Exploiting the Type I-B CRISPR Genome Editing System in Thermoanaerobacterium aotearoense SCUT27 and Engineering the Strain for Enhanced Ethanol Production

ABSTRACT

Thermoanaerobacterium aotearoense strain SCUT27 is a potential industrial biofuel-producing strain because of its broad substrate spectrum, especially the ability to co-use glucose and xylose. The bottleneck hindering the development of strain SCUT27 is the lack of selective markers for polygene manipulation in this thermophilic bacterium. In this study, the endogenous type I-B clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated (Cas) system was developed for multiplex genome editing of strain SCUT27. The protospacer-adjacent motif was identified by in silico analysis and verified with orotidine-5′-phosphate decarboxylase (pyrF) or lactate dehydrogenase (ldh) as the editing target. The type I-B CRISPR/Cas system was functional in strain SCUT27 with 58.3% to 100% editing efficiency. A multiplex genome editing method based on thymidine kinase (tdk) as a negative selection marker was developed, and strain SCUT27/Δtdk/Δldh/ΔargR, in which ldh and the arginine repressor (argR) were knocked out successively, was successfully obtained. Strain SCUT27/Δtdk/Δldh/ΔargR exhibited prominent advantages over wild-type SCUT27 in ethanol production, with significantly improved ability to metabolize xylose.

IMPORTANCE Thermophilic microbes have attracted great attention as potential candidates for production of biofuels and chemicals from lignocellulose because of their thermal tolerance and wide substrate spectra. The ability to edit multiple genes using the native type I-B CRISPR/Cas system would speed up engineering of Thermoanaerobacterium aotearoense strain SCUT27 for higher ethanol production from lignocellulosic hydrolysates. Here, we produced a mutant strain, T. aotearoense SCUT27/Δtdk/Δldh/ΔargR, using the native CRISPR/Cas system. The engineered strain showed satisfactory performance with improved ethanol productivity from various lignocellulosic hydrolysates. Our data lay the foundations for development of this thermophilic microbe into an excellent ethanol producer using lignocellulosic hydrolysates. The methods described here may also provide a reference to develop multigene editing methods for other microorganisms.

INTRODUCTION

Lignocellulosic biomass is a globally available, sustainable feedstock rich in sugars that can be converted into biofuels and specialty products by appropriate processing (1). Thermoanaerobacterium aotearoense strain SCUT27, which can co-use glucose and xylose, is a natural ethanol producer and a promising cell factory for the conversion of lignocellulosic hydrolysates to biofuels and other bioproducts at 40°C to 70°C (2, 3) (Fig. 1). To enhance ethanol production, the lactate dehydrogenase gene (ldh) was deleted from strain SCUT27; the mutant exhibited increased cell yield (~31.0%) and ethanol yield (~2.2-fold), along with higher acetate and ATP levels, compared with those of the wild type (3). Arginine repressors (ArgRs) are involved in several key metabolic pathways; previous studies have found that strain SCUT27/ΔargR₁₈₆₄ exhibited improved co-use of glucose and xylose and an increased ethanol yield because of enhanced xylose use and higher cell energy level (4, 5). On the basis of the results described above, multiplex genome editing targeting these key genes in T. aotearoense strain SCUT27 has great potential for facilitating high ethanol production from lignocellulosic hydrolysates. However, several problems hinder metabolic engineering in strain SCUT27, such as the limited available selective markers and the absence of efficient genetic engineering tools for successive gene deletion.

FIG 1.

The clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated (Cas) genome editing system in Thermoanaerobacterium aotearoense SCUT27. (A) An efficient genetic modification tool is imperative to engineer strain SCUT27 for biofuel production from lignocellulosic biomass. (B) Schematic of the endogenous type I-B CRISPR/Cas system in strain SCUT27. Created with BioRender.com.

Clustered regularly interspaced short palindromic repeats (CRISPR) and the CRISPR-associated (Cas) system are an RNA-guided adaptive immune system in bacteria and archaea that acts against invading foreign DNAs, such as phages and plasmids (6, 7). One study has reported that approximately 47% of genome-sequenced bacteria and 87% of sequenced archaea harbor CRISPR/Cas loci, all of which share similar features, consisting of identical direct repeats (DR) separated by variable sequences called spacers, along with a suite of associated Cas genes (8). The CRISPR system can be divided into two classes and six types (33 subtypes) based on the Cas proteins and the architecture of the CRISPR/Cas loci (8). The type II CRISPR/Cas9-based editing system is well understood and has been widely applied with high efficiency in a broad range of organisms (9–13). Although some thermotolerant Cas9s have been found, the use of the CRISPR/Cas9-based editing system in thermophiles has rarely been reported because of the severe toxicity of Cas9 protein to cells with low efficiency of transformation and editing (9, 10). In addition, the type II system has complex protospacer-adjacent motif (PAM) sequences and shorter single guide RNAs (sgRNAs) that limit the selection of editing locations and lead to potential off-target effects (10, 14). To make gene editing easier, much research has focused on using endogenous CRISPR/Cas systems for genome editing. Pyne et al. discovered the native type I-B CRISPR/Cas system in Clostridium pasteurianum and successfully deleted cpaAIR (15). Zhang et al. exploited the CRISPR/Cas system in Clostridium tyrobutyricum and realized the deletion of spo0A and pyrF (orotidine-5′-phosphate decarboxylase) by a single transformation, along with the integration of an alcohol dehydrogenase-encoding gene to replace cat1 for N-butanol production (16). Zhou et al. successfully deleted spo0A and aldh by customizing the endogenous CRISPR/Cas system in Clostridium butyricum, with editing efficiency of up to 100% (17). Liu et al. found an endogenous subtype II-A CRISPR/Cas system in Pediococcus acidilactici with which native plasmids were depleted for enhanced cell growth, and an l-lactate dehydrogenase (LDH)-encoding gene was integrated into the chromosome for improved lactic acid production (18).

In this study, we report the native type I-B CRISPR/Cas system in T. aotearoense strain SCUT27. In silico CRISPR array analysis and in vivo toxicity assay were performed and revealed that the motif TTA at the 5′ end of predicted protospacers was a functional PAM site. With a constitutive strong promoter (Pkan) for CRISPR mini-array expression, ldh was successfully deleted with editing efficiency up to 100% but with low transformation efficiency. To improve ethanol production, a multigene editing system was developed based on plasmid curing, using both ldh and argR, which were knocked out from strain SCUT27; the resulting strain, SCUT27/Δtdk/Δldh/ΔargR, was assessed for ethanol production in culture with xylose-rich lignocellulosic hydrolysates.

RESULTS

Identification and activity test of the type I-B CRISPR/Cas system in T. aotearoense strain SCUT27.

Recently, type II CRSIPR/Cas9 system tools have been reported for genome editing of thermophiles based on thermostable Cas9 derived from Geobacillus stearothermophilus, Acidothermus cellulolyticus, and Geobacillus thermodenitrificans T12 (9, 10). The Cas9 of G. thermodenitrificans T12 was tried for genome engineering in SCUT27 with plasmids pLY1_ldh-HA12 and pLY2_ldh-HA12 (Table 1) containing the thermostable cas9 gene, an sgRNA-expressing module targeting ldh, and a repair template. However, no transformants were obtained after electrotransformation, suggesting that this CRISPR/Cas9 system was not suitable for use in strain SCUT27, probably because of high toxicity of the potent Cas9 ribonucleoprotein (RNP). Therefore, we tried to exploit the native type I-B CRISPR/Cas system for genome editing in T. aotearoense SCUT27.

TABLE 1.

Strains and plasmids used in this study

Strains/plasmids	Description	Source
Strains
SCUT27	Wild type of T. aotearoense SCUT27 (GDMCC 60765)	Li et al. (28)
SCUT27/Δldh::ermR	Deletion of ldh (V518_0188) in SCUT27	Cai et al. (3)
SCUT27/ΔargR::kanR	Deletion of argR (V518_1864) in SCUT27	Qu et al. (5)
SCUT27/Δldh	Deletion of ldh in SCUT27	This study
SCUT27/Δtdk::kanR	Deletion of tdk (V518_2650) in SCUT27	This study
SCUT27/Δtdk	Deletion of tdk in SCUT27	This study
SCUT27/Δtdk/Δldh	Deletion of ldh in SCUT27/Δtdk	This study
SCUT27/Δtdk/Δldh/ΔargR	Deletion of argR in SCUT27/Δtdk/Δldh	This study
E. coli DH5α	Host cells for plasmids amplification	Tiangen
Plasmids
pIKM1	Original plasmid for cloning and expression with AmpR and KanR	Mai et al. (32)
pBluescript II SK(+)	Primordial suicide plasmid	Stratagene
pLY1	Insert GeoCas9 gene under promoter PadhE in pIKM1	This study
pLY2	Insert GeoCas9 gene under promoter Pkan-RSpbuE in pIKM1	This study
pLY1_ldh-HA12	Plamid used for ldh deletion	This study
pLY2_ldh-HA12	Plamid used for ldh deletion	This study
pKQ1	Inset Pkan and terminator (298 bp of sequence located at the 3′ end of the endogenous CRISPR array) in pIKM1	This study
pKQ1_pyrF1	Plamid used for pyrF (V518_1373) inference	This study
pKQ1_pyrF2	Plamid used for pyrF inference	This study
pKQ1_pyrF3	Plamid used as a control for targeting pyrF with GGC as PAM	This study
pKQ1_ldh-HA12	Plasmid used for ldh deletion	This study
pKQ2_ldh-HA12	Pkan in pKQ1_ldh-HA12 was replaced with Pcat1	This study
pKQ3_ldh-HA12	Pkan in pKQ1_ldh-HA12 was replaced with PadhE	This study
pKQ4_ldh-HA12	Pkan in pKQ1_ldh-HA12 was replaced with Pkan-RSpbuE	This study
pKQ1_argR-HA12	Palsmid used for argR deletion with sgRNA_argR	This study
pKQ1_ldh-HA12::tdk	Overexpress tdk based on pKQ1_ldh-HA12	This study
pKQ1_argR-HA12::tdk	Overexpress tdk based on pKQ1_argR-HA12	This study
pBlu_tdk-HA12	Plasmid used for tdk deletion	This study
pIKM1-Pldh-ldh	Overexpress ldh with Pldh	This study
pIKM1-PadhE-ldh	Overexpress ldh with PadhE	This study
pIKM1-Pcat1-ldh	Overexpress ldh with Pcat1	This study
pIKM1-Plac-ldh	Overexpress ldh with Plac	This study
pIKM1-Pkan-RSpbuE-ldh	Overexpress ldh with Pkan-RSpbuE	This study

CRISPRCasFinder, which is a web-based tool to identify CRISPR and the presence of Cas genes (19), was used here to find any native CRISPR/Cas system on the chromosome of strain SCUT27 (GCF_000512105, Thermoanaerobacterium aotearoense SCUT27 assembly genomic.fna). The results indicated that strain SCUT27 possesses a native CRISPR/Cas system, which was assigned to class 1 type I-B. Seven major CRISPR loci were found (Table S2). However, intact Cas genes (V518_0414 to 0422) were identified only in the upstream and downstream regions of the largest CRISPR array, which contained 55 spacers (Fig. 2A; Table S2). The other CRISPR arrays not coupled with Cas genes were defined as orphan CRISPR. One study reported that the longest array created by numerous spacer acquisition events and coexpression with a complete functional set of Cas genes might constitute an active CRISPR unit within a bacterial genome (20). Thus, this longest CRISPR array in T. aotearoense SCUT27 (positions 14334 to 18020 in NZ_AYSN01000014.1), including 55 spacers along with DR, was used for further research.

FIG 2.

Characterization of the type I-B CRISPR/Cas system in T. aotearoense SCUT27. (A) Structure of the only intact type I-B CRISPR/Cas locus in strain SCUT27, including cas6–cas8a–cas7–cas5–cas3 (effectors), cas4–cas1–cas2 (adaptation), and the longest array in the genome, containing 55 distinct spacers (diamonds) separated by 30-nt direct repeats (rectangles). (B) Identification of putative protospacers by in silico analysis. Spacer–protospacer mismatches are indicated in red. Putative protospacer-adjacent motifs (PAM) are indicated in yellow. (C) Identification of PAM for the CRISPR system in T. aotearoense SCUT27 (red box).

For the CRISPR/Cas system to recognize protospacers, a functional PAM at the 5′ end of the protospacer is necessary. Our primary approach to PAM identification was to explore the putative original sequences of the invading elements that ended up as CRISPR spacers during the “adaptation” stage in strain SCUT27. The search program CRISPRTarget (21) was used to analyze the CRISPR spacers by aligning these sequences with sequences in various databases, such as those of phage and plasmids, as described by Pyne (15). However, there was no obvious pattern of PAMs associated with the 55 spacers in the largest CRISPR array in strain SCUT27. Thus, next, we searched the closest related strain, Thermoanaerobacter thermosaccharolyticum M0795, which has 266 spacers, because of its similar Cas operon (Thethe_02655 to 02664) and DR motif GTTTTTAGCCTACCTATAAGGAATTGAAAC to those in T. aotearoense strain SCUT27. This revealed a high-frequency of the PAM motif TTA at the 5′ end of protospacers, based on the results of spacer–protospacer matching analysis (Fig. 2B and C). The predicted PAM sequence TNA in Thermoanaerobacter sp. was reported using various in silico tools (22–24). The PAM site TNA was also defined in the reported subtype I-B CRISPR/Cas system in Clostridia, such as Clostridium pasteurianum, Clostridium difficile, Clostridium tyrobutyricum, and Clostridium thermocellum (9, 15, 16, 25). On the basis of this analysis, the PAM site TTA in T. aotearoense strain SCUT27 was chosen and applied in the following experiments.

Toxicity assay was performed to evaluate the activity of the native CRISPR system of T. aotearoense strain SCUT27. An artificial mini-array cassette composed of 30-bp DRs and a 37-bp spacer was constructed (Fig. S1A to C), expression of which would cause the RNP to mediate a DNA strand break at the target site. Similar to most bacteria, strain SCUT27 does not encode proteins responsible for nonhomologous end joining. If the CRISPR/Cas system was functional, DNA cutting caused by the active RNP could not be repaired and hence no transformants would be observed. Here, the mini-array on the plasmid was designed to target the pyrF gene (V518_1373) in strain SCUT27 (Fig. S1). The 921-bp pyrF open reading frame (ORF) contains a total of 44 potential PAMs (TTA) on double strands of DNA. Two spacers (37 nucleotides [nt] downstream of TTA) were chosen to target the pyrF locus, and one nontarget spacer (PAM motif GGC) was designed as a negative control (Fig. S1A). In addition, the transformation efficiency of empty vectors was also determined (53.2 ± 6.7 CFU/μg DNA), and it showed the natural competence of this strain (26). With a strong promoter Pkan for sgRNA expression, we observed a 21-fold decrease in electrotransformation efficiency of strain SCUT27 (from ~48.3 to 2.3 CFU/μg DNA, with PAM TTA against GGC), indicating that CRISPR-mediated interference was happening and approximately 95% cell death occurred (Table 2; Fig. S1D). The background colonies (escape mutants) in the experimental groups may result from spontaneous mutations in the Cas operon (specifically Cas3), which make it lose function (9, 27).

TABLE 2.

Effect of spacers of different genes on transformation efficiency

Plasmid	5′-pAM	Spacer sequence in sgRNA (protospacer)	Transformation efficiency (CFU/μg DNA)
pKQ1_pyrF1	TTA	TACGTAGCTCCAACAACTGCACCTATAGAACTATATC	2.5 ± 0.9
pKQ1_pyrF2	TTA	CGTCATGTATATATGGTGTCATGGTATCTTCTCCCAT	2.1 ± 0.7
pKQ1_pyrF3	GGC	ATTATTTTTCTCAGTATTTTGGCCTCCTCTTTATACG	48.3 ± 11.5
pKQ1_ldh-HA12	TTA	CATACAAGATTTCTGGACTTCCATGGGGCAGAGTTTT	6.2 ± 4.4
pKQ1_argR-HA12	TTA	TAGAGGAGCTTAAGCTAAATAAAATTATAGCTACACA	3.1 ± 1.9

Endogenous type I-B CRISPR-based deletion of ldh from strain SCUT27.

The native CRISPR system was functional in T. aotearoense SCUT27, and therefore we constructed the gene editing plasmid pKQ1_ldh-HA12 targeting the ldh gene (V518_0188). A repair template was introduced downstream of the CRISPR mini-array, consisting of two homologous arms (HAs) and a KpnI digestion site, replacing 805 bp in the ldh locus to inactivate the ldh gene (Fig. S2A). The resulting ldh mutant could be identified easily by its distinct advantages in growth and ethanol production (3).

Transformation of pKQ1_ldh-HA12 into strain SCUT27 resulted in 12 to 65 colonies (~6.2 CFU/μg DNA) on plates containing 50 μg/mL kanamycin (MTCK). Colony PCR was performed with randomly chosen colonies to screen for ldh deletion mutants. As shown in Fig. 3A, 9 of 12 colonies (75%) had a clean deletion genotype (Δldh), two colonies had a wild-type genotype, and one colony showed a mixed genotype. We suggest that some mutation of key Cas proteins led to the ineffective function of the editing system in the false-positive colonies. Overall, however, with this native CRISPR system, desirable mutants (i.e., Δldh) could be reproducibly obtained with a high editing rate (varying from 58.3% to 100%) within 2 to 4 days of transformation (similar efficiency to that of traditional homologous recombination). The colony PCR products were sequenced and the target locus was edited as expected (Fig. S2A). The fermentation performance of the mutant was analyzed by high-performance liquid chromatography (HPLC); no lactate was detected, as was also observed for strain SCUT27/Δldh::ermR, previously constructed in our laboratory by homologous recombination (Table S5) (28).

FIG 3.

Optimization of the editing plasmid pKQ1_ldh-HA12 for deletion of gene ldh. (A) Gel electrophoresis analysis of ldh deletion with primers ldh-c-F and ldh-c-R distal to the entire recombination region. The resulting band is 1,616 bp if the deletion is successful and 2,421 bp for the wild type. (B) Relative expression strengths of various promoters and editing efficiency with the same sgRNA_ldh under the control of the corresponding promoters. The left y axis shows lactate dehydrogenase activity (bars), and the right y axis shows the editing efficiency (orange points). Δldh was the negative control, and the wild-type and Δldh::Pldh-ldh strains were used as positive controls. + indicates culture with 1 mM adenine; − indicates culture without adenine. (C) The efficiency of transformation and CRISPR editing with homologous arms (HAs) of various lengths and extra addition of HAs. The left y axis shows the transformation efficiency (bars), and the right y axis shows the editing efficiency (blue points). − indicates no HA addition. The data are the means and standard deviations of three replicates (*, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001).

Endogenous CRISPR editing system with various promoters for sgRNA expression.

Although the type I-B CRISPR/Cas system in T. aotearoense SCUT27 showed efficiency in CRISPR-mediated killing and a high editing rate was observed, only a few transformants were obtained. The low efficiency of transformation was possibly due to the high toxicity of CRISPR-induced DNA cutting with sgRNA expression under Pkan. Thus, three other constitutive promoters (PadhE, Pclo1313_1194, and Pcat1) were used to regulate the expression of the sgRNA, aiming to improve the transformation rate. The expression strength of all the promoters was characterized by LDH enzymatic activity (Fig. 3B and Fig. S3A).

Pclo1313_1194, a strong promoter in C. thermocellum, did not work in strain SCUT27 (Fig. 3B). PadhE was slightly weaker than Pkan, and its efficiency of transformation (~5.5 CFU/μg DNA) and that of editing (~75%) were close to those of Pkan (Fig. 3B, Table S3). Pcat1 was stronger than Pkan. However, both the transformation efficiency and editing efficiency of Pcat1 were lower than those of Pkan, indicating that stronger expression of sgRNA might cause much greater cleavage and result in higher toxicity to cells, along with an increased proportion of escape mutants.

Notably, through enzyme activity testing, we found that the LDH activity of strain SCUT27/Δldh::ermR::ldh with ldh complemented by pIKM1 was nearly equal to that of the wild type (Fig. 3B). It could be deduced that there was generally only one copy of pIKM1 in SCUT27, which was consistent with Walker’s resequencing data (plasmid copy number was approximately one in C. thermocellum) (9) and also with the low efficiency of transformation in our CRISPR/Cas experiments in strain SCUT27. As shown in Fig. 4, with a constitutive promoter to express the sgRNA, cells would die because of double-strand breaks (DSBs) where recombination did not occur, no matter how many plasmid copies were in the cell. When recombination did occur, if a cell contained only one copy of the editing plasmid, it would also die because of plasmid self-cutting and hence loss of resistance to kanamycin; only cells with several plasmid copies would survive. To overcome this problem, we looked for an inducible promoter to control DSBs in the initial stage of the transformation. With an inducible promoter, the cells containing one plasmid copy should be survivors on MTCK plates lacking the inducer (Fig. 4). Then, the inducer could be added to activate the expression of the sgRNA and cause cleavage; recombinant colonies could finally be picked up from nonselective plates.

FIG 4.

Effects of different types of promoters for single guide RNA (sgRNA) expression on the transformation of editing plasmids. (A) Constitutive promoter: cleavage (lightening) was performed by immediate expression of the CRISPR mini-array after transformation. On MTCK plates, transformants without homologous recombination, or recombined but containing just one copy of the plasmid, died; only transformants with several plasmid copies and homologous recombination could survive. (B) Inducible promoter: the transformants with one copy of the plasmid could be survivors on the MTCK plate in the absence of inducer. After induction, the recombined transformants could survive on nonselective plates, while those without homologous recombination died.

To test this hypothesis, the lactose-inducible promoter Plac (16) was tried. However, in strain SCUT27, nearly the same LDH activity was detected on expressing ldh under the control of Plac in the presence and absence of lactose (data not shown). Finally, a thermostable riboswitch element was adapted from a previous study (29) and we constructed the adenine-controlled inducible structure Pkan-RS_pbuE to control the expression of sgRNA to avoid cutting in the initial stage after transformation. LDH activity assays showed that the riboswitch-mediated inducible structure Pkan-RS_pbuE was strictly controlled by adenine (Fig. 3B). With sgRNA expression under the control of Pkan-RSpbuE, the number of transformants was obviously increased on MTCK plates (to approximately 42 CFU/μg DNA). Then, the colonies were induced with 1 mM adenine. However, the clones on the plates were mostly false-positives, indicating a low editing efficiency (Fig. 3B), even after passing five times for enrichment. Thus, the inducible promoter used here for the expression of the CRISPR mini-array might not be strong enough for editing.

Other attempts to improve the efficiency of transformation and editing.

The highly efficient cleavage by the CRISPR system resulted in there being few living colonies on plates, which was likely due to a low level of homologous recombination (9). One study has reported that in CRISPR/Cas genome editing of clostridial organisms, serial transfer (1:20 dilution) and subculturing in liquid medium could promote homologous recombination and increase the probability of successful gene editing (16). Thus, we cultivated bacterial solution with 5, 10, and 20 rounds of serial passaging in MTCK medium after the electric shock. However, the editing efficiency of the transformants was between 15.0% and 58.3%, lower than that without serial passaging (i.e., sgRNA expression under Pkan).

The effects of HA lengths on the efficiency of transformation and editing were evaluated. The length of the HAs was set to 250, 500, 750, and 1,000 bp, respectively. As shown in Fig. 3C, the transformation efficiency improved as the length of the HAs was increased, and the highest editing efficiency was observed when the length of the HAs was 750 or 1,000 bp. The addition of excess HAs also had an obviously positive effect on transformation efficiency; however, the editing efficiency of the transformants was relatively low (Fig. 3C). The extra healing DNA probably increased the opportunity for homologous recombination, leading to the higher transformation efficiency, because HAs carrying selection marker (PCR products) also served for genetic modification (30). However, the reason for the low editing rate is unknown.

Plasmid curing and argR deletion in Δldh mutant for higher ethanol production.

As described above, single gene deletion was achieved using the endogenous CRISPR/Cas system of T. aotearoense strain SCUT27. Next, multiplex genome editing was explored. Considering the limitations of both the selective makers and thermostable shuttle vectors that are available for strain SCUT27, plasmid curing is a must for continuous genetic engineering. The editing plasmid pKQ1_ldh-H12 in SCUT27/Δldh needed to be eliminated to acquire a marker-free mutant. First, the mutant clones were transferred into medium without kanamycin for 5 to 20 generations of transfer, aiming to spontaneously eliminate the plasmid. However, no colony without plasmid was obtained, probably because of the stable replication region of the vector. Then, chemical and physical methods were used in the early exponential growth phase for plasmid curing, such as application of 4 μg/mL isonicotinic acid hydrazide or 0.002% sodium dodecyl sulfate (31, 32). Physical processing, such as repeated freeze-thawing, heating at sublethal temperature, electroporation, or integration of several methods together, was also tried but did not work.

On the basis of thymidine kinase (tdk) as a negative selection marker (30), however, a method for efficiently screening plasmid-cured cells and multigene editing was established. FUDR (5-fluoro-2′-deoxyuridine) was the agent for negative selection (Fig. S4A), and its toxicity toward strains SCUT27 and SCUT27/Δtdk::kanR was tested. As shown in Fig. S4B, in the presence of 50 μg/mL FUDR, no wild-type colonies grew while many Δtdk colonies were observed, indicating that TDK could be used as a selection marker. In addition, tdk deletion had no significant influence on the growth of strain SCUT27 (Fig. S4D; Table S4). Thus, SCUT27/Δtdk could be used as a starting strain for engineering, without the need for tdk complementation. As shown in Fig. 5, strain SCUT27/Δtdk was first picked by the traditional method through homologous recombination without any positive markers and verified by PCR (Fig. S4C). Then, the mutant Δtdk/Δldh was obtained via electrotransformation of SCUT27/Δtdk with pKQ1_Δldh-H12::tdk. Through successive transfer (three generations) and spreading on the plates of 50 μg/mL FUDR (MTCF), Δtdk/Δldh colonies lacking plasmids were picked with a positive rate of approximately 25% (Fig. S5).

FIG 5.

General workflow of multiplex genome editing in T. aotearoense strain SCUT27. The suicide vector pBlu_tdk-H12 was used for deleting tdk from the wild-type strain. The shuttle vector pKQ1_ΔX-H12::tdk (X in this study was ldh or argR) was used for CRISPR-mediated recombination for deleting gene X from strain SCUT27/Δtdk or other engineered strains, with plasmid elimination for successive editing.

The mutant of strain SCUT27 with argR inactivation had a good performance in growth, ethanol yield, and energy level, and its ability to use xylose and lignocellulosic hydrolysates was improved compared to that of the wild-type strain (4, 5). Thus, argR was chosen as the target gene for the second round of editing (using pKQ1_ΔargR-H12::tdk). By the same process as that described above, the mutant SCUT27/Δtdk/Δldh/ΔargR was obtained successfully (Fig. S2B and Fig. S6), indicating that multiplex genome editing was feasible in T. aotearoense strain SCUT27.

Glucose and xylose use by the mutants.

Glucose and xylose are the two major fermentable sugars in various lignocellulose hydrolysates (33). Here, three glucose/xylose quality ratios (1:0, 2:1, 0:1) were investigated in flask fermentation with T. aotearoense SCUT27 strains wild type, Δldh, ΔargR, and Δldh/ΔargR. As shown in Fig. 6 and Table S5, compared with those of the wild-type, Δldh, and ΔargR strains, the maximum optical density at 600 nm (OD₆₀₀) of strain Δldh/ΔargR increased by 130% to 135%, 1% to 20%, and 27% to 71%, respectively, and the ethanol production increased by 273% to 320%, 21% to 54%, and 20% to 107%, respectively. These results suggested that knockout of both ldh and argR improved both the cell growth and ethanol production. The carbon distribution in strain SCUT27/ΔΔldh/ΔargR notably flowed to ethanol, with the yield of ethanol (0.36 to 0.40 g/g sugar) increased by 70% to 122%, 11% to 27%, and 1% to 31% compared to that of the wild-type, Δldh, and ΔargR strains, respectively. With xylose as the sole substrate or when it was mixed with glucose, strain SCUT27/Δldh/ΔargR exhibited better performance in sugar use and an obviously shortened lag phase of xylose use caused by carbon catabolite repression compared to the single gene-deletion mutants. These data suggest that T. aotearoense strain SCUT27/Δldh/ΔargR is a potential candidate for ethanol production from various pretreated lignocellulosic hydrolysates.

FIG 6.

Fermentation results for T. aotearoense strain SCUT27 (wild-type) and Δldh, ΔargR, and Δldh/ΔargR mutants with ~25 g/L of different carbon sources: (A) glucose, (B) xylose, or (C) mixed glucose:xylose = 2:1. The data are the means and standard deviations of three replicates. A significant difference in ethanol production was observed between strains Δldh/ΔargR and Δldh or ΔargR (*, P ≤ 0.05; **, P ≤ 0.01).

Fermentation kinetics of strain Δldh/ΔargR with various lignocellulosic hydrolysates.

Hydrolysates of rice straw (RSH), sorghum straw, peanut straw (PSH), wheat straw, and soybean straw pretreated with dilute H₂SO₄ were chosen to assess the ethanol production of strain Δldh/ΔargR. Xylose was the main sugar in the hydrolysates, with some other sugars (e.g., glucose, arabinose, and cellobiose) also present (34, 35). All the hydrolysates were diluted with water to give an initial xylose concentration of approximately 15 g/L. As shown in Fig. 7, as expected, improved ethanol production was obtained using strain Δldh/ΔargR, along with higher sugar use ability, with all the hydrolysates compared to those obtained using the wild type. The ethanol productivity of strain Δldh/ΔargR was improved 169% to 739%. Nearly all the xylose in hydrolysates except RSH was exhausted, which might be attributable to enhanced activity of xylose isomerase and xylulokinase as well as the higher energy level of the cells for xylose transport (5). Both the wild-type and Δldh/ΔargR strains had the ability to tolerate lignocellulose-derived inhibitors in the hydrolysates, such as weak acids, furan derivatives, and phenolic compounds (34, 36). However, strain Δldh/ΔargR had much greater resistance to the RSH and PSH (Fig. 7), possibly because of improved ability to eliminate reactive oxygen species, resulting from the inhibitors (4), and an enhanced ATP level, which enabled greater synthesis of heat shock proteins, pumping of cytoplasmic protons, and transformation of inhibitors (37). Our results suggest that T. aotearoense strain SCUT27/Δldh/ΔargR is a promising bioethanol producer with prominent advantages for coping with the harsh environments of various lignocellulose hydrolysates.

FIG 7.

Fermentation profiles of T. aotearoense strains SCUT27 and SCUT27/Δldh/ΔargR with dilute acid-pretreated biomass hydrolysates in serum bottles.

DISCUSSION

Over the past decade, the CRISPR/Cas system, which is an adaptive immune system in bacteria and archaea, has been repurposed for genome editing and transcriptional regulation in various species (38). However, the majority of the applications are based on the type II CRISPR/Cas9 system, which is hard to apply in thermophiles because of the need for thermostable proteins and the high toxicity of Cas9 (39). After an unsuccessful attempt to use the type II CRISPR/Cas9 system in T. aotearoense strain SCUT27, we turned to the endogenous type I CRISPR/Cas system, which has been explored in a few species of bacteria for efficient genome editing and transcriptional regulation (9, 15–17, 25, 40). Compared with the type II CRISPR/Cas9 system, the native CRISPR/Cas system with longer spacer sequences (~37 nt versus 20 nt for Cas9) decreases the probability of off-target effects (14) and thus avoids undesired mutations.

In this study, we successfully developed and applied the type I-B CRISPR/Cas system for efficient genome editing in strain SCUT27. Because of the deficiency of spacers in strain SCUT27, in silico analysis of the CRISPR arrays was also performed in T. thermosaccharolyticum M0795 to identify potential PAM sites that the two strains may share. Subsequent toxicity assay demonstrated that the bioinformatically predicted PAMs worked effectively in strain SCUT27. With the constitutive promoter Pkan to drive transcription of the synthetic CRISPR mini-array on a plasmid lacking HAs, the transformation efficiency with the target sgRNA (for pyrF) was ~21-fold lower than that with nontarget sgRNA (Table 2). In genome editing, ldh was successfully edited as the templates designed. Desirable mutants could be reproducibly obtained with a high editing efficiency (58.3% to 100%) (Fig. 3; Table S3).

Although high editing efficiency was displayed by the CRISPR/Cas system in strain SCUT27, the overall transformation efficiency was low for practical application as a genome editing tool. Thus, several promoters with different strengths to express sgRNA, or the riboswitch-controlled promoter Pkan-RS_pbuE induced by adenine to control the transcription of sgRNA, were chosen, aiming to improve the editing efficiency. However, we could not simultaneously achieve a high transformation rate and high editing efficiency (Fig. 3B; Table S3); a similar phenomenon was reported in other studies (9, 15, 16). In addition, it was found that use of long HAs or supplementation of HAs had an obviously positive effect on the transformation efficiency, but the editing rate was significantly decreased with the extra addition of HAs (Fig. 3C). However, depending on different genome editing purposes, a tradeoff can be engineered between transformation efficiency and genome editing efficiency.

Multiplex genome editing is a significant and necessary technology to obtain engineered strains with better performance; this cannot be realized easily by traditional homologous recombination. In C. tyrobutyricum, an artificial CRISPR mini-array on one plasmid with spacers for deletion of two genes (spo0A and pyrF) was constructed. However, the editing efficiency of the double deletion vector was significantly lower than that of a single deletion vector (16). Furthermore, as the number of editing loci increases, the size of the single plasmid (containing HAs) increases, increasing the difficulty of vector construction and transformation. To edit multiple loci in a genome efficiently, successive editing with one plasmid, including one spacer, would be a better choice. In strain SCUT27, because there are limited positive selection markers, plasmid curing is necessary for successive editing, and there is an urgent need for a novel way to efficiently screen cells lacking plasmids (Fig. 5). Here, on the basis of the starting strain SCUT27/Δtdk, we obtained strain SCUT27/Δtdk/Δldh in one round of editing. This mutant was plasmid-cured and further edited by the deletion of argR using the CRISPR/Cas system. This is the first report of successive gene editing in T. aotearoense using the type I-B CRISPR/Cas system. The resulting mutant, Δldh/ΔargR, showed improved performance in substrate use and ethanol production and yield, especially with xylose as the substrate (Fig. 6; Table S5). Using five kinds of lignocellulosic hydrolysate, the ethanol concentrations and productivity of strain Δldh/ΔargR were 9.27 to 10.45 g/L and 0.27 to 0.38 g/L/h, respectively (Fig. 7).

The applications of the endogenous CRISPR/Cas system for genome editing in T. aotearoense strain SCUT27 could potentially be much wider than just fragment deletion or insertion. This technique could be applied to introduce other types of mutation, such as point mutations, for which the HAs on the editing plasmid should be designed to introduce changes in the PAM to a nonfunctional motif. Alternatively, substitutions could be introduced into a seed region, the first eight nucleotides of the protospacer, which is crucial for CRISPR targeting (20, 27). In addition, CRISPR interference (CRISPRi), which allows repression of the expression of target genes, has already been performed with the subtype I CRISPR/Cas system (27, 41). The subtype I-B CRISPR/Cas system of Haloferax volcanii lacking the cas3 and cas6 genes was used for gene repression to various extents (41). The type I-E CRISPR/Cas system with deleted Cas3 was also coopted for programmable transcriptional repression in Escherichia coli and showed the strongest repression in targeting promoter regions (40). This strategy offers a simple approach to convert many endogenous type I CRISPR/Cas systems into transcriptional regulators, expanding the available toolkit for CRISPR-mediated genetic control while providing chances for genome-wide screens and pathway engineering.

In conclusion, the protocol of using the endogenous type I-B CRISPR/Cas system for multiplex genome editing described in this study extends the genetic toolbox for T. aotearoense SCUT27 and may be adapted to other microorganisms that carry active endogenous CRISPR/Cas systems. The deletion of both ldh and argR using the CRISPR/Cas system successfully engineered strain SCUT27 for enhanced ethanol production from lignocellulosic hydrolysates.

MATERIALS AND METHODS

Strains and culture conditions.

The strains used and constructed in this study are listed in Table 1. E. coli DH5α used for plasmid construction and propagation was selected and grown at 37°C aerobically in Luria-Bertani (LB) broth or on LB-agar plates supplemented with 50 μg/mL kanamycin or 100 μg/mL ampicillin. T. aotearoense strain SCUT27 and its mutants were cultivated in optimized MTC medium (pH 7.0) (3, 42) at 55°C with shaking at 150 rpm anaerobically in serum bottles. For the selection of transformants, solid MTC medium containing 50 μg/mL kanamycin (MTCK), 1 mM adenine (MTCA), or 50 μg/mL FUDR (MTCF) was used when required and put into an anaerobic chamber (Shellab, USA) for predeoxygenation for approximately 12 h before use. Lignocellulosic hydrolysates were prepared as described in a previous study (5), and 20 g/L of CaCO₃ was added to maintain the pH during the fermentation. The concentration of xylose in hydrolysates was adjusted to ~15 g/L.

Plasmid construction.

The plasmids constructed in this study are listed in Table 1, and the oligonucleotides used are listed in Table S1. Fragments were obtained by PCR using KOD DNA polymerase (Toyobo Co., Ltd.). Restriction endonucleases were purchased from New England Biolabs (Ipswich, MA). Plasmids for editing or overexpression were constructed by double-enzyme digestion and Gibson assembly with a ClonExpress MultiS one step cloning kit (Vazyme, Nanjing, China). The recombinant plasmids were transformed into E. coli DH5α cells (Weidi, Shanghai, China) for amplification and confirmed by colony PCR with rapid Taq DNA polymerase (Vazyme, Nanjing, China) and DNA sequencing (Sagan Co., Shanghai, China). Plasmid isolation and purification were performed using TIANGEN kits.

Pkan was cloned from the upstream region of the kanamycin-resistance gene from template pIKM1. Pcat1 was amplified from the cat1 (CTK_C06520) promoter region in the genomic DNA of C. tyrobutyricum ATCC 25755. P_{Clo1313_1194} was amplified from the Clo1313_1194 promoter region of C. thermocellum (43). The adenine-responsive riboswitch RSpbuE fragment was amplified from Bacillus subtilis (44), and the construction of Pkan-RSpbuE is shown in Fig. S3B. PadhE and ldh were cloned from T. aotearoense strain SCUT27. Every promoter was fused with the ORF of ldh through splicing by overlap extension (SOE) PCR and then cloned into the multiple cloning site of pIKM1 by Gibson assembly, generating vectors pIKM1-Pldh for LDH expression in strain SCUT27/Δldh::ermR. The resulting LDH enzyme activity was tested to characterize the promoter strength. As a control, a 225-bp segment upstream of the ldh coding sequence was amplified as the potential promoter for ldh complementation.

For genome engineering in strain SCUT27 by CRISPR/Cas9, plasmids pLY1 and pLY2 were constructed, containing the thermostable cas9 gene from G. stearothermophilus, respectively, under the control of PadhE and Pkan-RS_pbuE to regulate the toxicity of Cas9 expression. In addition, an sgRNA-expressing module (the motif as described in reference 10) targeting gene ldh under the control of the constitutive promoter Pcat1 was inserted into these two plasmids, as well as a repair template for ldh deletion through homologous recombination. The resultant plasmids pLY1_ldh-HA12 and pLY2_ldh-HA12 were then transformed into cells of strain SCUT27.

The construction of all editing plasmids with the native CRISPR mini-array was based on plasmid pKQ1, which was inserted into the promoter of kanamycin resistance gene Pkan and native putative terminator fragment Ter (298 bp of sequence located at the 3′ end of the largest CRISPR array in SCUT27) into the multiple cloning site of pIKM1. To obtain the killing or editing vectors with the mini-array, promoter Pkan and terminator Ter with the DR and targeting spacer sequences between them were amplified to form an ~950-bp CRISPR expression cassette (Pkan–DR–Spacer–DR–Ter) by SOE PCR followed by Gibson assembly. The repair templates, each containing two HAs (flanking ldh [V518_0188] or argR [V518_1864]), were amplified with different lengths of 250 to 1,000 bp and introduced into the SalI and EcoRI sites of pKQ1. For plasmid curing, the fragment tdk with the putative promoter and terminator (amplified with primers tdk in F/R) was inserted upstream of Pkan on the editing plasmid.

Single guide RNA design and CRISPR/Cas toxicity assays.

By identifying the PAM site (TTA) in the locus of the target gene, the spacer of the sgRNA was chosen with a 37-nt adjacent sequence downstream of the PAM, as shown in Table 2. The length was the same as the average for the spacers in the native CRISPR arrays. In a toxicity assay of the native CRISPR system, a sequence of 37 bp on the sense strand in the pyrF locus (V518_1373) with a GGC motif (instead of TTA) was used as a control for the nontarget spacer. Pkan was chosen for mini-array expression; it is not present in the genome of strain SCUT27, which would avoid unwanted homologous recombination.

Toxicity assay was used to assess functional DNA cutting by the type I-B CRISPR system in strain SCUT27 and was characterized by bacterial survival rate (transformation efficiency). Transformation of the wild-type strain by plasmids with target sgRNA or nontarget sgRNA to the pyrF-coding region was employed as described above. All transformations were done with biological duplicates, and the toxicity (cells-killed rate) was calculated using the following equation: (CFU of nontarget sgRNA − CFU of target sgRNA)/(CFU of nontarget sgRNA) × 100%. The editing plasmid used here was modified to include a repair template consisting of two HAs. The CRISPR editing efficiency was determined using the following equation: (CFU of true-edited mutant on MTCK plates)/(CFU on MTCK plates with plasmid) × 100%.

Transformations and mutant screening.

All plasmids were isolated from E. coli DH5α. Plasmids were transformed into T. aotearoense SCUT27 (OD₆₀₀ of ~0.8), and cells were recovered in MTC medium overnight, as described previously (3, 32). Clones that could grow on MTCK plates were picked into 2-mL centrifuge tubes containing 1.0 mL MTC medium with 50 μg/mL kanamycin and cultured at 55°C for 24 h in an anaerobic chamber (Shellab). Then, they were transferred into 20-mL serum bottles containing 10 mL medium for 24 h at 55°C and retransferred into 100-mL serum bottles containing 50 mL medium to enrich the mutants; adenine or FUDR was added when required. A series of subcultures (5% vol/vol inoculum) was carried out to enrich the strains with desirable homologous recombination or plasmid elimination before selection on plates. The last step was to identify mutants by colony PCR and sequencing of the PCR amplicon when required, using primers shown in Table S1.

Fermentation kinetics and analytical methods.

Flask batch fermentation of strain SCUT27 and its mutants was performed at 55°C and 150 rpm in 100-mL serum bottles containing 50 mL of MTC medium with various carbon sources such as 25 g/L glucose, 25 g/L xylose, or 25 g/L glucose/xylose mixture (mass ratio = 2:1) to mimic lignocellulosic hydrolysates, or with various types of lignocellulosic hydrolysate. During the fermentation, samples were taken every 4 h to detect the OD₆₀₀ of strains and residual sugar, ethanol, acetic acid, and lactic acid concentrations in the broth. Cell density was monitored using a spectrophotometer (PERSEE T6, Beijing, China). The concentrations of glucose and xylose, and production of ethanol and other metabolites, were measured by HPLC (Waters, Milford, USA). The experiments were performed independently at least three times, and data are expressed as the mean and standard deviation (SD). Statistical tests were performed using Statistical Package for the Social Sciences (SPSS ver. 19, IBM Corporation, USA). To compare differences between groups of experimental data, P values were determined by Tukey’s test in a one-way analysis of variance for multiple comparisons.

Enzyme assays.

T. aotearoense SCUT27 and its LDH-overexpressing mutants were cultured anaerobically in modified MTC medium at 55°C with 20 g/L glucose. When OD₆₀₀ reached ~0.8, 5 mL of cell culture was collected by centrifugation and the cell pellet was washed twice with potassium phosphate buffer (PBS) and resuspended in 400 μL of PBS for breaking by ultrasonication on ice. The supernatant of the cells was obtained by centrifugation at 12,000 × g at 4°C for 5 min, and the protein concentration was measured using a Pierce bicinchoninic acid (BCA) protein assay kit (Thermo Fisher Scientific, USA). LDH activity was measured in the direction of pyruvate reduction following our previous study (5). The volume of the LDH reaction system was 210 μL, containing 100 μL of 5 mM NADH, 100 μL of 22.7 mM pyruvate, and 10 μL of cell extract. The reaction was performed at 55°C for 15 min, and NADH consumption was measured at 340 nm. Specific activities are denoted in U/mg protein of cell extract, where one unit of enzyme activity corresponds to one unit value decline of A₃₄₀ per minute (5).

Data availability.

Data will be made available from the corresponding author on reasonable request.