26 June 2012

Do You Need Structure To Prosecute Fragments?

As pointed out here, the majority of people think you need to have structure to successfully move fragments forward.  I, as has been noted previously, disagree and vehemently so.  However, if you fall in the category where you need that crutch you may be familiar with INPHARMA.  X-ray crystallography is the workhorse of structures, but in many cases (far more as we get into more and more complex targets), X-ray fails.  Old-fashioned solution NMR (with 15N, 13C, etc.) is just not fast enough to be a viable tool.  A few years back, INPHARMA was introduced to try to bridge the gap between the two methods.
INPHARMA (INter-ligand mapping for PHARmacophore MApping) is based upon robust ligand-based screening data.  Two weakly binding, and competitive, fragments are put into solution with the target.  If the compounds are weak enough (>10uM) it is possible to detect a NOE between protons on the two ligands.  This NOE is mediated by the active site, so the compounds must be binding in the active together (see below).  It is then possible to run very intense computational methods to determine the orientation of the ligands in the binding site (if the structure of the target is known).  

In this paper, Isabelle Krimm looks at INPHARMA's ability to discriminate different binding sites on a single target.  For this she uses, glycogen phosphorylase (a type 2 diabetes target) that has four distinct binding sites: active site, inhibitor site, allosteric site, and new allosteric site. 
Cpd 1 and 2 bind the inhibitor site, where Cpd 3 binds the new allosteric site.  Cpds 4 and 5 are proposed analogs of 2, Cpds 6 and 7 are analogs of 3, and Cpds 8 and 9 are "frequent-hitters".  NOESY experiments were recorded for the six fragments in the presence of 2 and 3.  All compounds exhibited intramolecular NOEs upon binding to the target.  Additionally, Cpds 2 and 4/5 showed intermolecular NOEs, as did Cpd 3 with 6, 7, 8, and 9.  Cpds 4 and 5 did not shows NOEs with 3, nor did 6, 7, 8, and 9 show NOEs to Cpd 2.  For fragments 4-7, competition data support that these are inter-ligand NOEs and that Cpds 8 and 9 are non-specific binders.  No intermolecular NOEs were seen to Cpd 1.  Cpd 1 has a IC50 of 1 uM, while Cpd 2 is 100uM.  This supports theoretical calculations that the two competitive binders must have binding constants no more than 8x different.  

In this paper, Lee et al. demonstrate the use of hyperpolarization to increase the sensitivity of INPHARMA. Dynamic Nuclear Polarization (DNP) uses electrons to transfer magnetization to nuclei leading to orders of magnitude increases in signal.  Lee et al. use DNP to increase the signal in INPHARMA significantly by hyperpolarizing one of the two competitive compounds, in a method they call HYPER-BIPO-NOE (Hyperpolarized binding pocket NOE)[which may be one of the worst not-even-acronyms ever]. 
1D HYPER-BIPO-NOE spectra, a) in full scale and
b) expanded to show transferred signals. Stacked spectra are, from top
to bottom: hyperpolarized ligand 1 with ligand 2 and protein (HYPERBIPO-
NOE); hyperpolarized ligand 1 with ligand 2, but without protein
(Control 1); hyperpolarized [D6]DMSO/D2O, with only ligand 2 and
protein (Control 2); Thermal spectrum of the HYPER-BIPO-NOE
sample (Thermal). The resonance from DMSO, which was suppressed
using presaturation, is designated by *.
This tremendous increase in signal can be extraordinarily useful for INPHARMA applications because they are able to obtain the inter-ligand NOE in a single scan.  The authors then go on to demonstrate that HYPER-BIPO-NOE-INPHARMA data is similar to STD-INPHARMA and can be used to generate binding poses. 

These two papers show that INPHARMA can be a useful tool to orient fragments similarly in a binding pocket.  INPHARMA requires that you have a weak binder that binds in a known site on the target.  For the current generation of targets this may exclude INPHARMA from being used.  It is also important to note that the choice of mixing time (which can range from 70-800ms) is a critical parameter.  The incorrect choice of the mixing time can lead to poor signal intensity and false negative results.  DNP, while turnkey if you have enough money, is not common in industry at all.  I would be surprised if HYPER-BIPO-NOE-INPHARMA actually gets any traction there at all.  There is also a poll with this post, please read and answer.


Dan Erlanson said...

This is cool. In terms of advancing fragments without structure, the tangentially related NMR technique SAR by ILOE could work, though of course one needs to be careful to avoid artifacts. There is also the FtsZ story we discussed a few years back.

Anonymous said...

For a small FBDD biotech like the one I work for it is instrumental to work with targets that are fully fit for the lead generation technology we apply. This enables us to work efficiently and stay competitive, so structures are a must. If you are desperate for a particular target and have the time and money to spend I guess there is nothing stopping you from working with fragments without structures, but perhaps other lead generation approaches could more efficient in that case?

Morten G said...

Annnnnnnnd I'm confused.
Doesn't "It is then possible to run very intense computational methods to determine the orientation of the ligands in the binding site (if the structure of the target is known)." mean that you need a protein structure?
Which can be very tricky - especially with human proteins since you might not even be able to over-express or purify them.

Dr. Teddy Z said...

Morten G...for the very intensive structure calculations placing the fragment in the binding site, yes you do need a structure. However, if you observe an INPHARMA-NOE then you know your fragment is binding where the other fragment is, and if you know where that binds, possibly from structure previously determined, then you don't need structure. However, you could also get binding site information from native ligands, etc, and thus would not a priori need structure.
However, this has always been my problem with INPHARMA, it basically requires that you have structure at some point, which in my eyes, means that it is not going to be very useful for the current generation of targets (mostly non-structurally enabled).

Dr. Teddy Z said...

Anonymous...there are non-XRay structural techniques which can still give some structural data and guide FBDD. However, it all depends on your technology, and there will be different strokes for different folks.

Anonymous said...

Excellent post - thanks for putting it out there to remind us of all of the excellent methods available to us. I am all for using whatever means necessary to get at information that can help drive decisions on hypotheses, compounds to be made, etc. The very difficult part of using less common methods lies in convincing medicinal chemists who were raised on X-ray structures to accept the data derived from expts like these as 'high resolution' information.