Transcription activator-like effector nucleases (TALENs) are fusion proteins, which combine a TAL effector DNA-binding domain and the DNA-cleavage domain of the FokI restriction enzyme. TALENs can introduce targeted DNA double-strand breaks (DSBs) in the plant genome (Christian et al. 2010; Char et al. 2015). DSBs are repaired by non-homologous end joining or homologous recombination pathways, which result in deletions, insertions, or substitutions at the break site to knock out or knock in target genes (Waterworth et al. 2011). TALEN technology for genome editing has been utilized for many crop species such as maize, rice, wheat, and soybean with a wide variety of targeted traits (reviewed by Malzahn et al. 2017 and Zhang et al. 2018). TALENs are considered to have some advantages over the clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated 9 (Cas9) system, such as specificity of binding to target DNA with fewer off-target effects and a less restrictive requirement of target DNA sequences (Nishizawa-Yokoi et al. 2016). Several successful TALEN-mediated null segregants with desirable mutations were able to be created in T₁ and T₂ generations (Haun et al. 2014; Char et al. 2015; Shan et al. 2015). A study aimed to increase the oleic acid content of soybeans was carried out by introducing mutations in two fatty acid desaturase 2 genes (FAD2-1A and FAD2-1B) using TALEN technology (Haun et al. 2014). These investigators obtained three mutant lines with heritable FAD2-1 mutations and absence of the TALEN transgene, thereby increasing oleic acid content to 80%. Shan et al. (2015) edited the betaine aldehyde dehydrogenase gene (BADH2) using TALENs to generate six BADH2 mutations in the T₁ generation, and four of them were confirmed to have efficiently transmitted to the T₂ generation. However, some undesirable effects may be encountered during TALEN editing in plants, such as multiple copy numbers of TALENs, position effects of TALEN integration sites, partial deletions of TALENs, variable expression of TALEN arms, mosaic patterns, and mutation chimerism. These potential effects can produce unpredictable results, which can lead to low selection efficiencies and increased time to obtain desirable targeted mutagenesis (Li et al. 2012; Christian et al. 2013; Char et al. 2015; Shan et al. 2015; Nishizawa-Yokoi et al. 2016).

We not only face many technical challenges with genome-editing tools, but also public acceptance of the technology and government regulatory policies for its adoption. As of 2019, 24 gene-edited crops covering 16 crop species have been ruled outside the scope of traditional genetic-modification policies by the US Department of Agriculture. Cultivation of a high oleic acid soybean created by the Calyxt seed company using TALENs technology began in 2018, and its high oleic acid soybean oil was commercially marketed in the United States (reviewed by Park et al. 2019). In contrast, the Court of Justice of the European Union has ruled that gene-edited crops should be subjected to the same stringent regulations as conventional genetically modified (GM) organisms (Callaway 2018). On the one hand, they were concerned that genome-edited crops could be involved in such potential risks as increased weediness or gene flow, resistance evolution, and herbicide carryover to rotational crops, which has been observed on herbicide-resistant GM rice in some regions of the US, Brazil, and Italy (reviewed by Lassoued et al. 2019). On the other hand, the importance of potential unintended effects of the modifications with adverse consequences may have been overstated, considering more than 20 years of experience on the safety assessment of GM crops (reviewed by Kleter et al. 2019). Some genome-editing technologies create small mutations in the host genome that might also be achieved by traditional breeding and mutagenesis. The critical argument for risk assessment of new plant breeding technologies (NPBTs) mainly concentrates on detection methods for unintended consequences; thus, untargeted metabolomics could be incorporated as part of the assessment protocols for future biotech crops (Christ et al. 2018). Genome editing provides an invaluable tool for high-precision molecular breeding of crops. However, it requires a large amount of scientific research and data collection involving technical applications and omics data outputs for edited events, which helps alleviate public concerns regarding food and environmental safety due to gene-editing technology.

Here, we report the development of bialaphos resistance (bar)-knockout rice from a herbicide-resistant rice line (Ba15) via targeted knockout of bar using the TALEN method. We selected bar-mutated null segregants and examined transcriptome changes in these lines compared with their donor variety. The resistance gene, bar, which has been isolated from Streptomyces hygroscopicus, encodes the enzyme phosphinothricin acetyltransferase (PAT) that inactivates L-phosphinothricin (L-PPT) by transferring the acetyl group from acetyl-coenzyme A to the free amino group of L-PPT, yielding N-acetyl-L-PPT (Abdeen and Miki 2009). We targeted bar not only because glufosinate herbicide-resistant GM crops account for approximately 46% of the total cultivation area of GM crops worldwide (James 2018), but also because it is the most extensively used selectable marker gene in scientific literature. Through vector construction, transformation methods, progeny selection, and transcriptome analysis of TALEN-mediated bar mutant lines, we hope that our results can highlight potential problems in the application of this technique, and provide an approach to identify unintended effects resulting from both predictable and unpredictable off-targets.

Binary vector construction

The TALEN-encoding binary vector pPZP-TaleLR was constructed targeting the herbicide resistant gene bar in this work. The T-DNA construct consisted of the TALE-L and TALE-R expression cassettes, which were both driven by a double CaMV 35S (2 × P35S) promoter with nopaline synthase (Tnos) terminator for TALE-L and potato protease inhibitor II (TPinII) terminator for TALE-R, respectively. A plasmid pTOPO-35SBAR was cloned and digested by EcoRI, and then inserted the expression cassette of P35S:: bar::Tnos into pPZP-TaleLR to construct pPZP-TaleLR-bar binary vector.

Plant materials and biolistic rice transformation

A herbicide-resistant transgenic fixed rice line (Ba15) was used as the recipient variety for the TALEN vector transformation. The Ba15 rice line was developed by Dr. Myung-Ho Lim's group and confirmed to have a single copy of T-DNA harboring the CaMV 35S:: bar :: Tnos expression cassette inserted into an intergenic site on chromosome 4 downstream of the 2.912-kb Os04g0553300 gene in the Dongjin rice variety (Oryza sativa L.).

The sterilized mature rice seeds were sown on N6 medium containing 2,4-D herbicide in a dark room at 28℃. After four weeks, the compact embryogenic calli were transformed using a PDS1000/He particle bombardment system (Bio-Rad, Hercules, CA, USA) with a particle diameter of 0.6 mm and helium pressure of 1100 psi. The plasmid DNA of TALEN-encoding T-DNA binary vector (pPZP-TaleLR-bar) was mixed at 1:1 molar ratio prior to bombardment. After bombardment, medium containing a low dose of 2 ppm PPT was used for selection, but PPT-free medium was used for plantlet regeneration. After 3-4 months of cultivation, T₀ transgenic seedlings were transferred into soil at the GMO greenhouse of the National Institute of Agricultural Sciences, and T₁ seeds were harvested from each line.

TALEN construct and bar mutation analyses

Genomic DNA was extracted from young leaves by the CTAB method. PCR strategies were designed for TALEN construct confirmation using two sets of PCR primers, namely a forward primer F1 (5′-CTACAGCGGTGGGTA CAATCT-3′) and reverse primer R1 (5′-CCCGATCTA GTAACATAGATGACAC-3′) for TALE-L, and F1 and a reverse primer R2 (5′-GCAGGGATAAAAGCAATCTA TGTAA-3′) for TALE-R. The F1 primer was used for both TALE-L and TALE-R detection because the two nuclease cassettes possess regions of identical nucleotide sequence. R1 and R2 were designed to anneal to specific sequences in TPinII and Tnos terminators, respectively. The predicted PCR product sizes are 600 bp and 778 bp for TALE-L and TALE-R, respectively. PCR was performed following the procedures provided with the Ex Taq DNA polymerase (Takara BIO, Japan).

The bar sequence containing the TALEN-target site was amplified by PCR using Phusion High Fidelity DNA Polymerase (Thermo Fisher Scientific, Inc., Waltham, MA, USA) and the following primers: bar-F-5′-ATGAGCC CAGAACGACGCC-3′ and bar-R-5′-TCAGATCTCGG TGACGGGC-3′. To confirm the presence of sequence mutations, eluted and 3′-A-tailed PCR products were subcloned into the pGEM-T vector (Promega), and ten positive clones per line were then sequenced by Sanger sequencing. Furthermore, PAT protein expression in T₀, T₁, and T₂ plants was evaluated by Immuno-strip tests (Agdia Inc. USA).

Microarray and differential expression analyses

For two transgenic bar-mutated rice lines (R6 and R9), the T₁ generations were reproduced and three lines (R6-2, R9-15, and R9-20) were selected for T₂ plant production. Finally, RNA from three TALE-free bar-knockout lines (6-2-4, 9-15-4, and 9-20-2; T₂ generation) and the recipient herbicide-resistant Ba15 at the same growing stage was isolated for microarray analysis. RNA was isolated using TRIzol/chloroform (Invitrogen, Carlsbad, CA, USA) following the manufacturer’s protocols. DNA removal and concentration measurements were implemented using a TURBO DNA-Free DNase kit (Ambion, Austin, TX, USA) and Nanodrop ND 1000-spectrophotometer (Thermo Fisher Scientific, Wilmington, NC, USA), respectively. Total mRNA from each sample (0.5 mg each) was subjected to microarray analysis following the description by Chae et al. (2017).

The platform-formatted microarray data processing was carried out on the CLC genomic workbench software, version 9.0 (www.qiagenbioinformatics.com/). Differentially expressed genes (DEGs) with an absolute value of Log2 fold-change > 2, and a false discovery rate (FDR) P-value ≤ 0.05 were identified from each bar-knockout line relative to the recipient Ba15. By using the “Set up experiment” function, we classified the four samples into two groups, the “bar-knockout” group and “recipient” group. DEGs from the bar-knockout group were extracted compared to those from the recipient group based on a filtering parameter of absolute value of Log2 fold change > 2 and FDR P-value ≤ 0.005. Afterward, functional enrichment analysis involving gene enrichment and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways was performed to identify significantly enriched DEGs in gene ontology (GO) terms and metabolic pathways. Data were compared with the whole-transcriptome background using a gene functional annotation tool (https://david.ncifcrf.gov/tools.jsp) at a Bonferroni-corrected P-value ≤ 0.005.

TALEN design and recovery of rice plants with bar-mutations

The paired TALEs (TALE-L and TALE-R) with binding sequences of 5′-CCGCCGTGCCACCGAGGC-3′ and 5′-AGTGGTTGACGATGGTGC-3′ allowed TALE-L targeting 30 bp to 47 bp (18 nt) and TALE-R targeting 65 bp to 83 bp (18 nt) from the start codon of bar with a 17-bp DNA spacer were designed by using the TAL Effector Nucleotide Targeter 2.0 program (Doyle et al. 2012) as shown in Fig. 1A. The TALE architecture contains 136 aa of the N-terminus (truncation at residue 152 of the original 288-aa N-terminus), 17.5 repeats of RVD (repeat variable di-residues) arrays targeting the given DNA sequences, and a 231-aa C-terminus. In addition, we inserted two nucleotide sequences encoding a FLAG-tag and nuclear localization signal (NLS) upstream of the N-terminus, and the heterodimeric FokI endonuclease cleavage domain downstream of the C-terminus of the TALEs (Fig. 1B).

Figure 1. Schematics of TALE structure and TALE backbone transformation vector for herbicide-resistant bar gene knockout. (A) Simple diagram of TALE targeting of bar gene showing TALE-L and TALE-R with their repeat variable di-residue (RVD) encoding NN, HD, NI, and NG to recognize nucleotides G, C, A, and T, respectively. (B) The TALEN architecture. NLS, nuclear localization signal. (C) Transformation vector construction of pPZP-TaleLR-bar.

To produce TALEN-based bar-knockout rice transgenic lines, we transformed the binary vector of pPZP-TaleLR-bar (Fig. 1C) into the recipient herbicide resistant rice line Ba15 by particle bombardment. The transformed callus was grown on low-dose PPT selection medium. As a result, a total of 14 bar-mutated T₀ rice plants were recovered from 41 transgenic rice plantlets, accounted for 34.1% of TALEN editing efficiency on bar. Most TALEN-mediated mutation sites were observed on non-target sites of the T₀ transgenic plants, such as start codon (1-3 bp) and stop codon (544-552 bp) of bar, resulting in frameshift and deletion mutations. Single nucleotide polymorphisms (SNPs) involving A/G and G/A substitutions at 4 bp, 26 bp, 145 bp, and 437 bp of bar were observed in the T₀ plants of R6, R33, R28, and R32, respectively (Supplementary Table 1). In addition, we suspected a possibility that some T₀ plants harboring heterozygous or chimeric bar-mutations. Thus, immuno-strip test for PAT protein expressions were performed and results indicated that seven plants of R1-3, R9, R14, R33, and R41, three plants of R6, R7, and R16, and four plants of R17, R24, R28, and R32 showed normal, downregulated, and absent PAT protein expression among 14 of bar-mutated T₀ plants, respectively. Furthermore, failure of PAT protein ex-pressions were also detected on other eight T₀ plants of R15, R20, R21, R26, R29, R31, R35, and R37, which were confirmed without bar sequence mutations (Fig. 2A).

Figure 2. PAT protein expression for bar-mutated rice lines in T₀ (A), T₁ (B), and T₂ (C) generations. Red asterisks indicate transgenic lines selected for further analysis.

Transmission of TALEN-induced bar-mutations to T₁ and T₂ generations

To see whether the TALEN-mediated bar mutations were transmitted to the next generation, we harvested 11, 13, 40, and 31 T₁ seeds from the R1, R3, R6, and R9 transgenic plants, respectively, and sowed all of them in the greenhouse. Analysis of PAT protein expression revealed changes in the T₁ plants from R6 and R9, but no differences in the T₁ plants from R1 and R3. For T₁ progenies from R6, we detected 26 plants failure in PAT protein expression, 10 plants with downregulated PAT expression, and four plants with normal PAT expression, respectively. For T₁ progenies from R9, we found three plants failure in PAT protein expression, but six and 22 plants respectively showed downregulated and normal PAT protein expression (Fig. 2B).

In addition, bar sequence mutations were analyzed for four T₁ populations and results indicated that a total of 16 and 13 sequence mutation types were respectively identified in nine and four T₁ progenies from R6 and R9, but no mutations in T₁ progenies from R1 and R3. To compare mutation types and sites in T₀ plants, T₁ progenies showed de novo mutations and multi-site mutations per line (Table 1). Among mutations in T₁ progenies, two SNPs in R6 (G to T at 58 bp and G to A at 60 bp) and one SNP in R9 (G to A at 57 bp) were identified within the spacer sequence between the two TALEN recognition sites. In addition, we found G/T and A/G substitutions with high frequencies that occurred on random sites in the bar target.

Table 1 . TALEN-mediated bar mutations in T₀ R6 and R9 plants and their T₁ generations.

Mutant lines	Generation	Mutation types	Mutation site (bp)	PAT expressionz)	TALE-L/Ry)	Generation	Mutation sites (bp)	Amino acids	PAT expression	TALE-L/R
R6	T0	Start codon deletion/insertion	‒1/‒3	＋/‒	‒/＋	T1 (R6,1-40)	＋1/‒1/‒2/‒3	Frame shift	＋/‒	‒/＋
		SNP	A/G (4)				G/T (3), A/G (4), G/T (58), G/A (60)x), G/A (199), G/T (235), A/T (552)	Substitution	‒
		Stop codon deletion	‒				‒2/‒3/‒4/‒5/‒8	Deletion	＋
R9	T0	Start codon deletion/insertion	‒1	＋	＋/＋	T1 (R9, 1-31)	‒1/‒3	Frame shift	＋/‒	＋, ‒/＋
		SNP	‒				G/T (3), G/A (22), G/A (57), G/T (137), G/A (167), G/A (219), G/A (412), T/G (550)	Substitution	‒
		Stop codon deletion	‒1				‒1/‒4/‒5	Deletion	＋

^z)Active (＋) and inactive (−) PAT protein expression.

^y)Presence (＋) and absence (−) of TALE-L and TALE-R constructs.

^x)Underlined characters indicate mutations in the TALE:FokI cleavage site.

We selected three T₁ plants of R6-2, R9-15, and R9-20 with complete disruption of PAT protein expression and high seed-setting rates for T₂ plant productions. Twenty-seven of T₂ seeds derived from R6-2, 23 seeds from R9-15, and four seeds from R9-20 were sown in the greenhouse. We analyzed PAT protein expressions by immuno-strip and results indicated that these 54 of the T₂ plants had successfully inactivated bar gene expression (Fig. 2C). Three T₂ lines of R6-2-4, R9-15-4 and R9-20-2 were selected for bar sequence analysis, and results revealed each one SNP of G/T at 235 bp for R6-2-4 and G/A at 167 bp for R9-20-2, causing amino acid substitutions of Ala to Ser and Arg to His, respectively. For the R9-15-4 line, a 1-nt deletion was observed in the start codon of bar, causing a frameshift mutation.

Generation of TALEN-free bar-mutant rice lines

To obtain rice lines harboring the bar mutations but not the TALEN construct, the PCR-based assay was used as follows: primer pair F1/R1 to amplify the C-terminus of TALE-L and T-nos, F1/R2 to amplify the C-terminus of TALE-R and T-PinII (Fig. 3A). For R6 (T₀) and R6-1-40 (T₁) plants, only the TALE-R construct and not the TALE-L construct was detected. As shown in Fig. 3B, five T₁ plants for R6 and one T₁ plant for R9 produced a single PCR product of approximately 3-4 kb, which should have been amplified by the F1 forward primer in the TALE-R construct and the R2 reverse primer in the Tnos of the bar expression cassette. TALE-L/R constructs were both confirmed in the R9 (T₀) plant. Seven T₁ plants were selected for TALE construct confirmation, and the results indicated that one line (R9-18) contained both TALE L/R constructs, whereas another line (R9-12) was TALE-DNA free. Only the TALE-R construct was detected in five other lines (Fig. 3B). Ten T₂ plants derived from R6-2, R9-15, and R9-20 were confirmed free of TALE-L/R DNA (Fig. 3C), suggesting the T-DNA construct had been eliminated from T₁ to T₂ by segregations.

Figure 3. PCR strategy for confirmation of TALE-DNA constructs in bar-knockout transgenic rice lines R6 and T9, and their T₁/T₂ generations. (A) Simple diagram of TALE construct and PCR design. (B) and (C) TALE construct analysis for T₀, T₁, and T₂ plants of R6 and R9, respectively.

Transcriptome profiles in TALEN-mediated bar-knockout transgenic rice lines

Using microarray analysis, transcriptome profiling was performed on three transgenic rice lines (R6-2-4, R9-15-4, and R9-20-2) and the recipient variety, the herbicide-resistant transgenic rice Ba15. The three lines were confirmed to be TALEN-DNA-free and to have inactive bar expression. As shown in Fig. 4A, we identified 1,523 DEGs from R9-15-4, 1,090 DEGs from R9-20-2, and 978 DEGs from R6-2-4, compared to their recipient line Ba15. A hierarchical clustering of all DEGs revealed that most were line-specific. To better understand transcriptome changes in the TALEN-mediated transgenic lines, line-specific DEGs overlapping in two or three TALEN-mediated bar-knockout lines were extracted by using an adjusted P-value cut off of P ≤ 0.005 (Fig. 4B). The results revealed that 144 DEGs including 93 upregulated and 51 downregulated DEGs were identified from the bar-knockout group relative to the recipient group.

Figure 4. Differentially expressed genes (DEGs) detected from three bar-knockout lines compared to their recipient variety Ba15. (A) Hierarchical clustering of total DEGs based on a threshold of Log2 fold change > 2 and false discovery rate (FDR) P-value ≤ 0.05. (B) Upregulated and downregulated DEGs based on a threshold of Log2 fold change > 2 and FDR P-value ≤ 0.005.

Gene set enrichment analysis revealed 72 enriched DEGs that were classified into the two GO terms “cellular component” and “molecular function.” In the cellular component category, the GO term “plasma membrane” (GO: 0005886) involved 20 DEGs, including 10 upregulated and 10 downregulated DEGs with the highest enrichment scores. Sixteen DEGs (10 upregulated and 6 downregulated DEGs) that belonged to the GO term “extracellular region” (GO: 0005576) showed significant enrichment scores. In the molecular function category, the most significantly enriched GO term was “response to polysaccharide binding” (GO: 0030247) involving five DEGs comprised of four upregulated and one downregulated DEG (Table 2 and Table 3). Furthermore, DEG analysis revealed significantly downregulated bar expression with Log2 fold changes from -12.263 to -13.009 in the bar-knockout lines R9-15-4, R9-20-2, and R6-2-4. These results were consistent with PAT protein expression levels assessed by immuno-strip tests. Furthermore, we did not detect significantly enriched DEGs which were involved in KEGG pathway and closely related to transcription factors (Table 3).  

Table 2 . Gene Ontology (GO) annotation and GO enrichment analysis at P ≤ 0.005 of DEGs detected from the bar-knockout rice group compared to the recipient group.

Gene Ontology	Terms	Cluster frequencyz)	P-value
Cellular Component (CC)	Plasma membrane (GO:0005886)	27.78%	0.003531717
	Extracellular region (GO:0005576)	22.22%	0.004769093
Molecular Function (MF)	Polysaccharide binding (GO:0030247)	6.94%	0.00029600

^z)Cluster frequency was calculated as ratio of enriched gene numbers of each term to a total of 72 DEGs that were enriched for GO terms.

Table 3 . Enriched DEGs detected from bar-knockout transgenic rice group compared to the recipient herbicide-resistant line Ba15.

SEQ_ID	Log2 fold changes			Descriptions	GO termsz)	TAIRy)
	R9-15-4	R9-20-2	R6-2-4
Bar	−12.263	−12.419	−13.009	A bialaphos-resistantgene	-	TALEN-target
Os03t0392600-01	−9.128	−8.358	−8.999	Peptidase S10, serine carboxypeptidase family protein	CC	AT2G27920
Os09t0358000-00	−8.633	−6.202	−8.516	Similar to OsD305	CC	AT1G51890
Os11t0695000-01	−7.744	−7.639	−6.600	Similar to Leucine Rich Repeat family protein	CC	AT3G47570
Os09t0356800-01	−7.271	−3.446	−7.168	Protein kinase, core domain containing protein	CC	AT1G51850
Os11t0514500-01	−6.766	−8.222	−6.396	Sorghum bicolor leucine-rich repeat-containing extracellular glycoprotein precursor	CC	AT5G21090
Os02t0483000-00	−6.648	−5.603	−5.616	Similar to fasciclin-like arabinogalactan protein 8	CC	AT3G12660
Os11t0641500-01	−4.364	−4.436	−5.929	Cupredoxin domain containing protein	CC	AT3G09220
Os03t0184550-01	−3.596	−2.900	−2.921	Similar to Dihydroflavonol-4-reductase	CC	AT4G33360
Os10t0343400-01	−3.207	−3.590	−2.886	Cellulose synthase family protein	CC	AT3G03050
Os10t0142600-00	−2.762	−3.686	−4.315	Protein kinase, catalytic domain domain containing protein	CC, MF	AT1G21270
Os02t0740700-01	−2.512	−3.015	−2.516	Peptidase M10A and M12B, matrixin and adamalysin family protein	CC	AT1G24140
Os11t0115350-01	−2.227	−3.853	−3.681	Similar to Non-specific lipid-transfer protein 2	CC	AT2G38540
Os12t0228700-00	−2.156	−3.158	−3.029	Similar to 32 kDa protein	CC	AT1G73040
Os04t0175600-01	2.005	2.702	2.029	Similar to o-methyltransferase (EC 2.1.1.6) (Fragment)	CC	AT5G54160
Os12t0583300-01	2.390	2.923	2.099	Peptidase aspartic, catalytic domain containing protein	CC	AT2G03200
Os05t0318700-01	2.527	3.501	4.265	Similar to Resistance protein candidate (Fragment)	CC	AT3G51550
Os09t0339000-01	2.718	4.084	2.400	Protein kinase, core domain containing protein	CC	AT5G10530
Os11t0605100-01	3.197	4.683	5.049	NB-ARC domain containing protein	CC	AT3G14470
Os02t0111600-01	3.245	3.698	2.199	Serine/threonine protein kinase-related domain containing protein	CC, MF	AT1G21230
Os12t0431100-01	3.366	2.913	2.211	ATPase, AAA-type, core domain containing protein	CC	AT5G40010
Os07t0539900-01	3.497	3.157	2.956	Similar to Beta-1,3-glucanase-like protein	CC, MF	AT4G26830
Os01t0944500-00	3.682	3.588	3.094	Glycoside hydrolase, family 17 domain containing protein	CC, MF	AT3G57240
Os01t0660200-01	3.759	4.487	2.663	Acidic class III chitinase OsChib3a precursor (Chitinase) (EC 3.2.1.14)	CC	AT5G24090
Os06t0566300-00	3.867	2.056	2.389	Similar to zinc transporter 4	CC	AT1G10970
Os01t0713200-01	4.312	3.960	2.744	Similar to Beta-glucanase	MF	AT3G57260
Os06t0143950-00	4.396	5.334	6.007	Non-protein coding transcript	CC	AT1G79990
Os02t0550800-01	4.446	4.496	3.998	Ammonium transporter family protein	CC	AT2G38290
Os12t0628600-01	4.760	3.073	2.335	Similar to Thaumatin-like pathogenesis-related protein 3 precursor	CC	AT4G11650
Os07t0131375-00	6.228	5.106	5.720	Protein kinase, catalytic domain domain containing protein	CC	AT2G37710
Os11t0214700-00	7.629	3.431	3.181	Plant disease resistance response protein domain containing protein	CC	AT5G42510
Os01t0382000-01	8.453	6.368	6.203	Similar to Pathogenesis-related protein PRB1-2 precursor	CC	AT4G33720

^z)CC, cellular component; MF, molecular function.

^y)Means The Arabidopsis Information Resource.

In 2018, the first TALEN-edited high oleic acid soybean with approximately 80% oleic acid, which had been developed by Calyxt in the USA, was cultivated, and high oleic soybean oil was successful commercialized and released on the US market (Reviewed in Park et al. 2019). However, safety assessment of gene-edited crops is still a controversial subject in Europe and most Asian countries. In the present study, we edited the herbicide-resistant gene bar using TALENs to produce several bar-knockout rice lines from the herbicide-resistant transgenic line Ba15, and then evaluated transcriptome changes in these rice lines. We believe these results will provide useful information on the safety assessment of GM crops obtained from NPBTs, such as those using TALENs and CRISPR-Cas9.

In this report, we identified some challenging problems. First, we generated 41 T₀ plants comprised of 14 lines with bar sequence mutations and eight lines with inactive bar expression but without bar sequence mutations. Sequence analysis of the T₁ plants with inactive or reduced bar expression indicated that only nine plants from R6 and four plants from R9 showed bar sequence mutations. Although only a small number of positive clones were sequenced and more mutations could exist, we still suspect that there is a strong probability that off-target effects occurred within the CaMV 35S promoter in the bar expression cassette. To support this viewpoint, we sequenced the CaMV 35S promoter of 20 T₁ progenies from the R6 line, in which bar sequence mutations were not detected. One deletion, four SNPs, and one MNP (multiple base substitution) were observed in four of the lines (Supplementary Fig. 1).

Second, most mutations in the T₀ and T₁ plants occurred in non-target sites of the bar gene. Only three SNPs from R6 and R9 were detected within the spacer (48 bp to 64 bp) between the two TALE DNA-binding sites. Additionally, start and stop codon deletions in the bar gene accounted for a large proportion of the total mutation types. TALEN-induced mutagenesis generally depends on TALEN architecture, transformation methods, and plant species. The endonuclease organization and DNA-binding domain, TALE DNA-binding orientation, and the requirement of a thymine at position zero (T₀) had been shown previously to achieve more efficient DNA cleavage in the target region (Miller et al. 2011; Christian et al. 2012; Beurdeley et al. 2013). The TALE structure used in this study consisted of 17.5 RVD repeats with a 136-aa truncated N-terminus and FokI domain in the C-terminus in order to recognize the 17-bp DNA spacer of the two binding sites. The TALE: FokI fusion proteins bind to adjacent DNA sites starting at 5′ T0, with each located on the sense or antisense strand of the DNA in a tail-to-tail orientation. Such architecture has been studied for a wide range of DNA spacers of 10-35 bp with a high efficiency of cleavage and genome editing (Juillerat et al. 2014; Schreiber and Bonas 2014). In addition, TALE-L and TALE-R were both under control of the CaMV 35S promoter, with the aim to generate constitutive and equal expression of the TALENs. Hence, the TALE architecture should be able to edit the bar gene effectively. In addition, we considered another factor that could possibly cause non-target mutations. In the early stages of this study, we used a transformation vector that only contained the TALE expression cassette. After three days’ co-culture with Agrobaterium harboring TALE-DNA transgene cassette, rice calli were grown on selection-free medium. However, due to very low selection efficiency, we decided to modify the TALEN transformation vector to harbor a bar expression cassette. We adopted a low-dose PPT selection (half the PPT concentration commonly used) to identify positive calli on the selection medium. In theory, TALE:FokI should edit all bar binding sites in the rice genome, including the original site in the recipient line and the transformed site afterward. However, we suspect the rice calli that survived on the low-dose PPT selection medium only carried small mutations aimed to maintain partial function of the PAT protein. Rice calli with targeted mutations of bar such as large InDels were eliminated during the selection process. These would provide only a partial representation of the bar mutations occurring on non-target sites.

Finally, many de novo mutations and sites were identified from self-pollination-derived T₁ progenies, but were not inherited from the T₀ plants. Most of the T₁ lines had more than one mutation site. In addition, we analyzed bar sequence mutations in three selected bar-knockout T₂ plants that were TALE-DNA-free, and the results indicated that these lines were still heterozygous carrying 25-50% of the wild type bar gene sequence. Furthermore, we failed to detect the TALE-L construct in the T₀ plant of the R6 line. We considered the possibility of partial TALE-system transformation where only TALE-R was transformed into the rice callus. Likewise, all of the T₁ progenies from R6 were confirmed to possess the TALE-R structure, without segregation of the TALE construct. The R9 line showed somewhat different results for the TALE construct. Both TALE-L and TLAE-R were confirmed for the T₀ plant. An irregular segregation was observed in seven T₁ progenies in that one line (R9-18) contained both TALE constructs one line (R9-12) contained no TALE DNA, and the other five lines only contained the TALE-R construct. To produce these bar-knockout rice lines, we used particle bombardment for delivery of the vector DNA into the rice genome. We suspect that R6 and R9 contained multiple integrated copies of the TALE-DNA cassette, and some copies had a high probability of being partial vector integrations into the recipient rice line, such as only involving the TALE-R expression cassette. Particle bombardment remains the principle direct DNA transfer technique and has been reported to integrate T-DNA highly efficiently into plant genomes. However, the tendency to generate large transgene arrays containing rearranged and broken transgene copies is a noticeable issue (Altpeter et al. 2005). Multi- and/or partial-copies of T-DNA integrations may generate non-target and chimeric mutations in T₀ plants. The TALE-R expression cassette was detected in all T₁ plants from R6 and most of the T₁ plants from R9, which could cause ongoing cleavages in the T₁ progenies to produce many de novo mutations. Likewise, heterozygous lines accounted for a large proportion of the T₂ plants. Christian et al. (2013) reported TALEN-mediated mutations in Arabidopsis. They found that the continuous occurrence of mutations in individual cells at various stages of plant development may cause a mosaic pattern in mutant plants. To obtain homologous plants with the desired editing, one of the challenges arises from variable nuclease expression in transgenic lines due to variations in TALE copy number and position effects. A study on the creation of fragrant rice by TALEN technology to disrupt OsBADH2 gene expression was reported (Shan et al. 2015). These investigators achieved mutant segregations with TALEN-DNA-free lines in T₁ and T₂ generations. However, they found that chimeric mutations in a single target site might result from delayed cleavage in embryogenic cells that did not participate in gamete production. Moreover, some new InDel mutations among T₁ and T₂ offspring could be contributed to continued TALEN cleavage of the target. Combining our findings with those of the previous study, we believe that low copy numbers of TALEN transgene integrations and Agrobacterium-mediated transformation methods may allow the acquisition of mutated null segregants in early generations, making it similar to the development of conventional genetically modified (GM) plants to some extent.

Many studies have focused on the possible occurrence of off-target edits in mutant plants edited by TALENs or the CRISPR/Cas9 system, which have been reported to occur at a low frequency in plants (Zhang et al. 2014). In the present study, potential off-targets in the whole rice genome were searched using the two TALE-DNA binding sequences on the platform of the PROGNOS online program (Fine et al. 2013). We identified 384 potential off-target sites for the bar gene with parameter settings of a 6-bp mismatch and 10-bp to 30-bp spacers. However, none of these hits were located in exons, introns, 3′-untranslated regions, 5′-untranslated regions, or promoter regions of the rice genome (data not shown). The predictable off-target sites can be detected by in Silico predictions using online analysis platforms, but it still remains a challenge to identify unexpected effects which can be triggered by unpredictable off-target edits. Therefore, transcriptome profiling and comparisons should be an effective strategy for monitoring unintended consequences of gene-edited crops. Herein, we also performed microarray analysis on three bar-knockout null segregants that were derived from T₂ generations of the two TALEN-based mutated T₀ rice plants R6 and R9. Although hierarchical clustering showed a number of DEGs, most of them were line-specific. We identified 72 enriched DEGs in the bar-knockout group versus the recipient group. Gene set enrichment analysis indicated two GO terms (“cellular component” and “molecular function”) involving 31 DEGs that had significant enrichment scores, but the changes in expression of these DEGs were irregular in each GO term. Furthermore, these DEGs were not involved in the KEGG pathways and response to transcription factors. Hence, we conclude that the transcriptome changes in the bar-knockout lines may have resulted from the in vitro cell culture and transformation processes. Thus, it can be seen that TALEN-mediated editing occurred only on the bar gene and CaMV 35S promoter and had little effect on the whole transcriptome of rice.  

This work was supported by a grant (PJ01432206 and PJ01194401) from the National Institute of Agricultural Sciences (Rural Development Administration, Republic of Korea).

The authors declare that there is no conflict of interest.