In-cell DEL screening delivers hits against a helicase-like polymerase domain not amenable to in vitro selection

The helicase-like domain addressed in this study belongs to a DNA polymerase with an established role in the DNA damage response and is the subject of substantial therapeutic interest in oncology. It is also widely regarded as a difficult target for ligand discovery.

Affinity-based DEL selection in vitro requires purified protein that has been labelled with DNA. A selection of this type was performed on the purified target in plastic, but no hits were obtained.

An outcome of this kind is difficult to interpret and readily misattributed. A target that is not presented in a well-defined, native-like state during selection may appear to lack ligandable chemical matter when the limitation lies in how the selection was configured. A cell-based selection format was therefore adopted.

The selection was repeated using cellular Binder Trap Enrichment® (cBTE), a method for conducting DNA-encoded library selections inside living cells. The approach removes the requirement for purified protein, and binding is evaluated against target folded by the cellular machinery.

The target was expressed and specifically DNA-labelled in living Xenopus laevis oocytes. Lib054, a YoctoReactor®-generated DNA-encoded library of 407 million compounds, was microinjected directly into the cytosol, where binding proceeded in a crowded and physiologically relevant environment. Genuine binders were trapped together with the target, while non-binders remained unassociated and contributed only low-count background.

Library properties (Lib054)

• 407 million compounds, trimeric YoctoReactor format, with complete correspondence between barcode and compound
• 1,271 building blocks covering 349 chemical scaffolds
• Mean physicochemical properties: MW 541 Da, cLogP 0.95, Fsp3 0.54, HBA 6, HBD 3
• 99.7% of compounds in compliance with Lipinski’s rule of five

Selection was performed in a single round against four target constructs. Recovered library DNA was amplified, sequenced and decoded, and the number of observations per compound was determined. A mathematical threshold of eight or more observations (MT = 8+) was applied to the resulting signal distribution, separating hits from random background and yielding 138 primary hits.

The 138 hits were analyzed as a pairwise Tanimoto similarity matrix and organized by hierarchical clustering. Nine distinct chemical series were resolved (blocks 1–9, Figure 1), together with a number of singletons, providing several structurally independent starting points rather than a single chemotype.

Twenty-seven compounds, selected to represent the identified series and the more promising singletons, were resynthesized off DNA and characterized in biochemical assays. Five were confirmed as active: three members of chemical series 3 and two singletons. Three of the five showed sub-micromolar potency, with the most potent compound below 100 nM. Confirmed actives were further characterized by orthogonal biophysical methods, including surface plasmon resonance (SPR) and differential scanning fluorimetry (DSF).

The confirmation of three actives within a single clustered series indicates that related analogues were already present among the primary hits, which is of practical value for subsequent structure–activity analysis.

Both selections were performed against the same target using the same library. The in vitro selection on purified protein returned no hits, while the in-cell selection returned nine chemical series and confirmed actives extending into the nanomolar range. The difference is attributable to the conditions under which the target was presented to the library rather than to the chemical tractability of the target itself.

This distinction is of particular relevance to proteins that are difficult to prepare in a well-defined state for in vitro selection, including helicases, polymerases, repair enzymes and nucleic-acid-binding proteins more generally. Because cBTE requires no purified target protein, the preparative steps concerned are absent from the workflow.

Key points

• An unsuccessful in vitro selection does not establish that a target lacks ligandable chemical matter, as the outcome may reflect the conditions of target presentation.
• cBTE requires neither purification nor immobilization, and is therefore applicable to targets that are difficult to prepare in a defined, native-like state.
• Nine chemical series obtained from a single round of selection provide multiple independent starting points for medicinal chemistry.
• Sub-micromolar and sub-100 nM potencies were obtained from primary hits without optimization.