Drugs Against Bugs

This paper, from a consortium  of academics in Holland, Belgium, and Switzerland and a UK pharmaceutical company, reports on inhibitors for parasite specific phosphodiesterases (PDE). Trypanosomal disease is a major health obstacle in Africa; if the diseases enters into stage 2 (CNS penetration by the parasite) the patient dies. Trypanosoma brucei has two subspecies: T.b. gambiense and T.b. rhodesiense which makes the search for a pan-anti-trypanosomal agent even harder.  The current treatment options: 

...are limited and suffer from suboptimal dosage regimens and/or severe toxicity. During the second stage of the disease, the only choices are the arsenic-containing drug melarsoprol and the ornithine decarboxylase inhibitor eflornithine, used as monotherapy or in combination with nifurtimox (NECT).  However, eflornithine is not effective against T. b. rhodesiense.

Mapping of the T. brucei genome lead to the identification of trypanosomal PDEs.  Two of these PDEB1 and B2 play a pivotal role in proliferation.  Knocking down one or both TbrPDEB1 andTbrPDEB2 simultaneously leads to an arrest of parasite cell division, lysis of the parasites, and elimination of the infection in vivo (in an infected mouse model). Of supreme interest, the "P-pocket" (found in the crystal structure of  Leishmania major crystal structure) was conserved in their homology model. 

The HTS found compound 1 and in their follow up to chemotypes related to this, the came up with compound 6b.  Rolipram itself was inactive (>100uM) despite fitting nicely in the pocket (according to modeling).  However, the catechol moiety was shown to be the key for activity.  Replacing it with a less "worrisome" moiety led to a significant decrease in potency.  Compound 8d (R1, Obenzyl, R2=cycloheptyl) had 500 nM potency that maintained the same ligand efficiency, which indicates the atoms added were reliably efficient, but not super-efficient.  It also had anti-proliferative effects without cytotoxicity; the authors guess that this is due differences in cell-permeability compared to the cyclopentyl compounds. 

The benzylcatechol chemotype was then chosen for further optimization.  The docking shows that the benzyl moiety is at the entrance of the P-pocket (figure 3a) which should afford access to it through growing.  Table 2 shows the results of their efforts here with both the cycloheptyl and isopropyl (which were more ligand efficient).  The docking suggested that they should be able to grow into the p-pocket from the 4 position (Figure 3a).  Compound 20b (49nM) was the best inhibitor and as shown in the docking (Figure 3b) this is most likely due to interactions in the P-pocket.  Interestingly, they also propose that it may be displacing water from this pocket causing some sort of increase in affinity.  

The cLogP of the cycloheptyl compounds was a real issue affecting solubility and may have affected the antitrypanosomal activity.  20b, with a solubilizing tetrazole group, concentration-dependently worked in the anti-trypanosomal assay with a IC50 of 520nM.  It was also found to be very potent against PDE4A-D enzymes.  It should also be noted that 20b is 40 heavy atoms, which is 0.18LEAN. However, 20b was a good inhibitor of  T. b. rhodesiense (60nM).  It was tolerated by a human fibroblast cell line (250 fold sensitivity index)  Lastly, they were able to show that cAMP accumulated in the cells, indicating the target specific effects.

Often, I am very harsh on academic drug discovery, but this paper is an excellent example of what can be accomplished when it is done right.

 

Entropy-enthalpy transduction: time to throw up our hands?

Thermodynamics has come up several times on Practical Fragments. The binding of a ligand to a protein can be dissected into enthalpic and entropic components. Very simplistically, enthalpy underpins directed, often polar interactions, while entropy plays the dominant role in non-directed, hydrophobic interactions. Ligands that bind primarily through enthalpic interactions (such as hydrogen bonds) have been suggested to be more selective and “best in class”. Historically, a key theoretical advantage of FBLD is the notion that linking two fragments can provide an entropic advantage to the combined molecule compared with the isolated fragments. However, as discussed recently, reality sometimes cocks a snook at theory.

One stumbling block in trying to apply thermodynamics rationally is enthalpy-entropy compensation, a perverse trick of the universe in which, when you improve the enthalpy of an interaction, you may worsen entropy, and vice versa. For example, if you introduce a hydrogen bond into a protein-ligand interaction, the precise positioning required may cause increased rigidity, at an entropic cost.

Now a new paper in Proc. Nat. Acad. Sci. USA from Michael Gilson and colleagues at UCSD suggests that things may be even more complicated. They analyze a previously published 1 millisecond molecular dynamics simulation of the small (58 residue) protein BPTI. There are three main conformational states (or clusters), each with similar overall energies. However, the researchers find that the different conformational states have very different global enthalpies and entropies. Worse, very tiny perturbations, such as the distance between two side chain atoms, can cause one state to shift to another, in turn dramatically changing the overall thermodynamic signature.

In practice, this means that when you measure the thermodynamics of a ligand binding to a protein, the enthalpic and entropic changes observed could have more to do with subtle changes in the global conformation of the protein, or even changes in solvent binding, than to the ligand-protein interaction itself.

The researchers call this phenomenon entropy-enthalpy transduction (EET):

The thermodynamic character of a local perturbation, such as enthalpic binding of a small molecule, is camouflaged by the thermodynamics of a global conformational change induced by the perturbation, such as a switch into a high-entropy conformational state.

The researchers argue that EET could occur in many protein systems, so experimentally determined values of entropy and enthalpy for ligand binding are actually unreliable indicators of the local thermodynamic driving forces we normally try to influence.

Although the researchers develop a sophisticated mathematical framework to describe EET, at the end of the article I’m left wondering, is there any hope of using thermodynamics for practical drug discovery?

Fragments vs PI3K – AstraZeneca’s turn

Last year we highlighted a paper from AstraZeneca in which they used virtual screening to identify fragments that inhibit the p110β isoform of phosphoinositide 3-kinase (PI3K), a potential anti-cancer and antithrombotic target. In two recent papers in Bioorg. Med. Chem. Lett., Fabrizio Giordanetto and colleagues describe the optimization of one of these fragments to a potent, selective molecule with in vivo efficacy.

The first paper describes the initial fragment-to-lead work. A variety of changes to fragment 1 were explored, including adding a lipophilic phenyl group to increase potency (compound 2). At the same time, modifications were explored in the central pyrimidinone ring. Although compound 3 was less active than compound 2, it also had a considerably lower logD. Many additional changes were explored, and ultimately one of the most potent compounds was compound 16. The isomeric compound 22 was less potent, but had significantly better solublility and stability in a microsomal assay.

The second paper describes subsequent optimization, ultimately yielding compound (S)-21. There’s a lot of good medicinal chemistry that I can’t do justice to here, so definitely check out the two papers themselves. Compound (S)-21 is potent, selective against the related kinase p110α, and shows good activity in a dog model of platelet aggregation without causing an increase in bleeding time.

One of the nice things about this work is the fact that the researchers used a fragment-hopping approach and were not focused on potency to the exclusion of all other properties. Although one could argue that this is simply good medicinal chemistry, it can sometimes fall into the category of what Mike Hann has memorably christened “unknown knowns,” a trap this team avoided.

MK-8931, BACE1 inhibitor, enters Phase 2/3 for Alzehimer’s

Perhaps no other single target so successfully demonstrates the potential of fragment-based approaches as BACE1, a challenging aspartyl protease implicated in Alzheimer’s Disease. Practical Fragments has previously written about efforts from Merck, Lilly, Pfizer, Evotec, and Amgen (there’s also the Astex-AstraZeneca collaboration).

Today Merck announced that their MK-8931 has entered a Phase 2/3 clinical trial, a randomized placebo-controlled study which will run for 78 weeks and enroll up to 1700 patients in the phase 3 portion.

There have been concerns lately as to whether or not BACE1 is a viable target for Alzheimers; it will take large studies like this to answer that question. Practical Fragments wishes them luck.

Is this a Fragment?

Fierce Biotech has a listing of the top 15 (potential) late stage blockbusters in development.  I think these lists Fierce put out are very interesting.  The last one on the list is alpharadin.  Its a potential treatment for bone metastases and Bayer just came up with data that it could also be a broad spectrum treatment for prostate cancer.  We typically don't discuss late stage development; so few of us have actually been there its like Oz.  What makes this so interesting to me is the chemical structure of alpharadin: radium-223 dichloride.  RaCl2.  Three atoms or 297 g/mol. 

I realize this is not a fragment in the true sense of what most of us do.  I found it fascinating nonetheless.

FBDD Down Under

Fragment-Based Drug Design Down Under was held at Monash University in Melbourne, Australia earlier this month. The first dedicated FBDD conference in this country was full of enthusiasm: I had the impression many of the 100 or so participants, most of them Australian, were happily surprised to meet so many other fragment aficionados. With 18 oral presentations, nearly as many posters, and a lively panel discussion I can only touch on some of the broader themes here, so please weigh in with your own observations.

Fragment library design received considerable attention, which was nice as this is an area that is all too often ignored in conferences. Pete Kenny's name came up a couple times in helping to put together the CSIRO fragment library. Craig Morton gave an excellent overview of the SVIMR fragment library and some of the challenges constructing it: of roughly 1600 fragments purchased, 450 were either not sufficiently soluble in water or DMSO or not sufficiently pure to be included. David Chalmers of MIPS presented an analysis of the physicochemical properties of approved drugs, noting that roughly three quarters are ionizable, with potential implications for library design.

Three dimensional fragments have been much discussed lately, and Martin Drysdale of the Beatson Institute described a UK consortium, 3Dfrag.org, to put together a library of 3-dimensional fragments, as defined by having a principal moment of inertia closer to a sphere (think adamantane) as opposed to a plane (benzene) or a rod (2-butyne). The project is still in its early stages, with about 200 fragments acquired thus far. Martin also described an interesting collaboration with the Broad Institute to use existing DOS-derived fragment-sized molecules for screening.

There were several talks on fragment screening methods, especially NMR and SPR. In an intriguing comparison, Jerome Wielens of SVIMR described parallel efforts on HIV integrase, both using essentially the same (Maybridge) library. STD NMR screening produced more than 50 hits, ultimately yielding 15 co-crystal structures, while SPR screening (also discussed later by Tom Peat of CSIRO) produced 16 hits and ultimately 6 crystal structures, yet few of the hits were in common. There were differences in the protein constructs and pH, and some of the NMR hits may have been artifactual, while the use of a reference protein in SPR may have weeded out some true positives. All of which underlines the fact that using multiple biophysical methods is ideal.

STD NMR came under scrutiny from others as well: San Lim of MIPS described compounds that showed a signal when screened in mixtures but not when tested individually, and Martin Drysdale discussed one target that gave a 36% hit rate using the technique, leading him to pick SPR as a primary screening method. Still, there are some interesting possibilities: Thomas Haselhorst of Griffith University discussed using STD NMR not just for screening membrane proteins but for screening viruses, cells, and even fungal spores!

Markku Hämäläinen of GE Healthcare discussed the use of both SPR (specifically Biacore) and ITC. In the case of SPR, he termed one class of problematic compounds “selective promiscuous binders”: for example, a positively charged protein may cause negatively charged fragments to aggregate around it, giving anomalously high signals. Using a positive control and setting a maximum Rmax in fitting the data can help weed these out and provide more accurate dissociation constants. In a collaboration with Merck Serono on a kinase target, 105 hits from a 1920-fragment library gave an 80% confirmation rate when tested in ITC, and 41 of 48 produced co-crystal structures.

But as we are increasingly seeing, Biacore is no longer the only name in the SPR game: Olan Dolezal of CSIRO described Bio-Rad’s ProteOn instrument and found that, while it was less sensitive than Biacore, its higher throughput made it an attractive primary screening instrument.

There were also a couple interesting talks on in-situ methods for fragment assembly, including MS-based methods described by Sally-Ann Poulsen of Griffith and click-based methods discussed by William Tieu at the University of Adelaide. One of the problems with assembling a high-affinity molecule in situ is product release: a molecule made in situ might bind so tightly it never leaves the protein, which essentially stops production once a stoichiometric amount of the inhibitor is made. In Tieu’s case, the problem was cleverly overcome by introducing a mutation to lower the affinity.

Finally, Jonathan Baell of MIPS gave an excellent (though disturbing) talk on PAINS – a topic which is still unfortunately not sufficiently appreciated. In one illuminating example, he found that an in-house screen of a histone acetyltransferase produced only a single legitimate hit, along with a plethora of PAINS.

One common theme both in the presentations and offline discussions was the relative lack of chemistry support; definitely a pity, since there are certainly plenty of chemists looking for new opportunities. Of course, funding chemistry is a problem not unique to the Southern Hemisphere.

Australia is clearly a new world for fragments, and it will be fun to see how the field develops there. And on a personal note, I found Aussies to be some of the warmest, most genuine people I have met in any country. I definitely look forward to finding an excuse to return.

The End of the Trilogy

If you haven't had enough computational fragment papers, here is one more.  In this paper, Zhao et al. set out to find potent inhibitors of EphB4, a receptor tyrosine kinase.  This is not novel target space; the inhibitor dasatinib is already on the market.  This paper is an extension of the groups previous computational method, ALTA (Anchor-based Library Tailoring).  In ALTA, 1. small, mainly rigid fragments are docked. 2. Compounds with the most favorable binding energy are used to select  compounds which contain that fragment.  3.  Fragment is then flexibly docked.  In this paper, they add explicit solvent molecular dynamics to this process. 

The figure below shows the approach to selecting fragments for screening.  Most of the 563,000 fragments are 150-300 Da, have fewer than 5 rotatable bonds, and no formal charge.  They then selected a kinase focused collection by retaining only those with molecular weight smaller than 300 Da, a maximum of three rotatable bonds, more than one ring [Emphasis mine], and the capability to form two hydrogen bonds with the backbone polar groups of the so-called hinge region. For the latter criterion acidic CH groups (e.g., in aromatic rings) were also considered as donors.  The one ring criterion is because single ring anchors don't give enough energy of binding AND to open up IP space.

I find it strange that they consider an aromatic hydrogen capable of H-bonding for the purpose of calculating free energies.  I would like to know what impact this additional factor had in picking the compounds; it is not explained in the paper, nor is it explained in the SI.  

This led to the three active compounds shown in the table below.  Previous work by them showed that the hydroxy at position 5 of compound 1 (Compound 7) would generate a significant increase in binding energy, through two additional hydrogen bonds.  [And excuse my pedantry here, but there is no position 5, is there?  Aren't position 3 and 5 here the same and indistinguishable?  I am not the worlds best chemist, but I do know an equivalent position when I see one.]  

 Modeling confirmed this (prior to the initiation of chemistry).  This was confirmed by compound 7 being 50x more potent than the parent compound 1.  It is nice to see that this potency correlates with 2.5 kcal/mol or the addition of 2 additional hydrogen bonds, as was predicted.  They then co-crystallized the compound with the target EphA3, despite EphB4 being the actual target.  32 of 36 residues in the actives site are the same, including the 100% identity for those involved in binding 7. The 1.7A structure confirmed the predicted mode of binding.

 

It was then tested against related Y-kinases: 0.338 μM for Src, 0.864 μM for Abl1, 1.38 μM for Lck, 1.62μM for EGFR, while no inhibition was observed for IGF1R.Thus compound 7 has higher affinity for EphB4 than for these five tyrosine kinases. It also showed cellular activity.  

To sum up the state of the art of computational FBDD:

You can find fragments that fit in an active site, even if you have to model the active site.