The Protease Inhibitory Properties of the Alzheimer’s &Amyloid Precursor Protein*

We have expressed the 57-amino acid Kunitz domain of the Alzheimer's beta-amyloid precursor protein (APP751) as a bacterial fusion protein. The protease inhibitory properties of the purified fusion protein, BX9, were virtually identical in all respects tested to those of purified secreted APP751. Both proteins strongly inhibited pancreatic trypsin (Kis = 0.2 and 0.3 nM) and less well epidermal growth factor-binding protein (Kis = 1 and 3.5 nM), alpha-chymotrypsin (Kis = 3 and 6 nM), and the gamma-subunit of nerve growth factor (Kis = 8 and 9 M). Neither protein appreciably inhibited plasma and pancreatic kallikreins, thrombin, lung tryptase, neutrophil elastase, or cathepsin G. The remarkable similarity of the protease inhibitory profile of BX9 to that of secreted APP751 suggests that proper intramolecular disulfide bond formation has occurred in the bacterial fusion protein and leads to the conclusion that the amyloid precursor protein Kunitz domain is a relatively specific inhibitor of only a few trypsin-like arginine esterases.


Kunitz
domain of the Alzheimer's &amyloid precursor protein (APP751) as a bacterial fusion protein.
The protease inhibitory properties of the purified fusion protein, BX9, were virtually identical in all respects tested to those of purified secreted APP751. Both proteins strongly inhibited pancreatic trypsin (Kis = 0.2 and 0.3 IIM) and less well epidermal growth factor-binding protein (Kjs = 1 and 3.5 nM), cY-chymotrypsin (Kis = 3 and 6 nM), and the y-subunit of nerve growth factor (Kjs = 8 and 9 M). Neither protein appreciably inhibited plasma and pancreatic kallikreins, thrombin, lung tryptase, neutrophil elastase, or cathepsin G. The remarkable similarity of the protease inhibitory profile of BX9 to that of secreted APP751 suggests that proper intramolecular disulfide bond formation has occurred in the bacterial fusion protein and leads to the conclusion that the amyloid precursor protein Kunitz domain is a relatively specific inhibitor of only a few trypsinlike arginine esterases.
The @-amyloid precursor protein (APP)' in Alzheimer's disease is a membrane-spanning protein (Dyrks et al., 1988), two forms of which, APP751 and APP770, contain a 57-amino acid insert with striking homology to the Kunitz family of protease inhibitors (Tanzi et al., 1988;Ponte et al., 1988;Kitaguchi et al., 1988). We have previously shown (Oltersdorf et al., 1989) that the secreted form of APP751 (sAPP751) with this Kunitz protease inhibitor domain is protease nexin-II, which was first isolated from human fibroblasts (Knauer and Cunningham, 1982)  and Lu-chymotrypsin were determined by active site titration (Chase and Shaw, 1967;Bender et al., 1966). The concentrations of EGF-BP, y-NGF, and mGK22 were estimated from their known extinction coefficients (Blaber et al., 1989). The concentrations of neutrophil elastase and cathepsin G were also estimated from their extinction coefficients (Travis et al., 1978).  Fig. 2 shows that the purified fusion protein, BX9, migrated as a -25-kDa band on reducing sodium dodecyl sulfatepolyacrylamide gel electrophoresis, in good agreement with its predicted size. The inhibition of pancreatic trypsin (5 nM) by purified BX9 and sAPP751 is shown in Fig. 3. To establish the molar ratio of inhibition, trypsin (100 nM) was titrated with sAPP751 (O-80 nM), and the result is shown in the inset of Fig. 3. The linear regression plot gives a molar equivalence point of 0.9:1 (sAPP751:trypsin), indicating that a 1:l inhibitory complex is formed. (A correction factor of 1.1 was then used to calculate active sAPP751 concentration.) The remarkable level of superimposibility of the two curves in Fig.  3 leads to the calculation of very similar K, values, 0.2 nM for sAPP751 and 0.3 nM for BX9 (see Table I  tions of sAPP751 (Fig. 4, top) or BX9 (Fig. 4, bottom) were then used to calculate inhibition constants. The compiled Ki values calculated with both BX9 and sAPP751 are listed in Table I. For either inhibitor, the order of inhibitory potency is trypsin > EGF-BP > cu-chymotrypsin > r-NGF > plasmin > mGK-22. Neither inhibitor appreciably inhibited thrombin, lung tryptase, pancreatic kallikrein, plasma kallikrein, or the granulocyte serine proteases elastase and cathepsin G (KiS 2 1 PM). The strong inhibition of pancreatic trypsin by these inhibitors is consistent with the presence of arginine at the P1 position of the APP Kunitz domain (residue 301 of APP751). EGF-BP and -r-NGF (Taylor et al., 1974;Blaber et al., 1989) are also arginine-specific esterases, and their relatively strong inhibition by the Kunitz domain is therefore not unexpected. Surprisingly, mGK-22, which is closely related to both of these proteases (Drinkwater et al., 1987) and was earlier throught to be an epidermal growth factor-binding protein (Anundi et al., 1982;Blaber et al., 1987), is rather poorly inhibited (Ki -180 nM), suggesting that specific interactions other than P,-S1 must play an important role in determining inhibitor potency. The interaction with a-chymotrypsin (which does not have affinity for basic residues at P,) is also likely explained by favorable non-P& interactions. It should be noted that a-chymotrypsin is also inhibited (Ki = 6 nM) by bovine pancreatic trypsin inhibitor (Fritz and Wunderer, 1983), which is otherwise selective for trypsin-like proteases.
The poor inhibition of plasmin and the lack of inhibition of thrombin, tryptase, and pancreatic and plasma kallikreins by the APP Kunitz domain suggest that this inhibitor has restricted selectivity even for proteases with trypsin-like specificity. With the exception of that for pancreatic trypsin, the K, values do not suggest a very potent inhibitor but indicate that potential target proteases will likely be arginine-specific esterases. A human homolog of either EGF-BP or r-NGF has not yet been reported, and, in fact, the human kallikrein gene family seems to have far fewer members (Schedlich et al., 1987) than the mouse kallikrein family to which these curious arginine esterases belong. It is thus difficult to predict what the identity of the "physiologically relevant" protease, with which the APP presumably interacts, might be. Another possibility, of course, is that a true trypsin-like protease is expressed in the same tissues, including brain, where the APP is expressed.
The bacterially expressed fusion protein inhibitors, BX9 and sAPP751, have identical qualitative inhibitory profiles toward all the proteases tested in this study and, in addition, generate very similar K, values with the proteases examined.