Inverse drug screens: a rapid and inexpensive method for implicating molecular targets

Abstract

Identification of gene products that function in some specific process of interest is a common goal in developmental biology. Although use of drug compounds to probe biological systems has a very long history in teratology and toxicology, systematic hierarchical drug screening has not been capitalized upon by the developmental biology community. This "chemical genetics" approach can greatly benefit the study of embryonic and regenerative systems, and we have formalized a strategy for using known pharmacological compounds to implicate specific molecular candidates in any chosen biological phenomenon. Taking advantage of a hierarchical structure that can be imposed on drug reagents in a number of fields such as ion transport, neurotransmitter function, metabolism, and cytoskeleton, any assay can be carried out as a binary search algorithm. This inverse drug screen methodology is much more efficient than exhaustive testing of large numbers of drugs, and reveals the identity of a manageable number of specific molecular candidates that can then be validated and targeted using more expensive and specific molecular reagents. Here, we describe the process of this loss-of-function screen and illustrate its use in uncovering novel bioelectrical and serotonergic mechanisms in embryonic patterning. This technique is an inexpensive and rapid complement to existing molecular screening strategies. Moreover, it is applicable to maternal proteins, and model species in which traditional genetic screens are not feasible, significantly extending the opportunities to identify key endogenous players in biological processes.

FIG. 1

This diagram illustrates arrangement of possible targets (in this case, a subset of ion transporters) in a hierarchical tree. The groups are arranged from the broadest classes at the top to more specific subfamilies at the bottom. The individual colors of each element serve only to demarcate different levels (hierarchical layers) of the tree.

FIG. 2

Sample decision trees for testing the involvement of calcium (A) and chloride (B) signaling in a POI. Panel (a) illustrates the detailed logic for traversing a data tree to identify a calcium conductance. Panel (b) simply illustrates the corresponding drug and transporter trees for chloride.

FIG. 3

A more complex strategy for probing neurotransmitter effects in a POI. Oval nodes represent specific functional modules (proteins or physiological conditions) whose involvement is to be tested. For reasons of space, lower tiers have been left out of this diagram. For example, the serotonin receptor 1 family is subdivided into types 1a, 1b, 1d, 1e, 1f, 1p, and 1s. Green labels indicate an example of reagents that can be used to test the node to which they are attached. Arrows between two nodes indicate suggested screen paths to be taken depending on the outcome of each result. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]