Surface plasmon resonance (SPR) technology is expanding strategies for studying and analyzing binding kinetics, offering real-time, label-free analysis. Researchers at the “Developments in Protein Interaction Analysis—DiPIA2007” conference sponsored by Biacore (www.biacore.com) and GE Healthcare (www.gehealthcare.com), held in Phoenix in May, frequently reported using SPR as a primary screen for compound and fragment libraries with NMR in a validation role.
Takaaki Miura, Ph.D., research scientist at Chugai Pharmaceutical (www.chugai-pharm.co.jp), uses SPR to provide structural information on target molecules in the mechanistic study of proteins. SPR delineates the hot spots for ligand binding, revealing conformational changes and reaction mechanisms.
Dr. Miura said he was skeptical that structure could be detected by SPR, so he investigated the binding of substrates and their analogues of the enzyme UDP-N-acetylglucosamine pyrophosphorylase (UDP-GlcNAc)—a target for type 2 diabetes—to immobilize kinase as a free form. The project succeeded in establishing the essential determinants for binding and showing the enzyme’s possible reaction mechanism.
X-rays showed that the human form of that enzyme has a large, hydrophilic binding pocket. Using that information, Dr. Miura developed a fragmentation approach to determine the spots most important for binding, asking why the UDP group would not bind to AGX2. Reaction mechanism assays testing binding capability showed there was no binding of the upper form of UDP, and that GlcNAc-1P was essential for binding.
To determine whether binding occurred sequentially, UDP binding was tested both in the presence of MG2+ and in its absence. “Results indicated that the binding of UDP and GlcNAc-1P is ordered sequentially,” Dr. Miura said. In this same series of tests, he also saw dose-dependent binding. Last April, a colleague using 3-D x-ray techniques found a conformational change in the enzyme.
A second experiment bound an allosteric inhibitor to its target molecule, kinesin spindle protein (Eg5), to immobilize kinase in the presence of ATP. Eg5 is involved in mitosis, so it is important in cancer drug targeting, Dr. Miura said.
That study showed that an SB-751992 deriviative bound to free Eg5. “Dissociation is extremely strong, so we used an analog with less reactivity,” Dr. Miura noted, and found that it bound to the apo form, creating an ATP-noncompetitive form. The reaction was characterized by slow on/off rates and conformational changes, which were confirmed with NMR. The experiment also showed that monastrol and SB-751992 bound to the same site of Eg5.
Information Delivery in Real Time
One of the challenges in drug discovery and validation is getting biophysical information to the medicinal chemists in real time, according to James Murray, head of structural sciences at Vernalis (www.vernalis.com). SPR is delivering label-free screening and affinity rankings of fragments from Vernalis’ SeeDs (structural exploitation of experimental drug startpoints) library.
“Fragments often recapitulate aspects of known inhibitors in experiments,” Dr. Murray said. “The same fragment in three kinases will have different binding models, therefore, there are three different strategies for evolving, essentially, the same thing.”
These compounds can be screened using either NMR or SPR. In terms of advantages, NMR screens small molecules and receptor-based entities, compares behavior directly, offers fast identification, uses pools of about 12 compounds, and offers spectral resolution.
SPR can bind small molecules and receptors to chip surfaces, and screen individual compounds of about 500 µM, rather than pools. Throughput is about 1,400 fragments at values of 120 kDa per day. As such, it offers “a good primary screening technique for use as the first step in a fragment screening campaign,” Dr. Murray said.
Hits were described by the SeeDs library and validated by NMR. Using SPR as the initial screen resulted in hits with 1:1 binding, thus streamlining the process by eliminating “enormous” data errors common when NMR was used as the initial screen. Consequently, medicinal chemistry work could begin earlier than otherwise possible.
For the hits-to-leads/lead-optimization project, SPR provided a fast start and powerful information, including “on/off rates in discovery time,” he elaborated. “Seeing a sensograph makes a big difference and changes the way people think.”
“But,” he added, “there were frequent issues with regeneration. The compound often locks protein onto the NTA chip for incomplete removal.” That protein could be removed well with trypsin. He also reported that he gets better data interpretation in a high salt environment.
Although many of the speakers focused upon binding parameters, Robert Copeland, Ph.D., vp enzymology and mechanistic pharmacology, at GlaxoSmithKline (www.gsk.com), advocated a new approach to compound optimization based on residence time. The premise is that residence time is more important than binding parameters like affinity or target inhibition and can provide benefits in terms of target selectivity and duration of the pharmacologic effect.
“Residence time isn’t the only factor to look at,” but is proving to be an epiphany for many, Dr. Copeland said. Although “the concept is well-grounded in the literature, there have been no robust, systematic studies. Consequently, translating the concept to management and medicinal chemists is new, but it resonates with medicinal chemists.”
“When optimizing compounds, we often think of protein-drug interactions in terms of equilibrium constructs or equilibrium between bound and unbound states in small molecules,” but the most important feature, he insisted, is the time a drug spends on the molecule.
Usually, we think in terms of closed systems, in which the drug and target come into contact over time, or in terms of open systems, in which the drug flows in and is removed. But, Dr. Copeland said, “These are unrealistic ways to think about pharmacology.
“Measuring equilibrium in a closed system is a good way to screen or optimize drugs, but it doesn’t account for kinetics,” Dr. Copeland said. Likewise, two-step binding with target isomerization is more common, he noted, but causes researchers to “grossly underestimate affinity.”
“Additionally, there are a number of situations where one can have molecules with identical affinities but very different offerings. So, is it affinity of residence time that makes an effective drug?” he asked.
“You don’t need much residence time,” Dr. Copeland said. “A 30-minute half-life can have a great effect on the dynamics of a drug.” A large local concentration of drug in the viscous cytosol provides rapid rebinding, which leads to a long duration of molecular activity of the target. An HIV study showing viral replication and dissociation half life, he said, showed a “very strong correlation between dissociation half life and target inhibition (IC50) for viral replication. In terms of cellular activity, it is the dissociation half life that is most important.”
Another study, this time of HER2 receptors, showed a slight correlation between cell proliferation and affinity. The same cell-proliferation inhibition data, when plotted against dissociation half life showed “a great correlation.” It is also important, though, to view the inhibition curve over time, as inhibition wanes.
So, he asked, “Why do some drugs have long residence time and some do not?” A study of Tykerb (in Phase III/IV trials) showed conformational adjustments. Specifically, the ATP binding site closed down because the ATP cleft activity loop, the activation loop, folded over the binding site, Dr. Copeland explained.
Biacore’s Markku Hamalainen, senior scientist, chemometrics, also de-emphasizes the importance of binding affinity. In fragment screening, “affinity isn’t the most important issue. Ligand efficiency is key,” Dr. Hamalainen said. Rather than using structural analysis, which he terms “demanding and expensive,” he advocated an initial screening using SPR, followed by in silico screening with validation using x-ray crystallography or NMR.
One project he described, conducted by AstraZeneca (www.astrazeneca.com), used sensograms for data quality control, assaying four substances, including wild type thorombin, mutant thorombin, and GST. Affinities could be measured for 62 of the 80 fragments. Results showed that in-line mutant and reference proteins give invaluable information on binding specificities, but that secondary reactions seem slow.
“With thorombin, the P2/P3 pocket is shallow, so fragment binding is hard to identify,” Dr. Hamalainen said. Therefore, he used a concentration series of assays for a high confidence rate and ran 40 compounds in eight hours. He found an “excellent correlation between enzyme inhibition” and direct binding data, and found that SPR “is more sensitive for analysis of low affinity fragments.” Using Biacore’s A100 SPR analysis system, he said, can run about 1,400 fragments at 120 kDa per day with parallel analysis of multiple targets.
Information on Interactions
Work like that is being advanced by the just-launched Biacore X100, which “provides unique information on interactions” in real time, without using labels to determine binding association and dissociation, selectivity, affinity, kinetics, active concentration, and thermodynamics. The system excels at capturing low-affinity interactions, according to Michael Murphy, Ph.D., application scientist at Biacore.
Thuy Tran, a Ph.D., student at Uppsala University’s division of biomedical radiation sciences, used the X100 in an analysis of the binding kinetics—particularly affinity—on the in vivo HER2 tumor targeting properties of Affibody® molecules.
Tran performed a standard kinetics analysis to compare the binding of Affibody molecules to the EGF receptor family member, HER2. “I spent two to three days on regeneration issues and was running out of HER2 protein and time,” Tran recounted.
Moving to the X100, she used the custom assay wizard for single cycle kinetics, which allowed her to run five concentrations on a single construct, maximizing both the sample and her time. In validation tests, “I got the same binding affinity,” she said, “but saved time because no regeneration was needed.”