Flow injection electrospray MS
ultrafiltration are used to separate protein-bound ligands from non-ligands free in solution in order to increase the signal for ligands over background of non-ligands. Sampling and analysis takes place only at the end of the final round of selection. In the deconvolution/retesting phase, 10-30 compounds are included per tube, and 10% of the initial ("R0") volume is sampled as well as the entire volume at the end of R3. Included from  with permission from SAGE Publications.
size is set to be as large as possible to minimize the quantity of protein required and increase the throughput for screening the maximum number of compounds, while still trying to maintain a condition of excess free target. By screening @2700 compounds per mixture, the combined small molecule concentration is @4 mM, or @400 times in excess of the target protein. However, in a diverse small molecule mixture of 2700 compounds, very few compounds are anticipated to have Kd < 10 mM, or even KD < 100 mM, so that the probability of competitive binding leading to the loss of a high affinity ligand is very low. For example, in small molecule screening using an NMR affinity screening method , the frequency of compounds with KD < 1 mM is @0.25% . Therefore, if there are on average @7 very weak ligands (0.25% of 2700) per mixture, in aggregate these are in equimolar concentration with the target protein, and the protein will still be mostly unbound.
Multiple rounds of selection are carried out in order to increase the signal over background. When 90% of the volume is filtered, the initial equilibrium bound fraction of each compound is retained, in addition to a constant residual 10% (unbound) from the remaining volume. Though unbound ligands are being depleted during filtration, the initial equilibrium quantity of bound ligand is maintained because the protein concentration is also increasing at the same rate. For example, a ligand with KD = [protein] will be approximately 50% bound initially. As half the volume has passed through the filter, half of the free ligand has passed through (or 25% of the total), but now [protein] = 2KD, so 66% of the remaining 75% of the ligand will be bound, which is equal to 50% of the original ligand still bound. In other words, the use of ultrafiltration results in a continuous equilibrium such that the relative enrichment can be achieved on the basis of equilibrium rather than dissociation rate, particularly for weak binding compounds with Kd values in the low micromolar range (which typically equilibrate on a timescale that is faster than the volume reduction). After each round of selection, the volume is restored to the initial volume, but so is the initial protein concentration. Successive rounds of selection result in exponential enrichment of li-gands such that the final concentrations will be inversely correlated to the KD of each ligand (i.e., compounds with the highest affinity, or lowest KD, will be the most abundant).
By adjusting the target concentration, the screening stringency can be altered. Given the starting concentration of each compound in the mixture and the post-selection processing for mass spectrometric detection, the ASMS method is designed so that compounds that cannot bind (i.e., those that have KD > 10 x [protein]) are just below the limit for detection in the mass spectrometer, whereas those with the desired affinity (KD @ [protein]) will be >10x above the background as the only remaining peaks. In practice, for the majority of the library a compound with affinity equal to the protein concentration will be robustly identified, while a weaker binder will show less consistent results. However, compounds with weaker KD values, on the order of three-fold above the protein concentration, can also be readily observed when the compounds are especially well extracted and/or ionized in the mass spectrometer.
Parallelization of the processing of individual filter units can lead to extremely high throughput. Our compound library is split into two sets, which we screen in duplicate against every target. In the past a single replicate for either set was screened at the bench in one day. It then took two days or more for mass spectra to be acquired and analyzed due to experimental, equipment, computational, software, and database limitations. Our first step to improve ASMS throughput was to design a methodology that allows for both replicates to be performed simultaneously, greatly reducing the total bench time per target; the entire library can now be screened in duplicate in two days. Time-consuming steps were eliminated by the addition of more automation and by setting absolute time limits for each stage. We have greatly accelerated the data handling by processing and analyzing data in parallel on three or four computers simultaneously. We have improved our custom ASAE.NET automated picking software, resulting in faster analysis and better communication with our databases (data not shown) . Furthermore, we now stagger target screening such that two bench scientists and one mass spectrometrist can screen four or five targets at one time. Such process enhancements allow entire screens of a library of approximately 500 000 compounds to be completed in 2.5 weeks. Finally, switching from electrospray mass spectrometric analyses to LC-ESI-MS has afforded a nearly ten-fold increase in compound sensitivity and resolution. Such an enhancement in sample analysis suggests that we may be able to lower our protein and compound concentrations even further to help reduce our total protein consumption.
After affinity selection, an organic solvent extraction step separates ligands from the protein and prepares them for electrospray mass spectrometric analyses in both positive and negative ionization modes. The protocol was experimentally selected for efficient extraction of the widest range of drug-like  and lead-like [41, 42] compounds in the compound collection. The mass spectra of samples are processed and either inspected visually or by the aid of ASAE.NET analysis software (data not shown) . Peaks that stand out by comparison with the local background are identified as primary hits. In addition, spectra obtained with other compound mixtures are examined to determine whether the m/z ratio of identified peaks are unique to a particular mixture. Peaks with the same m/z ratio in spectra from multiple compound mixtures are generally artifacts, such as contaminants in the protein preparation. To ensure that hits are not missed, peaks are picked even if they are barely enriched over background. The false positives inherent in the noise near background are easily eliminated in the subsequent deconvolution step. The peaks of interest are converted into a list of potential ligands (hits). Each peak, however, corresponds on average to six massredundant compounds, with only one typically being responsible for the apparent binding. Therefore, only @17% of the primary hits are expected to demonstrate binding in subsequent retesting and deconvolution experiments. The primary screen for MurF ligands utilized 45 mixtures of approximately 2700 compounds each and was run in a single day. A duplicate screen was run on a second day. In the MurF screen, 434 peaks were identified as potential hits from the first experiment, ranging in monoisotopic mass from 249.09 Da to 773.50 Da. The number of peaks in each of the 45 mixtures ranged from one to 35. In the duplicate screen, 390 peaks were identified as potential hits, with 157 peaks overlapping between the duplicate screens. Compounds from the overlapping peaks were assembled into a primary hit list of 1147 compounds for subsequent retesting and confirmation.
Was this article helpful?