Screening MiniFrags by NMR

Small is becoming big. Five years ago we highlighted MiniFrags, consisting of just 5-7 non-hydrogen atoms; FragLites and MicroFrags soon followed. Screening these tiniest of fragments at high concentrations can thoroughly explore hot spots on a protein and identify favorable molecular interactions. But because they are so extraordinarily small, experimental methods for screening them have been mostly limited to crystallography. In a new J. Med. Chem. paper, Annagiulia Favaro and Mattia Sturlese (University of Padova) turn to the most venerable of fragment-finding methods, NMR.

 

The researchers started with the 81 reported MiniFrags and removed those with aqueous solubility less than 250 mM or without protons observable by NMR (such as phosphate). The remaining 69 fragments were dissolved directly in phosphate buffer, mostly at 1 M concentration, though lower solubility fragments were dissolved at 250 mM. Importantly, the pH of each sample was carefully adjusted to 7.1 to ensure that any signals correspond to MiniFrag binding and not to changes in experimental conditions.

 

As a test case, the researchers chose the antiapoptotic target BFL1. This protein is related to BCL2, the target of venetoclax, which was discovered using SAR by NMR. BFL1 has a hydrophobic cleft with five subpockets and has been studied by NMR. Like other BCL2 family members it is a difficult target, as we noted earlier this year.

 

The actual screen was done using chemical shift perturbation (CSP) detected by two-dimensional ¹H-¹⁵N HMQC. Fragments were screened at 100 mM, a 5000-fold excess above the protein concentration. Hits were confirmed at 20 mM (more on that below). As with the library preparation, pH was carefully controlled.

 

At such high ligand concentrations, any impurities could become a problem: a 2% contaminant would be present at 2 mM. To weed these out, the researchers performed WaterLOGSY experiments. These only produce a signal at ligand to protein ratios much lower than 1000 to 1, so any hits could only come from impurities.

 

Even at high concentrations, CSPs caused by weak fragments are small, so the researchers developed an analysis method to identify those that shift more than at least one standard deviation from the average. CSPs can shift in any direction on a two-dimensional map, but any one protein-ligand interaction should shift signals in the same direction. Here is where the 20 mM confirmation experiment comes into play: a “cosine similarity” assesses whether two CSPs are in the same direction and thus likely to be real.

 

Screening BFL1 led to 53 hits, a hit rate of 78%, similar to crystallographic screens of MiniFrags against other targets. Forty percent of MiniFrags bound to multiple sites on the protein; only 11 (16%) bound to a single site. The five subpockets were each liganded by 6-17 MiniFrags. In subsequent experiments, the researchers were also able to measure binding of two different fragments to different pockets simultaneously, akin to SAR by NMR.

 

This is an interesting approach, but while fragments with >5 mM dissociation constants have been advanced to drugs, the utility of a 100 mM binder remains to be seen. That said, the technique could be a boon for understanding protein-ligand interactions, and I look forward to seeing it applied more broadly. In particular, screening the same set of MiniFrags on the same protein by NMR, crystallography, and computational methods could be quite informative.

Detailing hot spots with atomic consensus sites

Practical Fragments has written frequently about hot spots, regions on proteins that are predisposed to bind ligands such as drugs. Determining whether a protein has a hot spot can help prioritize a target for screening, and one of the more established computational approaches to do so is FTMap, which we wrote about most recently just a couple months ago.

 

While FTMap can tell you whether a protein has one or more hot spots, it provides few further details, such as which regions might prefer a hydrogen bond donor or acceptor. This has now been addressed in a new J. Chem. Inf. Mod. paper by Sandor Vajda and collaborators at Boston University, Stony Brook University, and Acpharis. (Diane Joseph-McCarthy presented some of this work at the CHI DDC conference earlier this month.)

 

The original version of FTMap started with a collection of 16 very small molecule probes: these were docked all over a protein, with hot spots being identified as consensus sites where many probes bound. To get more information about each hot spot, the researchers have extended the method – now called E-FTMap – by increasing the number of probes to 119 covering key functional groups. For example, whereas FTMap included dimethyl ether as a probe, E-FTMap also includes 2-methoxypropane, 2-methoxy-2-methylpropane, and tetrahydropyran. If all these probes bind with the oxygen in the same part of the hot spot, this suggests a predilection for a hydrogen bond acceptor, and also provides information about nearby hydrophobic contacts.

 

By using a sufficiently diverse group of virtual probes, E-FTMap is able to more finely detail hot spots, tallying the “atomic consensus sites” within them. This is reminiscent of an approach we wrote about several years ago, though that method used just three different probes.

 

To benchmark E-FTMap, the researchers took 109 fragment-to-lead pairs with published crystallographic information and assessed whether the program could identify interactions that had been experimentally observed. The results were encouraging and far superior to the original version of FTMap. The highest ranked atomic consensus sites generally overlapped with appropriate atoms in fragments and leads. Interestingly, the results for fragments were better than those for leads, and the researchers suggest this is because the fragment “core is responsible for the bulk of the binding free energy in a ligand and that larger ligands bind by forming additional interactions at weaker hot spots that surround the fragment binding site.”

 

Next, E-FTMap was tested against five proteins for which between 31 and 353 fragment-bound crystal structures were available. Here too the program was broadly successful, though some fragments bound regions of the protein that E-FTMap overlooked, particularly in cases where there were conformational changes. This is not surprising given that the program assumes the protein remains rigid. (Other computational approaches such as SWISH, which we wrote about here, are starting to account for protein flexibility.)

 

E-FTMap looks qualitatively at specific atomic interactions, and one question I had was how well the atomic consensus sites matched up with binding affinities of known fragments; perhaps some crystallographically identified fragments bind so weakly one would not expect to find them computationally, as we discussed here and here. This hypothesis might be tested by focusing on comparisons with experimentally characterized fragments with the highest ligand efficiencies.

 

Also, I was struck by the fact that the virtual probes in E-FTMap are roughly the size of MiniFrags or MicroFrags, and I couldn’t help but wonder how well the atomic consensus sites from the virtual screens would correlate with the binding modes of these tiniest of fragments.

 

One nice feature of E-FTMap is that it can be accessed through a simple web server, so if you’re interested in these and other questions you can test it for yourself. If you do, please share your experiences.

Hot spots real and imagined

Practical Fragments has written several times about “hot spots”: regions on proteins where small molecules and fragments readily bind. Knowing whether your target protein has a hot spot can help you decide whether to pursue the target in the first place. A variety of computational approaches have been developed for finding hot spots, most of which start with a crystallographically determined structure. In a new J. Chem. Inf. Mod. paper, Sandor Vajda and collaborators at Boston University and Stony Brook University ask whether computational models of proteins can also be used for one of the more popular methods, FTMap.

 

The researchers started with a set of 62 proteins, each of which had a published crystal structure bound to a fragment (MW < 200 Da) as well as to a larger molecule. The predicted structures of these proteins were then downloaded from the AlphaFold2 (AF2) site, and these models were truncated to correspond to the residues seen in the crystal structures to facilitate comparisons. The computational models were quite similar to the experimental models, particularly when comparing the positions of the peptide backbone atoms which define the overall shape of the proteins.

 

Next, the researchers applied the program FTMap, which computationally explores the surface of proteins using a set of 16 very small probes such as ethanol. Hot spots are regions where lots of probes bind, and the “hotness” of these spots correlates with the number of bound probes. FTMap assessed hotness on the AF2 structures and the crystallographicaly determined structures. (Before running FTMap, the bound ligands in the crystal structures were computationally removed.) Additionally, the researchers ran FTMap on unliganded crystal structures for the 47 proteins where these had been reported.

 

FTMap was broadly successful at finding the hotspots defined by bound fragments, succeeding 77% of the time starting with either the fragment-bound or unliganded structures and 71% starting with the AF2 models. Implementing stricter criteria (demanding the experimental fragment binding site be the top hot spot, for example) reduced the success to 56% for the crystallographic starting points and 47% for the AF2 models.

 

The paper discusses several examples in detail, in particular the two where the AF2 models were most different from the experimental models. Both of these were large, multidomain proteins. When AF2 models of just the ligand-binding domains were used, the models were significantly improved. This seems to be a generally useful hack: generating truncated AF2 models for other proteins also improved the performance of FTMap.

 

The utility of AF2 models for docking has been the subject of some debate, with some arguing that even though the overall protein folds may be accurate, local side chain conformations may be wrong, and a single side chain rotation may make the difference between ligand binding or not. This paper suggests that hot spots are not too sensitive to these subtleties, and that AF2 models can be used for finding hot spots.

Ensembles of fragment structures guide selectivity

Scientists generally want structural information when a project begins, and ideally that structural information comes from crystallography. Most of us who have been doing drug discovery for a while can remember seeing the first structure of a favorite molecule bound to a target protein and being inspired, reassured, or sometimes confused. But as crystallography becomes increasingly high throughput, it is now not uncommon to obtain dozens or even hundreds of structures. What to do with all this bounty? In a recent open-access J. Chem. Inf. Model. paper, Mihaela Smilova, Brian Marsden, and collaborators at University of Oxford, the Cambridge Crystallographic Data Centre, and Exscientia describe one application.

 

Back in 2016 we wrote about a computational approach called hotspot mapping, which uses three small fragment probes (aniline, cyclohexa-2,5-dien-1-one, and toluene) to virtually explore potential binding sites and map hydrogen bond acceptors, donors, and apolar interactions. The idea was to predict binding sites and the key interactions likely to drive affinity. The new paper focuses not just on affinity, but on selectivity.

 

The approach starts by taking multiple structures of the same protein bound to various ligands, especially fragments. Ligands and water molecules are then removed, and hotspot mapping is conducted for each structure. Then, all the hotspot maps are combined to generate an “ensemble” hotspot map, which in theory should give a more complete picture of potential attractive and repulsive interactions than a single structure.

 

To assess selectivity, the ensemble hotspot map of one protein is “subtracted” from that of another. If the proteins are very closely related, this “selectivity map” might be empty: all the interactions for one protein would be present in the other. But if there are differences, they become very apparent.

 

Several retrospective case studies are provided. In the first, ensemble hotspot maps were generated from the closely related bromodomains BRD1 and BRPF1, using 23 and 26 fragment-bound structures, respectively. The selectivity map clearly shows the potential for a hydrogen bond donor on a ligand to bind to the backbone amide of a serine in BRD1; the corresponding residue in BRPF1 is a proline, incapable of making this interaction. And indeed, an examination of the literature revealed that this interaction had previously been used to generate inhibitors of BRD1 that were 15-fold selective over BRPF1.

 

The kinases p38α and ERK2 are also closely related, but selectivity maps generated from five p38α structures and 17 ERK2 structures revealed a hydrophobic pocket in the former but not in the latter. This pocket had previously been used to generate selective inhibitors of p38α. Similarly, 28 structures of CK2α and 32 structures of PIM1 were used to generate a selectivity map that also revealed a hydrophobic pocket that can form in the former protein and had been used to generate selective inhibitors.

 

Generally, the more structures available, the more informative the selectivity maps are. The researchers note that though they only used five p38α structures, the fragments were chosen to be diverse (and interestingly all of them made interactions in the hydrophobic pocket). Also, while some protein flexibility can improve the maps, too much is a problem. (For the kinases, only DFG-in structures were used, for example.)

 

This method is a nice synthesis of experimental and computational techniques. A skeptic might argue that it doesn’t provide fundamentally new information: in the examples provided, the selectivity features had already been found and exploited by medicinal chemists. But the automated process and the clear output may speed things up, especially for newer targets, and indeed the researchers note that it is being applied in-house at Exscientia.

 

Perhaps most importantly, if you’d like to try it yourself, the code is freely available here. Happy mapping!

Reverse micelle encapsulation for measuring low affinities

NMR is among the more sensitive fragment-finding techniques: the starting point for clinical compound ASTX660 had low millimolar affinity at best. Now, three papers by A. Joshua Wand and colleagues at University of Pennsylvania have taken sensitivity to a new level, enabling the detection of fragments that bind hundreds of times less tightly. (Derek Lowe recently wrote about one of them, and I highlighted a talk last year.)

All three papers focus on a method called reverse micelle encapsulation, in which an aqueous solution of protein and ligand is encapsulated in nanoscale reverse micelles measuring less than 100 Å in diameter. At this size, each micelle will contain at most just a single protein and a few thousand water molecules. Because of the small volume, the protein concentration – and that of any fragments – will be extraordinarily high. The micelles have polar groups pointed inwards towards their watery interior, and their hydrophobic tails point out towards solvent, typically pentane. The overall water content of the sample is typically around 2%.

Various NMR techniques can be used to study the proteins. Although the reverse micelles are larger than the proteins themselves and thus would be expected to tumble more slowly, the low viscosity of the pentane solvent makes up for this, providing high-quality spectra.

The primary paper, in ACS Chem. Biol., focuses specifically on fragments. To establish that the technique can detect weak interactions, the researchers show that they can measure the 26 mM dissociation constant of adenosine monophosphate to the enzyme dihydrofolate reductase.

Next, they turned to the protein interleukin-1β (IL-1β), an inflammatory target with no reported small-molecule binders. One challenge of the method is that hydrophobic fragments could partition into the micelles or even diffuse into the pentane, thus reducing their concentration. To avoid this, the researchers assembled a library of 233 very polar, water-soluble fragments with cLogP values < 0.5. A 2-dimensional NMR screen (¹⁵N-TROSY) using standard conditions (100 µM protein and 800 µM fragment) yielded no hits.

In contrast, NMR screening using reverse micelles with the protein at an effective concentration of 5 mM and fragments at 40 mM yielded 31 hits. Chemical shift perturbations (CSPs) were used to determine where they were binding. Ten of the fragments didn’t show clear binding to specific sites on the protein, but the remaining 21 did, with all but one binding to multiple sites. Of these, 13 also showed non-specific interactions with other regions of the protein. Altogether, the fragment binding sites covered 67% of the protein surface, with the receptor-binding interface particularly well-represented.

Concentration-dependent CSPs were used to determine dissociation constants, which ranged from 50 mM to over 1 M. An SAR-by-catalog exercise was able to improve the affinity of one fragment from 200 mM to 50 mM at one site, though it also binds three other sites with slightly weaker affinity.

The second paper, also in ACS Chem. Biol., uses IL-1β but focuses on the interaction of even smaller molecules such as pyrimidine, methylammonium, acetonitrile, ethanol, N-methylacetamide, and imidazole. Not surprisingly, the dissociation constants are even weaker, averaging 1.5 – 2.5 M.

Finally, a Methods in Enzymology paper goes into depth on how to actually run the experiments, including details on choosing detergents and making the micelles. At high fragment concentrations, for example, pH needs to be carefully controlled.

Five years ago we asked “how weak is too weak” for a fragment. In terms of practicality, I’d say that these fragments qualify. Indeed, the ligand efficiency for the best fragment mentioned above is just 0.15 kcal mol^-1 atom^-1.

But the findings do raise the almost philosophical question of what exactly constitutes a small molecule binding site. Astex researchers reported several years ago that most proteins have more than one, and their more recent work with MiniFrags suggest on average 10 sites at high enough concentrations. Similar results were also reported earlier this month from Monash. Whether or not the fragments from such screens turn out to be immediately useful, they could certainly advance our understanding of molecular recognition.

Crystallographic vs computational fragment screening

Several recent Practical Fragments posts have touched on crystallographic screening: from ultra-high concentration screening of “MiniFrags,” to an extensive analysis of fragment structures in the protein data bank, to an open-source effort to develop new antibiotics. A new paper in Phil. Trans. R. Soc. A by Tom Blundell and collaborators at University of Cambridge, the Diamond Light Source, University of Oxford, and several other institutes provides a useful synthesis and an interesting comparison with computational approaches.

The researchers were interested in the bacterial protein PurC, also known as SAICAR synthetase, which is essential for purine biosynthesis and is sufficiently different from its human orthologue to be an attractive antimicrobial target. The protein has an extended binding site that can accommodate ATP as well as its substrate CAIR and an aspartic acid. Using a traditional screening cascade, 960 fragments were screened at 5 mM in a thermal shift assay, resulting in 43 hits. Each hit was then soaked at 10 mM into crystals of PurC, resulting in 8 bound structures, all of which occupy the ATP-binding pocket. Isothermal titration calorimetry revealed dissociation constants as good as 178 µM, with a ligand efficiency of 0.39 kcal/mol/atom.

Next, the researchers ran a computational screen, Fragment Hotspot Maps. This confirmed the main fragment-binding site. Indeed, the crystallographically-identified fragments even make the hydrogen-bonding interactions predicted by the model. However, the computational approach also identified three other hot spots, two in the active cleft and one on the rear of the protein. There was also a “warm spot” next to the ATP-binding site. Are these real, or computational artifacts?

To address this question, the researchers screened fragments at a much higher concentration at XChem, and processed the data using the PanDDA software we’ve previously described. They screened two libraries of fragments at 30-50 mM: 125 “shapely” fragments and 768 “poised” fragments designed for rapid follow-up chemistry. The 8 hits from the first crystallographic fragment screen were also included. This exercise yielded structures for 35 fragments, 60% of which bound in the ATP-binding site, including all 8 of the previously identified ones. Most of the other fragments bound in shallow pockets or near crystallographic interfaces; only one of the other hot spots predicted computationally had a bound fragment, and that was present at low occupancy. Some hits made new interactions around the ATP-binding site, but none bound in the predicted warm spot. Unfortunately, the proportions of fragment hits coming from the two libraries are not broken out.

So in summary, both computational and crystallographic screening correctly identified the “hottest” hot spot, but each approach also identified additional sites that were not confirmed by the other. The researchers ask, “are these sites truly hot spots… or are they weak binding sites routinely seen in crystals?”

This is indeed the key question, and it would be interesting to see whether other computational approaches – such as FTMap or SWISH – are able to shed light on the matter.

Tiny fragments at high concentrations give massive hit rates

Screening fragments crystallographically is becoming more common, especially as the process becomes increasingly automated. Not only does crystallography reveal detailed molecular contacts, it is unmatched in sensitivity. At the FBLD 2018 meeting last year we highlighted work out of Astex taking this approach to extremes, screening very small fragments at very high concentrations. Harren Jhoti and colleagues have now published details (open access) in Drug Discovery Today.

The researchers assembled a library of 81 diminutive fragments, or “MiniFrags”, each with just 5 to 7 non-hydrogen atoms. Indeed, the fragments adhere more closely to the “rule of 1” than the “rule of 3.” Because the fragments are so small, they are likely to have especially low affinities: a 5 atom fragment with an impressive ligand efficiency of 0.5 kcal mol^-1 per heavy atom would have a risibly weak dissociation constant of 14 mM. In order to detect such weak binders, the researchers screen at 1 M fragment concentrations, almost twice the molarity of sugar in soda! Achieving these concentrations is done by dissolving fragments directly in the crystallographic soaking solution and adjusting the pH when necessary. Although this might mean preparing custom fragment stocks for each protein, it avoids organic solvents such as DMSO, which can both damage crystals and compete for ligand binding sites.

As proof of concept, the researchers chose five internal targets they had previously screened crystallographically under more conventional conditions (50-100 mM of larger fragments). All targets diffracted to high resolution, at least 2 Å, and represented a range of protein classes from kinases to protein-protein interactions. The hit rates were enormous, from just under 40% to 60%, compared to an average of 12% using standard conditions.

Astex has previously described how crystallography often identifies secondary binding sites away from the active site, and this turned out to be the case with MiniFrags: an average of 10 ligand binding sites per protein. In some cases protein conformational changes occurred, which is surprising given the small size and (presumably) weak affinities of the MiniFrags.

All this is fascinating from a molecular recognition standpoint, but the question is whether it is useful for drug discovery. The researchers go into some detail around the kinase ERK2, which we previously wrote about here. MiniFrags identified 11 ligand-binding sites, several of which consist of subsites within the active site. Some of the MiniFrags show features previously seen in larger molecules, such as an aromatic ring or a positively charged group, but the MiniFrags also identified new pockets where ligands had not previously been observed. The researchers argue that these “warm spots” could be targeted during lead optimization.

One laudable feature of the paper is that the chemical structures of all library members are provided in the supplementary material. Although it would be easy to recreate by purchasing compounds individually, hopefully one or more library vendors will start selling the set. If MiniFrag screening is standardized across multiple labs, the resulting experimental data could provide useful inputs for further improving computational approaches, as well as providing more information for lead discovery.

Universal fragments for discovering hot spots and aiding crystallography

A couple years ago we highlighted a paper from Eddy Arnold’s group at Rutgers University in which crystallographic fragment screening revealed over a dozen secondary ligand binding sites on HIV-1 reverse transcriptase (RT). Shockingly, the fragment 4-bromopyrazole bound to every single site, which led us to ask “is this a privileged fragment or a promiscuous binder? And as for the sites with no known functional activity, are these useful?” The Arnold group asked themselves these same questions, and provide answers in a new paper in the open-access journal IUCrJ.

The researchers first considered whether 4-bromopyrazole is special. They collected about 20 halogenated aromatic fragments and soaked these into crystals of HIV-1 RT at concentrations ranging from 20 to 500 mM. Of these, 4-iodopyrazole also bound at multiple sites, but most of the others – even closely related molecules such as 3-bromopyrrole or 4-bromothiazole – did not bind to any.

Next, the authors extended these observations to other proteins. When they soaked their molecules into crystals of the endonuclease from the 2009 pandemic influenza strain, they found that 4-bromopyrazole bound to four sites, including two of three identified in a previous crystallographic fragment screen. In one case, a phenylalanine side chain shifted to open up a new hydrophobic binding site. A similar and previously unobserved shift occurred with a tyrosine side chain when 4-bromopyrazole was soaked into the protein proteinase K. Thus, this fragment is able to identify otherwise cryptic binding sites.

Interestingly, the 4-bromopyrazole binding sites could be strikingly dissimilar, ranging from hydrophobic to mildly electropositive to strongly electronegative. The researchers note that the halogen can form either hydrophobic or polar interactions. Also, one pyrazole nitrogen can act as a hydrogen-bond acceptor while the other can independently act as a donor, and these interactions can be with the protein directly or through bridging water molecules.

Last week we highlighted work from Astex suggesting that secondary binding sites in proteins are common, but in most of those cases the proteins had only one additional site, and only a couple had five or six. In contrast, 4-iodopyrazole bound to 21 sites in HIV-1 RT, although only five of these had sufficiently good electron density to allow the entire fragment to be built. (That is, crystallography only clearly revealed the location of the iodine atom in the others.) How many of these sites are bona fide hot spots, and how many could be predicted using computational techniques such as FTMap?

This is all quite interesting, but, as we asked previously, is it useful? The researchers provide two applications.

First, 4-bromopyrazole may be a general probe to assess whether a protein is ligandable. Soaking crystals of the catalytic core domain of HIV-1 integrase in 500 mM of 4-bromopyrazole revealed no binding sites, in sharp contrast to HIV-1 RT, endonuclease, and proteinase K. Integrase also showed a very low hit rate in a general fragment screen, and a plot of binding sites vs fragment-screening hit rate for three proteins showed a linear correlation. Obviously this is a tiny data set, but if it holds up it could be an easy experimental way to assess the difficulty of targets.

Second, the bromine or iodine atoms in the pyrazole fragments could be used in single-wavelength anomalous dispersion phasing, a useful approach for solving crystal structures. The researchers demonstrated this experimentally for HIV-1 RT, endonuclease, and proteinase K, and suggest that 4-bromopyrazole and 4-iodopyrazole could be inexpensive and helpful additions to a “crystallographer’s toolkit.”

Thus, unlike other frequent-hitters such as PrATs, 4-halopyrazoles might be promiscuous yet specific – and useful. I look forward to seeing whether these "universal fragments" catch on.

Secondary ligand binding sites are common

Anyone who has been exposed to much crystallography will have seen examples where a ligand binds somewhere besides the active site of a protein. This is probably all the more likely in the case of fragments, both because fragments are soaked at high concentrations (and thus weaker ligands can be detected) and also because, being less complex, fragments will be able to bind to more sites. In some cases, such as FPPS and HCV NS3, ligands that bind at these “secondary sites” could be advanced to potent allosteric inhibitors. But how common are such sites? This is the question addressed by Harren Jhoti and Astex colleagues in a paper just published in Proc. Nat. Acad. Sci USA.

The researchers were privileged to have 5590 crystal structures of 24 proteins with at least one bound ligand from crystallographic fragment screens. Careful analysis to exclude buffers and molecules bound at crystallograpic interfaces left them with 53 sites total, with each protein having a fragment bound in at least 1 site; one had 6 (still far from the record 16 sites in HIV-1 RT discussed here). Importantly, 16 of the targets had at least 2 ligand binding sites, with an overall average of 2.2. This number of secondary sites is likely a lower bound, as some sites may have been blocked by crystal packing.

What can be said about these ligand binding sites? The researchers compared the sequence conservation between orthologous proteins from different organisms and found that primary binding sites are more conserved than the overall protein sequences. This is expected because, since the proteins likely have similar functions, there are more evolutionary constraints on the active site residues surrounding the primary sites. Interestingly though, the secondary sites were also significantly conserved, suggesting that they too may have some sort of function.

Protein mobility was also examined computationally, with the thought being that functional binding sites should be more rigid than the overall surface of the protein so as to minimize entropic costs of ligand binding. This turned out to be the case for all primary ligand binding sites, but it was also true for most of the secondary sites. Surprisingly, and in contrast to previous results, there were no differences in normalized B factors (roughly, temperature-related motions) for residues in either primary or secondary binding sites compared with surface residues in general.

Comparing the physical properties of the primary and secondary sites revealed that both were more lipophilic than the rest of the protein surface. Ligands tended to be slightly more buried in primary binding sites than in secondary sites, but there didn’t seem to be any differences among the ligands themselves, though the twelve shown in the paper are mostly “flat.”

These combined results suggest that the majority of proteins have multiple sites capable of binding to small molecule ligands. The researchers note that most of their examples are enzymes, so it may not be fair to extrapolate to other protein classes. That said, many GPCRs also have multiple ligand binding sites.

Secondary binding sites have several things going for them. First, allosteric sites provide a means to target proteins in which the primary binding site is problematic, perhaps because it is too closely related to other proteins. Allosteric sites can also be useful for targeting viral or cancer targets in which resistance is an issue, as in the case of ABL001. Finally, secondary sites provide an opportunity to develop not just inhibitors, but activators.

Of course, just because a fragment binds at a site doesn’t necessarily mean that the site is ligandable. Indeed, HSP70 appears to have 5 sites, yet by all accounts is an extremely difficult target. Four of the proteins (including HSP70) are described in some detail in the paper, with protein-fragment structures deposited in the protein data bank. It would be interesting to see how the secondary sites score as potential hot spots using software such as FTMap.

Still, knowing that secondary binding sites are the norm rather than the exception gives new impetus to look for them. It also suggests new areas of biology to explore. Molecular complexity is one thing, but it pales in comparison to biological complexity.

Fragments vs Pim-1

The three members of the Pim family of serine/threonine kinases help cancer cells proliferate and survive, and are thus interesting as potential anti-cancer targets. Structurally the kinases are also intriguing because they have a proline residue in the “hinge” region of the purine binding site, differentiating them from the other 500+ kinases in the human genome. Two recent papers describe different fragment-screening approaches against Pim-1: one case rediscovers fragments of a previously reported compound and the other identifies a new series of potent inhibitors unlike other reported kinase inhibitors.

In the first paper, published in Acta Cryst. Section D, researchers from Bayer Schering Pharma AG performed a crystallographic screen of just 36 fragments against Pim-1. Despite the small library size, a dozen of the fragments produced interpretable electron density, although only 4 of these showed activity better than 10 mM. Fragment 3, with an IC50 of 130 micromolar in an enzymatic assay, was the most potent. Interestingly, a close analog of this cinnamic acid fragment was also part of a Pim-2 inhibitor previously identified through HTS by a group from Boehringer Ingelheim. The Bayer folks synthesized this molecule (compound 1), confirmed that it is also potent against Pim-1, and found that, crystallographically, it overlays beautifully with the fragment.

The second paper, published in Bioorg. Med. Chem. Lett., describes how researchers at Genzyme used a fragment-based approach to develop potent Pim-1 inhibitors. The researchers used SPR to screen a library of about 1800 fragments at 75 micromolar concentration and then used a biochemical assay to characterize the active molecules. Benzofuran-2-carboxylic acids such as compound 10 (below) were particularly interesting: not only did they have high ligand efficiencies, they don’t look anything like typical kinase inhibitors. X-ray crystallography revealed that the bromine atom is binding in a hydrophobic pocket, while the acid is making hydrogen bond contacts with the catalytic lysine and other residues. A related fragment with comparable potency had a methoxy substituent in the 7-position, and by transforming this into a positively charged moiety the researchers were able to improve the potency to low nanomolar with a dramatic increase in ligand efficiency.

What’s also interesting is that if you overlay the two fragments from the two papers, they superimpose almost exactly on top of one another, with the carboxylic acid moieties making the same interactions – clearly a hot spot.

On a side note, as most readers are probably aware, sanofi-aventis has recently acquired Genzyme. Practical Fragments wishes all the folks in Massachusetts well during the integration.

Ligand efficiency hot spots

Hot spots are regions of a protein with a particular predilection for binding to small molecules – thermodynamic sinkholes, so to speak. Discovering one of these can get you to potent molecules very quickly. In an effort to better understand hot spots, Iwan de Esch and colleagues at VU University in Amsterdam and collaborators at Beactica have deconstructed a potent ligand for nicotinic acetylcholine binding protein (AChBP), a model protein for ligand-gated ion channels, which are implicated in a variety of neurological diseases. They report their results in a recent issue of J. Med. Chem.

The researchers started with the previously reported quinuclidine compound 6 (see figure) and fragmented this into 20 analogs. They tested these in a surface plasmon resonance (SPR) assay as well as in a more conventional radioligand binding assay; the agreement between these very different assay formats was excellent, further validating the utility of SPR as a useful tool for discovering fragments.

Not surprisingly, some of the fragments have higher ligand efficiencies than the larger, more potent molecule, suggesting that there is a hot-spot that recognizes the core fragment 25 (which is structurally related to nicotine). This concept of “group efficiency” has been described previously, and can be useful for optimizing fragments. For example, compound 22, without a basic nitrogen atom, has the lowest ligand efficiency in the bunch; presumably, simply adding the nitrogen would give a sizable boost in potency.

However, one needs to be cautious. The researchers use computer docking to develop models of how each of these fragments bind, but as we have seen before, isolated fragments do not always recapitulate the binding modes of fully elaborated molecules. Still, particularly in the absence of structure (as is the case with many ion channels), exercises such as this could provide useful ideas for what to do with fragment hits.

Hot spots for fragments

Although most people try to advance fragments to more potent molecules, some have taken the reverse approach: starting with potent binders and deconstructing them into fragments (see for example here, here, and here). A recent, thorough example in J. Med. Chem. shows how isolated fragments do not necessarily bind in the same manner as they do in fully elaborated molecules.

In this paper, Isabelle Krimm and colleagues at the Université de Lyon in France applied “fragment-based deconstruction” to inhibitors of the anti-cancer target Bcl-xL. This protein is one of the great success stories in fragment-based drug discovery: ABT-263, which is in multiple clinical trials, was discovered by researchers at Abbott using SAR-by-NMR. In that work, fragments were identified binding near each other on the protein (site 1 and site 2) and subsequently linked together. Very extensive medicinal chemistry eventually led to the picomolar inhibitor now in clinical testing.

Krimm and colleagues dissected 9 inhibitors of Bcl-2, including ABT-263, into 22 different fragments and studied their binding by NMR. They first used ligand-observed NMR (WaterLOGSY and saturation transfer difference, or STD) and found that 19 fragments interacted with the protein. When they then turned to protein-observed NMR (proton-15N heteronuclear single quantum correlation, or HSQC), only 13 fragments caused changes to the spectra of Bcl-xL, suggesting that the other six bound too weakly to detect. In fact, the most potent fragment has an affinity of just 2.7 mM, so it is not surprising that some of the fragments were undetectable.

The nice thing about protein-observed NMR is that it can provide insight into where on the protein the fragments bind, and in this case the researchers found that 12 of the 13 fragments that caused NMR shifts in the protein bind to site 1, despite the fact that structures and modeling suggest that some of these fragments should be binding in other sub-sites. (The 13th fragment appears to bind to multiple sites on the protein surface.) In other words, the binding modes of the isolated fragments are not the same as the binding modes of the fragments when assembled.

The authors conclude that fragments “will interact with their preferred binding site, which can be different from the site they occupy when they are included in the larger molecule.”

Interestingly, one of the fragments studied by Krimm (2,3-dihydroxynapthalene) was also tested at Abbott, but found to bind in site 2. The reason? In the Abbott study, this fragment (and a number of others) were tested in the presence of a fragment that binds to site 1. It seems that site 1 is a thermodynamic sink, or hot spot. Unless this site is filled, other fragments will bind there, even if they could also bind elsewhere on the protein. The implication is that, if you want to find fragments that bind to a new site on your protein, it may be worth screening in the presence of a fragment known to bind to an existing site.