Zinc fingers are a class of DNA-binding proteins that were so named “because they contained zinc and gripped or grasped DNA” as linear, repeating units like fingers on a hand.1 The canonical Cys2His2 type of zinc finger domain is characterized by an antiparallel ß-sheet linked to an α-helix. Amino acid residues on the α-helix (commonly denoted as position -1, 1, 2, 3, 5 and 6) are known to confer binding specificity to three to four bases of DNA.2,3 Each ~30 amino-acid-long finger can be linked in tandem with other fingers to bind longer sequences of DNA, or with other proteins such as nucleases and activation/repression domains to modulate gene expression upon binding.
These handy proteins are highly abundant, comprising an estimated 3% of genes in the human genome.1 Zinc fingers are also the most frequently occurring DNA-binding domain observed in human transcription factors.4 Their potential for targeting any DNA sequence coupled with their small size and natural occurrence make zinc finger proteins an attractive gene therapy modality. Despite a firm understanding of single-finger design principles and the development of user-friendly prediction tools (see ZF Tools,5 for example), multi-finger design remains challenging because DNA binding specificity and affinity can be greatly affected by the presence of adjacent fingers. In this SNiP, Ichikawa et al. applied machine learning in hopes of developing a proverbial ‘Zinc Finger Rosetta Stone’ for designing zinc fingers capable of targeting any genomic sequence.
Rather than trying to capture the binding preferences of all possible permutations of multiple zinc fingers in tandem–which would be technically impossible with current technology considering each finger could be modified at six helical positions with 20 amino acids–Ichikawa et al. used an elegant strategy that focused on examining the impact of the zinc finger to zinc finger interface on DNA binding. Initial training was performed on data from 10 three-finger libraries where the six helical residues of two fingers was held constant, but that of the C-terminal third finger was completely randomized. Each library was screened 12 times against the complete 64 possible NNN DNA sequence targets. The final model incorporated data from an additional 200 two-finger libraries, screened against 6-bp targets! Read the publication to see how the resultant “ZFDesign” model performed for creating sequence-specific zinc finger nucleases, activators and repressors.
Title: A universal deep-learning model for zinc finger design enables transcription factor reprogramming
Authors: David D. M. Ichikawa, Osama Abdin et al.
Journal: Nature Biotechnology, Preprint, 26 Jan 2023.
DOI: 10.1038/s41587-022-01624-4
Product Usage: Transcriptional activator KLF6 was “reprogrammed” with six zinc fingers that were predicted by ZFDesign to bind a tetO reporter sequence. A plasmid encoding this reprogrammed transcription factor was transfected into a HEK 293T reporter cell line using TransIT®-LT1 Transfection Reagent.
Query our Citations Database to discover more ways TransIT® transfection propels zinc finger research.
References
The TransMission
Feedback or questions? We’d love to hear from you. Email techsupport@mirusbio.com or call us at 888.530.0801.