Linking lessons on EGFR

From the earliest days of FBLD, fragment linking has been an aspirational goal. Snapping together two low affinity binders to get an additive or even superadditive boost in affinity is conceptually appealing but, as our poll showed, not always easy. In a new open-access J. Med. Chem. paper, Hung Nguyen, David Heppner, and collaborators at The State University of New York at Buffalo and elsewhere propose a general workflow.

 

Epidermal growth factor receptor (EGFR) is a prominent cancer target with multiple approved drugs. It has been studied for decades and thus makes an excellent model system; we wrote about some of Heppner’s work on this protein a couple years ago. As with most kinases, many known ligands bind in the so-called hinge-binding region where the adenine of ATP normally sits. But some ligands have been identified binding to an adjacent pocket, and in a 2024 open-access Comm. Chem. paper the Heppner group reported linking ligands at these two sites.

 

The new paper systematically expands on this work, diving deep into the SAR of compounds binding to the oncogenic L858R/T279M EGFR mutant. Specifically, the researchers vary connecting points as well as the length and composition of the linker.

 

 

As expected, the point of connection can make the difference between improved activity and dead compounds. Linker length matters too, as shown for compounds 4, 5, and 6, where the “Goldilocks” two-carbon linker (compound 5) binds orders of magnitude more tightly than the one- or three-carbon linker. Remarkably, compounds 4 and 6 are less active than one of the initial fragments! Linker composition is also important, but perhaps less than one might expect: the amide-containing linker in compound 8 is only marginally more potent than compounds 9 and 10.

 

The researchers characterize several of their linked molecules both crystallographically and using molecular dynamics. They argue that “maximal flexibility is likely the most important design criterion for linking molecules that occupy adjacent pockets.” That is consistent with this post from 2012 but doesn’t quite fit with the fact that the amide-containing compound 8 is better than either of the (presumably more flexible) compounds 9 and 10.

 

The paper is actually more of a fragment-merging than a fragment linking story: the central phenyl ring is common to both starting compounds, so it is not really possible to assess the extent of additivity. Also, activity is based on inhibition rather than thermodynamic measurements, so it is impossible to say whether the improved affinities of the linked molecules are due to entropic, enthalpic, or some combination of factors.

 

Still, this is a nice contribution to the literature of fragment merging/linking, and the general guidelines are sound: assess multiple connection points and linker lengths before refining the linker itself.

Automating high-throughput NMR

NMR was the first practical method used to identify fragments, described thirty years ago in this famous SAR by NMR paper. That headstart made NMR the go-to method for finding fragments, and as recently as ten years ago it was still in the lead, according to our poll in 2016. But as other methods have become faster and more effective, the relative popularity of NMR has declined, to fourth place in our 2024 poll. Now, a new paper in J. Med. Chem. by Bradley Falk and colleagues at Bristol Myers Squibb reports improvements that might help it move up the ranks.

 

One of the challenges of NMR is that it requires so many steps. Protein-detected NMR requires isotopically labeled protein, but even ligand-detected NMR involves measuring spectra for small molecules in the absence and presence of protein. The specific NMR settings may need to be tweaked for any given molecule, and the spectra are then analyzed, usually manually, to decide whether the molecules are likely binders.

 

The new paper focuses on automating virtually everything in ligand-detected NMR, from sample preparation through NMR runs to analysis. It is meant to be more general than a fragment screen with a curated library where each member is already well understood. First, novel compounds are examined in a band-Selective Optimized Flip Angle Short Transient (SOFAST) NMR experiment, which is particularly sensitive (down to 20 µM ligand). Testing in both DMSO-d₆ and assay buffer provides solubility and quality control data. This is an essential step: as we noted some time ago more than a third of compounds in some vendors’ catalogs fail these quality checks. In addition, a CPMPG SOFAST experiment in assay buffer reveals whether the compounds aggregate, a hazard we wrote about here.

 

Next, an automated workflow funnels appropriate compounds to the next step, in which compounds are assessed for protein binding using CPMG or STD NMR. Four separate samples are tested: low or high concentrations of compound in the absence or presence of protein. Again, automated approaches are used to choose the specific NMR parameters and analyze the results. As the researchers note, “human input is reserved for evaluations that require integrating multiple lines of evidence.” The output places compounds into three bins, “advance to med chem,” “advance with orthogonal confirmation,” or the garbage bin.

 

The researchers describe an example using the model protein human carbonic anhydrase II (hCAII), which we’ve written about previously. The known ligand benzenesulfonamide was tested and passed into the “advance to med chem” bin, despite the fact that, with a K_D of 0.8 µM, it is more potent than most ligands studied by NMR.

 

One thing missing from the paper is a summary of overall time and throughput. The experimental section states that the CPMG-SOFAST experiment itself takes half an hour, and indeed the researchers note that the platform is “instrument-time limited rather than sample preparation- or data analysis-limited.” In terms of absolute throughput, it probably can’t compete with other biophysical methods such as SPR. The automation and automated decision-making is the real advance. In 2019 I satirically imagined a fully autonomous drug discovery engine. As artificial intelligence becomes ever more intelligent, that post may prove prescient. If so, it will feature automated tools such as those described in this paper.

Covalent fragments vs RAB27A/B: towards a chemical probe

Last week we highlighted an example of adding a warhead to a fragment-derived, non-covalent protein binder to generate a covalent inhibitor. This is probably the most common method for identifying covalent ligands. But as we noted in 2019, the reverse “covalent-first” approach can be just as effective. A recent open-access paper in J. Am. Chem. Soc. by Elena De Vita, Edward Tate, and collaborators at Imperial College London and University of Nottingham provides another success.

 

The researchers were interested in the closely related proteins RAB27A and RAB27B, two of the more than 60 members of the Rab family, which is itself part of the large Ras superfamily. RAB27A/B are potential antitumor and antiviral targets, but the only reported chemical probe, Nexinhib-20, sets off PAINs alerts.

 

The researchers had previously found two non-conserved reactive cysteine residues, C123 and C188, near the co-called effector binding region, a shallow pocket involved in protein-protein interactions essential for RAB27A/B function. They screened a library of 960 acrylamide-containing fragments, each at 500 µM, in an assay that detects free thiols. Versions of RAB27A in which either C123 or C188 was mutated to a serine were used so the researchers could assess whether compounds were selective for one cysteine. They also screened the compounds against glutathione, the abundant cellular thiol present in all mammalian cells, and focused on those compounds that were at least five-fold more reactive against the protein than against glutathione. The best compound was IMP-1704, which modified C123 much more efficiently than C188.

 

A crystal structure of IMP-1704 bound to RAB27A confirmed that the compound binds in the “WF pocket” of the effector region, with the phenyl ring displacing a tyrosine residue and a sulfonamide oxygen atom making a hydrogen bond to a lysine residue. The researchers also obtained a crystal structure of IMP-1704 bound non-covalently to a C123A mutant of RAB27A and found that it binds in a similar manner. In other words, IMP-1704 does not require the acrylamide warhead to bind the protein, though with no reported affinity it might be quite weak.

 

The crystal structures suggested a fragment growing strategy from the pyrrolidine ring, leading to IMP-1712, which modified the protein twice as rapidly. Interestingly, the enantiomer of this compound was inactive. Further SAR and fragment-growing ultimately led to the most potent compound in the paper, IMP-2552. As with IMP-1712 the enantiomer was inactive, providing a useful negative control

 

Although the covalent efficiency of IMP-2552 is relatively modest, the researchers used an analog with a clickable alkyne probe to assess what proteins it binds to in cells. RAB27A was one of the top protein hits, and engagement with RAB27A could be competed with the non-tagged IMP-2552. Coimmunoprecipitation experiments demonstrated that IMP-2552 could block interaction with an effector protein, while the enantiomer was inactive. IMP-2552 was also active at 10 µM in a phenotypic assay, while the enantiomer was not.

 

Importantly, the researchers also profiled Nexinhib-20, the previously reported inhibitor. In contrast to IMP-2552, Nexinhib-20 reacted with both C123 and C188 and also reacted very rapidly with glutathione, suggesting generic rather than specific reactivity. Moreover, in the phenotypic assay at 10 µM, “Nexinhib-20 showed profound cytotoxicity and cell death.” Hopefully these findings will make people reconsider before using Nexinhib-20 in their assays, or when reading conclusions drawn from using this compound.

 

Overall this paper is a nice example of screening for covalent ligands and optimizing them. It is also another application of covalent fragment screening applied to a GPTase; we’ve written previously about KRAS. In contrast to KRAS and several other covalent programs, the initial hit here is rule-of-three compliant.

 

And although improvements in potency and selectivity are still needed to formally designate IMP-2552 a chemical probe, the molecule appears to be the best we currently have, particularly when used with its negative control.

From fragment linking to covalent RAD52 inhibitor

Targeting proteins involved in DNA repair is a promising approach against cancer. However, finding effective small molecule inhibitors can be difficult because the protein-DNA interface tends to be very polar, with few classically ligandable pockets. In a new J. Am. Chem. Soc. paper, Nicolas Bocquet, Olivier Querolle, Ulrich Lücking, and collaborators at FoRx Therapeutics, WuXi AppTec, Diamond Light Source, University of Bern, Sandexis, and Weizmann Institute of Science provide a proof of concept.

 

The researchers were interested in RAD52, which mediates several DNA repair functions. The N-terminus of the protein forms an oligomeric ring structure, and each subunit contains a solvent-exposed groove that binds to single-stranded DNA. Since DNA is negatively charged, the researchers conducted a crystallographic screen against a library of 320 carboxylic acid containing fragments. This identified three related molecules, of which compound 3 had the highest occupancy. Because the compounds bind at the DNA-binding interface, the researchers developed a fluorescence polarization (FP) DNA-displacement assay and found that the fragments disrupt DNA binding at high micromolar concentrations.

 

Interestingly, the crystal structure revealed two copies of compound 3 bound to each subunit of RAD52: each positioned where a single nucleotide would normally bind, with the carboxylic acids making similar contacts as the phosphates. In other words, the fragments mimic two consecutive bases of a single-stranded oligonucleotide. Naturally, this suggested a linking approach.

Linking two close analogs of compound 3 led to compound 5, with a satisfying two orders of magnitude improvement in activity in the FP assay, which the researchers suggest is largely due to lowering the entropic penalty. Linking directly off the thiazole moieties produced similar results, and a macrocycle had slightly better activity, though not so much better as to justify the synthetic hassle. A crystal structure of compound 5 revealed a small hydrophobic cleft, and this was exploited with a phenyl moiety in compound 11, a high nanomolar inhibitor in the FP assay.

 

A crystal structure of compound 11 bound to RAD52 confirmed that the compound binds as designed. The structure also revealed a cysteine residue (C64) in close proximity to the phenyl moiety. Adding an acrylamide warhead yielded a covalent inhibitor, and further optimization ultimately led to FORX-864, which has low nanomolar activity in the FP assay after 4 hours. The researchers appropriately measured the k_inact/K_I, which came in at a respectable 230,000 M^-1s^-1, better than some covalent clinical compounds. Removing the warhead led to a dramatic loss in potency (IC₅₀ = 27 µM in the FP assay).

 

In a cellular assay to measure single-stranded annealing (SSA), FORX-864 showed roughly 1 micromolar activity, while a compound lacking the warhead was inactive. FORX-864 was also inactive in a cell line in which RAD52 had been knocked out, suggesting on-target activity. The researchers profiled the reactivity of FORX-864 in an activity-based proteomic profiling (ABPP) assay and found that it was reasonably selective. Intravenous dosing in mice showed encouraging pharmacokinetic properties, though with two carboxylic acid moieties oral dosing might be challenging.

 

Overall this is a nice example of liganding a difficult target. Given that FORX-864 mimics a dinucleotide, it may bind to other DNA- or RNA-binding proteins. However, covalent binding is required for high potency, and it is unlikely these other proteins will have cysteine residues in the appropriate position, which should help with selectivity. FORX-864 certainly appears to be the best chemical probe for RAD52 to date.

Focal molography for assessing protein-ligand interactions

The list of technologies used to find and validate fragments is long, with more than a dozen in our 2024 poll. But it’s always fun to learn about new approaches. At the Novalix conference earlier this year I learned of “focal molography,” and this is the topic of a recent paper in J. Med. Chem. by Nicholas Favalli, Andreas Frutiger, Sebastian Oehler, and colleagues at Philochem and lino Biotech.

 

The approach is described in detail in this Wikipedia article. Briefly, it is a biophysical technique based on changes in diffraction. A sensor chip contains a matrix of “molograms,” each of which is a lens constructed of micropatterned ridges and grooves. The ridges of each mologram are functionalized with unique single-stranded oligonucleotides. These capture complementary strands, each attached to a molecule to be screened. The grooves are coated with oligonucleotides without small molecules to act as a control in the subsequent readout. Proteins are then flowed over the chip. Protein binding to the small molecules of a specific mologram changes the topography of the ridge, and so the diffraction. This change can be measured in units of “coherent mass density” (pg/mm²). Because only specific binding to the ridges is detected (the grooves provide background controls for each mologram), the technique is suitable for even crude samples such as cell lysates.

 

An obvious application for focal molography is to confirm the binding of hits identified from DNA-encoded libraries (DEL), since the small molecules already have an oligonucleotide tag. This is the focus of the new paper.

 

The researchers started by assessing three known binders to the model protein carbonic anhydrase IX (CAIX). Surface plasmon resonance (SPR), the most frequently-used chip-based biophysical method, was used as a benchmark in two formats, one in which the small molecule was bound to the SPR chip, which is most comparable to focal molography, and one in which the protein was bound to the SPR chip.

 

SPR and focal molography agreed in rank order for the three compounds, but focal molography sometimes suggested tighter binding. For the highest-affinity compound, SPR gave a K_d of 28 nM for compound bound to the chip and 450-750 nM for protein bound to the chip, while focal molography gave a K_d of 0.82 nM. Agreement was better for the weaker binders, one of which was mid-nanomolar and the other low micromolar. The kinetic constants k_on and k_off also differed between focal molography and SPR by as much as two orders of magnitude.

 

Next, the researchers examined hits from a DEL screen against CAIX, some of which were selective against the related protein CAII. A total of two dozen molecules were tested by both focal molography and fluorescence polarization (FP) as an orthogonal technique. Again, the rank orders of affinities for the compounds to CAIX and CAII agreed while the absolute values varied, with focal molography usually suggesting ten-fold tighter binding. (Though the FP assay often bottomed out above the K_d values measured by focal molography.)

 

Finally, the researchers examined the binding of a known high-affinity binder to the oncology target fibroblast activation protein (FAP). In this case the measured affinities were close to the reported value.

 

This is an interesting approach, and according to the researchers “focal molography can support up to ~500 duplicate kinetic measurements per day on a single regenerable chip.” The observation that affinity measurements seem to be tighter than other methods does raise concerns. I wonder if the density of the molecules on the molograms leads to rebinding, as is sometimes seen when SPR chips are too heavily functionalized. Alternatively, since CAIX is a homodimer, perhaps avidity effects played a role. Also, the researchers suggest that the technique is best used for compounds with affinities better than 10 µM, which reduces utility for many fragment programs.

 

If you’ve used focal molography, how has it worked for you?

Twentieth-Fourth Annual Discovery on Target Meeting

Cambridge Healthtech Institute held its annual Discovery on Target (DoT) meeting in Boston last week. Attendance may have set a record, with over 1000 participants. There were sessions on antibodies, conjugates, and radioligands, but I’ll just touch on some of the small molecule themes most relevant to this blog.

 

Covalent approaches

As with last year, covalent approaches to drug discovery were well-represented. Dhirendra Simanshu (Frederick National Laboratory) described several drugs birthed in the NCI RAS Initiative, including ones targeting the G12C mutant of KRAS and C242 on the RAS-binding domain of PI3Kα. He also described how various KRAS mutants cause resistance to drugs, finding that most of them affect the K_I term (binding) more than the k_inact term (bonding). Continuing on the KRAS theme, Jack Sloan discussed efforts at BMS to discover inhibitors targeting both the on- and off-states of KRAS G12C.

 

Kasia Handing (Tango Therapeutics) presented both covalent and non-covalent approaches against VRK1, a synthetic lethal target for glioblastoma. A biochemical screen yielded covalent inhibitors that were characterized crystallographically, but they turned out to be too non-selective to progress. In the meantime, an in-cell DNA-encoded library (DEL) screen led to more advanceable molecules.

 

DEL was indeed a major topic of the conference, and here too covalent approaches are being explored. Elizabeth D’ambrosio described how GSK is identifying covalent binders from DEL screens, a process which includes testing for time-dependence as well as aggressive washing steps under denaturing conditions to quench unreacted molecules and wash away any sticky, non-covalently bound compounds. And Xiaojie Bruce Lu (Chinese Academy of Sciences) described applying DEL to a variety of targets, including some covalently targeting non-cysteine residues.

 

One of the advantages of DEL is being able to screen large libraries, and increasing throughput is also of interest to Harvard’s Steve Gygi, who is screening covalent fragments against cells and cell lysates. In 2024 we described how he was using tandem mass tag (TMT) multiplexing to analyze 18 samples in a single mass-spectrometry run, and he has now extended this to 35 samples. What’s more, Steve is combining 16 different fragments in each sample and introducing clever deconvolution approaches to further accelerate the process.

 

Not everyone is targeting cysteine residues. Mark Tye described the discovery of Revolution Medicines’ zoldonrasib, which uses an aziridine warhead to target the G12D mutant of KRAS. And Balyn Zaro (University of California San Francisco) described using ninhydrin derivatives to target arginine. Most organic chemists are familiar with ninhydrin’s reaction with primary amines such as the N-termini of peptides, but Balyn has been able to optimize selectivity for arginine and has used the resulting chemical probes to identify more than 7000 arginine residues in cells, about 5% of which are hyperreactive. Interestingly, some cell lines seem to have a greater proportion of reactive arginine residues than others.

 

Direct-to-Biology (D2B) approaches

Another recurring theme was direct-to-biology, in which crude reaction mixtures are screened directly in various assays without purification. Indeed, Nicholas Bland of Domainex said that after looking at the agenda of the meeting he considered skipping the introductory slides – and this was on the first official day of the conference. Nicholas described using D2B for optimizing PROTACs as well as for improving fragment hits against the GPCR A_2A receptor.

 

Jack Sadowsky, who also presented at DDC this past spring, described how he and his colleagues at Kimia are using D2B in combination with machine learning to develop selective inhibitors of kinases while also improving their in vitro ADME properties. This effort entailed making and testing tens of thousands of compounds from a virtual space of some 80 million molecules.

 

But even small-scale D2B can be valuable. Trevor Sherwood described how he and his colleagues at BMS used D2B to make and test 663 molecules to improve the potency of an EP4 agonist identified from a high-throughput screen. This involved multiple reaction types, including photoredox chemistry, for which they built a custom-designed reactor to evenly heat and illuminate the microtiter plates. The D2B work was done in parallel with medicinal chemistry, but interestingly the development candidate BMS-986526 came directly from the D2B work and has entered a phase 1/2a trial for rheumatoid arthritis.

 

There were other talks and posters on D2B that I missed, but I did attend two breakfast roundtables on the topic that were quite lively despite happening at 7:30 – a testament to the interest in the topic.

 

“Traditional” fragment-based drug discovery

At a plenary panel moderated by Drug Hunter’s Dennis Hu on drugging difficult targets, Biogen’s Klaus Urbahns asked the other panelists if they were still doing fragment-based lead discovery. John Tallarico of Novartis and Ian Storer of AstraZeneca immediately said yes. And in a separate presentation, Kris Borzilleri described how she and her Pfizer colleagues used fluorine-detected NMR and fragment optimization to identify ligands for an orphan GPCR.

 

Debanu Das described using high-throughput crystallography screening at XPose Therapeutics, now part of Mid-Atlantic BioTherapeutics, to identify inhibitors of DNA damage response pathway proteins. Although the initial hits had no detectable activity, SAR by catalog was able to get to nanomolar inhibitors of the protein APE1, which is overexpressed in ovarian cancer, in a matter of weeks.

 

But perhaps the most robust validation of FBDD was provided by Steve Fesik of Vanderbilt, who described a decades-long effort to drug the MYC pathway, which is probably dysregulated in almost all cancers. We described some of his published work on WDR5 here, but he is now making progress against MYC itself. This has not been easy: MYC is intrinsically disordered, and SAR-by-NMR screens against both the transactivation domain and the DNA-binding domain failed. However, MYC forms a structured leucine-zipper heterodimer with the protein MAX, and screening this complex yielded hits at the DNA binding interface. (Earlier this year we described work from Novartis on a different leucine zipper.) The hits have been optimized to low nanomolar binders that disrupt DNA binding, unlike a slide full of other (alleged) MYC inhibitors, including high-profile commercially available compounds such as MYCi361 and MYCi975.

 

I’ll close – hopefully before you stop reading. If you attended, what talks or posters stood out? And if you wish you had attended, mark your calendar for Oct. 18-21 next year, when DoT returns to Boston in time for peak leaf-peeping season.

Expanding the Scope of chemical space at Novartis

Without a good library, it is difficult to find a good hit, so library design is one of the most import aspects of a screening campaign. The size of a library also impacts how it can be screened. A virtual or DNA-encoded library might be billions (or even trillions) of compounds, but for most physical screens, such huge numbers would be unfeasible. Generally, fragment screens sample more chemical space but require more resources, while screens of larger molecules can be higher throughput. In a new (open-access) J. Chem. Inf. Model. paper, Johanna Jansen, Charles Wartchow, and colleagues at Novartis describe a middle ground, the “Scope Concept.”

 

The researchers, many of whom have fragment and biophysics experience, were interested in molecules that might be larger than the rule-of-three guidelines for fragments, but still small enough to sample reasonable swaths of chemical space. To this end they developed the Scope library. To be included, compounds need to have molecular weights between 200-400, measured or calculated solubilities of at least 100 µM, at least two moieties capable of forming hydrogen bonds, and a “Scope ring system,” meaning two rings connected by a 0-3 atom linker. Additionally, undesirable substructures were removed, and in-house compounds had to pass quality control. Originally compounds were required to have log P < 5, but the solubility limit weeded these out anyway. Some warhead-containing molecules were included, though the fraction is not disclosed.

 

The Novartis compound collection was queried using these criteria to choose a 10,000-compound core Scope set as well as a 40,000-compound extension set designed for higher throughput screens. While these numbers are not large in the context of high-throughput screening libraries, they are larger than most fragment libraries according to our polls.

 

The Scope libraries proved to be popular at Novartis: 18 screens were performed in the first year across three sites. Half of these were biochemical screens. Of the rest, six used SPR, two used DSF, and one used native mass spectrometry (nMS). Most of the biochemical screens used the full 50,000 library, while the biophysical screens used the 10,000 core set and sometimes expanded to the full set. For four of five targets screened using a biochemical assay, primary hit rates were similar or slightly higher for the Scope library than a diversity library, though the Scope library was screened at a 50-100 µM concentration as opposed to 10-50 µM.

 

Three targets are discussed in some detail. The first, trypanosomal CLK1 kinase, had a freakishly high primary hit rate of 21% when screened at 100 µM, compared to 2.5% for a diversity set screened at 12.5 µM. Another kinase had an 11% hit rate, while a third had a more reasonable 1.2% hit rate when the ATP-binding site was blocked, suggesting binding to this site. The researchers now screen kinases against lower concentrations of the Scope compounds due to the high ligandability of kinases in general.

 

The chemical structures of half a dozen hits are shown, with IC₅₀ values ranging from 30 nM to 1.1 µM. Interestingly two of the molecules are covalent modifiers, and both are somewhat large for fragments, with 19 or 20 non-hydrogen atoms.

 

Another target screened was the molecular hub protein 14-3-3, which we wrote about most recently just a few months ago. A screen of the 10,000-compound Scope core set looking for stabilizers of 14-3-3σ and a peptide from the estrogen receptor yielded 86 hits, of which three were fully validated in orthogonal assays. One of these yielded a crystal structure and turned out to be a covalent modifier at a cysteine residue previously shown to be reactive. As with CLK1, the hit was on the larger side, with 22 non-hydrogen atoms. A separate publication by Colin Skepper and colleagues in ACS Med. Chem. Lett. describes this and three other covalent hits derived from diversity sets. Here too, the hits ranged from 20 to 24 non-hydrogen atoms.

 

The native mass spectrometry screen mentioned above was run against the main protease (MPro) of SARS-CoV-2. The 10,000 compound core set was screened in pools of four, with each compound at 25 µM. Of 61 primary hits, five validated with IC₅₀ values < 25 µM. In contrast to many inhibitors of this protein, these were mostly non-covalent.

 

The Scope Concept is a nice addition to the field of library design. As we noted last year, covalent fragments may need to be larger than non-covalent fragments, and it is interesting that two of the three case studies identified covalent modifiers. I hope Novartis publishes a follow-up in a few years with lessons from additional screens.

Design vs randomness in lead optimization across a frustrated landscape

The goal of lead optimization is to improve the properties of a molecule such that it can become a drug or a chemical probe. Early in a program, particularly one based on fragments, the focus is often potency, but eventually drug metabolism and pharmacokinetics (DMPK) parameters become important. I always enjoy reading optimization stories and seeing clever use of design principles. But how effective is the design process? Or, to put it another way, how often are improvements just luck? This is the subject of an open-access paper just published in Nature by Bryan Roth, James Fraser, Brian Shoichet, and collaborators at University of California San Franciso, University of North Carolina Chapel Hill, and several other institutes.

 

The goal was to create a “random background” against which to compare directed efforts. To do so, the researchers chose six protein targets with which they had experience: the GPCRs α₂AAR, µOR, and CB2; the transporter SERT; and the soluble enzymes AmpC and Mac1. For each protein, they then chose one to six  ligands, 18 total, with IC₅₀, K_I, or EC₅₀ values ranging from 1 nM to 43 µM. These ‘parent ligands’ were then analyzed systematically to see how they could be modified one atom at a time by adding a methyl group, a halogen, or a phenol, or replacing an aromatic ring carbon with a nitrogen. A decade ago, we wrote about the importance of synthetic tractability, and this was a key consideration here. Ultimately the researchers made 257 molecules, with each parent giving rise to between 5 and 31 analogs. These analogs were then tested in a variety of assays.

 

Breaking things is usually easier than improving them, and that turned out to be the case here: 30% of changes caused a drop in activity of at least ten-fold. Perhaps more surprisingly, 29 of the 257 analogs (11.3%) had at least ten-fold better activity than the parent molecules. Such increases were widespread: only one target (AmpC) did not see a ten-fold increase, and 10 of the 18 parent molecules had more potent analogs. In other words, the odds of improvement for any given chemical series were better than even, even without using any structural insights from the protein.

 

This latest work stands in good company. Back in 2012 we highlighted an analysis of the effects of methyl substituents across thousands of examples covering more than 100 proteins. That work came to similar conclusions, with ten-fold improvements in 8% of cases. A more recent paper looked at 633 published examples where added chlorine atoms improved activity by at least ten-fold. This latest Nature paper finds methyl substituents to be slightly more likely to give ten-fold improvements than chlorine atoms, while adding fluorine was much less likely to yield such improvements, and swapping an aromatic carbon for a nitrogen never did.

 

One might assume that gaining potency by adding a methyl group or a chlorine atom is due solely to increased lipophilicity, but this did not seem to be the case as assessed by calculating LipE/LLE values.

 

Weaker ligands were slightly more likely to be improved by making random changes, which makes sense if you consider that a more potent molecule is likely to be highly complementary to its binding site. Similarly, larger pockets were more likely to see ligand activity improve with random changes.

 

The researchers crystallographically characterized 22 analogs of varying affinities derived from three parent ligands against the protein Mac1 to look for general lessons. Although the analyses proved interesting, “many of the effects of the small perturbation analogs could be explained post hoc from their structures; however, fewer were easily anticipated.”

 

Of course, as noted above, affinity is just one component of a chemical probe, let alone a drug, so the researchers also assessed several in vitro DMPK parameters: microsome stability, plasma stability, plasma protein binding, solubility, hERG inhibition, and membrane permeability. Similar to the activity measurements, some analogs had improvements in one or more properties, while others saw declines. Unfortunately, the analogs with improved potency often had decreased values for in vitro DMPK parameters; none of the 29 analogs with ten-fold improved potency had combined improvements for stability, permeability, and plasma protein binding. As the researchers ruefully conclude, “the changes in in vitro PK were, at best, orthogonal to affinity or potency fold change, and most were, if anything, anti-correlated with it.” This “frustrated landscape” is all too familiar to medicinal chemists.

 

Making and testing all these molecules took a tremendous amount of work (the paper lists over 30 authors), and the researchers naturally wondered whether computational methods would have saved them the effort. They used free energy perturbation (specifically FEP+, from Schrödinger) to calculate affinities and three additional programs to calculate in vitro DMPK parameters. Although the affinity correlations were strong, there was still considerable variation for individual compounds: of 34 analogs predicted to increase in binding energy by more than 1 kcal/mol (about five-fold), only 14 actually did so, while 8 showed decreased activity. Meanwhile, while plasma protein binding and permeability predictions correlated with experimental values, stability and solubility “were essential uncorrelated.”

 

As of now, computers are still no substitute for the good old-fashioned design, make, test cycle. This paper suggests that, in addition to careful design, it may also be worth introducing some randomness into your analogs, particularly when they are easy to make. Sound advice indeed.

Covalent allosteric inhibitors of NLRP3

NLRP3 has received considerable attention as an inflammatory disease target. The protein normally exists in an auto-inhibited conformation, but in response to various stimuli it changes shape to help form a multimeric complex called the inflammasome. This in turn activates various inflammatory pathways, including cell death. Several drugs targeting NLRP3 have gone into the clinic, but none have been approved. In a new open-access Br. J. Pharmacol. paper, Brian Cook, Gabriel Simon, and colleagues at Vividion describe a covalent inhibitor that binds to a previously unknown site.

 

As we wrote nearly a decade ago, Vividion pioneered the use of chemoproteomics to identify covalent binders in cells and cell lysates. While exploring the mechanism of the previously reported NLRP3 inhibitor MCC950, they found that cysteine 280 became resistant to modification by electrophiles in the presence of the ligand. MCC950 does not contain an electrophile, arguing against direct interaction with Cys280 and suggesting instead that binding causes a conformational change to the protein which makes Cys280 less accessible. In other words, protein inhibition correlated with Cys280 being less exposed.

 

A screen of Vividion’s library yielded no hits against Cys280 but more than 300 against cysteine 463. Interestingly, modification of C463 also caused a decrease in reactivity of Cys280, suggesting a conformational change similar to the one caused by MCC950. Thus, the researchers pursued Cys463 binders.

 

Two parameters were used to guide compound optimization, target engagement and promiscuity. The first number, TE₅₀, measures the extent of covalent engagement of a given cysteine after 1 hour in cell lysates. The second, P₅₀₀, measures the percentage of all measured cysteines in the sample labeled by at least 40% after incubation with 500 µM compound for 1 hour. One of the initial hits, VVD-124476, gave a low micromolar value for TE₅₀ but labeled some 20% of the measured proteome. Indeed, it was previously reported as a covalent ligand for a completely different protein. To be useful as a NLRP3 inhibitor, the researchers needed to reduce promiscuity, and the most straightforward approach is to reduce general reactivity.

 

To decrease reactivity, the researchers swapped the acrylamide warhead with a butynamide (VVD-142397). Butynamides are reported to be less reactive than acrylamides, a result corroborated at Vividion with chemoproteomic data on 400 matched molecular pairs. The acrylamide to butynamide swap improved selectivity some 20-fold, though it also decreased potency against NLRP3. Increasing the core ring size regained activity, and rigidifying the molecule led to further improvements. Further optimization ultimately led to VVD-338213 and several related molecules.

 

Although the researchers do not report k_inact/K_I values, a back of the envelope calculation puts it around 1200 M^-1s^-1 for VVD-338213, a potency where cell activity might be expected. Happily this turned out to be the case: the molecule blocked IL-1β secretion from human cells at lower concentrations than MCC950. Importantly, in cells where Cys463 had been mutated to alanine, MCC950 was still active while the covalent molecules were inactive.

 

In addition to promising cell activity, many of the covalent molecules had reasonable stability in whole blood, good permeability, and low efflux. (I do wish the ligand efficiency reactivities of the molecules were reported; after all Vividion, introduced this metric.) Some of the molecules also showed good brain penetration. In vivo pharmacodynamic studies are complicated by the fact that Cys463 is not conserved in mice, but fortunately the researchers could run studies on transgenic mice in which human NLRP3 had been introduced. In these animals, IL-1β was decreased by both VVD-338213 and by MCC950.

 

The researchers were also able to obtain a cryo-EM structure of one of their molecules bound to NLRP3, which confirmed covalent binding to Cys463. More importantly, it revealed that the binding site is in a cryptic pocket some 20 Å from where MCC950 binds. (One wonders if the computational technique we discussed last week would have been able to predict this pocket.) Similarly to the WRN inhibitor Vividion has taken into the clinic, the molecule appears to make only hydrophobic contacts with the protein, with no polar interactions. Mechanistically, the “Cys463 ligands act as molecular doorstops, stabilizing the inactive conformation and preventing the structural rearrangements necessary for transition to the active inflammasome disc.”

 

This is a lovely paper and a case study in transforming a highly reactive molecule into a selective molecule with in vivo activity. It is interesting that the starting molecule, with 28 non-hydrogen atoms, flagrantly violates the rule of three. Last year we asked whether covalent fragments need to be larger than their non-covalent counterparts, and this paper provides another example where the answer is yes. Whether or not this chemical series ultimately advances, it is nice to have another chemical probe for this target.

Computational prediction of cryptic pockets

Last week we highlighted how NMR-based fragment screening was able to identify cryptic pockets on a small protein. Experimental methods are (still) the gold standard, but they take considerable effort and resources. In a new open-access J. Am. Chem. Soc. paper, Neha Vithani, David LeBard, and collaborators at OpenEye, Bayer, Genentech, and Khartis provide a computational approach.

 

The researchers defined six categories of cryptic pockets: 1) side-chain motion, 2) loop motion, 3) secondary-structure motion, 4) secondary-structure change, 5) interdomain motion, and 6) pre-existing deep pockets with no access to the surface of the protein. They collected a set of 21 proteins with 26 cryptic pockets overall (some proteins contain more than one) whose apo (unbound) and holo (ligand-bound) structures were available in the Protein Data Bank. This “Encore Dataset” contains all six types of cryptic pockets, though some are represented by only two or three examples.

 

To computationally identify cryptic pockets, the researchers used weighted ensemble molecular dynamics (WEMD) simulations on the apo structures. The details are quite technical but involve focusing on “normal modes” of protein movement. To find pockets, xenon is included in the simulation; these large atoms will fill pockets that open up during the course of the simulation, so detecting bound xenon reveals pockets.

 

WEMD successfully identified 24 of the 26 cryptic pockets in the Encore Dataset. Interestingly, both failures were for the same protein, hepatitis C polymerase, one a preexisting deep pocket and the other a secondary-structure change from order to disorder. (This latter category is exemplified in one of the cryptic pockets we discussed last week, in which crystallography revealed that a helix became disordered upon ligand binding.) The researchers hypothesize that these types of order-to-disorder transitions may be particularly difficult to detect computationally because the models are designed to identify secondary structures, thus overstabilizing them.

 

So WEMD is able to correctly predict most cryptic pockets in the Encore Dataset, but does it also predict cryptic pockets that don’t exist? Paul Samulsen famously quipped that economists had “predicted nine out of the last five recessions,” which actually looks pretty good compared to WEMD, which predicted more than 30 potential cryptic pockets on the protein biotin carboxylase. For comparison, the software program fpocket, which we wrote about last year, predicts 13 for this protein.

 

To rank which pockets actually exist, the researchers turned to another computational technique called Target X, which was described in this Research Square article earlier this year. This proved reasonably successful at ranking experimentally observed pockets near the top, with the top two predicted biotin carboxylase pockets corresponding to known pockets.

 

This is an interesting paper, but I am left with more questions. First, while the Encore Dataset is nice, why didn’t the researchers (also) use CryptoSite, a previously published set of 93 proteins with cryptic sites, which we discussed in 2024? More importantly, how can drug hunters decide whether to target a specific cryptic pocket? For instance, does WEMD and Target X allow one to assess whether cryptic pockets can support high-affinity ligands? These would be interesting questions for a follow-up paper.

Fragments find cryptic sites on ARID1B

We’ve written previously about the concept of synthetic lethality, in which changes to one protein make cancer cells more dependent on another. Inhibiting this second protein allows selective targeting of cancer cells while sparing normal cells. One such opportunity is the pair of proteins ARID1A and ARID1B, which assemble into a large complex that helps remodel chromatin to regulate transcription. ARID1A is mutated in some 9% of cancers, and ARID1B can partially compensate. CRISPR knockdown studies show that these cancer cells are dependent on ARID1B. ARID1B is huge (2236 amino acids!) and predicted to be mostly disordered – not the easiest target for small molecule drugs. However, a small 109-residue AT-Rich Interactive Domain (ARID), which binds to DNA, had previously been crystallized. This domain is the focus of a new Proc. Nat. Acad. Sci. USA paper by Michael Holliday and colleagues at Relay Therapeutics.

 

Computational assessment using Schrödinger’s SiteMap did not show any ligandable pockets, but molecular dynamics revealed that the N-terminal helix might open to reveal a shallow hydrophobic groove beneath it. Thus encouraged, the researchers performed a 2D NMR (¹⁵N-SOFAST-HMQC) screen of 480 fragments and obtained two hits, both with weak but measurable affinities. Interestingly though, the NMR signals are consistent with slow kinetics, suggesting that the molecules bind to rare conformations of the protein. Chemical shift perturbation (CSP) mapping showed that the two fragments bind near one another but not exactly at the same site.

 

Crystallography revealed that compound A-1 bound to a previously cryptic tunnel of ARID that forms by the displacement of a tryptophan side chain – a motion not seen in the molecular dynamics simulation. Preliminary medicinal chemistry improved the potency to mid-micromolar affinity.

 

Compound B-1 proved recalcitrant to crystallography, but cocrystallization with the more potent compound B-2 yielded a high resolution structure. This molecule did in fact bind where the N-terminal helix normally binds, as predicted by molecular dynamics; the helix itself became disordered.

 

But wait, there’s more! To find additional chemical matter, the researchers conducted a virtual screen of 150,000 compounds against ARID. Of the several dozen molecules tested by 2D-NMR, two showed CSPs similar to those caused by compound A-1, and their binding modes were confirmed by crystallography. But one of the molecules that showed CSPs similar to those caused by compound B-1 also caused CSPs in a different region of the protein, and crystallography confirmed binding to yet a third site.

 

In order for the compounds to be selective for cancer cells they need to be selective for ARID1B over ARID1A, which is 83% identical in the ARID domain. NMR experiments revealed that the molecule that binds to the third site binds with similar affinity to ARID1A, while compound A-2 binds about 5-fold less tightly. Happily, compound B-2 showed no binding to ARID1A even at 10 mM. Thus, the researchers have found three starting points for optimization, and evidence that selectivity is possible.

 

As the researchers acknowledge, the compounds will require “substantial additional optimization,” not least for affinity. Intriguingly though, compounds A-2 and B-2 do bind sufficiently close to one another that a fragment linking or merging approach may be warranted.

 

So in the end, ARID1B ARID presents two cryptic sites and a third more conventional (albeit challenging) site. The presence of multiple small-molecule binding sites is consistent with this paper from 2016, but still surprising in such a small protein. As we noted then, just because a site binds fragments does not mean it will support high-affinity ligands. Will ARID1B turn out to be like KRAS, which ended up being druggable, or like IL-1β, which has resisted high-affinity small molecules? The current paper provides good starting points toward answering this question.

Fragments vs DHX8: towards a chemical probe

RNA helicases present an intriguing but challenging opportunity for drug discovery. Their importance for cell growth makes them attractive cancer targets. But to function, they must undergo considerable conformational changes, and this dynamism makes it difficult to find high-affinity inhibitors. In a new open-access J. Med. Chem. paper, Gurdip Bhalay, Rob van Montfort, and collaborators at The Institute of Cancer Research (ICR), Merck KGaA, and Proteros describe their efforts against DHX8.

 

DHX8 is one of 14 or so DEAH-box helicases and helps release mRNA from spliceosomes. Previous research had found it to be important for the cellular stress response as well as malignant cell transformation and progression, and siRNA-knockdown experiments showed that cancer cells were particularly dependent on DHX8 activity. Thus, the researchers set out to find a chemical probe.

 

The 2000-fragment ICR library was screened (at 300 µM) against the helicase domain of DHX8 using both surface plasmon resonance (SPR) and thermal shift assay (TSA). In addition, the full-length yeast homolog of DHX8, Prp22, was also screened by SPR to look for fragments that might bind outside the helicase domain. The TSA screen yielded 18 hits, while the SPR screen yielded 91, of which 63 also bound Prp22, while another 17 fragments bound Prp22 more strongly than the helicase domain of DHX8.

 

Eight-point dose response curves were collected for the hits, and 10/18 TSA hits gave a concentration-dependent response on both DHX8 and Prp22, along with 42/108 SPR hits, while another 5 SPR hits only bound to Prp22. Surprisingly, there were only 2 hits in common between the TSA and SPR hits; the researchers note the importance of using multiple orthogonal screens.

 

Next, the confirmed hits were tested in three ligand-observed NMR assays: CPMG, WaterLOGSY, and STD. Reassuringly, 39 were active in all three assays. Competition experiments suggested that the fragments did not bind in the ATP-binding site, but some of them were competitive with RNA.

 

Crystallography was attempted on several of the fragments, and compound 1 yielded a structure revealing that it did in fact bind in the “RNA-binding tunnel” of the helicase domain, with the carboxylic acid making contacts similar to a backbone phosphate. Although the dissociation constant (assessed by SPR) was > 1 mM, compound 1 was active in an ATP-hydrolysis assay and an RNA fluorescence anisotropy (FA) binding assay with triple-digit micromolar activity (values for the FA assays are shown here).

 

Much of the paper is devoted to the optimization of this fragment, which, true to the reputation of the helicase family, turned out to be difficult. Testing related compounds led to compound 6, with low micromolar activity, but despite extensive efforts it was impossible to improve the affinity much further. Thus, the researchers reexamined their larger compound collection for other similar compounds and identified compound 24. Although weaker than compound 6, a crystal structure revealed that the phenyl ring binds deeper into the RNA-binding tunnel, providing new opportunities for fragment growing. Structure-based design led to compound 34, with low micromolar activity, and further optimization ultimately led to compound 53, with submicromolar activity.

 

The mechanism of DHX8 requires the RNA-binding tunnel to open and close right where compound 53 sits, so its inhibitory activity can be attributed to blocking this motion. In addition to biochemical activity, the molecule also showed low micromolar activity in a cellular target engagement assay. However, no other cellular data are reported.

 

This is a textbook example of fragment finding and growing against a difficult target. The fact that the researchers returned to fragments after the initial series reached an affinity plateau is a useful reminder that it is all too easy to get stuck in a local minimum, and you may need to take a step back to make progress. Although compound 53 is just on the edge of what may be considered a chemical probe in terms of potency, it is the first reported nanomolar inhibitor of DHX8. The researchers also mention a second series derived from a different fragment, and it will be interesting to see if this yields even better molecules.

Site-specific HDX-MS for increased resolution

Most methods to characterize protein-ligand binding interactions have drawbacks. Crystallography, obviously, requires your protein to crystallize, while NMR and cryo-EM both have size limitations and are resource-intensive. Hydrogen-deuterium exchange mass spectrometry (HDX-MS), which we wrote about earlier this year, is theoretically faster and can apply to more proteins, but the technique provides limited resolution. Now a new open-access paper in J. Am. Chem. Soc. by Derek Wilson and collaborators at York University, the Structural Genomics Consortium, and SCIEX have improved the resolution down to single amino acids.

 

As we wrote in March, HDX-MS relies on the exchange between deuterium and protons on proteins; slower exchange is associated with more stable secondary structures. The exchange is detected by digesting the protein with proteases and determining the mass of the resulting peptide fragments using MS; incorporation of deuterons increases the mass of the peptides. If, for example, a peptide has three deuterons incorporated, the mass would increase by 3. However, in conventional HDX-MS, one can’t tell exactly where the deuterons have exchanged.

 

Last year the researchers described a method in J. Am. Soc. Mass Spectrom. to do just this using some clever mass spectrometry experiments. I won’t go into details as the paper is open access, but suffice it to say that it involves electron-activated dissociation (EAD) to fragment the peptides while preventing the scrambling of the protons and deuterons that normally limits resolution.

 

In the new paper, the approach was applied to the oncology target WDR5, which we previously discussed here. The researchers applied site-specific HDX-MS to three small molecule and two peptide ligands. All five of these ligands had previously been characterized by conventional HDX-MS as well as surface plasmon resonance (SPR), and four had also been studied crystallographically.

 

Reassuringly, the results from site-specific HDX-MS were mostly consistent with conventional HDX-MS, but there were some surprises. For example, the site-specific approach revealed protection from exchange for one amino acid (which makes a hydrogen bond to the ligand) and enhanced exchange for the adjacent amino acid on the same peptide. In conventional HDX-MS, the “peptide contains offsetting uptake increases and decreases, resulting in a ‘net-zero’ signal.”

 

The researchers also found that, for another residue, the change in HDX-MS signal correlates with the affinity of the ligand, as assessed by SPR. The mechanism is unclear since the concentrations of ligands were far higher than their affinities, but the results are intriguing.

 

The new paper describes just a single, well-behaved target, but if site-specific HDX-MS applies more broadly, it would offer a significant advance in obtaining structural information. The researchers note that the data analysis is currently “manual and prohibitively cumbersome,” a problem that should be solvable with automation. I look forward to seeing it applied prospectively to more targets, particularly if it works for weak binders such as fragments.

Filtering fluorescent frauds from fragments for PTP1B

Protein tyrosine phosphatase 1B (PTP1B) is a long-standing target for diabetes and other diseases. But as we’ve written previously, it is challenging to drug due to its small and highly charged active site, which is conserved among phosphatases. In a new open-access Drug Design Dev. Ther. paper, Frank Boeckler and colleagues at Eberhard Karls Universität Tübingen screen fragments against PTP1B and demonstrate the importance of distinct, rigorous controls.

 

The Boeckler group previously developed a fragment library enriched in halogens (HEFLib, which we discussed here) and a separate covalent fragment library (CovLib). These, as well as two other small libraries, were screened against PTP1B using a fluorescence activity assay. Compounds were incubated with the enzyme for an hour at 2 mM in the presence of a substrate (DiFMUP, 6,8-difluoro-4-methylumbelliferyl phosphate) that releases a fluorescent reporter (DiFMU) when hydrolyzed by the enzyme. A molecule that inhibits the enzyme should lead to less fluorescence. Samples were run in triplicate, and detergent was included to avoid aggregation artifacts. A total of 515 fragments were tested, leading to 56 preliminary hits, defined as molecules that decreased fluorescence by more than 50%.

 

A major challenge with fluorescence assays is that the compounds themselves can interfere with the readout. To assess this, the researchers retested their hits both before and after incubation of PTP1B with substrate. A fluorescence artifact should give similar activity in both conditions, while a true inhibitor should only cause a decrease in fluorescence when added at the beginning of the assay, before the protein has had a chance to process substrate. Unfortunately, only 11 hits passed this test.

 

To further confirm mechanism, the researchers developed an HPLC-based assay that separates the substrate DiFMUP from the product DiFMU. Because this assay detects PTP1B activity using separation rather than fluorescence, it should be immune from fluorescence artifacts. Only three of the 56 preliminary hits confirmed in this assay. Dose-response experiments revealed them to be quite weak, with IC₅₀ values above 5 mM. All three are potentially covalent inhibitors, and two showed reactivity with glutathione. PTP1B has a reactive active-site cysteine so covalent inhibitors are not surprising, though it remains to be seen how specific they are.

 

This paper is a useful reminder of the importance of validation: an initial 11% hit rate dropped to just 0.6%. Full experimental details for all the assays are provided, as are the structures of the hits, while SMILES strings of all 515 compounds screened are listed in the supplementary material. If you’re running a fluorescent fragment screen for the first time, this paper is worth consulting.

Merck’s journey from flatland

Back in 2009 we highlighted a paper, “Escape from flatland,” suggesting that molecules with higher fractions of sp³-hybridized (ie, non-aromatic) carbons are more likely to become drugs. This hypothesis has been challenged on methodological grounds, and an analysis last year concluded that “Fsp³ may not have been a useful metric to optimize.” Now a new J. Med. Chem. paper by Olivia Garry and colleagues at Merck weighs in on the debate.

 

The researchers calculated Fsp³ (number of sp³-hybridized carbons divided by total carbons) for millions of small molecules (MW < 1200) synthesized at Merck from 2000 through 2024. Interestingly, mean Fsp³ values dropped from 0.34 to 0.31 between 2000 and 2006 or so, coinciding with the increasing popularity of metal-mediated cross coupling reactions during that period. Since then, Fsp³ has risen to 0.40 in 2024, which could be due to the focus on “escaping flatland” as well as the increasing availability of saturated building blocks and new synthetic methodologies.

 

Consistent with the original publication, the average Fsp³ increases as molecules progress through development, from 0.34 overall to 0.36 for those that go into rat pharmacokinetic studies and to 0.38 for preclinical candidates – close to the 0.39 found for drugs approved since 2009 according to the study last year.

 

A key correlation proposed in the original flatland publication was that solubility increased with Fsp³, and this was observed again here. Of 435,280 compounds having Fsp³ between 0.00-0.70 with measured solubility data, 55% of those with more aromatic character had low solubility, while only 15% of those with Fsp³ > 0.65 were poorly soluble. The correlations held for solubility at neutral pH as well as at pH 2. Molecules with lower Fsp³ also tended to have more aromatic rings, and this has previously been shown to lower solubility.

 

In contrast to solubility, lipophilicity (logD) did not correlate with Fsp³ among 503,780 compounds. Apparent permeability showed a complex and noisy correlation for 16,570 compounds, leading the researchers to conclude that “Fsp³ is not a good parameter to optimize P_app.”

 

Inhibition of three cytochrome P450 enzymes, CYP 3A4, 2C8, and 2C9, was examined for 51,760 compounds. Weak trends were observed for the latter two, with highly aromatic compounds being more likely to inhibit the enzymes, but no trend was seen for CYP 3A4.

 

Inhibition of human ether-a-go-go-related gene (hERG) is a major red flag for compound progression, and here there was a correlation, with higher Fsp³ compounds showing lower inhibition. This trend held for both aminergic and nonaminergic molecules, though as with solubility these trends could be driven by aromatic ring count, which has independently been correlated to hERG inhibition.

 

Since the 2009 flatland publication, many solvents have been spilled to replace phenyl rings with more shapely, non-aromatic moieties. The researchers looked at matched molecular pairs for internal compounds to see what effects these changes had on solubility. Simply saturating a phenyl to a cyclohexyl ring tended to decrease solubility, while the best solubility improvements resulted from cyclopropyl or isopropyl substitutions. But these swaps did little for shapeliness: of the eight substitutions for which sufficient data existed, cyclopropyl and isoprproyl had the lowest increase of average Fsp³. As the researchers note, “using Fsp³ without consideration of other properties is not a good strategy to optimize solubility.”

 

Overall this study suggests that leaving the comforts of flatland may be worthwhile. Molecules with higher Fsp³ values are sometimes more challenging to synthesize, but this itself can lead to new intellectual property space.

 

It’s worth emphasizing that the correlations between Fsp³ and solubility, CYP inhibition, and hERG inhibition are limited. In the end, the researchers recommend “considering Fsp³ in combination with other factors in medicinal chemistry optimization as its standalone effect on specific properties is modest.” Just as general risk factors won’t predict whether any given person will contract a disease, metrics won’t predict whether any given molecule will succeed (or fail) as a drug. 

Ligand affinities for isolated proteins vs in cells and cell lysates

Earlier this year we highlighted a study that compared the affinities of kinase inhibitors for isolated proteins vs those same proteins in cells and found differences – sometimes dramatic. A new open-access paper in J. Am. Chem. Soc. by Blake Peterson and colleagues at Ohio State University addresses the same question, and explores insights from cell lysates as well.

 

To measure affinity in cells, the earlier paper used NanoBRET, which relies on a fusion protein consisting of the kinase and a luminescent protein. It is possible that this artificial protein may behave differently from the native protein, so the Peterson group developed the Fluorescent Probe Cellular Binding Assay (FPCBA). This technique makes use of a fluorophore conjugated to a ligand for the protein of interest. The protein of interest is overexpressed, the binding of fluorescent probe causes cells to become fluorescent, and the level of fluorescence can be quantified using flow cytometry. Adding a (non-fluorescent) ligand displaces the fluorescent probe, which is rapidly effluxed, and the loss of fluorescence can be used to calculate the affinity of the ligand.

 

For the sake of brevity I’m skipping lots of important details, such as detailed characterization of efflux and use of a second fluorescent protein to correlate expression of the target, but these are described extensively in the paper. The researchers focus on a fluorescent probe based on dasatinib, an approved drug that binds to dozens of kinases.

 

The researchers compared the affinity of dasatinib and imatinib, another approved drug, to the kinase ABL1 using both FPCBA and NanoBRET. The results were reassuringly similar, though dasatinib bound 3 or 4-fold more tightly in the PBCBA assay than the NanoBRET assay while imatinib bound up to 2-fold less tightly, which might be due to changes in the regulatory mechanism of the ABL1 fusion protein required for NanoBRET.

 

Whole cell assays tend to be lower throughput, so cell lysates are often used instead. For kinases, a common assay makes use of “kinobeads.” Native kinases are captured from cell lysates on beads and quantified using mass spectrometry. An inhibitor blocks binding of the kinase to the bead, and this reduction in binding can be used to calculate the affinity. The researchers measured the affinity of dasatinib and imatinib against a score of kinases using both FPCBA and kinobeads.

 

Of 21 kinases, 14 showed similar affinities (within four-fold) for dasatinib by both FPCBA and kinobeads. Six of the remaining seven showed up to 13-fold tighter binding in FPCBA than kinobeads, which the researchers attribute to the kinases being in a more active state in cells than in lysates. (A similar phenomenon has been seen for the helicase WRN.)  In contrast, the kinase SRC showed 11-fold weaker activity in cells than lysates, which the researchers attribute to autoinhibitory regulation.

 

The situation becomes even more discordant when comparing affinities in cells to the affinities of isolated recombinant kinase domains using the KINOMEscan assay. Since a picture is worth several hundred words (the length of this blog post), I’ve graphed the results. While possible to see a correlation, it is also possible to see dragons in the shapes of clouds.

 

The affinity of dasatinib was generally higher (often by more than 10-fold) in the KINOMEscan than the FPCBA assay, which is not surprising given the lack of ATP in the biochemical assay. But there were exceptions, most notably EPHA4, which binds slightly more tightly to dasatinib in cells than in the KINOMEscan assay. Interestingly, the affinity in the cell lysate assay was lower; the researchers suggest that in cells the kinase is in a more active form than it is in lysates, perhaps due to membrane localization.

 

In the end, this paper is another useful reminder that assays with isolated proteins do not necessarily translate to cells. And of course, activity in cells does not necessarily translate to activity in vivo. Understanding why is part of what makes drug discovery both frustrating and exciting.