Highly active zinc-finger nucleases by extended modular assembly

Abstract

Zinc-finger nucleases (ZFNs) are important tools for genome engineering. Despite intense interest by many academic groups, the lack of robust noncommercial methods has hindered their widespread use. The modular assembly (MA) of ZFNs from publicly available one-finger archives provides a rapid method to create proteins that can recognize a very broad spectrum of DNA sequences. However, three- and four-finger arrays often fail to produce active nucleases. Efforts to improve the specificity of the one-finger archives have not increased the success rate above 25%, suggesting that the MA method might be inherently inefficient due to its insensitivity to context-dependent effects. Here we present the first systematic study on the effect of array length on ZFN activity. ZFNs composed of six-finger MA arrays produced mutations at 15 of 21 (71%) targeted loci in human and mouse cells. A novel drop-out linker scheme was used to rapidly assess three- to six-finger combinations, demonstrating that shorter arrays could improve activity in some cases. Analysis of 268 array variants revealed that half of MA ZFNs of any array composition that exceed an ab initio B-score cutoff of 15 were active. These results suggest that, when used appropriately, MA ZFNs are able to target more DNA sequences with higher success rates than other current methods.

Figure 1.

The drop-out linker scheme. SphI, BsmI, and HindIII sites were introduced into the zinc-finger coding region as silent mutations. Digestion with one restriction enzyme followed by ligation allows the full set of arrays to be created in parallel in one day.

Figure 2.

Activity of the CS series of ZFNs determined by a SSA assay. The number of fingers in the left and right arrays for each ZFN are indicated. Note that the CS3-1, 5-1, 6-1, and 6-2 series started with four- or six-finger arrays instead of three-finger arrays to reduce the assembly effort required prior to the invention of the drop-out linker strategy. Based on the data, the missing ZFNs seemed likely to be inactive and were not tested subsequently. (-) SSA reporter only as a negative control. (Ctrl) GZF3-L3 + GZF1-R3 control nuclease to which all activities are normalized. (Error bars) The standard deviation of normalized duplicates from at least two experiments. (*) P < 0.01 compared to the negative control based on a one-tailed homoscedastic t-test. (Black arrows) ZFNs used in genomic cleavage assays.

Figure 3.

Functional selection of ZFNs using the Surveyor assay in HEK293T cells. ZFN CS2-1-L5 + R3, as a positive control, and all 16 combinations of ZFN T2-X6 were assayed in duplicate. The number of fingers in the left and right arrays for each ZFN are indicated above each pair of lanes. Numbers below each lane indicate the percentage of modified alleles averaged over the duplicates. (Arrows) Appearance and position of the Surveyor cleavage band. (Black box) ZFN pair T2-X6 L5 + R4 that was used for Illumina sequencing analysis.

Figure 4.

The ab initio B-score as a predictor of relative affinity and activity of MA ZFNs. (A) The B-scores are shown for each Barbas zinc-finger module (Bhakta and Segal 2010). B-scores were summed and compared to the measured affinities of (B) seven three-finger arrays (Sander et al. 2009); (C) 16 three-finger arrays (Segal et al. 1999); and (D) six six-finger arrays of mixed composition (MS Kim et al. 2011; Shimizu et al. 2011). (E) Receiver operating characteristic (ROC) analysis of four predictors of ZFN activity (≥8% SSA activity) was performed for the 92 array variants of the CS ZFN series using the Daim package of the R statistics program. (Comb.B) The combined B-score of the modules in the left and right arrays. (Comb.GNN) The combined number of GNN modules. (Comb.fns) The combined number of fingers. (Comb.ddG) The combined ΔΔG of the GNN modules (Sander et al. 2009).

Figure 5.

Influence of “disruptive” interfinger linkers on extended-MA ZFN activity. ZFN activity was determined by the SSA assay for ZFNs with long linkers (TGSQKP) inserted between fingers 2 and 3 and fingers 4 and 5. (Std Link) All canonical linkers (TGEKP). The array configuration for the Std Link ZFN was GZF3-L3 + GZF1-R3, CS2-1 L5 + R3, CS5-1 L4 + R4, CS6-2 L5 + R4, CS7-3 L4 + R6, and T2-X6 L5 + R4.

Figure 6.

An analysis of affinity, specificity, and cytotoxicity. (A) EMSA was used to determine the affinity of CS2-1 three- and five-finger arrays for their specific (cognate) target as well as a nontarget (e.g., the L3 array on the R3 binding site). A large ratio of nontarget:specific binding is an indicator of good specificity. (B) In vitro binding specificity was also determined using the Bind-n-Seq target site selection assay (Zykovich et al. 2009). The binding preference of CS2-1-L3 appears to differ from the intended target site at three positions (red boxes). (C) Cytotoxicity was assessed as a decrease in the percentage of ZFN-expressing cells on day 5 compared to day 1. To follow only those cells expressing nuclease, a GFP expression vector was cotransfected with the indicated ZFN expression vectors. (GFP) Cells cotransfected with GFP and empty ZFN vector as a positive control. (GZF3-wt) A nuclease known to be cytotoxic (Szczepek et al. 2007) as a negative control. (Error bars) The standard deviation of normalized duplicates from at least two experiments. (*) P < 0.00001 compared to the GFP-only positive control based on a one-tailed homoscedastic t-test. (D) Cleavage activity is a summary of the SSA data from other figures and is shown here for reference.

Figure 7.

The utility of the B-score for extended-MA. Distribution of combined B-scores for successful (red bars) and unsuccessful (blue bars) ZFNs based on the 268 array combinations in this study. The optimal cutoff of 15 is indicated (green line).

Figure 8.

Spectrum of sequences targetable using extended MA. (A) An image from the UCSC Genome Browser shows potential extended-MA and CoDA ZFNs sites found within a 60-kb region at the human 9p21 locus (hg18, chr9:22060000-22119999). Each vertical black bar represents a target site. (B) A comparison of the transcript targeting capabilities reported for CoDA and calculated from the data in this study for extended MA.