Stopping Virulence...One Fragment at a Time.

The best way to not get an infectious disease is vaccinate.   Streptococcus pneumoniae is repsonsible for a million deaths world-wide every year.  For Streptococcus pneumoniae, there are a numbers of vaccines on the market.  These vaccines are bacterial polysaccharides either naked or conjugated to a protein.  They are highly effective, but don't cover all serotypes (there are ~100).  And sometimes a novel serotype arises.  So, if you do get infected treatment is key.  Beta-lactams are the first line of defense, but multi-drug resistance is on the rise, so alternate forms of treatment are needed. Targeting virulence factors has become a recent line of research.  Pneumococcal surface antigen A (PsaA) is strictly conserved surface-exposed lipoprotein expressed by all known pneumococcal serotypes and is essential for colonization and pathogenesis.  PsaA is an integral part of an ATP-binding cassette(ABC) transporter protein complex known as the PsaBCA permease, which is involved in manganese (Mn2+) transport across the bacterial cell membrane. (See there's always a metal involved in cool biology.)  This makes PsaA a good target for pneumococcal infections.  In this paper, a group from down under presents their results using fragments to target PsaA.

They custom built a fragment library (via outsourcing) ~1500 fragments.  This struck me as unusual, if not unique.  Typically, academics make their own or just buy one off the shelf.  I would love to hear why this path was chosen.  In the SI, they do say they used "relaxed" Ro3, but the only relaxation seems to be on the MW.  Have other academics gone this route?  I would love to know more (you can be anonymous in the comments, hint hint).  These were docked into the PsaA metal binding site (Figure 1) based on 3D shape and electrostatic similarity. These were then scored using FlexX. 

Figure 1.  Structure of PsaA. 

The top 300 fragments were manually inspected and then subjected to a cluster analysis.  The 60 most diverse fragments were then tested in a competitive Zn-binding assay.  Zn is a irreversible inhibitor of PsaA and the assay uses this to test for compound binding.  10 of the 60 fragments exhibited greater than 15% inhibition at 100 microM.  Two of these compounds showed greater than 50% inhibition at 1mM (Cpd 15 and 58, Figure 2.)

Figure 2.  Fragments with greater than 50% activity at 1 mM.  Hydrogen bond acceptors are shown in red, H-bond donors in Blue.

So, with crystal structures available, the authors decided to inspect the docked poses rather than actually trying to obtain a structure of the fragments bound to the protein. So even though docked fragments can, and do tend to, keep their original locations, experimental data is key to confirming in silico predictions.  The made 31 compounds around 15, and one that replaced the p-nitro, o-methoxy phenyl with o-hydroxypphenyl was the best 15h (28 microM, pIC50/HAC=0.37).  To that end, they tried to soak apo-crystals with cpd 15h and were unsuccessful due to limited compound solubility and affinity for the target.  They did not attempt soaking compound 58, which they was unable to be further "optimized" with simple SAR.  Cpd 15h did have antimicrobial activity: significant growth inhibition at 180 ug/ml and total growth inhibition at ~800 ug/ml.  They did a further round of optimization.

This is an example of real FBDD approach, in contrast to just using the words.  However, I think this is really a MPU (minimal publishable unit).  If we are lucky, we can expect to see future papers coming out describing their success (or failure) against this target. 

Charting new chemical space for kinase inhibitors

Since the advent of imatinib, kinase inhibitors have become a thing in drug discovery, with more than two dozen already approved. Indeed, kinases are the targets of more than a third of reported fragment-derived compounds to reach the clinic. Given that all 500+ human kinases bind ATP, you would think that the chemical space would be pretty well picked over by now. As Hongtao Zhao and Amedeo Caflisch at the University of Zurich show in a recent Bioorg. Med. Chem. Lett. paper, this is not the case.

The researchers started by extracting all 26,668 kinase inhibitors with MW < 600 Da and IC₅₀ or Ki < 10 µM from the ChEMBL database; three quarters of these were better than 1 µM. These have been tested in aggregate against 367 kinases, of which 88 have more than 100 reported inhibitors!

The molecules were then deconstructed into 10,302 ring-containing fragments, such as benzene (7.1% of kinase inhibitors), 2-methylaminopyrimidine (3.5%) and N-methylmorpholine (2.3%), as well as more obscure structures. In fact, more than half (53%) of these fragments were not found within 7.5 million commercial compounds in the ZINC database. In other words, many fragments that form a part of known kinase inhibitors are not represented among commercial compounds, despite many vendors offering “kinase inhibitor libraries”.

What about the reverse question, analyzing commercial molecules for new kinase inhibitors? The researchers focused on possible “hinge-binding” fragments – those that have at least one hydrogen bond donor and one acceptor in close proximity to one another so as to be able to interact with a conserved region of kinases. Not surprisingly, more than half of the fragments (5681) found by deconstructing the kinase inhibitors fit this description. More interestingly, 196,904 potential hinge binders resulted from deconstructing the ZINC compounds, of which only 1% had been reported as kinase inhibitors.

Digging into the data more deeply, the researchers classified hinge binders as monocyclic, bicyclic, and multicyclic. This analysis revealed that the overlap between kinase inhibitors and commercial compounds was particularly low for multicyclic fragments. This intuitively makes sense: medicinal chemists often turn to ring construction to fix all manner of problems, both pharmaceutical and IP-related, so the under-representation in commercial compounds is likely because medicinal chemists introduce rings into simpler starting molecules. Also, from a molecular complexity standpoint, multicyclic ring systems may be less likely to bind to a wide variety of proteins than simpler monocyclic fragments.

More than five years ago Practical Fragments highlighted a paper from Abbott describing their efforts to generate novel hinge binders. As this and related analyses show, there is still plenty of chemical space left to explore and exploit.

Just Because its called "Fragment-Based"...

When my parents were young and just starting out (the late 60s) they needed a vacuum cleaner.  So a vacuum cleaner salesman came to the house eager to make the sale.  This was the era of the Space Race, plastics, and so on.  So, it was cool to be associated with this.  The eager young vacuum cleaner salesman showed my parents the fine, sleek design of the vacuum cleaner (it was a ELECTROLUX).  It was long and sleek, looking like a rocketship (or a dachshund).  It came with a lot of nozzle attachments.  One in particular was shaped to be very narrow, and get in between the couch and wall for example.   He was particularly proud of this piece: the AEROspace tool.  He even wrote it down as such.  You must have a "AEROspace" tool for your vacuum.  It was an example of great marketing, associate yourself with something very popular to make something mundane appear special.     

So, this paper comes along from Moffatt Cancer Center and USF targeting ACK1 (aka TNK2).  This paper purports to have a "innovative fragment approach" (mix and match).  I love novel approaches to libraries.  So, let's dive in. There is a good deal of work that has been done with ACK1 by Amgen, OSI/Astellas, and others.  Dasatinib and Bosutinib also show activity against ACK1 also.  Based upon all of this previous work and the knowledge of the pyrimidine core they decided to approach the target as laid out in Figure 1.

Figure 1.  Library Design Approach

So, this paper doesn't interest me, although they do come up with some potent compounds, from a what they discovered aspect, rather from a more philosophical aspect. What does it mean to do fragments?  This harkens back to the Safran Zunft challenge.  To me, FBDD is about using simple, small molecules.  Pyrimidine series 9 does not fit any definition of a fragment (Cpd 8 would, but it was never tested AFAIK).  What they did was identify a variety of fragments which would be inputs for creating a small library of lead-like compounds.  However, for this to be "Fragment-based" I would think that they would tested each individual component and prioritized chemistry based upon that.  Or maybe they could have made R3=H.  They don't report Ligand Efficiencies (cue Pete Kenny).  This is simply not "Fragment-based" anything.  Nor, do I think this approach is novel.  Nor do they explain how this is novel.  

So, I think we have entered the time when anything that uses a fragment in the design process is fragment based.  Based on this line of thinking, Nicolaou's total synthesis of Taxol is "Fragment-based". Beware those talking the talk, but not walking the walk.

Predicting protein ligandability and conservation of fragment binding modes

Say you have a protein target, and you want to know whether you will be able to find small molecules that bind to it. A fragment screen can give you a good idea as to the likelihood of success: if you find lots of different fragments with high affinities (say, better than < 0.1 mM), your protein is likely to be highly “ligandable.” On the other hand, if you get very few fragments, and most of them are weak (> 1mM), be prepared for a slog.

Of course, it would be even better if you didn’t have to do a physical screen at all, and two recent papers show how a computational approach may be sufficient. The first, by Dima Kozakov, Sandor Vajda, and their collaborators at Boston University and Acpharis is a detailed how-to guide in Nature Protocols. The second, in Proc. Nat. Acad. Sci. USA by Dima Kozakov, Adrian Whitty, and Sandor Vajda and their collaborators at Boston University, Northeastern University, and Acpharis, addresses some interesting questions about fragment binding.

The main program is called FTMap (also highlighted here); it and several related programs are accessible through a free web server. It is remarkably easy to use: just provide a protein data bank (PDB) ID or upload your own structure and away it goes.

The program works by docking a set of 16 virtual probes (such as ethanol, acetonitrile, acetamide – the largest molecule is benzaldehyde) against a protein and looking for “hot spots” where many fragments cluster. Previously the researchers demonstrated that known ligand-binding sites in proteins tend to be computational hot spots, where at least 16 probes bind. (Note that due to their small size, multiple probes of the same type – acetone, for example – can bind within the same hot spot simultaneously.) In other words,

The strongest hot spot tends to bind many different fragment structures, acting as a general “attractor.”

On the other hand, a hot spot with fewer probe molecules is unlikely to have enough inherent binding affinity to bind to ligands with low micromolar or better affinity.

A related program is called FTSite, which focuses on more thoroughly characterizing the best binding sites. Other programs allow for protein side chain flexibility, docking custom probes, or docking against ensembles of protein models such as generated by NMR structural methods.

The PNAS paper goes further to ask about ligand deconstruction. Specifically, why is it that when a larger ligand is dissected into component fragments, sometimes the fragments recapitulate the binding modes seen in the larger molecule, and sometimes they do not? The answer:

Because a substantial fraction of the binding free energy is due to protein-ligand interactions within the main hot spot, a fragment that overlaps well with this hot spot and retains the interacting functional groups will retain its binding mode when the rest of the ligand is removed.

The researchers support this assertion by examining eight literature examples in which structural information was available for fragments and larger ligands (some of which we’ve covered here, here, and here). In cases where the isolated fragments overlapped with 80% of atoms in probe molecules within a given hot spot, the fragment binding mode remained conserved. Also, these fragments tended to have high ligand efficiency values.

This is neat stuff, and it will be fun to see how general it is. I’m especially happy to see that all of the software is free and open access. Even though I’m hardly a computational chemist, I tried playing around with it and found it remarkably fast and easy to use. So if you have a protein with no known ligands, FTMap can find hot spots, and if they’re particularly promising, this should embolden experimental work.

When Fragments don't deliver...

In the olden days (1980s), during the cold war, Russia was "a riddle wrapped in a mystery inside an enigma".  Kremlin Watching was serious and important thing. When I write up papers, I do the same thing but trying to figure out what the actual story is.  We all know a lot more happened than is written down in 10-20 pages of an article.  This paper has me really doing it; so follow along.

Tuberculosis is a scourge caused by a mighty nasty bug.  People have been using fragments to try to combat it for a long time: 2009 and 2014: targeting pantothenate synthesis and biotin synthesis. AstraZeneca join the party (just as Entasis spins out) with this paper.  In it, they describe their NMR fragment screen combined with a HTS biochemical screen targeting thymidine synthesis.  All the TK inhibitors are TMP or thymidine analogs.  The HTS of 120,000 compounds lead to multiple 1-30 uM active site binding (confirmed by HSQC NMR) inhibitors.  Compound 1

Cpd 1.  3.6 uM, 0.46 LE, 3.54 LLE.  

Figure 2.

was chosen as the basis for the hit to lead campaign.  Modeling suggested that the pyridone core is a thymidine mimic (Figure 2). This novel core allowed to reach sub micromolar potency within 10 compounds of the original hit.  The pyrimidine core was also potent, but not as much as the pyridone.  Pyranones were inactive, as was any other group but the cyano at the 2 position. Crystallography was a key to verifying the binding mode of the compounds.  One point of this is that verified means within 1 A of the predicted pose.  SAR led to the fused pyridinone, a 2 nM inhibitor, which nonetheless had no cellular activity.  The propose that this is due to the ionic nature of the compound, but ureas, amides, and sulfonamides did not afford the desired activity. 

Figure 3.  Fused Pyridinone showing X-ray Contacts

So, as is becoming a very common theme in fragments, they decided to use fragments to try to discover an alternate scaffold.  Using TROSY (HSQC for big proteins), they screen 1200 fragments in pools of 6.  Those fragment hits, termed FRITs which is a first for me (I think I like it.), with a LE greater than 0.25 were followed up by X-ray crystallography.

Figure 4.  Napthyridinone FRIT.  590 uM, LE=0.3. 

Figure 4. shows the best FRIT and its crystal contacts.  Combining this with the knowledge from the cyanopyridinone series, a virtual library was created and docked.  Hidden in their description, it appears that the library was passed by real chemists to prioritize the cpds.  Kudos.  With very limited SAR, they achieved significant potency (Figure 5), but still without cellular potency. 

Figure 5. 200 nM, LE=0.34.  

But, WAIT, this series wasn't advanced any further because the cyanopyridinone was in "advanced lead generation".  Why, you ask?  Well, the oxidized form of Cpd 1 had exhibited moderate cellular activity.  While they don't say it, I would imagine that this means that in doing the analytical work on the compound they found a portion that had oxidized, cleaned it up, and then tested the "bad" part.  I would love to know if this is how it happened.  I would hate to learn they had planned on an oxidized compound all along.

So, on to sulfone and sulfoxides of Cpd 1.  Knowledge from the cyanopyridinone series was used to select appropriate substituents, which seems to indicate a timeline of how things happened or a "we've got nothing left to try" issue.  Again, I would love to know which.  Both the sulfones and sulfoxides showed cellular activity with increase in IC50.  And again X-ray showed that the binding mode was retained, with the sulfoxide adjacent to Arg95.  This then caused them to go back and look at the cyanopyridinones again and realize that the sulfone/sulfoxides might have just the right physicochemical properties.

I think this is a really good paper, and hopefully indicates that more work on this target and with these series are coming.So, I don't know if the fragments failed, or if something better came along.  I would think the latter, but it could be the former.  Again, I would love to know.

Fragments vs Factor VIIa

The blood coagulation cascade involves several serine proteases, many with an appetite for arginine-containing peptides. The polar, basic guanidine moiety of arginine tends to wreak havoc on the pharmacokinetic properties of small molecules, sparking an intensive search for replacements. A few months ago we described how researchers were able to use fragment screening to find an alternative moiety for one member of the blood coagulation cascade. In a recent paper in J. Med. Chem., Daniel Cheney and colleagues at Bristol-Myers Squibb report their work on another, factor VIIa.

The researchers started by filtering commercially available small molecules to look for those with ≤ 17 non-hydrogen atoms, ≤ 3 rotatable bonds, and without anything nasty. This computational work left them with 18,000 fragments. These were then clustered based on similarity, and 200 compounds were chosen by chemists as having the potential to bind in the deep S1 pocket, where the guanidine normally binds.

At the same time, the 18,000 fragments were computationally docked (using Glide) against several different crystal structures of factor VIIa; this “ensemble docking” was used to account for the protein flexibility observed in various structures. This led to a further 250 fragments being chosen.

The 450 fragments were then assessed in biochemical and STD NMR-based assays, and 41 were soaked into crystals of factor VIIa, resulting in 27 structures with fragments bound in the S1 pocket. Happily, 12 of these fragments were – unlike guanidine – neutral. All of them were quite weak (even by fragment standards), with Ki values ranging from 8-19 mM, though searching for related fragments led to some with slightly improved affinities. However, when examining the binding mode of fragment 7, the researchers realized they could use it to replace a more basic moiety in their existing lead series (17), yielding compound 18. Although this reduced the potency, it dramatically improved the permeability. Also, the researchers stated that they were able to subsequently improve the potency, with details to come in a subsequent paper.

This is another nice example of using fragments to fix part of a larger molecule, though it is not necessarily easy. The researchers note that other attempts to append new fragments onto their existing scaffold were unsuccessful, likely due to geometric incompatibilities. This paper is also an illustration of how long it can take to get things published. One of the authors gave a nice presentation on some of this work at an ACS meeting in 2012, and there’s a line in the paper referring to a publication that came out “shortly after completion of this work” – in 2006! Still, late or not, it is nice to see the story in print, with a promise of more to come.

More Notes from DDC 2015

Dan and I both gave our thoughts on the conference last week.  But there was more than just the talks.  There were roundtables.  I chaired one on using kinetics and thermodynamics to drive medchem for the PPI track.  It was a lively discussion.  It was agreed that dyed in the wool enzymologists are priceless.  Kinetics is useless if clearance is the driving force, so this becomes a PK/PD issue.  But does it always have to be?  Paul Belcher from GE shared the On/off rate map (Figure 1) that shows what realm of binding you are in depending on your reates.  Paul also mentioned that Tony Gianetti, formerly of Genentech, used HSA and SPR to assess a more realistic picture of how compounds interact in plasma.  In terms of earlier phase uses, one of the round table attendees mentioned that she had seen talks of people using kinetic data to drive medchem.  She couldn't recollect who, so if any of our astute readers have references please share.  We also discussed using kinetic data to rank compounds with similar IC50.  A question was raised whether or not kinetics can be a good surrogate for receptor occupancy? 

Figure 1.  On/off Rate Map: A = affinity limited efficacy, B= on rate limited efficacy, C= rapid off rate limited, D= slow off protected efficacy

So, what about thermodynamics?  By and large, this was viewed as retrospective only.  Paul from GE did share that they have an app note of using SPR to generate thermodynamic data (I can't figure out how to link it, so if you want it contact me (or Paul) and we can send it). 

The main thrust was that neither kinetics nor thermodynamics are used to make prospective medchem decisions, rather they are used to justify in retrospection. Specifically for PPIs, the consensus was that the focus should be on on rate because you have to the compound in there when you can (i.e. when the complex is "open" enough). 

Derek Cole of Takeda led one of the FBDD round tables: Practical Aspects of Fragment Screening. Here is a picture, courtesy of Bjorn Walse of Saromics.  

His notes are replicated below:

Round table became figure 8 with two tables, with 2-3 deep seats and 40 -50 participants. FBDD expertise from novice to experts, including Teddy Zartler, Dan Erlanson, Gregg Siegal, and Andrew Petros. Large attendance highlights the number of newcomers to FBDD, confirmed by Dan Erlanson during opening when 2/3 of attendees indicated this was their first CHI FBDD meeting. Very lively debate/discussion covering 4 primary targets.

1. Designing and building and storing libraries. Discussed size of library i.e. 1000 or 40K. Agreed that a good library of 1000 should yield lots of high quality hits. Best to keep HA low, 10 - 16 (majority in 12 - 14 range). Discussed 3D vs. flat fragments. Flat give higher hit rate and should be major part of library. 3D likely give lower hit rate but may yield very exciting hits. Discussed complexity and the need for fragments to have enough complexity, but not two pharmacophores. (ref. Astex work). IF just starting out, best to buy a vendor library, e.g. Maybridge or others, which are fully characterized.

2. Screening techniques. NMR and SPR most common. Both very good. Tm - fast, inexpensive and can correlate with x-ray. What to do if no biochemical activity. Might be fine if below sensitivity of biochemical assay, i.e. very small fragment, however if larger fragment, need to understand why not being detected.

3. Potential pitfalls. Make sure library is soluble above assay conditions, i.e. > 1 mM in aqueous buffer (1 - 2% DMSO). Check for aggregation. Run SPR clean screen.

4. Fragment hit follow up. Think of fragments as seeds to identify protein compatible pharmacophore. SAR by catalog of similar fragments or fragments which present a similar pharmacophore is of great value. May find fragments which are much more potent, efficient, or which crystallize (if original was unsuccessful). Good to design diverse library, but similarity in fragments is different than similarity in large molecules, small 1-atom changes can have profound effect on binding mode, potency, etc.

If I missed any other highlights, please add them in the comments, or email me and I can add it in.  

Sloppy science

As regular readers may have discerned, I’m favorably disposed to most of the papers I highlight. They may have flaws or inconsistencies, but, with rare exceptions, I generally just ignore particularly problematic publications. Last year Teddy introduced the term PAINS-shaming to draw attention to – how shall we phrase it? – less salubrious specimens. Building on this alliterative theme, today’s post is about sloppy science. A fundamental tenant of sound science is to consider alternate explanations for results. Ignore this at your peril.

An example was published in J. Cancer Prev. The researchers were interested in a mutant of isocitrate dehydrogenase 1 (IDH1), a hot cancer metabolism target. They screened 500 fragments in a functional spectrophotometric assay, with each fragment present at the very low concentration of 5-10 µM. One of these inhibited the mutant protein by 80% – pretty impressive for a fragment. Until you look at the structure: 2-(3-trifluoromethylphenyl)isothioazol-3(2H)-one (shown below).

Fifty years ago, researchers showed that this chemical class (isothiazolinones, also called isothiazolones) could react with thiols, like this:

Isothiazolinones have been categorized as PAINS, though they do not show up in the original computational filters. However, Pete Kenny has (repeatedly) stated that having a dubious structure should not automatically disqualify a compound from further investigation, so what else do we know about isothiazolinones?

Well, there’s this paper, which concludes a discussion of isothiazolinones by stating:

We could not develop these into useful compounds and ultimately the structure–activity relationship (SAR) was uninterpretable. Most insidiously, there were encouraging aspects of sharp SAR as there always are with these PAINS, but this is eventually overwhelmed by flat and nonsensical SAR. Unpredictable nonspecific cytotoxicity was manifest. We found our compounds to be rapidly reactive with thiols under assay conditions.

Of course, one could argue that this is anecdotal. But then there’s this paper, with the unambigious title “Isothiazolones; thiol-reactive inhibitors of cysteine protease cathepsin B and histone acetyltransferase PCA”. The first line of the abstract states:

Isothiazolones and 5-chloroisothiazolones react chemoselectively with thiols by cleavage of the weak nitrogen-sulfur bond to form disulfides.

The researchers go on to demonstrate this using both small molecules and proteins, and some of the compounds they investigate are structurally quite similar to the hit here.

So in all likelihood the fragment described in the most recent paper reacts with one or more cysteine residues in IDH1, of which there are several. It is notable that the researchers conducted their assay in the absence of added thiol reducing agents, so modification of the cysteines would effectively be irreversible under their assay conditions.

What we have here is the re-identification of a known thiol-reactive molecule without any acknowledgement or apparent awareness that the molecule is reactive. I have no problem with covalent inhibitors, but I do have a problem with a generically reactive molecule being touted “for a future lead development”, as the researchers state in the abstract. It took me just minutes to track down the references above, and the fact that neither the researchers nor the reviewers did so is inexcusable.

Granted, this paper is not published in a high profile journal, and the easiest response would be to ignore it. It is certainly not the only one of its kind. Doing so, however, implicitly endorses sloppy science. This paper will undoubtedly pad the resumes of the authors. Highlighting its problems will hopefully make others wary of wasting time with this new "selective inhibitor."