On Fri, 9 Jun 2000, Petr Kuzmic wrote:
> Not only HIV-1 protease, but long before that also pepsin as a "model
> aspartic protease". Various people spent years hunting for small
> inhibitors of /renin/ (as opposed to "rennin") for blood pressure
> regulation. Pepsin was used to model the relationship between substrate
> reactivity (Km, kcat, kcat/Km) and inhibitor potency (Ki). Oh yes, and
> then there was Bartlett's work on thermolysin, penicillopepsin, and
> carboxypeptidase A. Bartlett was showing these nice looking pictures
> showing correlations between kcat/Km and Ki as an indication that the
> inhibitors were "transition state analogs". Oh well.
I actually have solved two structures of HIV-1 PR with a
D25N mutant:substrate-bound a wild type:two cleaved product halves
complex. The substrate used is a synthetic cyclic peptide. The data was
collected under cryostatic conditions and has defracted to 1.6
angstroms. An overlay of these structures and a parent
hydroxyl-ethyl-amine inhibitor show there is no-wat this class of
inhibitors mimics the transition state, nor do they bind in the same
manner to substrates/products as they are more "puckered" about P1,
P1'. The structures suggest the mechanism of catalysis is subtle and not
spectacular, with the enzyme locking the N-terminal half to effect
efficient electron transfer. Indeed the N-terminal product is in an almost
identical position. From my observations I would say there isn't any
inhibitors to HIV-1 PR that would mimic the transition state, though
I'd be happy to compare any structure that is suggested can. I would
confidently say it would also be the case for the larger monomeric
aspartyl proteases.
> I believe that co-operative effects ("subsite interactions") are largely
> responsive for the failure of substrate vs. inhibitor correlations, but
> perhaps even more importantly, /nonspecific/ interactions might be even
> more important. We got our hands on one of those phenomena -- long
> range, nonspecific, electrostatic interactions in pepsin catalysis
> (http://www.biokin.com/papers/biblio.html#Kuzm9175). Those show
> clearly that extrapolating from substrate kinetics (a /dynamic/ event)
> to the binding of reversible inhibitors (a /static/ event) is very
> tricky.
Interesting, I'll check it out. From the Henderson paper and more
highlighted in Methods in Enzymology (1994) v241 pp 127-156 "Use of
Steady State Kinetic methods to Elucidate the Kinetic and Chemical
Mechanisms of Retroviral Proteases." By Meek et al. It is plausible
for HIV-1 PR there is a hierarchical classification system of the
substrates into several types (class 1, class 2, class 3 and mixed).
This system is largely ignored, but is extremely useful. I believe the
reason it is ignored is because it is described in an SIV paper, so isn't
easily found by a simple lit search on HIV-1 PR and is in a virology
journal - not the normal reading matter of enzymologists.
In this paper, the authors conclude SIV and HIV have similar cleavage site
specificities (most are homologous) and define a sequential ordering of
substrate cleavage. The classes are as follows and some show interesting
trends which you my not normally consider as a trend;
Class 1: Tyr/Phe~Pro (P1~P1' and possibly an aromatic in P1') with
hydrophobic residues in P2,P2' (Asp in P2 for non-HIV and Asn for HIV) and
Gly/polar groups in P3. These also resemble HTLV-I and FIV consensus
sequences except P2 has a hydrophobic preference (HIV also likes a
hydrophobic residue in P2).
Class 2: All contain Arg in P4 and the sequence Phe-Leu (for all
retroviral aspartyl proteases generally in P1'-P2', but can be changed for
HIV). The general type being ArgP4..Leu~Phe-LeuP1-P2'
(p51-p15). Interesting with respect to this class of enzymes is they like
P3 to be polar and P2 to be hydrophobic. Indeed immunodeficiency proteases
cleave either side of a hydrophobic residue based purely on the neighbours
(co-operative binding).
Class3: All contain Gln/Glu at P2' and prefer hydrophobics at P2, P1, P1'.
Mixed: the mixed sequence for HIV-1 PR is querky. It's sequence P5-P5' is,
Arg-Pro-Gly-Asn-Phe~Leu-Gln-Ser-Arg-Pro, so it contains the class 2
determinants (Phe-Leu and Arg 4 as 4th amino acid from Phe), but it has a
shifted cleavage sequence from class 2 because of that Class 3 Gln in P2'
complicating things (there is also an Asn at P2 which stabilises
hydrophobics into P1 so cleavage follows a class 3 cleavage sequence. Ie,
the class 3 factor and stabilising Asn causes an alteration in the
cleavage pattern and co-operative shifting.
Now Tozser et al 1991 (FEBS 279, 356-360) examined P4 (Ser) substitution
in a classic class 1 HIV-1 PR sequence and found bulky long chains and
branched chains were bad (hence Arg importance as a Class 2), and smaller,
more Ser-like and helix breaking (extended or beta preference) amino acids
here stabilise the hydrophobic and Class 1 and mixed cleavage type (I
believe Asn in P2 is exactly the same). This happens by increasing the
Kcat and/or decreasing the Km.
For interests sake, the numbers of each class for HIV-1 PR is thus class 1
(3), class 2 (1 - and rare) class 3 (2) mixed (2). Class 3 appear to be
the "preferred" class and hence best for inhibitor design.
Meek et al., takes all the classes into account and examines the product
inhibition of the different classes concluding that the kinetic mechanism
of HIV-1 PR is random, with products forming both binary and ternary
enzyme product(s) depending purely on the substrate used. For strong
substrates of HIV, once cleaved, one generally leaves and is an
uncompetitive inhibitor, whereas the other is held more tightly but is a
non-competitive inhibitor. Other scenarios exist (a non-competitive
inhibitor (product) that is released and a competitive inhibitor that is
held), and it is possible to make a non-native substrate where both
inhibitors are competitive which leads to a random release - though they
are within experimental error and an ordering may still exist.
Indeed if you put the two competitive product inhibitors together you can
reform a complete substrate (as has been done), but non/uncompetitive
inhibitors such as those with a Pro residue just can't reform (think of
the turn-like nature of the Pro and how hard it is to reform a bond to
that amino acid in terms of the position of a cleaved peptide that has
left and then tries to re-enter. The energy barrier is too high as the
substrate binding pocket essentially relaxes/collapses).
So what is a good inhibitor going to possess in terms of substrate
halves? Substrates with poor Kcat/Km values are generally good product
inhibitors, whereas good substrates with high Kcat/Km values show less
product inhibition. Why? Because a good product inhibitor is more
pre-ordered to form the end product of catalysis and so mimics the
transition state poorly and hence the energy transition is higher on
binding and through catalysis, but is then nicely relaxed and happy once
cleaved.
If you think in terms of what half of your inhibitor is doing what and
instead of trying to mimic the transition state per se, you consign
yourself to purely building a good substrate or a tight binding product
(or a combination?), and follow the rules (by the way, combining the rules
works *really* well too - I have a peptide inhibitor that is more potent
than non-peptides mixing the classes).
> > Indeed, following these rules you can make a *very* potent
> > inhibitor. Perhaps now in light of viral resistance we can return to real
> > drug design. Which brings me to a question - why do you think it is that
> > assays for HIV-1 PR/inhibitors conducted with mutant enzymes and the
> > compounds that select them fail to predict the said selection?
>> Can you elaborate a little bit what do you mean by "selection"? I must
> admit that I've not been following the work on mutant protease kinetics
> as much as I should.
>> Do you mean to say that HIV-Pr inhibitors toward which the virus becomes
> resistant are /still/ potent in vitro inhibitors of the protease?? I
> could just look up some recent papers from Jon Erickson's lab, but they
> are not easily accessible for me right now.
That is exactly what I mean. Especially in my hands. It is impossible to
predict selection via an assay system. I believe it is the nature of the
assay or the protease, which co-embraces the ligand.
Unfortunately the assay I have inherited is apparently flawed It is at the
wrong pH and salt concentration. This is significant as the pH optimum can
vary by as much as 2 units depending on the particular substrate and can
vary from 4-6, though it can operate from 3-7.5. All my assays are
conducted at pH 6.5 to be more cytosol-like.I believe cytosolic pH is not
relevant to HIV-1 PR activation and activity. The ionic strength has a
great effect on the catalytic efficiency (Kcat/Km) and varies with ionic
strength Km can change from milli--->micro in a transition from I= 0.1-2M,
with Kcat increasing with I. Our assay uses I=0.1. However, the
concentration inside a virion is likely to have a larger value of I and
lower pH value.
To highlight the failings of the assay, think of the optimum
crystallisation conditions which often have a semblance to the assay
conditions. The norm is to grow using 2-5mL of protein-substrate (pH 5.5
acetate buffer) over a reservoir containing 100mM acetate buffer (pH
5.5) and 35-60% saturated ammonium sulfate as the precipitant. If you look
at the structures of these mutant proteases, you will find they are only
subtley different, almost non-recognisable. I wonder if the nature of the
solution (hydrophilic) and the nature of the inhibitors (hydrophobic) does
not create an energetic incentive to form a complex despite the
disfavoured Kcat/Km?
Indeed, I'm wondering if it would be worth trying an assay with the
polyprotein precursors gag and gag-pol without and in the presence of an
inhibitor. I wonder how useful examining the protease totally in the
context of it's native environment would be?
Len...
