Structures of KLK protease substrate transition states

Kallikrein-related peptidases (KLKs) are 15 (chymo)trypsin-like serine proteases that are secreted to a varying extent in nearly all tissues of humans. Some KLKs are required for fertilization, tooth development, and desquamation of the skin, while they are also critically involved in various diseases, e.g. prostate cancer or atopic dermatitis. Moreover, KLK3 (PSA) is a widely used clinical marker for prostate cancer. Up to now, information on substrate recognition and catalytic mechanisms of the KLKs on the structural level is scarce. Thus, studies of the enzyme substrate transition states are desired. To this end, substrate analog complexes in crystals will be investigated by X-ray diffraction measurements. Crystallization conditions for all KLKs from 1 to 8 and KLK10 are known. As first stage of the project, KLKs will be purified according to established or recently developed protocols, using E. coli, L. tarentolae, and HEK293 cells. Glycosylated proteins can be trimmed with deglycosylating enzymes, in order to facilitate crystallization. Crystals will be grown and soaked with substrate analogs, in order to form complexes with the respective KLK. These complexes represent either non-covalent enzyme substrate ("Michaelis") complexes, tetrahedral transition states, acyl intermediates, or enzyme product complexes. For stabilization of the usually short-lived transition states, inactive mutants can be used, e.g. KLK2-Ser195Ala with hexa- to octapeptides. Also non-cleavable surrogates of the scissile peptide bond serve this purpose, such as reduced amides (-CH 2 -NH-). Tetrahedral transition states could be mimicked by phosphinates (-PO(OPhe)-) and difluoromethylene ketones (-CO-CF2 -). Eventually, aldehyde or boronic acid peptide inhibitors can mimick acyl intermediates, while product complexes may be captured in shock frozen crystals as well. Too labile complexes can be stabilized by disulfide tethering of a KLK- Cys mutant and a Cys-containing compound prior to crystallization. The best locations of such disulfides around the active site have been determined by modelling using known crystal structures. Based on the previous investigations, the second project stage will comprise the complex formation of active KLKs with pro-KLKs and other natural substrates that are available as commercial products or recombinantly expressed proteins. Among them are semenogelins, enamel matrix and corneodesmosomal proteins. As special case of very slowly turned over substrates Kazal-type inhibitors (LEKTIs) in complex with KLKs shall be crystallized. Selected substrate fragments can be chemically ligated to suitable synthetic compounds, in order to yield stable KLK transition states complexes, which may be supported by disulfide tethering. Ideally, a series of all transition state analogs should be generated with a full length substrate. Analysis of such complexes will greatly expand our understanding of the substrate recognition and cleavage by KLKs and serine proteases in general. Eventually, it can be expected that the methods and compounds of the proposal will open new ways for basic research and bioorthogonal or even pharmaceutical applications.

The project investigated structure and function of serine proteases from the family of human kallikrein-related peptidases (KLKs). Several of these family member are interesting in diseases, such as prostate or ovarian cancer and other pathological conditions. In particular the steps of the catalytic cycle of peptide or protein hydrolysis were to be analyzed. To this end, mainly those KLKs with known crystallization conditions were structurally determined and experiments on their function conducted. Structures of the supposed tumor suppressing KLK10 in an active and inhibited, but zymogen-like conformation were solved and gave surprising insights in an unusual activation and a novel Zn2+ inhibition mechanism. High disorder in some surface loops and the N-terminus supports a partially induced fit mechanism of substrate turnover. Similarly, two structures of KLK8 were determined, which acts in synaptic remodelling at neurons for memory formation. The comprehensive study included a thorough specificity profiling, mutational analyses, which located the stimulatory Ca2+ and inhibitory Zn2+ sites. Molecular dynamics calculations explained the substrate binding in comparison with the ligand-free and the inhibitor bound state. Moreover, the calculations showed that KLK8 possesses an allosteric surface loop network, of which some components are found in other human KLKs. Here, the basic mechanism appears to be more like the concept of conformational selection, which has been previously confirmed by our group for glycan free and glycosylated KLK2 in another project. A collaborative clinical and molecular biology study was performed, which proved the role of KLK6 and KLK8 as valuable biomarkers for ovarian cancer. The importance of glycans as regulators of enzymes inspired a review on protease glycosylation and its influence on the particular catalytic transition states of enzymatic activity, with various examples form the KLK field. Another cooperation employed computational biological methods to calculate the most likely substrate specificity of KLK7. In addition, an unexpected switch of the tetrahedral transition state was observed in extended molecular dynamics calculations, resulting in a reversed peptide backbone bound to the active site. This finding might serve as basis for the synthesis of new inhibitory compounds of pharmaceutical interest. Eventually, a cooperation on the NMR structure of the human inhibitor polypeptide SPINK6, which binds several KLKs, succeeded in solving it with two alternative conformations of the major active site binding residue, which resembles in one case a tetrahedral intermediate. Overall, the focus of the project shifted more to the conformational states of the KLK peptidases, which was stimulated by new insights based on the conformational selection mechanism. However, as it has been shown for various enzymatic systems, this mechanism can also transition into a mechanistic induced fit model.