The results of a protein design project are used to compare ... of human pancreatic secretory trypsin inhibitor (hPSTI) ... strong and specific human leukocyte elastase inhibitors ..... hydrolysis, thereby creating a 'permanent' inhibitor (Figure. 4).
Protein Engineering vol.8 no.l pp.45-52, 1995
Highly effective protease inhibitors from variants of human pancreatic secretory trypsin inhibitor (hPSTI): an assessment of 3-D structure-based protein design
M.Szardenings1, B.Vasel2, H.-J.Hecht2, J.Collins3 and D-Schomburg2-4 Departments of ^Molecular Structure Research and 3Genetics, Gesellschaft ftlr Biotechnologische Forschung, D-38124 Braunschweig, Germany 'Present address: Freie University Berlin, Institut filr Knstallographie, Takustrasse 6, D-14195 Berlin, Germany ''To whom correspondence should be addressed
Introduction Protein design projects in which the target protein is to be given a particular property depend initially on the ability to obtain and correlate reliable functional information with the variation of the natural sequences. Such projects are often limited by the available knowledge of the 3-D structure of primary sequences of natural analogues (Schomburg, 1989). We describe here the design of potent elastase inhibitors derived from the human pancreatic trypsin inhibitor (hPSTI) by discrete amino acid exchanges. hPSTI belongs to the family of Kazal inhibitors that exhibit extensive structural similarity, including the distribution of six cysteine residues known to be involved in disulfide bridges (Aviles, 1993). These include the PSTIs of mammals and the ovomucoids, which possess three protease inhibitory domains, the third having been studied extensively for a number of different ovomucoids for which there are also crystallographic data on free and protease-complexed structures (Laskowski and Kato, 1980; Bode and Huber, 1992, 1993). A large number of sequences are known for the third domain of ovomucoids and they can be correlated with their properties as serine© Oxford University Press
Human leukocyte elastase (HLE) is a protease of physiological importance. There is evidence that it is involved in several disorders, such as adult respiratory distress syndrome (ARDS), septic shock, multiple organ failure (MOF) and bums shock amongst others, which might be successfully treated by the application of inhibitors (Jochum et al., 1981; Duswald et al., 1982; Lang et al., 1989). Therefore, a strong binding and immunological tolerance (human origin, small molecular weight, relatively short residence time in the body) are considered to be the prerequisites for an ideal inhibitor with therapeutic potential. It was expected that variants of human PSTI could perhaps fulfil these requirements and so be developed for clinical application. The kinetic characteristics of a proteinase inhibitor useful in clinical treatment can be described as follows (Bieth, 1984). A simplified description of the reaction of peptide proteinase inhibitors of the Kazal type that satisfies the kinetic observations with ovomucoid third domain (Ernpie and Laskowski, 1982) can be given by: E +I
El
E + I*
(1)
with E = free proteinase, I = free inhibitor, ET = enzymeinhibitor complex and I* = cleaved inhibitor. A cleavage of I (inhibitor) is reversible. In the cleaved form it exhibits a reduced affinity for the protease (Tschesche et al., 1984). In fact, the rate of cleavage is in most cases slower than the dissociation rate so that El mainly undergoes dissociation 45
Downloaded from http://peds.oxfordjournals.org/ by guest on October 29, 2015
The results of a protein design project are used to compare different predictive strategies with respect to proteinprotein interactions. We have been able to generate variants of human pancreatic secretory trypsin inhibitor (hPSTI) optimized with respect to the affinity and specificity for human leukocyte elastase relative to trypsin and chymotrypsin, and in particular chymotrypsin. The extremely strong and specific human leukocyte elastase inhibitors were thus developed in three rounds of mutagenesis and two rounds of 3-D modelling; only 24 variants in total were synthesized, although variations at seven different amino acid positions were involved (i.e. from 207 possible variants). An excellent elastase inhibitor could be designed with the minimum of two amino acid exchanges. The value of structural modelling and actual structure determination is discussed in the light of the experimental results of the designed protein variants and the results of tertiary structure determinations of the free variant and the inhibitorprotease complex. Particular reference is given to the strategy to be followed in protein design projects in general and to the development of protease inhibitors in particular. Key words: elastase inhibitors/protease inhibitor/protein design/ structure-function relationship
protease inhibitors. It was assumed initially, as proposed by Laskowski's group, that this knowledge would facilitate the design of inhibitors with new proteinase specificities derived from members of this family by using, for example, 'the sequence to reactivity algorithm [which is] a combination of procedural rules and tables of empirical data which would allow one to predict the various interesting properties of a newly sequenced ovomucoid domain from the amino acid sequence alone' (Laskowski et al, 1989). A number of exchanges have been proposed that should determine the specificity of the inhibitors. Nevertheless, it proved difficult to turn hPSTI into a specific and strong inhibitor of elastase by simply relying on algorithms derived from primary sequences. Several differences between ovomucoid third domain and PSTI have been described (Bolognesi et al., 1982; Fujinaga et al., 1982; Read and James, 1986; Hechtef al., 1991). Most obvious is one additional amino acid between the first two N-terminal cysteines. But there are also structural differences. Only the loop between Cysl6 and Cys24 is structurally identical, whereas the rest of the PSTI molecule deviates with respect to interactions in the proteinase-inhibitor complex to such an extent that extrapolations based on ovomucoid data are largely invalid.
M^zardenlngs et aL
The reaction given above is, of course, an over-simplification of the reaction of two macromolecules that may be divided up in several intermediate steps on the molecular level (Quast et al, 1978). Unfortunately only the macroscopic constants, which tacitly circumscribe the accumulation of constants for individual steps in the docking and mutual structural adaptation involved in the interaction (Hecht et al, 1991), are accessible to direct kinetic measurements. Materials and methods Construction and expression of the variants The construction of synthetic PSTI genes and their expression has been described (Maywald et al, 1988; Collins et al, 1989, 1990). Site-specific mutagenesis with short oligonucleotide primers (Kramer et al, 1984; Szardenings and Collins, 1990) was used to construct variant PSTIs in a new vector (pMAMPF; Szardenings and Collins, 1990) which can be propagated in either plasmid or single-strand phage form. The vector uses the OmpA signal sequence for secretion, and synthesis is controlled by a X P L promoter. Dideoxy sequencing on plasmid single-strand DNA was as described (Sanger et al, 1977). Purification of the inhibitors 0.05% Tween-80 and 5.5% perchloric acid were added subsequently to the supernatant of a typical 1 1 batch fermentation (Szardenings and Collins, 1990). The precipitate was removed by centrifugation after 30 min and the solution readjusted to pH 8.0 by the addition of a saturated solution of Tris. Salts were removed by dialysis against water in a dialysis tubing (molecular weight cut-off 3000 Da). Further purification was carried out by filtration through a DEAE Sephacel ion exchanger. The flow-through and the first fractions obtained by washing with 20 mM Tris pH 8.0 were adjusted to pH 2.7 with citric acid and applied to an S-Sepharose fast flow column. The column was washed with 10 mM citric acid (pH 2.7) and most PSTI variants could be obtained by running a gradient (0-1 M NaCl) at -0.6 M NaCl. Depending on batch and particular variant, the PSTI already represented -50% of the entire protein content of the peak fractions, as judged by SDSPAGE. The protein was either purified further by an additional chromatography on a Mono-S column or the fractions were 46
dialysed directly against water and lyophilized. The final purification was carried out by reversed-phase chromatography on a ProRP column running a gradient from 0 to 60% acetonitrile containing 0.1 % (v/v) TFA, as described (Maywald etal., 1988). All columns were prepared from material obtained from Pharmacia. The correct removal of the ompA leader from the inhibitor had already been shown by sequencing the N-terminus of several variants (Maywald et al, 1988; Szardenings and Collins, 1990). Samples obtained at all stages of the purification after the acidic precipitation step could be used for initial determinations of the K\ values. No differences were observed compared with the purified inhibitor, thus confirming the stability of the inhibitor during purification. Measurement of kinetic constants Protease inhibitor tests and calculation of the kinetic constants are described (Chase and Shaw, 1969; Beatty et al, 1980; Bieth, 1980, 1984; Geiger and Fritz, 1984; Tschesche et al, 1984; Maywald et al, 1988). kon and /toff measurements and calculations were carried out as described (Beatty et al, 1980; Bieth, 1980, 1984). The substrates Sue-Ala-Ala-Pro-Phe-pNA (Bachem) for chymotrypsin (Merck AG, Darmstadt, Germany) and methoxy-Suc-Ala-Ala-Pro-Val-pNA (Behring Diagnostics, La Jolla, CA) for elastase were used in the kinetic measurements. Protease stocks were titrated as follows. Chymotrypsin was titrated with aprotinin (Trasylol, Bayer AG, Wuppertal, Germany). A stock solution of PSTI4 that is resistant against repeated freeze and thaw cycles was titrated with this chymotrypsin solution and used for the titration of human neutrophil granulocyte elastase stocks (a gift from Merck AG). The activity of protease stocks was routinely controlled as those of elastase stock solutions tended to decrease during storage. The measured Kt values for elastase were in most cases only apparent values [A^p)], as a new equilibrium between proteinase, inhibitor and substrate formed during the measurement. Measurements with elastase therefore required a 10 min preincubation in the presence of substrate. The reason for this was the relative high dissociation rate of the complex between elastase and most of the inhibitors. The K-, value was calculated according to Bieth (1984): (2)
with [S] = substrate concentration and Km = MichaelisMenten constant of the substrate (5.3X 10~5 M for the elastase substrate). The validity of the equation was checked for every variant by variation of the substrate concentrations (0.125-0.5 mmol). This gives the possibility of recalculating the Km value of the substrate and evaluating the accuracy of the applied correction if Klim>) and [S] are known for at least two independent measurement series. Apart from PSTI99P, the elastase A", values for all inhibitors had to be corrected. Possible cleavage of the inhibitors was checked by longer preincubation periods. No measurable change in the inhibition measurements was found, even when the reaction mixtures were incubated for two to three times the normal incubation period that resulted in maximum inhibition. Kinetic measurements were carried out with an Ultrospec II spectrophotometer using software purchased from Pharmacia. All kinetic constants were confirmed by multiple (at least three) independent series of experiments carried out on at least two independent preparations of the inhibitor.
Downloaded from http://peds.oxfordjournals.org/ by guest on October 29, 2015
yielding intact I and cleaved inhibitor accumulates only slowly. The K, of a proteinase inhibitor (K, = kon/kon for Kazal-type inhibitors) is often taken as a measure of the usefulness of an inhibitor in clinical applications. This can be misleading if either the association rate constant (kon) is too small (complex association kinetically hindered) or the dissociation rate constant (koft) is too high to cause significant inhibition in a biological system (Bieth, 1984). The measurement of one rate constant seemed to be sufficient, as the second can be calculated by A"; = kon/lcoff. In biological systems, the fast formation (high ^on rate) of the complex is desirable as the inhibitor might be degraded or excreted before significant amounts of the target enzyme are inhibited. High complex stability (low /toff rate), on the other hand, is even more desirable, because in these cases a single dose of inhibitor (a high dose in the case of low &
