Stability Matters in Drug Discovery
Simple Chemical Stability
Much of the material in my previous blog on solubility, especially relating to kinetic solubility, is relevant to stability measurement in terms of the techniques used. The same buffers and LC-UV analysis methods are applicable and MS can be useful to help identify any appearing peaks. In fact, solubility assays can potentially reveal stability issues and we should certainly be on the lookout in these assays for peaks appearing in samples that were absent in the standard.
There are many reasons to study stability in a simple buffer. Often it is to troubleshoot issues reported in other assays. Also, it may be to model biological stability, e.g. gastric stability in a much more simple system. Pretty much any pH or buffer system can be chosen. I would advise including 20% DMSO by default to mitigate against solubility problems as far as possible, and to always monitor peak area precision.
A convenient test compound concentration is 100µM which will give a good UV signal for most compounds with a suitable injection volume.
Experiments are most conveniently performed at ambient temperature simply by leaving the vials or plate in the LC autosampler but can be run at elevated temperature (37°C most notably) in a water bath with samples removed at intervals or even cooled, both of which are obviously less convenient (and make sure there is enough sample!)
Data Processing and Units
A semi-log plot of ln(peak area) against time is usually constructed to follow the disappearance of the test compound. The slope (after changing the sign as it will be negative) of this plot gives the first order rate constant (k’) and its reciprocal gives the half-life. I think half-life is the most understandable parameter to report and allows very convenient comparisons within an intuitive frame of reference. If the number of compounds being analysed in a study is such that it is impractical to analyse sufficiently frequently to determine a half life then compounds can be ranked based on % remaining after a relevant and discriminatory time point.
Note that in the types of chemical situations under discussion here, half lives can vary from days (months even) for the slow loss of a compound in a DMSO stock (actually quite a rare situation) to minutes, say, in a glutathione reactivity or microsomal stability assay.
Photostability (as a predictor/precursor of phototoxicity)
Among efforts that I have been involved in to correlate in vitro phototoxicity data with simple chemical measurements, and hence to build simple predictive models, two angles of attack have borne most success, namely an LC-based photostability assay and UV spectrophotometry. The objective of this work was to find ways to reduce the numbers of compounds needing to be sent for an in vitro photocytotoxicity (3T3 neutral red uptake) assay by reliably excluding compounds that would be negative in that assay. A certain rate of false positives was acceptable but false negatives had to be eliminated and criteria were set with this goal in mind.
The LC assay was carried out in 96 well plate format with appropriate UV transparent seals. The irradiation was via the long wavelength UV source of a TLC plate viewer at a fixed distance. LC analysis was performed before and after irradiation and the % change in peak area was calculated. Suitable controls are required to allow inter experimental comparisons to be made, i.e. to monitor consistency in the radiation delivered from plate to plate (and within plate). It would be possible to incorporate the control as an internal standard to further improve accuracy.
The rationale of the photostability assay was that the first step in a phototoxic event is absorbance of UV or visible radiation to form reactive species that are responsible for the toxicity. Hence a further elaboration of the photostability assay was to run it in the presence and absence of glutathione (GSH) which potentially removes reactive moieties (free radicals) thus promoting generation of more and a greater amount of decomposition of the test compound. The relative amount of decomposition in the presence/absence of GSH (the GSH Index) was then an additional parameter for interpretation of phototoxic risk.
One additional point to make about the photostability assay is that with the experimental set up used it was possible to observe the wells under irradiation and fluorescence (blue or green) was frequently observed in some of them. Invariably the fluorescing wells were associated with photo-unstable compounds and the more intense the fluorescence appeared to the eye the greater the instability. It may well be that a simple fluorescence spectrophotometry assay may be an even simpler predictor of phototoxicity.
The second approach involved recording UV spectra using accurately prepared solutions in UV cuvettes. As the spectral region of interest is above 270nm DMSO can be used as the solvent. There are several guidelines on potential phototoxic compounds based on UV absorbance measurement by the WHO and EMEA but their suggested permissible absorption co-efficient above 270nm is so low that these are of little use. Henry et al at Pfizer (Journal of Photochemistry and Photobiology B: Biology 96 (2009) 57–62) suggest a much more realistic limit of 1000 Lmol-1. Our approach was to concentrate very closely on the tail of the absorption maxima rather than the maxima themselves and to report the maximum wavelength for which the extinction coefficient was >100Lmol-1. For a particular project a very good correlation between this parameter and an in vitro phototoxicity flag was found. The key to the success of this assay was in accuracy of sample preparation and very careful blanking of the spectrophotometer baseline.
As is often the case in this type of work it can be very project specific and different assays work best for different projects or, in some cases, there just isn’t a correlation presumably because the phenomenon you're trying to predict (in this case phototoxicity) is occurring by some other mechanism not accounted for in the simpler model. Finally, in the case of in vitro phototoxicity, we’re trying to predict the result of an assay that is in itself only a model being used to predict an in vivo result. An in vitro flag is by no means a guarantee of a real problem.
Glutathione Reactivity Assay
Glutathione (GSH) is a very useful test molecule for potential covalent binders. A GSH reactivity assay can be used to screen molecules for their potency to act as electrophiles and hence bind covalently to nucleophilic sites such as thiol groups in proteins. If this activity is by design then the assay is effectively a potency (enzyme inhibitor) assay. If it is not desired, but there is a risk that chemotypes are set up to act in this way and bind irreversibly, then the assay is a toxic risk assessment.
Note that I am not talking about glutathione trapping as it is perhaps normally understood here in that there is no metabolic (or photochemical) activation of the molecules but rather any reactivity is intrinsic to the molecules under test.
Glutathione (reduced) is typically found in cells at concentrations of 1-10mM where it acts as an antioxidant. Hence, a typical assay protocol would be to incubate the test compound (50µM) with 5mM GSH i.e. 100X excess in, say, 25-50mM phosphate buffer (pH 7.4) and follow the disappearance of test compound over an appropriate timescale (up to 6h). DMSO (10%) is advisable for compound solubility.
As usual, the number of timepoints that can be analysed is somewhat dependant on throughput and also the goal of the assay. The parameter reported may be % remaining at a fixed timepoint for ranking compounds in a series, or the half-life for a somewhat more rigorous measurement.
Although cysteine is less biologically relevant (present in cells at 1000X lower (i.e. µM) concentrations) it is 50X more reactive than GSH. Hence, it could be used in place of GSH to differentiate between low reactivity compounds.
Chiral stability assays to study the racemisation of compounds are clearly of great importance. Initially, stability may be checked out in simple aqueous buffers at various pH values likely to be encountered by the molecule in the body (say 1.2 to 8). However, eventually plasma stability (or stability in other tissue homogenates) may be required. Microsomal or hepatocyte chiral stability may also be required. One thing that all these systems have in common is that they are highly aqueous whereas chiral separations are most commonly realised using normal phase systems, be they LC or SFC. Hence there is potentially a considerable advantage to having reverse phase or CE-based chiral separations in these circumstances in order to simplify the sample preparation needed e.g. simple protein precipitation. If this is not possible then an extraction procedure will be required prior to analysis.
Formulation Stability in Discovery
Formulations, be they suspensions or solutions, can present both short and long term stability issues both in terms of their physical and their chemical stability. I will present some information on formulation solubility and stability measurement in my next blog.
If any of the issues discussed in these articles are of interest to you, please feel free to contact me directly on firstname.lastname@example.org for further information.