Multigene Editing in the Escherichia coli Genome via the CRISPR-Cas9 System

Abstract

An efficient genome-scale editing tool is required for construction of industrially useful microbes. We describe a targeted, continual multigene editing strategy that was applied to the Escherichia coli genome by using the Streptococcus pyogenes type II CRISPR-Cas9 system to realize a variety of precise genome modifications, including gene deletion and insertion, with a highest efficiency of 100%, which was able to achieve simultaneous multigene editing of up to three targets. The system also demonstrated successful targeted chromosomal deletions in Tatumella citrea, another species of the Enterobacteriaceae, with highest efficiency of 100%.

INTRODUCTION

Metabolic engineering is widely applied to modify Escherichia coli to produce industrially relevant biofuels or biochemicals, including ethanol (1), higher alcohols (2), fatty acids (3), amino acids (4), shikimate precursors (5), terpenoids (6), polyketides (7), and polymeric precursors of 1,4-butanediol (8). An important example of a successful metabolic engineering project is the modification of E. coli to produce 1,3-propanediol, which was developed by Genencor and DuPont (9) and led to a commercial process. This industrially optimized strain required up to 26 genomic modifications, including insertions, deletions, and regulatory modifications. Such large numbers of genome editing targets require efficient tools to perform time-saving sequential manipulations or multiplex manipulations.

A wide variety of tools for targeted gene editing, which can be classified into homologous recombination and group II intron retrohoming, are available for E. coli (10, 11). The efficiency of introduction of mutations mediated by homologous recombination can be improved (i) by using counterselection markers, such as the typical sacB-based method (12), and (ii) by improving the frequency of homologous recombination by using phage-derived recombinases (RecET and λ-Red) (13–15), applying double-stranded (16, 17) or single-stranded donor DNAs (18), or inducing double-stranded breaks (DSBs) in a chromosomal target using I-SceI (12, 19, 20). The λ-Red recombinase method (13) and group II intron retrotransposition (21) leave scars in the genome that limit their application in allelic exchange. Of all the methods mentioned above, only single-stranded-DNA (ssDNA)-based gene modification mediated by λ-Red was further developed as a multiplex genome editing tool, known as multiplex automated genome engineering (MAGE) (22, 23), which greatly facilitates genome-scale engineering. However, the short ssDNA oligonucleotide-mediated MAGE has advantages in allelic exchange-based genome mutation but has challenges regarding targeted multiple gene insertions over a certain length (22).

The clustered regularly interspaced short palindromic repeats–CRISPR-associated system (CRISPR-Cas system) was used recently as efficient genome engineering technology in several prokaryotes and eukaryotes, including (but not limited to) E. coli (24), Saccharomyces cerevisiae (25), Streptomyces spp. (26), higher plants (27), Bombyx mori (28), Drosophila (29), and human cell lines (30–32). The type II CRISPR-Cas system from Streptococcus pyogenes uses a maturation CRISPR RNA (crRNA) and trans-activating crRNA (tracrRNA) guiding the nuclease Cas protein 9 (Cas9) to the target of any DNA sequence, known as a protospacer, with a protospacer-adjacent motif (PAM) present at the 3′ end (NGG in the case of S. pyogenes, where N represents any nucleotide) (33). In genome editing cases, the 20-bp complementary region (N₂₀) with the requisite NGG PAM matching genomic loci of interest was programmed directly into a heterologously expressed CRISPR array, and fused crRNA and tracrRNA as a single synthetic guide RNA (sgRNA) transcript obviated the need for processing the transcribed CRISPR array (pre-crRNA) into individual crRNA components (31).

In E. coli, the CRISPR-Cas9 system has been demonstrated to apply allelic exchange with efficiency as high as 65% ± 14% (24) and to control gene expression via a nuclease-deficient Cas9 protein (34, 35). No detailed method for applying the CRISPR-Cas9 system in precise genome editing, including gene insertions and knockouts, has been published. Therefore, we developed a CRISPR-Cas9 system-based continual genome editing strategy, including gene insertions and knockouts of both single and multiple (up to three) targets, and expanded the system to include Tatumella citrea, another species of the Enterobacteriaceae, for continual gene deletions.

MATERIALS AND METHODS

Strains, plasmids, and growth conditions.

The bacterial strains and plasmids used in this study are given in Table 1. E. coli DH5α was used as a cloning host, and E. coli MG1655 or T. citrea DSM 13699 was used in the genome engineering procedures. The genomic DNA of S. pyogenes strain MGAS5005, kindly provided by Xuesong Sun of Jinan University (Guangdong, China), was used to amplify the cas9 gene. E. coli or T. citrea (36, 37) was grown in LB medium (1% [wt/vol] tryptone, 0.5% [wt/vol] yeast extract, 1% [wt/vol] NaCl) at 37°C or 30°C. Ampicillin (100 mg/liter), kanamycin (50 mg/liter), spectinomycin (50 mg/liter), or chloramphenicol (25 mg/liter) was added as needed.

TABLE 1.

Strains and plasmids used in this study^a

Strain or plasmid	Characteristics	Source or reference
Strains
E. coli DH5α	F− endA1 glnV44 thi-1 recA1 relA1 gyrA96 deoR nupG ϕ80dlacZΔM15 Δ(lacZYA-argF)U169 hsdR17 (rK− mK+) λ−	TaKaRa
E. coli MG1655	K-12; F− λ− rph-1	CGSC 6300
Streptococcus pyogenes MGAS5005	M1 serotype, wild type	ATCC BAA-947
Tatumella citrea
DSM 13699		DSMZ
MGly1	MG1655 ΔcadA::cat	This study
MGly2	MGly1 Δcat::cadAp	This study
1655ΔcadA	MG1655 ΔcadA	This study
1655ΔcadAΔmaeAΔmaeB	MG1655 ΔcadA ΔmaeA ΔmaeB	This study
1655ΔmaeAΔmaeB	MG1655 ΔmaeA ΔmaeB	This study
1655ΔyjcS::ybaS	MG1655 ΔyjcS::ybaS	This study
1655ΔyjcS::evgAS	MG1655 ΔyjcS::evgAS	This study
1655ΔmaeB::gltPΔmaeA	MG1655 ΔmaeB::gltP ΔmaeA	This study
13699ΔtkrA	DSM 13699 ΔtkrA	This study
13699ΔtkrAΔglk	DSM 13699 ΔtkrA Δglk	This study
General plasmids
pKD46	repA101(Ts) bla araC ParaB-Red	13
pSU2718	p15A cat	52
pTrc99A	pMB1 bla lacIq	53
pIJ778	bla aadA FRT	54
pKD46K	repA101(Ts) kan araC ParaB-Red	20
pTrc99A-spec	pMB1 aadA lacIq	This study
Crispr-Cas9 system plasmids
pCas series
pCB001	p15A cm Pcas-cas9	This study
pCas	repA101(Ts) kan Pcas-cas9 ParaB-Red lacIq Ptrc-sgRNA-pMB1	This study
pCasΔcas9	repA101(Ts) kan ParaB-Red lacIq Ptrc-sgRNA-pMB1	This study
pTarget series harboring sgRNAs, with or without donor DNAs
pTargetF-pMB1	pMB1 aadA sgRNA-pMB1	This study
pTargetF-cadA	pMB1 aadA sgRNA-cadA	This study
pTargetF-kefB	pMB1 aadA sgRNA-kefB	This study
pTargetF-yjcS	pMB1 aadA sgRNA-yjcS	This study
pTargetF-cat	pMB1 aadA sgRNA-cat	This study
pTargetF-kefB-yjcS	pMB1 aadA sgRNA-kefB, sgRNA-yjcS	This study
pTargetF-cadAp	pMB1 aadA sgRNA-cadAp	This study
pTargetTΔR-ΔcadAp	pMB1 aadA ΔcadAp (430 bp)	This study
pTargetT-ΔcadAp	pMB1 aadA sgRNA-cadAp ΔcadAp (430 bp)	This study
pTargetT-ΔcadA	pMB1 aadA sgRNA-cadA ΔcadA (804 bp)	This study
pTargetT-ΔmaeA	pMB1 aadA, sgRNA-maeA ΔmaeA (708 bp)	This study
pTargetT-ΔmaeB	pMB1 aadA sgRNA-maeB ΔmaeB (829 bp)	This study
pTargetT-ΔmaeAΔmaeB	pMB1 aadA sgRNA-maeA, ΔmaeA (708 bp) sgRNA-maeB ΔmaeB (829 bp)	This study
pTargetT-ΔcadAΔmaeAΔmaeB	pMB1 aadA sgRNA-cadA ΔcadA (804 bp) sgRNA-maeA ΔmaeA (708 bp) sgRNA-maeB ΔmaeB (829 bp)	This study
pTargetT-ΔyjcS::ybaS	pMB1 aadA sgRNA-yjcS ΔyjcS (733 bp):: ybaS (1.3 kb)	This study
pTargetT-ΔyjcS::evgAS	pMB1 aadA sgRNA-yjcS ΔyjcS (733bp):: evgAS (4.5 kb)	This study
pTargetT-ΔmaeB::gltP	pMB1 aadA sgRNA-maeB ΔmaeB (829 bp)::gltP (1.7 kb)	This study
pTargetT-ΔmaeB::gltP ΔmaeA	pMB1 aadA sgRNA-maeB ΔmaeB (829 bp)::gltP (1.7 kb) sgRNA-maeA ΔmaeA (708 bp)	This study
pTargetT-ΔtkrA	pMB1 aadA sgRNA-tkrA ΔtkrA (978 bp)	This study
pTargetT-Δglk	pMB1 aadA sgRNA-glk Δglk (963 bp)	This study

^abla, ampicillin resistance gene; kan, kanamycin resistance gene; aadA, spectinomycin resistance gene; cat, chloramphenicol resistance gene; P_cas-cas9, the cas9 gene with its native promoter; P_araB-Red, the Red recombination genes with an arabinose-inducible promoter; P_trc-sgRNA-pMB1, sgRNA with an N₂₀ sequence for targeting the pMB1 region with a trc promoter; sgRNA-cadA, sgRNA with an N₂₀ sequence for targeting the cadA locus; sgRNA-cadA^p, sgRNA with an N₂₀ sequence for targeting the partial cadA fragment inserted inside the heterologous cat loci of strain MGly2. ΔcadA (804 bp), editing template with an 804-bp region homologous to the cadA locus; ΔcadA^p (430 bp), editing template with a 430-bp region homologous to the partial cadA fragment inserted inside the heterologous cat loci of strain MGly2; ΔmaeA (708 bp), editing template with an 804-bp region homologous to the maeA locus; ΔyjcS (733 bp)::ybaS (1.3 kb), editing template with an 804-bp region homologous to the yjcS locus with a 1.3-kb ybaS insertion; ΔyjcS (733 bp)::evgAS (4.5 kb), editing template with a 733-bp region homologous to the yjcS locus with a 4.5-kb evgAS insertion; ΔmaeB (829 bp)::gltP (1.7 kb), editing template with an 829-bp region homologous to the maeB locus with a 1.7-kb gltP insertion.

Plasmid construction.

All constructs used in this study are given in Table 1, and the sgRNA, primer, and N₂₀ sequences followed by the PAM used in this study are given in Tables S1 and S2 in the supplemental material. Plasmids and genomic DNA were extracted using the AxyPrep kit (Corning) according to the manufacturer's instructions. PCR used the polymerases Taq (Thermo Scientific) and KOD-plus-neo (Toyobo). Restriction endonucleases and T4 DNA ligase were purchased from Thermo Scientific.

The two-plasmid system, in which the cas9 gene and the sgRNA directing it to the targeted region were separated in the pCas and pTarget series, was used for genome editing as shown in Fig. 1. pCas in the two-plasmid system consisted of cas9, λ-Red, a temperature-sensitive replicon, and the sgRNA with a lacI^q-P_trc promoter guiding the pMB1 replication of pTarget. pCB001 was constructed by amplifying the cas9 sequence and the native promoter from S. pyogenes MGAS5005 with primers pA001 and pA002, followed by ligation to pSU2718, which was digested with PstI/XbaI. The kanR-repA101(Ts) fragment containing the kanamycin-resistant gene kanR and the temperature-sensitive replicon repA101(Ts) were amplified from pKD46K (21) by primers pA006/pA007, the lacI^q gene and the P_trc promoter (lacI^q-P_trc fragment) were amplified from pTrc99A by pA008/pA009, and the sgRNA-pMB1 sequence was amplified from pTarget, the construction of which is described below, by pA010/pA011. The λ-Red recombinase gene was amplified from pKD46 by pA012/pA013, and digested by XbaI. pCas was constructed by ligating the cas9 cassette digested from pCB001 by PstI/XbaI, with the PstI/BglII-digested overlap PCR product of kanR-repA101(Ts), the lacI^q-P_trc fragment, and the XbaI/BglII-digested λ-Red gene.

FIG 1.

Construction of the CRISPR-Cas two-plasmid system. The cas9 gene and the sgRNA directing it to the targeted region were separated in pCas and pTarget series. (a) pCas contains the cas9 gene with a native promoter, an arabinose-inducible sgRNA guiding Cas9 to the pMB1 replicon of pTarget, the λ-Red recombination system to improve the editing efficiency, and the temperature-sensitive replication repA101(Ts) for self-curing. sgRNA is displayed with its secondary structure (51). (b) pTarget was constructed to express the targeting sgRNA, with (pTargetT series) or without (pTargetF series) donor DNA as editing templates. Cas9, Cas9 endonuclease; pJ23119, synthetic promoter (38); N20, 20-bp region complementary to the targeting region (38); araC, arabinose-inducible transcription factor; pKD46K, a form of pKD46 in which the bla gene is replaced with the aadA gene that confers kanamycin resistance (21); pTrc99A-spec, a form of pTrc99A, in which bla was replaced by aadA, which confers spectinomycin resistance.

The pTarget series had two versions, pTargetT and pTargetF, which had donor DNA for recombination supplied in the plasmid pTarget and not supplied, respectively (Fig. 1b). pTargetF consists of the sgRNA sequence, the N₂₀ sequence, and the multiple restriction sites, with the donor DNA supplied as fragments. pTrc99A-spec was constructed by ligating the MluI/XhoI-digested pTrc99A framework, including the pMB1 replicon amplified by pA003/pA056 from pTrc99A, with the spectinomycin-resistant gene aadA amplified by pA054/pA055 from pIJ778. The sgRNA sequence with promoter pJ23119 and the multiple restriction sites was synthesized de novo as described previously (35) (GenScript) and was inserted into NdeI/XhoI-digested pTrc99A-spec (Fig. 1). The pTargetF series, used in target single-gene modification with a targeting N₂₀ sequence of gene loci of interest, was obtained by inverse PCR with the modified N₂₀ sequence hanging at the 5′ ends of primers and followed by self-ligation (38). pTargetF-kefB-yjcS consisting of double sgRNAs was achieved by BioBrick cloning with BamHI and BglII (39). sgRNA-yjcS with its promoter was digested from pTargetF-yjcS with BamHI/BglII and inserted into the BglII-digested pTargetF-kefB. The pTargetT series consisted of the sgRNA sequence, N₂₀, the multiple restriction sites, and the donor DNA used as the genome editing template. The editing templates had a 250- to 550-bp sequence homologous to each side (upstream or downstream) of the targeted region in the genome. pTargetT-ΔcadA, pTargetT-ΔmaeB, and pTargetT-ΔmaeA were constructed by inserting the editing template through overlap PCR of the three fragments amplified by primers pB014/pB015, pB027/pB030, and pB016/pB017 to form upstream editing templates and pB029/pB028, pB058/pB059, and pB060/pB061 to form downstream editing templates from the MG1655 genome. The sgRNA fragment amplified by primers pB019/pB018, pB025/pB018, or pB053/pB018 from pTargetF was inserted into the SpeI/SalI-digested pTargetF. pTargetT-ΔmaeAΔmaeB was constructed by inserting the fragment amplified from pTargetT-ΔmaeA by pB062/pB063 into the SalI/BglII-digested pTargetT-ΔmaeB. pTargetT-ΔcadAΔmaeAΔmaeB was constructed by inserting the fragment amplified from pTargetT-ΔcadA by primers pB064/pB065 into the HindIII-digested pTargetT-ΔmaeAΔmaeB. pTargetT-ΔyjcS::ybaS or pTargetT-ΔyjcS::evgAS was constructed by inserting the fragment joined by overlap extension PCR amplified using primers pB037/pB041 or pB037/pB045, pB040/pB042 (to form the ybaS fragment) or pB044/pB046 (to form the evgAS fragment), and pB043/pB036 or pB047/pB036 into the PstI/HindIII-digested pTargetF-yjcS. pTargetT-ΔmaeB::gltP was constructed by inserting the fragment overlapped by PCR and amplified by primers pB073/pB018, pB074/pB075, pB076/pB077 (to form the gltP fragment), and pB078/pB050 into SpeI/SalI-digested pTargetF-yjcS. pTargetT-ΔmaeB::gltPΔmaeA was constructed by ligating the sgRNA and editing template fragment digested from pTargetT-ΔmaeA by BamHI/SalI to BglII/XhoI-digested pTargetT-ΔmaeB::gltP.

For the control experiment, strain MGlyl was designed by inserting the cat gene amplified from pSU2718 by pB068/pB069 into the cadA loci of MG1655. Strain MGly2 was constructed by inserting a 275-bp fragment of cadA (cadA^p) amplified from MG1655 by pB070/pB071 in the cat loci of MGlyl to inactivate the chloramphenicol resistance activity by standard CRISPR-Cas system protocol (described below) using pCas and pTargetF-cat. pCasΔcas9 was constructed by digestion of a 1,435-bp fragment of cas9 from pCas by NdeI followed by self-ligation. pTargetF-cadA^p was constructed routinely as described above by inverse PCR with primers pB079/pB033. pTargetT-ΔcadA^p was constructed by inserting the pB066/pB067-amplified fragment (donor DNA) into the BglII/XhoI-digested pTargetF-cadA^p and the pTargetTΔR-ΔcadA^p missing the targeting sgRNA, which was constructed by inserting the pB066/pB067-amplified fragment into the BamHI/XhoI-digested pTargetF-cadA^p.

pCas and pTargetF were deposited in Addgene under the numbers 62225 and 62226.

Genome editing.

MG1655 and DSM 13699 competent cells harboring pCas were prepared as described previously (16, 36, 37). Arabinose (10 mM final concentration) was added to the culture for λ-Red induction according to the protocol. For electroporation, 50 μl of cells was mixed with 100 ng of pTargetT series DNA; electroporation was done in a 2-mm Gene Pulser cuvette (Bio-Rad) at 2.5 kV, and the product was suspended immediately in 1 ml of ice-cold LB medium. When the donor DNA was supplied in a PCR fragment, 100 ng of pTargetF series DNA and 400 ng of donor DNA were coelectroporated. Cells were recovered at 30°C for 1 h before being spread onto LB agar containing kanamycin (50 mg/liter) and spectinomycin (50 mg/liter) and incubated overnight at 30°C. Transformants were identified by colony PCR and DNA sequencing.

For control experiments, the strain MGly2 modified from MG1655 was used as the host. pCas and pCasΔcas9 with a cas9 deletion were cotransformed with pTargetT-ΔcadA^p and pTargetTΔR-ΔcadA^p with targeting sgRNA deletion, respectively; pCas was also cotransformed with pTargetF-cadA^p without a cat homologous fragment (Fig. 2A). Both λ-Red induction and noninduction were done by adding arabinose (10 mM final concentration) or not, according to the previous protocol. Cells were recovered at 30°C for 1 h before being spread onto LB agar containing kanamycin (50 mg/liter) and spectinomycin (50 mg/liter) or kanamycin (50 mg/liter) and chloramphenicol (25 mg/liter) and incubated at 30°C overnight.

FIG 2.

Effects of cas9, targeting sgRNA, donor DNA, and λ-Red in the CRISPR-Cas two-plasmid system. (A) Diagram of the experimental conditions. (a) cas9 was deficient in pCas; (b and c) targeting sgRNA (b) or donor DNA (c) was deficient in pTargeting series; (d and e) λ-Red with (RED+) (e) or without (RED-) (d) induction. (B) Mutation efficiency. The fraction of spectinomycin-resistant (spec) and kanamycin-resistant (kan) or chloramphenicol-resistant (cm) and kanamycin-resistant (kan) CFU calculated from total CFU was determined under the experimental conditions shown under the histogram and depicted in panel A. Data are means ± standard deviations from three independent experiments.

Plasmid curing.

For the curing of pTarget series, the edited colony harboring both pCas and pTarget series was inoculated into 2 ml of LB medium containing kanamycin (50 mg/liter) and IPTG (isopropyl-β-d-thiogalactopyranoside; 0.5 mM). The culture was incubated for 8 to 16 h, diluted, and spread onto LB plates containing kanamycin (50 mg/liter). The colonies were confirmed as cured by determining their sensitivity to spectinomycin (50 mg/liter). The colonies cured of pTarget series were used in a second round of genome editing. pCas was cured by growing the colonies overnight at 37°C nonselectively (13).

RESULTS

Establishment of a two-plasmid-based CRISPR-Cas9 system.

Two-plasmid systems were designed initially to use the CRISPR-Cas9 system, as reported for E. coli (24, 35), which separated cas9 and the sgRNA in pCas and pTarget series, respectively (Fig. 1). pCas was constructed by introducing the Cas9 protein from S. pyogenes MGAS5005 with its native promoter, the temperature-sensitive replicon repA101(Ts) from plasmid pKD46 for self-curing (13), the λ-Red gene under the control of the P_araB promoter, which is induced by l-arabinose (40), and an sgRNA containing an N₂₀ sequence targeting the pTarget pMB1 replicon (sgRNA-pMB1) under the control of an IPTG-inducible promoter, P_trc. The sgRNA targeting the genome loci of interest located in the pTarget series was expressed from a minimal constitutive promoter with a pMB1 origin of replication (Fig. 1b).

This CRISPR-Cas9 system was first tested for the effect of a deficiency of any of the four motifs cas9, sgRNA targeting the genome loci, donor DNA, and λ-Red gene. MGly2 was designed specifically as a control host modified from MG1655 to have a heterologous chloramphenicol cat resistance gene insertion at the cadA locus, which was inactivated by a DNA fragment [cadA^p] inserted inside the cat gene locus. Modified MGly2 colonies harboring pCas series (kanamycin resistant) with the cadA^p deletion were expected to retrieve chloramphenicol resistance activity and thus survival on agar containing chloramphenicol and kanamycin. The total CFU were calculated by growth on agar containing kanamycin and agar containing spectinomycin resulting from pCas and pTarget-ΔcadA^p (spectinomycin resistant) cotransformation into MGly2 (Fig. 2A). A deficiency of cas9 in pCas or targeting sgRNA in pTarget resulted in a low level of recombination efficiency (<5%) through λ-Red recombination and a low survival rate on chloramphenicol selection medium, as expected. A deficiency of cat homologous fragments (donor DNA) or the λ-Red gene without induction resulted in a very low survival rate, even in the absence of chloramphenicol selection compared to that without cas9, since most of the strains were killed by cas9 through introduction of dsDNA breaks into the chromosome. The CRISPR-Cas9 system using pCas and pTargetT-ΔcadA^p with cas9, targeting sgRNA, cat homologous fragments, and the λ-Red gene resulted in a 100% mutation rate and a relatively high survival rate (1.12E−05), which indicated that expression of the λ-Red protein increased the target site mutation rate by CRISPR-Cas9 significantly (24). In addition, using the CRISPR-Cas9 system by introducing dsDNA breaks into the chromosome increased the rate of recombination of the damaged DNA, as reported elsewhere (27). We counted 2.61-fold more colonies (1.58E−05/6.08E−06) after cotransformation with pCas and pTargetT-ΔcadA^p compared to a deficiency of the cas9 construct (Fig. 2B). Without the donor DNA, the few colonies observed on chloramphenicol selection medium likely resulted from an escape from the death effect of DSBs by alternative end joining (41).

The two-plasmid-based CRISPR-Cas9 system makes multiplex gene modifications continuously.

This CRISPR-Cas9 system was tested for (i) single, double, and multiple gene deletions and (ii) single and double gene insertions. For a single gene deletion, as shown for cadA, 86% ± 4% of the transformants showed the expected genotype and 100% of the cells lost pTargetT-ΔcadA (Table 2, experiment 1). We then doubled and tripled the number of editing targets. When MG1655 harboring pCas was transformed with pTargetT-ΔmaeAΔmaeB or pTargetT-ΔcadAΔmaeAΔmaeB, which were expected to perform a maeA-maeB double deletion or a cadA-maeA-maeB triple deletion, the mutation efficiencies were as high as 97% ± 4% and 47% ± 8%, respectively (Table 2, experiment 2 and 3).

TABLE 2.

Mutation efficiency of the CRISPR-Cas two-plasmid system^a

Expt no.	Host cell	Plasmid pTarget	Targeting genome locus of sgRNA	Donor DNA supplied in pTarget or in PCR fragment (F)	Length (bp) of homologous extensions (upstream, downstream)	Mutation efficiency (%)b	Plasmid pTarget curing efficiency (%)
1	MG1655	pTargetT-ΔcadA	cadA	pTargetT-ΔcadA	523, 281	86 ± 4	100
2	MG1655	pTargetT-ΔmaeAΔmaeB	maeA, maeB	pTargetT-ΔmaeAΔmaeB	250, 550	97 ± 4	ND
3	MG1655	pTargetT-ΔcadAΔmaeAΔmaeB	cadA, maeA, maeB	pTargetT-ΔcadAΔmaeAΔmaeB	250, 550	47 ± 8	ND
4	MG1655	pTargetT-ΔyjcS::ybas	yjcS	pTargetT-ΔyjcS::ybaS	373, 360	92 ± 0	ND
5	MG1655	pTargetT-ΔyjcS:: evgAS	yjcS	pTargetT-ΔyjcS::evgAS	373, 360	75 ± 18	ND
6	MG1655	pTargetT-ΔmaeB::gltP ΔmaeA	maeB, maeA	pTargetT-ΔmaeB::gltP ΔmaeA	250, 550	78 ± 26	ND
7	1655ΔcadAc	pTargetT-ΔyjcS::evgAS	yjcS	pTargetT-ΔyjcS::evgAS	373, 360	92 ± 7	100
8	MG1655	pTargetF-cadA	cadA	ΔcadA (F)	523, 281	69 ± 4	ND
9	MG1655	pTargetF-yjcS	yjcS	ΔyjcS::evgAS (F)	40, 40	6 ± 4	ND
10	MG1655	pTargetF-yjcS	yjcS	ΔyjcS::evgAS (F)	373, 360	28 ± 10	ND
11	MG1655	pTargetF-kefB-yjcS	kefB, yjcS	ΔkefB, ΔyjcS::evgAS (F)	250, 550	0	ND
12	DSM 13699	pTargetT-ΔtkrA	tkrA	pTargetT-ΔtkrA	483, 513	100 ± 0	100
13	DSM 13699 ΔtkrAc	pTargetT-Δglk	glk	pTargetT-Δglk	500, 500	94 ± 8	100

^aThe genome editing was performed with the CRISPR-Cas two-plasmid system with pCAS and pTarget, as shown. ND, not determined.

^bDetermined from triple electroporation experiments by colony PCR from 12 transformants for each mutation (agarose electrophoresis gels of colony PCR and relative sequencing results are shown in Fig. S1 and S2 in the supplemental material).

^cSecond round of genome editing.

We used pCas to perform a single insertion and a mixed gene insertion and deletion. High mutation rates (92% ± 0% and 75% ± 18%, respectively) were obtained when ybaS (1.3 kb) and evgAS (4.5 kb) were inserted into yjcS (Table 2, experiment 4 and 5). For mixed gene insertion and deletion, 78% ± 26% of the colonies showed the expected genotype for the deletion of maeA and the insertion of gltP (1.7 kb) into the maeB locus (Table 2, experiment 6).

Continual gene editing was tested (Table 2, experiment 7). When 1655ΔcadA, cured of pTargetT-ΔcadA, was transformed with pTargetT-ΔyjcS::evgAS, the mutation efficiency for the insertion of evgAS into the yjcS locus was relatively high, 92% ± 7%. pCas was finally cured by the end of the procedure by culture at 37°C overnight. In all our experiments, >90% of colonies regained kanamycin sensitivity, indicating successful clearance of the temperature-sensitive plasmid pCas, in accordance with published data (13). Agarose gel electrophoresis of colony PCR and the sequencing results are supplied in Fig. S1 and S2 in the supplemental material. This demonstrated the feasibility of performing multiple rounds of genome editing to engineer novel bacterial strains.

We did not attempt multiple gene deletions or insertions of more than three genes because the cloning procedure for pTargetT was complicated and time-consuming when multiple donor DNAs were included. The method will not have the level of efficiency needed for metabolic engineering of an industrially relevant strain. Thus, although the problem of low efficiency of gene insertion was solved and double or multiple gene deletions or insertions were achieved, a simpler procedure for genome editing is needed.

Simplified genome editing by a CRISPR-Cas9 system with donor DNA supplied as a fragment.

To simplify the cloning procedure for the pTarget series, the donor DNA was designed to be supplied in fragments. For single-gene editing, pTarget could thus be cloned simply by changing the N₂₀ sequence of the sgRNA when different genomic loci are being targeted, which could be done by inverse PCR with mutations incorporated into the primers (38), resulting in the pTargetF version (Fig. 1b). Double- or multiple-gene editing of the pTargetF series with double or multiple sgRNAs could be done easily by the BioBrick method (38).

By using the pTargetF series with donor DNA supplied as fragments, we obtained single-gene cadA deletion efficiency as high as 69% ± 4% when pTargetF-cadA and the fragments homologous to the upstream and downstream regions of the cadA locus (obtained by overlap PCR) were cotransformed into MG1655 harboring pCas (Table 2, experiment 8). For gene insertions, because λ-Red recombination can be obtained efficiently with homologous regions of ≥40 bp (13), we reduced the homologous length from 300 to 500 bp to 40 bp, which could be incorporated directly into the PCR primers for the donor DNA fragment. However, a very low mutation efficiency of 6% ± 4% was obtained when we inserted evgAS into the yjcS locus (Table 2, experiment 9). We extended the homologous length in the donor DNA to ∼400 bp for the same targeting site, and a higher insertion rate of 28% ± 10% was obtained (Table 2, experiment 10).

We attempted to perform double-gene editing (Table 2, experiment 11) by combined deletion of locus kefB with the insertion of evgAS into locus yjcS, but we obtained no double mutation.

Results of agarose gel electrophoresis of colony PCR and sequencing are supplied in Fig. S1 and S2 in the supplemental material.

Application of two-plasmid-based CRISPR-Cas9 system in Tatumella citrea for continuous gene deletion.

To evaluate the possibility of a broader applicability of the system described above, T. citrea DSM 13699, another member of the Enterobacteriaceae, was selected. Two genes, encoding a subunit of glyoxylate reductase (tkrA) and glucokinase (glk) were chosen as individual targets. The system fit DSM 13699 well without any modification, with 100% ± 0% tkrA deletion efficiency and 94% ± 8% second-gene glk deletion efficiency (Table 2, experiments 12 and 13). pTargetT-ΔcadA, pTargetT-ΔyjcS::evgAS, pTargetT-ΔtkrA, and pTargetT-Δglk were 100% cured (Table 2). The observed efficient genome editing of T. citrea without strain-specific backbone modification of the two-plasmid-based CRISPR-Cas9 system suggests a possible broader applicability of this system in various Enterobacteriaceae species.

DISCUSSION

In this study, we expanded the application of the CRISPR system from the published allelic exchange procedure (24) to targeted single or multiple gene deletions and insertions in E. coli and another Enterobacteriaceae species, T. citrea. Compared to published scarless genome modification methods, such as those involving sacB (12), I-SceI (12, 19, 20), and MAGE (22, 23), the CRISPR-based targeted genome modification method can perform multiple gene insertions or deletions, whereas sacB or I-SceI could be used to modify only single targets each time. ssDNA oligonucleotide-mediated MAGE was used successfully for multiple allelic exchange, but small-fragment (30 bp) insertion decreased mutation efficiency dramatically (12, 22). In addition, the CRISPR-based gene modification system offers unprecedented convenience and efficiency in design and manipulation. Targeting any site of interest requires the insertion of only a short spacer into a targeting sgRNA construct, pTargetF in this study, which can be achieved by inverse PCR and self-ligation within 2 days, with donor DNA supplied as PCR fragments (Fig. 1b). The manipulation time for the procedure was reduced to 2 days for each round of modification, and up to three gene targets can be modified simultaneously; an additional 2 days are required for the entire procedure (Fig. 3). For the metabolic engineering case that required 26 genomic modifications, as mentioned in the introduction, the total manipulation time can be 20 to 54 days. Metabolic engineering is based on the cell system network in which simple gene engineering might result in unexpected phenotypes, and with the rapid development of genome sequencing technology (42), more sequenced genotypes need to be illustrated biologically. This CRISPR-based time-saving genome modification method will be a powerful tool in the metabolic engineering field and will facilitate the output of genetically modified strains, thus increasing the likelihood of engineering complex strains. T. citrea is an important host for production of the industrially relevant vitamin C precursor 2-keto-d-gluconic acid (43). The application of the CRISPR-based gene modification system in T. citrea will greatly facilitate metabolic engineering of this strain compared to the only traditional homologous recombination-based gene knockout system as described previously (43, 44). The successful expansion of this system without any specific modification to T. citrea indicated its wide adaptability and flexibility in other Enterobacteriaceae species.

FIG 3.

Detailed diagram of continual genome editing with the two-plasmid system.

The challenge facing DSB-based, multiplex genome modification techniques might be the toxicity of simultaneous multiple chromosomal breaks and the high rate of nonhomologous end joining (NHEJ), which could lead to unintended rearrangements (10). However, E. coli lacks the NHEJ mechanism, although a small possibility of an alternative end-joining mechanism exists (41), and is highly reliant on a homology-directed repair system to repair DSBs in the chromosome (45, 46). Thus, the success of multiplexing depends on the fine-tuning of Cas9 activity and the rescue efficiency of homology-directed repair. Double-stranded, λ-Red-mediated recombination successfully rescued the low efficiency of the E. coli native homology-directed repair system and, thus, succeeded in multigene editing even when Cas9 was expressed constitutively, while single DSB generated by constitutively expressed Cas9 could not be repaired without induction of λ-Red (Fig. 2B). Originally, we used an arabinose-inducible promoter to express Cas9, and thus, some cells escaped cleavage on the chromosome caused by the induction efficiency of the arabinose promoter (47); as a result, the curing efficiency of pTarget cleaved by cas9 did not reach 100% (data not shown). We failed to clone the IPTG-inducible trc, which was expressed targeting sgRNA, and constitutively expressed cas9 in one plasmid, which might cause by leakage of the trc promoter. We did not investigate the possibility of using other inducible promoters for both cas9 and sgRNA or λ-Red to incorporate these into one plasmid, because the two-plasmid-based system has the advantage of a shorter total manipulation time than the inducible one-plasmid system if the number of targets is >2 (the two-plasmid system needs 2n+2 days, and the one-plasmid system needs 3n days).

For the CRISPR-Cas9 system given in Table 2, we can perform up to three gene deletions and mixed gene deletions and insertions with acceptable levels of efficiency (47% ± 8% and 78% ± 26%, respectively). However, cloning of pTargetT, which contains multiple targeted sgRNAs and donor DNAs, was both time-consuming and labor-intensive. Thus, donor DNAs supplied in fragments that can be cotransformed into the cell with sgRNAs contained in pTargetF reduced the amount of time and labor needed for the constructions. However, the efficiency decreased dramatically when increasing the batch targets numbers (0%) (Table 2, experiment 11) or decreasing the length of homologous extensions from 300 to 400 bp to 40 bp (6% ± 4%) (Table 2, experiment 9). This was because the efficiency of double-stranded, λ-Red-mediated recombination was not sufficiently high, or the transformation efficiency of the dsDNA in E. coli was low. The recombination efficiency might be improved by using ssDNA as the donor DNA, as λ-Red-like proteins also facilitate the recombination of smaller ssDNA fragments, such as those used in MAGE. If CRISPR and MAGE are combined, the challenge might be that the multiple, repeated sgRNAs in pTarget will lead to rearrangements by self-homologous recombination, as well as its limitation in gene insertion manipulation.

Off-target effects of Cas9 in human and murine cells have been reported (48, 49), and some methods have been applied to mitigate these effects, including cooperative use of offset nicking and a cas9 nickase mutant (50). To reduce the off-target effects of Cas9 in this study, an N₂₀ sequence was selected to ensure the last 12 bp was highly specific for the targets (24).