Archive for July, 2010

Scintillation Proximity Assay technology (SPA) provides a homogeneous assay format that is useful for receptor ligand binding assays. The homogeneous nature eliminates the separation steps necessary for filtration assays and enables optimization for automated high-throughput drug screening. Applying this technology to 384-well plate formats has allowed researchers to further increase assay throughput. However, not every assay adapts well to higher density plate formats, so another means of increasing assay throughput is required. In these cases we propose that screening with compound mixtures, or ‘compressed plate screening’, is a useful alternative. Compressed plates are not new to the screening community, but the complexity of data analyzis and cumbersome follow-up investigations of primary hits have kept many researchers from using them^(1,2). This article addresses these limitations including the issues of false positives and the preservation of assay sensitivity. Additionally, compressed plate generation using compression algorithms and the process of deconvolution for the purpose for data analyzis are discussed. We propose that compressed plate screening is feasible when primary hit rates are low (<1%) and when screening at low compound concentrations (1μM). Results indicated that more than 7-fold savings in time, money, and reagents after follow-up analyzis has been completed.

Compressed Plates

Testing mixtures of compounds in a well increases assay throughput, but a method for identifying active compounds from active mixtures is needed⁽³⁾. Using an algorithm to ‘compress’ 96-well microtiter plates in a ‘self-deconvoluting’ fashion, generated compressed plates in an 8:1 format. The 8:1 compressed plate contained 720 unique test compounds (18 or 20 compounds per well) and is prepared from eight 96- well plates containing 90 compounds each.

To reduce the imbalance in the number of compounds per well on the destination plate, the eight 96-well compound plates are reformatted into 9 x 10 matrices. These matrices are aligned to make two 18 x 20 matrices, or two ‘source plates’ (one source plate is shown in Figure 1).

Figure 1. One source plate (18 x 20 matrix).

Compounds in each 18 x 20 matrix are then combined to make row mixtures and column mixtures; each destination plate would have 36 row mixtures and 40 column mixtures containing 18 or 20 compounds per well. Row mixtures from source plates 1 and 2 (S1 and S2) were dispensed into the upper half of the destination plate (18 row mixtures each: S1R1–S1R18 and S2R1-S2R18). S1 and S2 column mixtures were dispensed into the bottom half of the destination plate (20 column mixtures each: S1C1-S1C20 and S2C1-S2C20). Wells H7-H12 are reserved for controls (Figure 2).

Figure 2. 96-well destination plate containing compound mixtures.

Each compound was present on the destination plate in two mixtures: one row and one column mixture. Both mixtures are otherwise unique, so that a positive assay result in two such wells identified one and only one compound. For example, if mixtures in wells A3 and E7 in the assay plate above were identified as active, the deconvoluting algorithm identifies the source plate or 18 x 20 matrix in which the compound resides. The algorithm also identified the original compound plate and the putatively active compound (identified by the intersection of the row 3 column 7 mixtures, Figure 3). Deconvolution of hits using the matrix layout was time consuming and error prone, so software was developed to automate the process.

Figure 3. Intersection of active column and row mixtures identifies compound.

An active row mixture and column mixture from a source plate was needed to identify an active compound; the intersection of row and column ‘uniquely’ identifies it. When more than one row and column mixture on a plate are active an individual compound has not been identified (Figure 4a). Assume compounds B11 and C10 were true actives. Compound C9 was falsely declared active because it was in column 3 (an active mixture because of compound B11) and in row 4 (an active mixture because of compound C10). A similar scenario exists for compound B12. The deconvolution program reported that all 4 mixtures (R3, R4, C3, and C4) were active in the assay, but could not decipher which combination of active mixtures constituted this activity. Therefore there was a configuration artifact which generated false positives. All four compounds had to be retested and only half (compounds in B11 and C10) were confirmed. This scenario becomes more complex as hit rates increased (Figure 4b). For effective screening, primary hit rates for 8:1 compressed plates should be <1% and the plate to plate hit rate variability should also be low.

Figure 4a. Configuration dilemma leading to false positives.

Figure 4b. Computer simulation deriving the % of false positives generated with increasing hit rate in compressed plates.

In addition to false positives generated with increasing hit rates, the presence of several compounds in a mixture also increased the likelihood of additive, synergistic, or antagonistic interactions^(2,3). Such interactions (in addition to pipetting errors) resulted in spurious hits, where a single row or column hit on a plate occurred without the corresponding column or row hit. In the absence of a correspondingly active mixture, this spuriously active well would not be flagged as a hit and would not need to be reconfirmed. The presence of several compounds in a mixture may reduce assay sensitivity, or the ability to identify an active compound within a mixture. Assay sensitivity was tested using 76 mixtures from an 8:1 compressed plate library using a receptor binding assay for serotonergic 5-HT_2C ligands (Figure 5). Hits were identified as compounds producing >50% inhibition of specific radioligand binding⁽⁴⁾. However, each of the 76 mixtures gave <50% inhibition in the assay and would not have been flagged as active in a screen. Mixtures were then spiked with 5-HT_2C ligands of varying affinity; MK212, mCPP, or metergoline⁽⁵⁾. In each of the inactive compound mixtures, these ligands were detectable, indicating that unknown mixtures do not adversely affect assay sensitivity. Additionally, the assay was able to distinguish between low, moderate, and high responses with compounds of varying affinity. Similar results were obtained using serotonergic 5-HT₇ and dopaminergic D₄ receptor binding assays (data not shown).

Figure 5. Measure of SPA 5-HT_2C receptor binding assay sensitivity.

Following compressed plate SPA assay development and validation, a screen was conducted for serotonergic 5-HT_2C ligands. Hits were identified as compounds producing >50% inhibition of specific radioligand binding. Fresh samples of these putative actives were obtained and tested individually. Confirmed active compounds were further evaluated for potency and receptor selectivity.

Results

Primary Screen

• 107,644 compounds were screened
• 88,744 compounds were screened in 8:1 compressed plates
• 18,900 compounds were screened in singleton plates (new compound acquisitions were not available in compressed plate format) Primary/Putative Hits
• 862 primary actives were identified (0.8% primary hit rate)
• 752 primary actives from compressed plates (0.8% compressed plate primary hit rate)
• 110 primary actives from singleton plates (0.6% singleton plate primary hit rate)

Confirmed Hits

• 369 compounds confirmed as actives (43% confirmation rate)
• 269 confirmed compressed plate hits (36% compressed plate confirmation rate)
• 100 confirmed singleton plate hits (91% singleton plate confirmation rate)

Final Hit Rate

• Overall confirmed hit rate for the screen: 0.3%e
• Overall compressed plate hit rate: 0.3%
• Overall singleton plate hit rate: 0.5%

The 36% confirmation rate for primary actives in compressed plates seems remarkably low in comparison to the 91% confirmation rate for hits identified in singleton plates. Later screens for serotonergic 5-HT₇ ligands and dopaminergic D₄ ligands revealed similar hit rates for compressed and singleton plate sets.

The low compressed plate confirmation rates resulted from the complexity of the plate configuration and the deconvolution process. The computer simulation in Figure 4b illustrated the relationship between hit rates and false positives; the 0.8% primary hit rate in the 5-HT_2C screen generated the predicted number of false positives.

Although 862 putative actives had to be retested in order to confirm 369 active compounds, compressed plates still provided considerable savings in time and money. The primary screening and confirmation of hits from the 88,744 compounds tested in compressed plates required 164 assay plates. In singleton plate format, 986 plates would have been required to assay the same compounds. This reflects better than 7-fold savings in time, reagents, plates, and compounds for assays using 8:1 compressed plates.

With primary hit rates of <1%, follow-up and data analyzis were manageable. Compounds discovered in the screen can not yet be disclosed, the identification of several known commercial 5-HT_2C ligands in this assay (Table 1) further validates the feasibility of screening successfully with compressed plates.

Table 1. Known compounds identified in the 5-HT_2C SPA screen.

Conclusions

Testing compounds as mixtures in SPA receptor binding assays can dramatically reduce screening effort.

Additionally 8:1 compressed plate formats can further reduce screening effort by more than 7-fold when screening at low compound concentrations, when compound mixtures do not adversely affect assay sensitivity, and when hit rates are low. The deconvolution program identified false positives due to complexity of the compressed plate format, therefore all putative hits from mixtures had to be confirmed as singletons. It worked well when hit rates and plate to plate variability are low, such that the complexity of analyzis and follow-up were manageable. Although confirmation rates are likely to be much lower for compressed plate hits than those identified in singleton plates, the savings achieved via combination of SPA technology and compressed plate formats are substantial. Compressed plates can therefore provide an excellent format for increasing assay throughput when assays are not easily adaptable to higher density plate formats.

Source: las.perkinelmer.com