Modularity of Serpins

An exciting application of protein engineering is the creation of proteins with novel functions by the retrofitting of native proteins. Such attempts might be facilitated by the idea of a mosaic architecture of proteins out of structural units. Even though numerous theoretical concepts deal with the delineation of structural “modules,” their potential in the design of proteins has not yet been sufficiently exploited. To address this question we used a gain of function approach by designing modular chimeric molecules out of two structurally homologous but functionally diverse members of the superfamily of serine-proteinase inhibitors, α1-proteinase inhibitor and thyroxine-binding globulin. Substitution of two of four α1-proteinase inhibitor modules (Lys222 to Leu288 and Pro362 to Lys394, respectively), identified by α-backbone distance analysis, with their thyroxine-binding globulin homologues resulted in a bifunctional chimera with inhibition of human leukocyte elastase and high affinity thyroxine binding. To our knowledge, this is the first report on a bifunctional chimera engineered from modules of homologous globular proteins. Our results demonstrate how a modular concept can facilitate the design of new functional proteins by swapping structural units chosen from members of a protein superfamily.

In all but the smallest proteins, crystallography has revealed that polypeptide chains form several more or less compact units. When loosely connected to the remaining molecule, such units are usually referred to as domains, which implicates the possibility of an autonomous existence (1). In many other cases, the mosaic nature of proteins is less obvious, and numerous concepts have been developed to facilitate the delineation of "modules" thought to rule the folding, function, and biological evolution of proteins (2)(3)(4)(5)(6)(7)(8). The increasing frequency with which functionally unrelated proteins are found to contain recurrent structural motifs suggests that the number of natural folds is limited (9,10) and that complex proteins have evolved by modular assembly (11). Such evolutionarily refined structural units are attractive candidates as building blocks for the design of novel proteins. This concept may be exploited for the in vitro recombination of homologous, i.e. structurally related, proteins.
Based on sequence similarities, an ever increasing number of homologous but functionally diverse proteins are recognized as members of the superfamily of serine-proteinase inhibitors (serpins). 1 They presumably evolved from a common ancestor at least 500 million years ago (12). Most of more than 100 known members of the serpin superfamily are true inhibitors of serine proteinases, best exemplified by the archetypical serpin ␣ 1 -proteinase inhibitor (␣ 1 PI). Serpins are fundamentally important in the regulation of major proteolytic cascades, such as blood coagulation, fibrinolysis, inflammatory response, and extracellular matrix turnover (reviewed in Ref. 13). However, some serpins have lost the inhibitory function and serve as transport proteins for small ligands, such as thyroxine-binding globulin (TBG) (14) and corticosteroid-binding globulin (15). TBG has an exceptionally high binding constant (K a ϭ 10 10 M Ϫ1 ) for thyroxine (T 4 ) and a binding energy close to a covalent bond (16).
The crystallographic structures of several serpins have been determined (reviewed in Refs. 17 and 18). Their highly compact single-domain structure has a scaffold of three crossed ␤-sheets (A-C). Inhibitory serpins are characterized by a reactive site loop (RSL) located between ␤-sheets A and C. Proteinase inhibition involves the incorporation of the cleaved RSL into the A-sheet. This structural rearrangement is accompanied by an increase in stability (stressed-to-relaxed transition (19)) and the generation of SDS-stable serpin-proteinase complexes (20). Although the individual serpins have become remarkably diversified by evolution, they share a common molecular pathology (21). Inhibitory dysfunction is caused by disturbances of the hinges of the RSL (P14 -P12 of the RSL and strand 1C) (22) or by prevention of insertion (23).
Although ␣ 1 PI has no known ligand, its sheet C and part of sheet B form a twisted ␤-barrel-like structure, characteristic of ligand-binding proteins. By affinity labeling the homologous regions have been shown to comprise the hormone-binding sites of TBG (24) and corticosteroid-binding globulin (25), both of which share 40% sequence identity with ␣ 1 PI.
So far, it has not been tested whether the inhibitory function and the ligand-binding function are mutually exclusive within the serpin scaffold. We now present a novel concept of a modular architecture of the serpin structure and construction of an ␣ 1 PI-TBG chimera with both inhibitory activity and high affinity T 4 binding.

EXPERIMENTAL PROCEDURES
Materials-␣ 1 PI M-type cDNA (26) was a kind gift from R. Foreman (Southampton, United Kingdom). A vector containing the full-length cDNA of TBG had been constructed previously (27). Vent DNA polymerase and restriction endonucleases were obtained from New England Biolabs. Spodoptera frugiperda Sf9 cells (ATCC catalog no. CRL 1711) and wild type baculovirus DNA were from Invitrogen. Liposomes for transfection and SF900 II insect cell culture medium were purchased from Life Technologies, Inc. Purified TBG and rabbit anti-TBG serum were generously donated by R. Gä rtner (Munich, Germany). Rabbit anti-␣ 1 PI serum, human leukocyte elastase (HLE, EC 3.4.21.37), and transthyretin were from Calbiochem. T 4 stock solutions and TBG concentrations were quantified using commercially available radioimmunoassays (Brahms Diagnostica, Berlin, Germany and CIS Bio Int., Gif-Sur-Yvette, France). Inhibition assays and active site titrations were measured on a Beckman DU 640 spectrophotometer.
Construction of Hybrid TBG-␣ 1 PI Transfer Vectors-Human TBG cDNA was subcloned via the KpnI and HindIII sites and human ␣ 1 PI cDNA via the PstI site into the transfer vector pBlueBac4 (Invitrogen). The splicing sites of the chimeras mapped to highly conserved regions (homology region H1 and H2) and to a putative permissive surface loop (splicing site RS, C-terminal to the RSL). The chimeric constructs were then generated by repeated cycles of two-step polymerase chain reaction overlap extension (28) with the linearized TBG and ␣ 1 PI plasmids or the gel-purified intermediate polymerase chain reaction products (P 1 T 2-4 , P 1 T 2 P 3-4 ) 2 as templates, respectively. The cDNAs were fused sequentially at homology regions H1 and H2 and splicing site RS with the primers listed in Table I. Primers P-N and T-C provided PstI and KpnI linkers, respectively, for subcloning into pBlueBac4. The correct sequence of the final products was confirmed by automated sequencing with fluorescent dye terminators (PRISM System 377, Applied Biosystems).

Generation of Recombinant Baculovirus and Expression in Insect
Cells-Sf9 cells (5 ϫ 10 6 log phase) maintained exclusively in serumfree medium were cotransfected with 1 g of linearized wild type virus and 4 g of each of the transfer plasmids by lipofection (29). ␤-Galactosidase-positive recombinant clones were selected by plaque assay and screened for wild type virus contamination by polymerase chain reaction (30). For protein expression, log-phase Sf9 cells from a spinner culture were seeded in tissue culture flasks and infected with recombinant virus at a multiplicity of infection of five. The medium was changed 12 h later and supplemented with 10 M 1-(L-trans-epoxysuccinylleucylamino)-4-guanidinobutane and 10 M pepstatin A (both from Roche Molecular Biochemicals). Forty-eight hours post infection, the culture supernatants were collected by centrifugation at 1500 ϫ g for 15 min, concentrated, and washed (0.1 M NaCl, 0.1 M Hepes, pH 7.4) by ultrafiltration (Centriplus 30, Millipore Corp.). Protein concentrations were determined by Scatchard analysis of T 4 binding and by densitometry of Coomassie Blue-stained gels using purified serum TBG as the standard.
Western Blotting-Samples were run on 10% continuous tris/glycine gels under denaturing, nonreducing conditions. For PAGE under native conditions, SDS was omitted from all buffers. Blotted nitrocellulose membranes were probed with rabbit anti-TBG or rabbit anti-␣ 1 PI antiserum as primary antibody, respectively, followed by enhanced chemiluminescence immunodetection with horseradish peroxidase-conjugated donkey anti-rabbit IgG (Amersham Pharmacia Biotech) as secondary antibody.
Inhibitor Assay-HLE was incubated at 37°C for 15 min with increasing amounts of recombinant proteins in assay buffer (see above). The residual proteolytic activity was calculated from the increase in absorbance (410 nm) after the addition of 0.5 mM N-methoxysuccinyl-A-A-P-V-p-nitroanilide (Calbiochem) as chromogenic substrate. Rates of substrate hydrolysis were constant over the 3-min period used to determine residual activities. The intercept on the abscissa of the plot of the fraction of enzyme remaining (E/E 0 ) versus the ratio of the initial inhibitor to initial enzyme concentration (I 0 /E 0 ) yielded the apparent stoichiometry of the reaction. Control reactions with supernatants of cells infected with baculovirus expressing TBG excluded endogenous HLE inhibitory activity, degradation of HLE, and substrate loss to endogenous proteinases.
T 4 Binding Assay-Parameters of T 4 binding to the recombinant proteins were measured by a method previously described in detail (31). Briefly, samples were diluted with 270 mM barbital buffer (pH 8.6) or phosphate-buffered saline (pH 8.0) and incubated with [ 125 I]T 4 (specific activity, 48.8 MBq/g, NEN Life Science Products) in the presence of increasing amounts of unlabeled T 4 . After equilibration, protein bound was separated from free [ 125 I]T 4 with anion exchange resin beads (M-400, Mallinckrodt), and the specific 125 I binding was determined. The affinity constants (K a ) and binding capacities for T 4 were calculated by Scatchard analysis (32).
Heat Denaturation-The functional stability of recombinant proteins was quantified by thermal denaturation in a water bath at 60 Ϯ 0.1°C for various periods of time or at various temperatures for 20 min, respectively. The samples were then cooled on ice and centrifuged for 15 min at 13,000 ϫ g to remove precipitated protein. Residual specific T 4 binding capacity or HLE inhibitory activity was expressed relative to controls kept at 4°C. The half-lives (t 1/2 ) of heat denaturation were calculated by least square analysis of semi-logarithmic plots of the remaining specific T 4 binding versus time of incubation.

RESULTS
Design and Construction of Chimeras-Based on the structure of ␣ 1 PI and guided by a distance analysis of its carbon backbone using a diagonal plot (8,33), four compact structural units of the serpin fold were identified (Figs. 1 and 2A). Modules 1 and 3 complement each other to form an ␣-␤ sandwich, while modules 2 and 4 constitute a discontinuous ␤-barrel fold. The segregation into these two subdomains becomes even more 2 The names of the chimeras illustrate the composition from modules (the subscript numbers) of TBG and ␣ 1 PI (preceding letter T or P), e.g. in P 1 T 2 P 3-4 , modules 1, 3, and 4 are ␣ 1 PI sequences, whereas the second module is a TBG sequence. The denotation of serpin secondary structure elements and their assignments to TBG are as described in Ref. 17.

TABLE I
Oligonucleotide primers for splicing by overlap extension polymerase chain reaction For the internal primers, the first letter of the primer names denotes their 5Ј-origin from the TBG or ␣ 1 PI sequence. H1, H2, and RS denote the locations of the corresponding splicing sites (bold letters) at homology regions H1 (␣ 1 PI numbering: Val 218 -Met 221 ), H2 (Pro 289 -Thr 294 ), and the RSL (Pro 361,362 ), respectively. N or C denote N-or C-terminal external primers, specific for the TBG or ␣ 1 PI coding sequences or their reverse complements, respectively (bold letters). To introduce the putative ligand-binding site of TBG into the ␣ 1 PI scaffold, its complete ␤-barrel was substituted by the TBG homologue to give chimera P 1 T 2 P 3 T 4 (Figs. 2B and 9). As controls, chimeras containing only module 2 of TBG in the ␣ 1 PI scaffold (P 1 T 2 P 3-4 ) or module 1 of ␣ 1 PI in the TBG scaffold (P 1 T 2-4 ) were constructed. The boundaries of the modules coincided with regions that are highly conserved throughout the serpins (homology regions H1 and H2) or matched to a permissive loop region (RS, C-terminal to the RSL), thereby reducing the probability of local structural disturbance in the chimeric proteins. The corresponding hybrid cDNAs, generated by repeated cycles of splicing by overlap extension polymerase chain reaction, were used to produce recombinant baculovirus by in vivo recombination in insect cells.
Expression of Recombinant Proteins and Reaction with HLE-Bv-␣ 1 PI, bv-TBG, and the three chimeras were efficiently secreted by the insect cells with similar expression levels of up to 5 g/ml after 60 h in serum-free medium. The structural integrity of the proteins was evident by their reaction with specific polyclonal anti-␣ 1 PI and anti-TBG antibodies, whereas there was no detectable cross-reactivity between bv-TBG, bv-␣ 1 PI, or wild type baculovirus with these antisera.
Chimera P 1 T 2 P 3-4 and, to a lesser extent, P 1 T 2 P 3 T 4 retained the inhibitory properties of ␣ 1 PI and formed SDS-stable complexes with HLE (Fig. 3). P 1 T 2 P 3-4 and P 1 T 2 P 3 T 4 showed significantly more cleaved inhibitor than bv-␣ 1 PI. The reaction of HLE with P 1 T 2 P 3 T 4 was slower than with ␣ 1 PI, as indicated by the large amount of uncleaved inhibitor at a molar ratio of one. In contrast to the stable reaction products of cleaved bv-␣ 1 PI, increasing HLE concentrations led to a loss of detectable P 1 T 2 P 3 T 4 -HLE complex concomitant with the disappearance of free P 1 T 2 P 3 T 4 (Fig. 4). As expected, bv-TBG and P 1 T 2-4 harboring the RSL equivalent of TBG behaved like pure substrates (Fig. 3).
Inhibitor Assay-The residual proteolytic activity of HLE preincubated with increasing amounts of the inhibitors showed a linear dependence characteristic for tight binding inhibition (Fig. 5). The stoichiometries of inhibition (SI), defined as mole of serpin required to inhibit 1 mole of HLE, were 1.3 for bv-␣ 1 PI, 2.1 for P 1 T 2 P 3-4 , and 11 for P 1 T 2 P 3 T 4 . These SI values were in agreement with the reaction products on the immunoblots (Figs. 3 and 4). Again, bv-TBG and P 1 T 2-4 showed no inhibition of HLE.
Analysis of T 4 Binding-Scatchard analysis of T 4 binding showed no detectable T 4 binding activity (K a Ͻ 10 6 M Ϫ1 ) for chimera P 1 T 2 P 3-4 and the bv-␣ 1 PI control. However, in P 1 T 2 P 3 T 4 , transposition of the complete ␤-barrel motif into the ␣ 1 PI frame created a high affinity T 4 -binding site (K a ϭ 1.7⅐10 8 M Ϫ1 ), comparable with the first binding site of transthyretin (Fig. 6B). The additional substitution of module 3 in P 1 T 2-4 increased the T 4 binding affinity to almost half of that of bv-TBG or human serum TBG (K a ϭ 0.5⅐10 10  and anti-TBG antibodies (lower panel). P 1 T 2 P 3-4 formed SDS-stable complexes with HLE similar to bv-␣ 1 PI but also showed a significant substrate reaction. Chimera P 1 T 2 P 3 T 4 also formed an HLE-inhibitor complex (detected with both antibodies), but most of the protein was cleaved. P 1 T 2-4 and bv-TBG showed pure substrate reactions. M Ϫ1 ) (Fig. 6A), although 35% of its residues differed from the wild type TBG sequence.
Heat Stability of the Chimeras-To examine the effect of module exchange on the stability of the chimeras, heat denaturation experiments were performed. Incubation of bv-␣ 1 PI and P 1 T 2 P 3-4 at 60°C caused a shift in electrophoretic mobility on native PAGE (Fig. 7) compatible with dimerization, as has been previously shown for human serum ␣ 1 PI (36). Bv-TBG and P 1 T 2-4 were not stable at 60°C. In contrast, P 1 T 2 P 3 T 4 exhibited neither multimerization nor loss of soluble antigen even at 80°C.
Consistent with the increased conformational stability on native PAGE, P 1 T 2 P 3 T 4 displayed no significant decline of T 4 binding after incubation at temperatures as high as 85°C for 20 min (Fig. 8B). However, its inhibitor function was lost at a slightly lower temperature than that of bv-␣ 1 PI, starting at 55°C (Fig. 8C). SDS-PAGE analysis of P 1 T 2 P 3 T 4 denatured at 65°C revealed that this material was still a specific substrate FIG. 4. Digestion of bv-␣ 1 PI and P 1 T 2 P 3 T 4 at different HLE-toserpin ratios. Samples were incubated with increasing ratios of HLE to inhibitor (E/I). Reactions were stopped after 20 min by denaturation at 95°C in 0.1% SDS. The fraction of complexed P 1 T 2 P 3 T 4 was smaller and less stable than that of bv-␣ 1 PI.
FIG. 5. Titration of HLE with ␣ 1 PI-TBG chimeras. HLE was titrated with each of the chimeras at 37°C. After incubation for 15 min, residual HLE activity was determined. For bv-␣ 1 PI, P 1 T 2 P 3 T 4 , and P 1 T 2 P 3-4 , linear titration curves were obtained irrespective of the substrate concentrations tested (0.1 and 1 mM, K m of 0.15 mM). The intercept with the abscissa yielded an apparent stoichiometry of 1.3 for bv-␣ 1 PI (E), 2.1 for P 1 T 2 P 3-4 (f), and 11 for P 1 T 2 P 3 T 4 (q). Bv-TBG (Ⅺ) and P 1 T 2-4 (OE) showed no inhibition of HLE even at a high molar excess.

FIG. 6. T 4 binding of ␣ 1 PI-TBG chimeras.
A, Scatchard analysis of T 4 binding of bv-TBG (Ⅺ) and human serum TBG revealed no significant differences in binding affinity (K a ϭ 1.2 Ϯ 0.11⅐10 10 M Ϫ1 ). The K a of chimera P 1 T 2-4 (OE) was only slightly reduced (0.5 Ϯ 0.14⅐10 10 M Ϫ1 ). B, the binding affinity of P 1 T 2 P 3 T 4 (q) was 70 times less than bv-TBG, but at 1.7 Ϯ 0.3⅐10 8 M Ϫ1 it was still higher than the second-best natural T 4 -binding protein, transthyretin (E) (K a ϭ 0.9⅐10 8 M Ϫ1 ). ␣ 1 PI and P 1 T 2 P 3-4 had no specific T 4 binding. The plots are representative of four independent experiments .   FIG. 7. Native PAGE analysis of heat-denatured chimeras. Samples were incubated for 30 min at the indicated temperatures, separated by native PAGE and probed with anti-␣ 1 PI (bv-␣ 1 PI, P 1 T 2 P 3-4 , P 1 T 2 P 3 T 4 ) or anti-TBG (P 1 T 2-4 , bv-TBG) antiserum, respectively. P 1 T 2 P 3-4 exhibited a behavior similar to bv-␣ 1 PI, with polymerization at 60°C and almost complete loss of immunoreactive material at 80°C. P 1 T 2 P 3 T 4 showed neither signs of polymerization nor loss of detectable antigen at 80°C, whereas P 1 T 2-4 showed a large proportion of pre-existing polymers and loss of detectable antigen at 60°C. No bv-TBG polymers were detectable, but antigenicity was lost after incubation at 60°C .   FIG. 8. Functional stability of chimeras. A, rate of thermal inactivation for P 1 T 2 P 3 T 4 , P 1 T 2-4 , and bv-TBG as determined by residual T 4 binding capacity. Proteins were heated at 60°C, aliquots were removed at the indicated time intervals and centrifuged, and the residual T 4binding activity was determined. Values are expressed as proteinbound T 4 relative to the basal levels and represent the means Ϯ SD for three independent experiments. Plots of the log binding capacities versus time of incubation were linear, indicative of an apparent firstorder process. Bv-TBG (Ⅺ) had a slightly reduced functional stability compared with human serum TBG (छ) (t 1/2 of 4.5 versus 7 min), whereas P 1 T 2-4 (OE) was rapidly denatured (t 1/2 ϭ 2 min). Note that P 1 T 2 P 3 T 4 (q) is essentially stable at 60°C with no significant loss of T 4 binding capacity within 30 min. B, heat denaturation profile illustrating the markedly increased functional stability of uncleaved P 1 T 2 P 3 T 4 comparable with bv-TBG cleaved by HLE (؋). C, functional stability measured by means of the residual HLE inhibitory activity. P 1 T 2 P 3 T 4 lost its inhibitory potency at temperatures slightly lower than bv-␣ 1 PI (E), whereas P 1 T 2 P 3-4 (f) was less stable in this assay.
for HLE but did not form a serpin-enzyme complex (data not shown).
The inhibitor function of chimera P 1 T 2 P 3-4 was also less stable than that of bv-␣ 1 PI and was completely inactivated at 55°C. The t 1/2 (60°C) of T 4 binding of P 1 T 2-4 was reduced to about one-third of the t 1/2 of bv-TBG (Fig. 8A). DISCUSSION Genetic engineering has become a mainstay in elucidating the still inadequately understood structure-function correlation of proteins. This information is critical for the understanding of the diversity of proteins and the design of new drugs. In recent years, research has moved from the substitution of single amino acids to the concept of a modular design of proteins. In some proteins, structural and functional units are readily obvious, e.g. the extra-and intracellular and the transmembrane domains of membrane-bound receptors. The identification of discrete units has been used for the successful construction of chimeric receptors (37). However, in chimeras of globular proteins so far only similar functions have been substituted (38 -41). In this study, we present a strategy to engineer bifunctional chimeras from integral parts of homologous proteins. Based on a concept of a modular architecture of the serpins (Fig. 9), we have combined two different functional properties of the serpin superfamily, proteinase inhibition and ligand binding, into one chimeric molecule. The inhibitory and ligand-binding characteristics of the chimeras are summarized in Table II.
In chimera P 1 T 2 P 3 T 4 the transfer of the T 4 -binding site of TBG into the ␣ 1 PI frame was achieved by substituting the ␤-barrel-like structure of ␣ 1 PI with its TBG homologue (modules 2 and 4). This chimera exhibited inhibition of and complex formation with HLE, characteristic of inhibitory serpins such as ␣ 1 PI. In comparison with the archetypical, evolutionarily refined ␣ 1 PI, it was a weaker inhibitor with a higher apparent stoichiometry of inhibition and a shorter half-life of its complex. In addition to proteinase inhibition, chimera P 1 T 2 P 3 T 4 also exhibited a specific, high affinity T 4 binding. Although its binding affinity was 70-fold lower than that of TBG, it was still higher than that of transthyretin, the next best natural T 4binding protein.
In contrast, the substitution of only module 2 and thus only part of the ␣ 1 PI ␤-barrel including the environment of the affinity-labeled Lys 253 (24) did not result in detectable T 4 bind-ing. Similarly, a chimera harboring only module 4 of TBG produced a dysfunctional, secretion-deficient protein (data not shown). Only the substitution of the complete ␤-barrel, comprising modules 2 and 4, was sufficient to transfer the high affinity T 4 -binding site. Consequently, both modules seem to participate in avid T 4 binding, in agreement with the demonstration that all parts of the T 4 molecule, and thus an extensive surface of interaction of T 4 with the binding cavity of TBG, are essential for its high binding affinity (42). Furthermore, the functional transfer of the T 4 -binding site of TBG into the ␣ 1 PI frame unambiguously locates the ligand-binding site to the ␤-barrel motif of the serpins. Surprisingly, P 1 T 2 P 3 T 4 remained in solution and retained its T 4 binding activity even at remarkably high temperatures (Fig.  8). Serpins tend to polymerize at elevated temperatures (43) and simultaneously lose their activity and escape immunodetection as a result of precipitation. Polymerization is thought to involve the insertion of the loop of one serpin molecule into either sheet A (44) or C (21, 45) of another molecule. Both models require detachment of strand 1 from the C-sheet (46). The extended RSL of chimera P 1 T 2 P 3 T 4, which is engineered to be 3 or 7 amino acids longer than in TBG or ␣ 1 PI, respectively (Fig. 10), most likely delays the heat-induced release of strand 1C from the C-sheet, compatible with the increased thermal resistance of an ␣ 1 PI variant with a C-terminal extended RSL (48).
The discrepancy in the functional stability of T 4 binding versus HLE inhibition of P 1 T 2 P 3 T 4 could be the result of a higher intrinsic stability of the ␤-barrel than the remaining molecule. Significant heat-induced unfolding might occur without affecting the ␤-barrel and thus T 4 binding. However, the cooperativity in the unfolding of serpins (19,49) argues against this possibility. More conceivably, a local structural rearrangement of the RSL is responsible for the observed loss of inhibitory activity at intermediate temperatures. During heat exposure the A-sheet of the serpins is supposed to open up and accept a portion of its own RSL. In P 1 T 2 P 3 T 4 this might distort the RSL near the scissile bond, resulting in a pure substrate behavior toward HLE, whereas the extension of the RSL prevents detachment of s1C and hence both polymerization and loss of T 4 binding. This limited structural transition of P 1 T 2 P 3 T 4 might resemble the spontaneous conversion of plas- FIG. 10. Sequence alignment of the RSL regions of TBG, ␣ 1 PI and P 1 T 2 P 3 T 4 . The active site residues of ␣ 1 PI and the cleavage site of TBG by HLE (47) are depicted in bold letters. Note that the RSL of chimera P 1 T 2 P 3 T 4 harbors a seven-residue C-terminal extension, compared with the ␣ 1 PI loop, including the HLE cleavage site of TBG. sec. struct., secondary structure elements.  minogen activator inhibitor-1 from an active to a latent conformation in vivo (50,51).
In conclusion, the successful construction of a bifunctional chimera clearly demonstrates that ligand binding and proteinase inhibition are not exclusive within the serpin structure and provides evidence for their proposed modular architecture. Moreover, because our approach does not rely on specific features of the serpins but rather uses general design criteria such as compactness of modules and sequence conservation at fusion points, it appears not to be limited to this protein superfamily. There are many examples in which unrelated functions have evolved within a conserved structural scaffold (52-54), occasionally recruiting different portions of a molecule as reactive centers (55,56). Thus the exchange of homologous modules offers vast possibilities for the design of chimeric proteins with new functional properties. Furthermore, the integration of two functions in one globular protein suggests the potential to introduce novel allosteric effects, e.g. modulation of enzymatic activities upon ligand binding.