assay, maspin was placed in the chamber on the opposite side of the membrane from the cells (9), which results in the establishment of a concentration gradient across the membrane. Thus, the observed differences for the R340Q mutant may be due to the specific experimental conditions of the different assays or to distinct mechanisms of action of maspin.
The C-terminal region of maspin is not required for adhesion-promoting activity since the maspin/ovalbumin chimera containing the C-terminal end (aa 346-375) of ovalbumin remains fully active. Previous studies showed that deletion of the P7' residue through the C-terminal results in the loss of maspin activity (9,15). In this deletion mutant, the internal β-strands B4 and B5 and the exposed β-strand C1 (Fig. 1) are missing. These results suggest the C-terminal region, which is a highly conserved region of serpins (27,28), is required for proper conformation of the molecule including the RSL.
Not only is the RSL of maspin required, it is sufficient to induce cell-ECM adhesion in both cell systems and to inhibit carcinoma cell invasion. Both the RSL peptide and the ovalbumin mutant containing the RSL of maspin can efficiently inhibit carcinoma cell invasion and induce increased adhesion of carcinoma cells to fibronectin and corneal stromal cells to several ECMs. The sufficiency of the RSL peptide is unusual because the serpin inhibition mechanism requires additional regions of the protein. This mechanism involves proteolytic cleavage at the P1 site, insertion of the RSL into β-sheet A and translocation of the attached protease to the opposite side of the molecule (29). Thus, the stimulation of cell-matrix adhesion and the inhibition of invasion probably do not involve the serpin mechanism of protease inhibition. However, a role of the RSL as a canonical protease inhibitor cannot be ruled out.
Conversion of inactive ovalbumin to a maspin-like molecule by swapping the RSL is atypical for the serpin system. A number of studies have explored the possibility of exchanging activities and/or specificities among serpins by swapping the RSL regions between inhibitory serpins such as α1antitrypsin and α1-antichymotrypsin (30). Except for the highly identical (90%) serpins SCCA1 and SCCA2 (31), replacement of the RSL is not sufficient to transfer the specificity of one serpin to another (30). However, the specificity of the inhibitory activity of serpins has been changed by site-specific mutagenesis within the RSL (32,33).
The use of the RSL in a role that does not involve the serpin mechanism of inhibition is also unusual. Although the RSL of non-inhibitory serpins such as pigment epithelial-derived factor (PEDF) and thyroxine-binding globulin (TBG) acts as a substrate rather than an inhibitor for proteases, the RSL is not a functional domain responsible for their activities. Cleavage of the exposed RSL does not affect the neurite-promoting function of PEDF (34). Since the hormone-binding site in TBG resides in β-sheets B and C, cleaved TBG still can bind to thyroxine (35). The RSL of cortisol binding globulin (CBG) is also not a hormone-binding domain, yet cleavage of the RSL by elastase reduces the binding affinity for cortisol (36).  respectively. Adhesion was assayed as given in Fig 2. (Error Bars = Standard Deviation, * p< 0.05 relative to the fibronectin control, ** p< 0.05 relative to the laminin control, *** p< 0.05 relative to the type I collagen control)  Table 1 Type Name Sequence * Restriction sites EcoR I or Bgl II (italic letters) at 5' overhang of sense or antisense, respectively of both ovalbumin and maspin (bold letters). … Sequences in bold letters complementary to ovalbumin or maspin (bold letters) with 5' overhang containing ovalbumin or maspin sequences (nonbold) for overlap PCR extension. ‡ Site-directed mutation (bold and underlined letters).