NEW YORK – The Broad Institute's David Liu and colleagues from the Broad, Harvard University, and Boston Children's Hospital published a study in Nature Biotechnology yesterday describing their development of a system for phage-assisted continuous evolution of base editors in order to improve their editing efficiency and target sequence compatibility.
The researchers then used this BE-PACE system to evolve cytosine base editors (CBEs) that could overcome target sequence constraints of canonical CBEs. One of these evolved CBEs, named evoAPOBEC1-BE4max, was up to 26-fold more efficient at editing cytosine in the GC context — a disfavored context for wild-type APOBEC1 deaminase, which is used to modify targets — while maintaining efficient editing in all the other sequence contexts the researchers tested. Another evolved deaminase, named evoFERNY, was 29 percent smaller than APOBEC1 and edited efficiently in all the sequence contexts that were tested.
One factor that limits the efficiency of the most commonly used CBEs is the native sequence context preference of APOBEC1, which deaminates GC motifs poorly, the researchers noted. A GC target positioned in the center of the editing window may be edited efficiently by APOBEC1-based CBEs, but other GCs are not. Previous studies have shown that CBEs incorporating different cytidine deaminases can edit GC targets more efficiently, but non-APOBEC1-CBE alternatives have shown lower average performance than APOBEC1-CBE across a variety of targets in human cells.
Given these challenges, the researchers attempted to use phage-assisted continuous evolution (PACE) — which performs dozens of generations of mutation, selection, and replication per day — to generate editors with improved target sequence context compatibility and higher activity. During PACE, an activity of interest is coupled to the propagation of M13 bacteriophages that encode a biomolecule with that activity. In the BE-PACE system, this circuit must be activated by a single-base change, respond to small numbers of editing events, and turn on rapidly enough to support phage propagation under continuous dilution. To meet these requirements, the team designed a gene circuit in which cytosine deamination occurred on the transcription template strand to revert an inactivating mutation in a protein.
The BE-PACE circuit resulted in a nearly ten-fold overnight phage propagation in a host-cell culture and more than 1,000-fold selectivity for base editor phages, the researchers said. In further validation experiments, they found that BE-PACE generated improved deaminases.
To address the sequence context limitations of APOBEC1, the investigators constructed a low-stringency circuit with a target that required editing of a GC to support maximal phage propagation. After successive PACE circuits, they isolated mutant phages that showed weak but measurable activity on the GCC1 circuit. This led to top-performing phage clones that showed up to 28-fold improvements in apparent activity when tested on the GCC1 circuit in a luciferase assay.
To determine whether the apparent activity improvements in the luciferase assay translated into improved mammalian cell base editing, the team then subcloned a panel of evolved deaminase variants into the architecture of the BE4max base editor and transfected them into HEK293T cells, along with guide RNAs targeting five genomic sites previously shown to undergo efficient editing.
"Under optimal plasmid dosing and conditions, we observed that editing efficiency at the center of the activity window reaches a maximum of about 60 to 80 percent, likely limited by CBE-independent factors such as transfection efficiency or cellular DNA repair processes," the authors wrote. "Notably, editing at positions away from the center of the activity window was improved for all evolved BEs, and editing values at these positions correlated with luciferase assay activity. Among APOBEC1 CBEs, evolved mutations including H122L and D124N resulted in a striking improvement in base editing of GC targets."
Based on these results, the researchers selected one high-performing evolved variant of each deaminase to characterize in depth: evoAPOBEC1, evoFERNY, and evoCDA1. They created BE4max variants and tested their editing activity, and characterized a panel of 24 CBE variants to dissect the roles of evolved mutations. They found that the critical H122L and D124N mutations evolved during APOBEC1 PACE are present in less than 1 percent of the 1,189 naturally occurring APOBEC sequences, and often co-occur.
"EvoAPOBEC1-BE4max dramatically outperforms the state-of-the-art CBEs BE4max and AncBE4max at GC targets, while showing similar or higher activity at non-GC targets," the authors concluded. "The base editing window of evoAPOBEC1-BE4max is very similar to that of BE4max. Despite its 29 percent smaller size compared to APOBEC1, evoFERNY performs comparably to APOBEC1 in CBEs at non-GC targets and more effectively on GC targets. EvoFERNY-BE4max shows further improvement on GC targets, giving similar or higher editing levels compared to evoAPOBEC1-BE4max."
However, the researchers added that in all their experiments, the evoCDA1-BE4max showed increased off-target editing. Therefore, they recommended that it be used in carefully chosen applications that would be most suited to its characteristics. "At well-edited sites, this CBE generates higher indel levels without increasing editing beyond plateau levels, instead showing an expanded window," they wrote. "By contrast, at poorly edited sites, it shows increased in-window editing levels without increased indels. These considerations suggest that evoCDA1-BE4max should be applied when off-target and bystander editing are not concerns and high efficiency is paramount."