Golden gate shuffling: a one-pot DNA shuffling method based on type IIs restriction enzymes

Abstract

We have developed a protocol to assemble in one step and one tube at least nine separate DNA fragments together into an acceptor vector, with 90% of recombinant clones obtained containing the desired construct. This protocol is based on the use of type IIs restriction enzymes and is performed by simply subjecting a mix of 10 undigested input plasmids (nine insert plasmids and the acceptor vector) to a restriction-ligation and transforming the resulting mix in competent cells. The efficiency of this protocol allows generating libraries of recombinant genes by combining in one reaction several fragment sets prepared from different parental templates. As an example, we have applied this strategy for shuffling of trypsinogen from three parental templates (bovine cationic trypsinogen, bovine anionic trypsinogen and human cationic trypsinogen) each divided in 9 separate modules. We show that one round of shuffling using the 27 trypsinogen entry plasmids can easily produce the 19,683 different possible combinations in one single restriction-ligation and that expression screening of a subset of the library allows identification of variants that can lead to higher expression levels of trypsin activity. This protocol, that we call 'Golden Gate shuffling', is robust, simple and efficient, can be performed with templates that have no homology, and can be combined with other shuffling protocols in order to introduce any variation in any part of a given gene.

Figure 1. DNA shuffling strategy.

(A) Two DNA ends terminated by the same 4 nucleotides (sequence f, composed of nucleotides 1234, complementary nucleotides noted in italics) flanked by a BsaI recognition sequence, B, form two complementary DNA overhangs after digestion with BsaI. (B) For shuffling, genes of interest are aligned, and recombination points consisting of 4 nucleotide sequences (f1 to fn+1) are defined on conserved sequences. Module fragments (core sequence, C1 to Cn, plus flanking 4 nucleotide sequences) are amplified by PCR and cloned in an intermediate cloning vector. Module fragment plasmids and the acceptor vector are assembled in one restriction-ligation with BsaI and ligase. S1 and S2, two different selectable markers. Z, lacZ alpha gene fragment.

Figure 2. Assembly of a GFP construct from 10 plasmids.

(A) Construct maps. Input modules contain a core region C flanked by BsaI restriction sites in opposite orientation composed of a recognition site (B, ggtctcn, B, ngagacc) and a 4 nucleotide cleavage site (boxes flanking the core region). pX-LacZ, acceptor vector. pGFPi, resulting construct. Restriction sites for AvrII and XmaI are shown as white arrows. (B) Ethidium bromide-stained gel with products obtained by restriction-ligation of the 9 input module plasmids. M: GeneRuler 1kb DNA Ladder Plus from Fermentas. Restriction-ligation was performed at 37°C for 3 (lane 3h) or 6 hours (lane 6h) or with 25 cycles (2 min 37°C+5 min 16°C, lane 25) or 50 cycles (lane 50), and without BsaI enzyme (lane nb). The arrow indicates the 1.17 kb linear assembled GFP gene product. (C) Ethidium bromide-stained gels of 72 minipreps digested with XmaI and AvrII (expected fragment sizes: 4.6 kb, 945 and 555 bp), obtained from restriction-ligations performed for 6 h 37°C (6 h), for 25 or 50 cycles (25×/50×), with normal ligase (nl) or high concentration ligase (hcl). Numbers indicate minipreps with an incorrect restriction pattern, and stars indicate constructs that consist of dimers (same restriction pattern as monomers). V, vector pX-lacZ.

Figure 3. Structure of incorrect GFP constructs and model for their formation.

Only the portion between modules 5 and 9 is shown. An additional inserted module (module 7) is shown in pink. Ligation of two DNA ends despite a mismatch in the overhangs leads to a plasmid that can be repaired or segregated in two different sequences. Both alternatives were in fact observed in sequenced plasmids pGFPi-24/34 and 64.

Figure 4. Shuffling of trypsinogen.

(A) Alignment of the aminoacid sequence of bovine cationic trypsinogen (BC), bovine anionic trypsinogen (BA) and human cationic trypsinogen (HC). Nucleotide sequence of the chosen recombination sites is shown. (B) Map of the 27 trypsinogen module plasmids, the acceptor vector, and of an example of one of the resulting shuffled construct obtained. B, BsaI restriction site. S, K: spectinomycin and kanamycin resistance genes. RB/LB, T-DNA right and left borders. AttB, Phage C31 recombination site, N tobamoviral 3′ non-translated region, T, Nos terminator. (C) Ethidium bromide-stained gels of 28 minipreps prepared from single colonies (1 to 24) or from 4 libraries (L1–4, approximately 700 clones in each) digested with XmaI (incorrect pattern 1, 2, 7 and 17).

Figure 5. Structure of incorrect trypsinogen constructs and models for their formation.

Additional inserted modules in clone ts2–40 are shown in pink, and a religated overhang with 3 nucleotides shown in green. For this clone, exonuclease removal of an A resulted in two complementary 3-nucleotide overhangs that could be ligated.

Figure 6. Activity assay of the shuffled trypsinogen constructs.

(A) Three constructs (in Agrobacterium) are coinfiltrated for each trypsinogen construct: the 5′ viral vector module (pICH30211), a trypsinogen construct, and an integrase construct (not shown). In planta recombination leads to formation of an assembled construct (1) which leads to viral expression of a fusion protein (2) containing a signal peptide (SP), Arabidopsis thaliana SUMO exons, and trypsinogen. The signal peptide is cleaved upon import through the ER (3), and trypsin is obtained by autocatalytic cleavage of the proprotein (red arrow). Grey boxes represent introns. (B) Activity and structure of some of the constructs obtained from the first round of shuffling (name, column 1 and activity, column 2), activity expressed relative to activity of bovine cationic trypsinogen (BC). Activity for the parents (BA, HC, BC) was also measured (from corresponding constructs infiltrated as a control). GFP is used as a negative control. The last 3 constructs (boxed) were used for a second round of shuffling. (C) Best construct obtained with the second round of shuffling.