Mapping protein conformations with fragments

Proteins can be remarkably dynamic, and, as we noted recently, different conformational states can reveal different pockets for small molecule ligands. But how can one survey and categorize all the possibilities? In a recent J. Chem. Inf. Model. paper, Doeke Hekstra and colleagues at Harvard University present a new tool for doing so.

 

High-throughput crystallographic fragment screens are becoming faster and more widely accessible, and the researchers wondered whether the information from these screens could be used to map protein conformational landscapes. To do so, they built a Python program called COLAV, short for COnformational LAndscape Visualization. This open-source tool can compile data from hundreds of protein coordinate files and then, for each protein, calculate the dihedral angles between backbone atoms, the pairwise distances between the alpha-carbon atoms, and the strain.

 

To a first approximation, dihedral angles capture local movements, while distances between alpha-carbons capture global movements, such as the distance between the N-terminus and C-terminus. Strain measurements are also local but can reveal particularly important features such as hinge movements. Also, while dihedral and pairwise distances can be calculated for single proteins, strain measurements are calculated after first aligning multiple structures.

 

Having calculated these three parameters for individual protein structures, COLAV can compare them across the selected set of structures using principal component analysis (PCA). These comparisons can reveal clusters with similar dihedral angles, pairwise distances, or strain.

 

The researchers provide two case studies. The first is the metabolic disease target PTP1B, which we recently wrote about here. This enzyme has been pursued intensively for decades, so the researchers were able to draw on 163 individual protein structures deposited in the protein data bank (PDB) as well as 187 structures from a high-throughput crystallographic fragment screen. PTP1B contains two flexible loops, each of which adopts one of two conformations, and COLAV successfully segregated all 350 structures into four clusters. Importantly, these four clusters were found whether the structures were pulled from the PDB (representing experiments conducted across multiple labs and years) or from the fragment screen, suggesting that a single crystallographic fragment screen can identify most or all of the conformational states available to a protein. This is particularly impressive given that most of the fragments bound in allosteric sites while most of the ligands found in the PDB bound in the active site.

 

Next, the researchers turned to the main protease (MPro) of SARS-CoV-2, the subject of intense and successful drug discovery efforts. They used 656 structures from the PDB and 631 structures from high-throughput crystallographic screens to perform COLAV analyses. Unlike PTP1B, discrete conformational clusters were not observed; rather a continuous band was seen, suggesting that the protein can assume myriad conformations. Here too though, the fragment screens were able to sample most of the conformations observed in the PDB.

 

The fact that a single high-throughput crystallographic screen can capture the conformations seen in hundreds of hard-won discrete protein-ligand crystal structures is encouraging, though of course the paper only describes two case studies. Also, as the researchers note, any structure that cannot be crystallized is not sampled. Since COLAV is free to use, it will be fun to see it applied to other proteins.

Fragments in the clinic: Lirafugratinib

With crystal structures of protein-ligand interactions becoming increasingly accessible, it is easy to forget that proteins do not exist as the static structures seen on page or screen. Indeed, back in 2018 we quoted Karplus quoting Feynman that “everything that living things do can be understood in terms of the jiggling and wiggling of atoms,” and even the smallest proteins have lots of atoms. In an open-access paper published in Proc. Nat. Acad. Sci. USA earlier this year, Heike Schönherr, David Shaw, and collaborators at Relay Therapeutics, D.E Shaw Research, Pharmaron, and Columbia University take advantage of these movements.

 

The researchers were interested in finding selective inhibitors of fibroblast growth factor receptor 2 (FGFR2), which is activated in many cancers. The four members of the FGFR family are so closely related that finding selective inhibitors is difficult. Inhibiting FGFR1 can lead to hyperphosphatemia, while inhibiting FGFR4 can cause diarrhea, side effects seen with the approved fragment-derived drug erdafitinib.

 

Although the structures of FGFR1 and FGFR2 are very similar, extended (25 µs) molecular dynamics simulations revealed that the so-called P-loop of the proteins behaved differently: in FGFR1 it became disordered, while in FGFR2 it remained more rigid. The researchers sought to take advantage of these differences with a covalent inhibitor.

 

The researchers started with a non-selective hinge-binding fragment, compound 1. Adding an acrylamide warhead led to a nanomolar inhibitor with modest selectivity for FGFR2. (All IC₅₀ values are measured after 30 minute incubations.) Growing the molecule into the so-called back pocket of the kinase led to compound 5, with nearly 100-fold selectivity for FGFR2 over FGFR1. 

 

 

The path from compound 5 to lirafugratinib (also called RLY-4008) looks straightforward but was anything but. First, the aryl acrylamide was a metabolic liability, so the researchers attenuated the reactivity by adding a methyl group. Mechanistic studies with this molecule revealed that while it had only a slightly better affinity (K_I) for FGFR2 than FGFR1, it had a k_inact value about 15-fold higher for FGFR2. Molecular dynamics studies suggested that the relevant cysteine in FGFR1 is locked in a position too far from the acrylamide to react, while the corresponding cysteine in FGFR2 may be able to more closely approach the acrylamide warhead.

 

Further optimization, guided by extended molecular dynamics simulations, led eventually to lirafugratinib with ~250-fold selectivity for FGFR2 over FGFR1 and >5000-fold selectivity over FGFR4. Remarkably, the noncovalent version of lirafugratinib, compound 11, shows dramatically lower affinity for both FGFR1 and FGFR2 and very little selectivity between them. The ligand seems to assume a different binding mode after covalent bond formation, which could explain these differences in selectivity.

 

Mouse studies of lirafugratinib showed tumor stasis or regression without increased serum phosphate levels. More importantly, early clinical data has shown “minimal hyperphosphatemia and diarrhea.”

 

This is a lovely example of structure and dynamics-based design (SDBD?). Commonly cited advantages of covalent drugs include improved potency and extended pharmacological effects, but this work shows that they can also achieve remarkable selectivity between closely related proteins, even when both proteins contain cysteine residues in the same location. Moreover, an open-access paper in Cancer Discov. that dives more deeply into the biology shows that lirafugratinib is selective across the kinome, inhibiting just two of 468 kinases other than FGFR2 by >75% at 500 nM.

 

The next time you’re trying to find a selective inhibitor for one member of a protein family, it may be worth taking a covalent approach, and paying close attention to dynamics along the way.

SAR by TR-HT-SAXS

Well that’s an acronym soup! SAR by NMR was the first practical fragment-finding method, and over the years Practical Fragments has covered lots of other techniques. Small-angle X-ray scattering, or SAXS, has not been among them. As the name suggests, this technique uses X-rays, typically produced at a synchrotron. However, unlike conventional crystallography, it doesn’t require crystalline material. Instead, proteins in solution are analyzed to provide information on their size and shape. The resolution is too low to assess small molecule binding, but suitable for observing dimerization or changes in conformation.

 

Time-resolved SAXS, or TR-SAXS, examines SAXS over time in response to a trigger. For example, you can rapidly add a ligand to a protein and watch for changes in conformation. And HT simply means high throughput. A recent Nature Chemical Biology paper from Chris Brosey, John Tainer, and collaborators at the University of Texas MD Anderson Center, Lawrence Berkeley National Laboratory, University of California Santa Cruz, and University of Arkansas for Medical Sciences Little Rock describes structure-activity relationships by time-resolved high throughput small-angle X-ray scattering (TR-HT-SAXS).

 

The researchers were interested in apoptosis-inducing factor (AIF), a mitochondrial protein with potential implications for cancer and other diseases. AIF normally exists as a monomer in complex with an FAD cofactor. Binding of NADH causes reduction of FAD to FADH^- and concomitant dimerization of the protein. Could fragments do the same, allowing dimerization on demand?

 

A library of 2500 fragments purchased from Life Chemicals was screened at 0.75-1.5 mM against the AIF-FAD complex using differential scanning fluorimetry (DSF), and those that raised or lowered the temperature by more than 1.7 ºC were further characterized by microscale thermophoresis (MST). This led to 32 binders and 7 negative controls, or molecules that did not confirm either by DSF or MST. (Side note: although many people discount compounds that give negative thermal shifts, the natural ligand NADH lowers the melting temperature of AIF by a whopping 10.8 ºC.)

 

Next, the fragment binders and negative controls were screened at 0.5-1 mM by TR-SAXS. Intense X-rays cause reduction of the FAD cofactor, but in the absence of NADH or other ligands the AIF protein remains monomeric. However, some fragments did cause dimerization of the protein during TR-SAXS. Interestingly, these fragments were structurally related to one another. Subsequent crystallography revealed that they bind where NADH normally binds and make some of the same interactions to induce protein dimerization. The paper includes much more detailed characterization, including mutagenesis, spectroscopic, and protein crosslinking experiments to further understand the mechanism.

 

TR-SAXS is an interesting addition to our toolbox of biophysical methods suitable for fragment screening. It does have some disadvantages, such as the need for large amounts of protein at high concentrations: 67 µM in this case. Also, the “HT” may be somewhat aspirational, with a current throughput of 100-200 compounds per synchrotron shift. Finally, the technique is probably best suited to well-characterized proteins where SAXS data can be carefully modeled. With these limitations in mind, it will be fun to see how generally TR-SAXS finds fragments that alter the conformation and multimerization of proteins.

Second harmonic generation (SHG) vs KRAS

Practical Fragments is currently running a poll on fragment-finding methods used by readers – please vote on the right-hand side. One biophysical method that perhaps we should have included is second harmonic generation (SHG). A recent paper in Proc. Nat. Acad. Sci. USA by Josh Salafsky, Frank McCormick, and collaborators at Biodesy, University of California San Francisco, and elsewhere describes the technique and its application to find fragments that bind to the oncogenic protein KRAS.

In SHG, two photons of the same energy are absorbed by a material which then emits a single photon with twice the energy. In the commercial instrument developed by Biodesy, a powerful 800 nm laser irradiates a dye, and the 400 nm photon it emits is detected. The intensity of the signal is exquisitely sensitive to the precise orientation of the dye. If a protein is labeled with an SHG-active dye and then immobilized on a glass surface, even subtle changes in conformation will be detected.

The researchers chose the G12D mutant form of KRAS, which is one of the most common variants and is associated with particularly aggressive tumors. They labeled the protein with a lysine-reactive SHG dye under conditions in which each protein would, on average, have one covalently-bound dye molecule (though some would have none and others would have more than one). Proteolysis and mass-spectrometry analysis revealed that the dye molecule labeled three different lysine residues, which the researchers viewed as a feature since a ligand causing a conformational change to any of the lysine residues would generate a signal. The researchers also demonstrated that the dye modification did not interfere with the ability of KRAS to bind to the RAS-binding domain of RAF.

Labeled KRAS was then immobilized and tested against several proteins known to bind it, including antibodies and the nucleotide exchange factor SOS. These produced SHG signals, presumably by causing conformational changes to KRAS, while non-binders such as tubulin did not.

Having established that the assay could detect binders, the researchers screened 2710 fragments at 250 and 500 µM, and obtained a whopping 490 hits. These were then triaged by screening at lower concentrations and performing dose-titrations, and 60 were then characterized by SPR.

Fragment 18, 4-(cyclopent-2-en-1-yl)phenol, showed binding by both SHG and SPR, and was further studied by 2-dimensional NMR (¹H-¹⁵N HSQC). This technique allowed measurement of the weak 3.3 mM dissociation constant. More importantly, it allowed the researchers to establish the binding location as being near the so-called “switch 2” region where SOS normally binds. This is the same region where a previous NMR screen had identified the slightly more potent fragment DCAI. The current paper confirmed that finding, though the researchers found evidence that DCAI may bind to other sites too. Docking studies using SILCS suggested that fragment 18 likely binds in a similar orientation as DCAI. Not surprising given the low affinity, the new fragment did not show functional activity in a biochemical screen.

SHG is an interesting approach, and the ability to rapidly assess protein conformational changes distinguishes it from other biophysical techniques. Site-specific labeling would produce more informative data on which regions of a protein move. However, I wonder if SHG is perhaps too sensitive, as evidenced by the large number of hits. Indeed, the researchers demonstrated that the promiscuous lipophilic amine mepazine also generated a strong SHG signal with KRAS. It would be interesting to do a head-to-head comparison with other similarly rapid techniques such as DSF or MST. Have you tried using SHG, and if so, how did it perform for you?

Fragments find flexibility in fascin 1

Protein flexibility can be both an opportunity and a barrier – quite literally, when a solid wall of protein seems to block opportunities for fragment growing. But like secret doorways, protein domains can yawn open to expose tunnels and cavities. An example of this was published earlier this year in Bioorg. Med. Chem. Lett. by Stuart Francis and collaborators at the CRUK Beatson Institute.

The researchers were interested in fascin 1, which increases the invasiveness of multiple cancers by helping pack filamentous actin into bundles important for cell migration. The team began their search for an inhibitor by performing a surface plasmon resonance (SPR) screen of 1050 fragments, generating an impressive 53 hits. Although a number of these were reported to bind to multiple sites on the protein, only one is discussed.

Compound 1 binds between two domains of the protein in a pocket that does not exist in unbound fascin. However, the fact that the pocket completely envelopes the fragment “hampered attempts to develop the series.” Fortunately, the researchers were following the patent literature, and when they characterized compound 2 (not a fragment, and reported by a different group), they discovered that while it binds in the same pocket as compound 1, additional conformational changes occur to accommodate the larger molecule.

Next, the researchers looked for analogs of compound 2 and also performed a virtual screen against the enlarged pocket. Of 110 commercial compounds tested, three gave dissociation constants better than 100 µM, including compound 3. The researchers recognized that compound 3 lacks the halogens found on both the original fragment and compound 2, and by adding these they were able to improve the affinity more than ten-fold. Further optimization ultimately led to BDP-13176, with mid-nanomolar affinity by SPR and ITC as well as activity in a functional assay. Although the molecule has reasonable solubility and stability against liver microsomes, it has low permeability and high efflux.

This is a nice structure-based design story, and while the fragment did provide some information about the binding site, one could argue that the real breakthrough came with determining the binding mode of compound 2. Indeed, without this information, it would have been all too easy to assume that the pocket was not ligandable. This is an important reminder that crystal structures usually only reveal one form of a protein. The system is also a good test case for modelers who want to see how their algorithms perform against a dynamic protein. Breakthroughs are often unexpected, and it is always worth making a few compounds that don’t look like they’ll fit.

Fifth NovAliX Biophysics in Drug Discovery Conference

Last week NovAliX held its biophysics meeting outside of Strasbourg for the first time. Naturally they chose Boston, one of the most European of US cities and a major hub of drug discovery. The event brought together 118 participants from 15 countries, roughly 80% from industry. Although the food and drink could not compare to France, the science and discussion were every bit as satisfying. With 30 talks and 22 posters I won’t attempt to be comprehensive, but as with last year just try to capture a few themes. 

One particularly noteworthy session was devoted to single particle cryo-electron microscopy (cryo-EM), which was recently reviewed in Nat. Rev. Drug Discov. by conference chairman Jean-Paul Renaud and a multinational team of experts. The approach involves flash-freezing a thin film of sample and using transmission electron microscopy to capture two-dimensional “projection” images of your target. If the protein is randomly oriented you can computationally combine thousands of individual images into a three dimensional structure. Although the technique has been around for decades, until recently the resolution was too low to be useful for structure-based drug design. Recent advances in hardware and computation have led to what’s come to be known as the “resolution revolution,” explained Gabe Lander (Scripps).

One advance is the 300 keV Titan Krios – a massive (and massively expensive) instrument that is so widely coveted that Gabe showed pictures of happy scientists hugging newly delivered crates. Indeed, of the ~1000 structures solved to < 4 Å resolution, the vast majority of them were solved on one of more than 130 Krios instruments throughout the world. But Gabe showed that high resolution structures can be obtained with more common 200 keV instruments, including a 2.6 Å resolution structure of aldolase (150 kD), a 2.9 Å structure of hemoglobin (64 kD), and a 2.9 Å resolution structure of alcohol dehydrogenase (81 kD) with bound NAD⁺ cofactor. Although only a handful of sub-2 Å structures have been reported, he thought these would become routine in the next few years.

Bridget Carragher (New York Structural Biology Center) described challenges and how to overcome them. Currently it takes at best eight hours to go from data to structure, but she thought getting this to under one hour would be achievable. Moreover, cryo-EM can be used to characterize different conformational or oligomeric states present in a single sample, as Giovanna Scapin (Merck) demonstrated with insulin binding to its receptor. Indeed, even simple visualization – without fancy computational processing – can provide useful information about protein aggregation, as demonstrated by Wen-ti Liu (NovAliX).

Although primary fragment screening still looks a long way off for cryo-EM, it should start to provide useful structural information for fragments bound to targets less amenable to conventional biophysical techniques, such as membrane proteins – the topic of another session.

Miles Congreve (Heptares) discussed how their stabilized “StaR” GPCRs can provide high-resolution crystal structures suitable for FBDD (see for example here). This has allowed them to discover less lipophilic, more ligand-efficient drug candidates against a variety of targets.

According to Anass Jawhari, it isn’t even necessary to make mutant GPCRs: Calixar has developed proprietary detergents that can stabilize full length adenosine A_2A receptor for a week – more than enough time to perform STD NMR screens of 100 fragments and identify 19 hits, some of which turned out to be functional antagonists. Matthew Eddy (University of Southern California) used two-dimensional NMR on this same protein to reveal dramatic differences in conformational dynamics when bound to agonists vs antagonists.

Indeed, conformational changes and dynamics were a running theme throughout the conference. Keynote speaker and Nobel-laureate Martin Karplus (Harvard) quoted fellow Nobelist Richard Feynman: “everything that living things do can be understood in terms of the jiggling and wiggling of atoms.” (As an aside, Martin’s MCSS method pioneered computational FBDD approaches, predating SAR by NMR.) Göran Dahl (AstraZeneca) described how large scale conformation changes well outside of the active site of PI3Kgamma were responsible for freakishly high selectivity of a class of inhibitors.

But how do you detect conformational changes? We’ve previously mentioned Biodesy’s SHG approach, and Parag Sahasrabudhe (Pfizer) described how this proved useful for classifying ligands for IL-17A. Gerrit Sitters (Lumicks) described a completely different “dynamic single-molecule” (DSM) approach, which involves trapping a single fluorescently labeled protein between DNA strands tethered to two microspheres. Changes in protein conformation caused by ligand binding change the distance between microspheres, and these can be detected to within 1 Å.

Kinetics is intimately linked to dynamics, but the factors responsible for slow binding and dissociation are still poorly understood. Chaohong Sun (AbbVie) examined an archive of 8000 data points and found that on-rates and off-rates each varied by more than five orders of magnitude. There was no correlation with ClogP of the ligands, though larger ligands were more likely to have slower kinetics. There were also significant target effects; on-rates were consistently slow for one target.

As we’ve previously discussed, off-rate screening (ORS) can be used to identify hits in crude reaction mixtures, and Menachem Gunzburg (Monash University) described how this technique is being used in hit-to-lead efforts. Lowering the temperature to 4 °C and adding 5% glycerol further slows dissociation, allowing weaker hits to be discovered.

At the extreme, irreversible inhibitors have an off-rate of 0, and Gregory Craven (Imperial College London) described quantitative irreversible tethering of electrophilic fragments to cysteine residues in proteins using a fluorimetric plate-based assay. As we’ve noted, one challenge with irreversible tethering is deconvoluting intrinsic reactivity from proximity-directed reactivity, which Gregory addresses using a reference thiol such as glutathione.

There is much more to say but in the interest of time I’ll stop here. If you missed the conference you have two chances next year: June 4-7 when it returns to Strasbourg, and November 20-22 when it will be held in Kyoto. And there are still excellent events coming up this year – hope to see you at one!