Introduction

Genetic/allelic variations in plants, either occur naturally or artificially, may cause mutations. Artificial mutagenesis methods include chemical/physical treatments and novel plant breeding technologies, such as genome editing. Genome editing system utilizes site-specific nucleases to introduce precisely targeted double-strand breaks (DSB), and the desired modifications are obtained by subsequent endogenous DSB repair machinery robustly. Those site-specific nucleases include zinc finger nucleases (ZFNs), transcription-activator like effector nucleases (TALENs), and clustered regulatory interspaced short palindromic repeats associated protein Cas9 (CRISPR/Cas9) (Lusser et al. 2012; Zhu et al. 2017). Among them, the CRISPR/Cas9 system is more compelling, and particularly, it can modify multiple plant genes concurrently (Feng et al. 2014; Xie et al. 2015; Ishizaki 2016; Zong et al. 2017). CRISPR/Cas9 system introduces a single guide RNA (sgRNA) to guide the Cas9 protein to target genomic DNA consisting of the protospacer adjacent motif (PAM) to generate double-strand breaks (DSB). Those DSBs in plants are repaired mainly through error-prone non-homologous end joining (NHEJ), which mostly generates insertion/deletion (INDEL) frame-shift mutations with only a few base pairs (bp) variation, leading to loss of function via premature translation termination (Ren et al. 2016; Pan et al. 2016; Zhu et al. 2017). However, Cas9 protein can also target homologous genomic sites of sgRNA concurrently, which may cause unintended off-target mutations with one to few bp variations (Jinek et al. 2012). These small mismatches in the genome cannot be identified through agarose-gel electrophoresis (Denbow et al. 2017), which averts the mutation screening and effects of subsequent functional analysis. Therefore, there is an urgent need to develop an effective, reliable, and inexpensive method for parallel analysis of CRISPR/Cas9-induced on- and off-target INDELs and to distinguish them from those that naturally occurred for plant science research.

Many different methods have been first developed in other model systems rather than plants to detect the mutations of target loci, all have their particular limitations (Zischewski et al. 2017). Most frequently used methods for INDELs identification include: (1) enzyme mismatch cleavage (EMC) assay; (2) fluorescence-based high-resolution melting analysis (HRMA) technique; and (3) modified migration-based heteroduplex mobility assay (HMA) (Thomas et al. 2014; Vouillot et al. 2015; Zischewski et al. 2017). EMC assay utilizes the most popular enzyme T7 endonuclease 1 (T7E1) or Surveyor nuclease to cleave heteroduplex DNA at mismatches with one or few nucleotides, and resulting mutation with these small mismatches can be analyzed by agarose gel electrophoresis (Vouillot et al. 2015). This method is easy to handle, cost-effective, suitable for large INDEL detection (Vouillot et al. 2015; Zischewski et al. 2017). But it is less sensitive; it cannot identify homozygous mutations, and it is not suitable for polymorphic locus analysis (Kim et al. 2011; Huang et al. 2012). HRMA method characterizes DNA samples based on their disassociation behavior and detects small sequence differences in PCR amplified sequences, just by direct melting. With the use of specific DNA dyes, high-end instrumentation and sophisticated analysis software, these differences are detected (Dahlem et al. 2012; Wang et al. 2015). HRMA is simple, rapid, sensitive, and compatible with high-throughput analysis, but it cannot detect comparatively large INDELs (> 100 bp) (Thomas et al. 2014; Zischewski et al. 2017). HMA takes advantage of the modified migration to separate re-hybridize PCR products in polyacrylamide gel electrophoresis (PAGE) (Ota et al. 2013). It is easy to operate, fast, cheap, suitable for detection of single nucleotide polymorphism (SNP) and small INDELs. But it cannot detect larger deletions (Ota et al. 2013; Zischewski et al. 2017). Other reported INDEL detection methods include PCR combined with ligation detection reaction (PCR-LDR) (Kc et al. 2016), restriction fragment length polymorphism (RFLP) (Kim et al. 2014), PCR based on two primer pairs (Yu et al. 2014), Tracking of Indels by Decomposition (TIDE) (Brinkman et al. 2014), CRISPR Genome Analyzer (CRISPR-GA) (Güell et al. 2014), and droplet digital PCR (ddPCR) (Findlay et al. 2016). Most of them are less sensitive, expensive, time-consuming, and not suitable for larger INDEL detection (Brinkman et al. 2014; Güell et al. 2014; Kim et al. 2014; Yu et al. 2014; Kc et al. 2016). In plant, similar methods have been developed recently. These methods include EMC (Nekrasov et al. 2013; Shan et al. 2014), HRMA (Denbow et al. 2017), annealing at critical temperature PCR (ACT-PCR) (Hua et al. 2017), PCR and Amplicon labeling-based method (Biswas et al. 2019), mutation sites-based specific primers PCR (MSBSP-PCR) (Guo et al. 2018), and cleaved amplified polymorphic sequence (CAPS) (Kohata et al. 2018). Although most of them are proved to be effective in certain cases, the applicability was adversely affected by intrinsic limitations, such as limited sensitivity and specificity, time/labor-consuming, and inapplicability for SNP detection (Nekrasov et al. 2013; Shan et al. 2014; Denbow et al. 2017; Hua et al. 2017; Guo et al. 2018; Kohata et al. 2018; Biswas et al. 2019). Most importantly, all of them lack multiplex capabilities; and none of them finds its application in natural variation discrimination.

In the past years, neither did basic functional analysis of candidate gene/sgRNA nor applied breeding use long-range PCR or long-read next-generation sequencing (NGS) methods, such as PacBio, to detect unintended genomic changes including off-targets in plants; therefore, several early reports claimed that the off-target events in CRISPR/Cas9 edited plants are rare; even it occurs, the frequency is very low (Feng et al. 2014; Zhang et al. 2014; Gao et al. 2015; Tang et al. 2018; Xu et al. 2015). However, increasing evidence showed unexpected high frequency of off-target mutagenesis in CRISPR/Cas9-induced Arabidopsis (Zhang et al. 2018) and rice (Endo et al. 2015; Li et al. 2016). Therefore, the detection of off-target is equally important to that of detection of on-target in plants, particularly for functional analysis. There are both bioinformatics compatible online-based tools (such as CRISPR.P, CCTop) and experimentation approaches (such as NGS, BLISS, BLESS, GUIDE-seq) to predict and identify putative off-target sites, respectively (Yan et al. 2017; Germini et al. 2018; Grohmann et al. 2019; Hahn and Nekrasov 2019). Online tools cannot stand alone without experimental validation, while most of proposed experimental approaches involve NGS, and generally are complex and time consuming. The most popular off-target identification approach is the amplification of silico predicted potential off-target sites and followed by Sanger sequencing. Nevertheless, it might overlook mutations at other alleles (Zischewski et al. 2017). EcoTILLING, a high-throughput-based method, has been reported to detect naturally occurred mutations, which, however, is labor/time-consuming (Rigola et al. 2009). In addition, pan-genomes might help to identify natural mutations in the genome of interest (Zhao et al. 2018; Li et al. 2019). Notably, none of above-mentioned methods could fulfill the simultaneous detection of both on- and off-targets and natural variations in plants. A recently developed method qEva-CRISPR can detect all types of on-target mutations in human cells with high sensitivity regardless of mutation type and several off-targets (Dabrowska et al. 2018); however, no corresponding method has been established in plant. Development of competent, reliable and inexpensive multiplex capabilities-based method would help to screen on- and off-target as well as natural mutations concurrently, at the early stages in pooled samples, which would accelerate further functional analysis or breeding in plants.

In this study, taking advantage of the multiplex capability of a previously reported multiplex ligation-based probe amplification (MLPA) method, we developed a new MLPA-based method that allowed us to identify CRISPR/Cas9-induced on- and off-target INDELs, and natural occurred INDELs in rice. The sensitivity, reliability, and applicability were analyzed using CRISPR/Cas9-induced mutants targeting different genes and different rice cultivars harboring natural occurred SNPs in semi-dwarf1 (SD1) loci.

Materials and methods

Plant materials

Several CRISPR/Cas9 edited mutant lines targeting on semi-dwarf 1 (SD1) and other genes and several rice (Oryza sativa) varieties harboring natural mutation in SD1 (including Kasalath, Xiushui, and Minghui63) were used in this experiment (Table S1). In addition, 9522 (Oryza sativa ssp. japonica) was used in this study as a reference. All above-mentioned rice lines, including mutants and wild types, were grown in the paddy field of Shanghai Jiao Tong University (30°N, 121°E), Shanghai, China, under natural rice growing conditions.

Plant genomic DNA extraction

Genomic DNA from leaf tissues was extracted as previously described with minor modifications (Murray and Thompson 1980). Leaf tissues were ground in the presence of liquid nitrogen and then incubated with lysis buffer (1.5 X CTAB) and RNase at 65 °C for 60 min. The liquid phase was collected after centrifugation, extracted again with phenol: chloroform and trichloromethane, and mixed with an equal volume of isopropyl alcohol to precipitate the genomic DNA. The pellet was washed twice with 70% ethanol, air-dried and then dissolved in ddH2O. The quality and quantity of the extracted genomic DNA were evaluated using both the NanoDrop 1000 UV/vis Spectrophotometer (NanoDrop Technologies Inc., Wilmington, DE, USA) for OD260/OD280 and OD260/OD230 and the electrophoresis on 1% (w/v) agarose gel in 0.5 × TBE with Gel Red staining. All extracted genomic DNA was stored at − 20 °C until used in the experiments.

Probe design

All oligonucleotide probes for MLPA-based method were designed according to a previously adopted strategy (Kozlowski et al. 2007; Marcinkowska et al. 2010), and synthesized by Invitrogen Co., Ltd. (Shanghai, China). Each multiplex ligation-based probe was composed of two half-probes: 5′ half-probe and 3′ half-probe. Each half-probe consisted of the primer-specific binding sequence (PBS-red), the stuffer sequence (SS-gray), and the target-specific hybridization sequence (THS-black). In addition, the control probes were chosen outside of the genomic region of interest, dispersing ideally on different chromosomes. In general, paired half-probes for mutants were designed to directly adjacent to the putative mutated-regions, while control probes were designed to locate in the genomic regions free of repetitive elements, SNPs, and small INDELs (Fig. 1; Kozlowski et al. 2007; Marcinkowska et al. 2010). Sequences and detailed characteristics of all oligonucleotide probes used in this experiments are listed in Table S1.

Fig. 1
figure 1

Schematic presentation of the principle and steps for the identification of mutations by the developed multiplex ligation-dependent probe amplification (MLPA)-based method. MLPA-based method consists of five steps (from top to bottom). a Denaturation. Double stranded genomic DNA (gDNA) is denatured to two single strands by heating. In general, on-target, off-target and natural variation-specific MLPA paired half-probes are designed to directly adjacent to the putative mutated-sites (target region). Therefore, any mismatches will avert subsequent ligation of paired half-probes and further PCR amplification. DSB, double strand breaks; PAM, protospacer adjacent motif. b Hybridization. A probemix is added to the denatured gDNA sample for hybridization under stringent conditions. Each MLPA probe consists of two half-probes; 5′ half-probe and 3′ half-probe and each half-probe composed of the primer-specific binding sequence (PBS, red), stuffer sequence (SS, gray), and the target-specific hybridization sequence (THS, black) that correctly hybridize to the adjacent target of the sample gDNA. WT, wild-type; MT, mutant. c Ligation. Correctly hybridized 5′ half- and 3′ half-probes are ligated successfully into a single longer probe. d PCR amplification. All successfully ligated probes can be used as template for PCR amplification using universal primer pair that sits at the far 5′- and 3′- ends of the ligated probes. One of the universal primer (herein is the forward primer F) is fluorescent-labeled (FamF). R, reverse primer. e Capillary electrophoresis. All fluorophore-labeled amplicons are separated by capillary electrophoresis based on the lengths of their different length of SS, which show corresponding chromatograph peaks at the expected positions. RFU, relative fluorescence units (color figure online)

Full size image

MLPA program

MLPA reactions were carried out in a 50 µl of reaction mixtures as suggested (https://www.mlpa.com). Five µl of sample genomic DNA (approximately 100 ng) was denatured by heating at 98 °C for 5 min and subsequent cooling down; the resulting denatured DNA was mixed with 3 µl of hybridization master mix to a volume of 8 µl, heated at 95 °C for 2 min, and hybridized at 60 °C for 16 h. Hybridized probes were mixed with 32 µl of Ligase-65 master mix to reaction volume 40 µl, and ligated at 54 °C for 15 min, and followed by heating at 98 °C for 5 min to inactivate ligase enzyme. Ligated probes were cooled down to room temperature and mixed with 10 µl of polymerase master mix that contained 2 µl of SALSA PCR primer mix (universal primer pair, one of which was fluorescently labeled forward primer), 0.5 µl of polymerase, and 7.5 µl of ultrapure water. The final PCR reaction volume was 50 µl. PCR amplification reaction was carried out by 35 cycles at 95 °C for 30 s, 60 °C for 30 s and a final stage at 72 °C for 60 s. The amplified fluorophore-labeled MLPA PCR products were separated with LIZ GS500 size standard by capillary electrophoresis on an ABI Prism 3130XL apparatus (Applied Biosystem, USA) and analyzed using Gene Marker software v2.6.4 (Soft Genetics, USA).

Results

Design of a MLPA-based system

The whole procedure of multiplex ligation-dependent probe amplification (MLPA)-based method included following five steps: DNA denaturation, probe hybridization, ligation, PCR amplification, and capillary electrophoresis (Fig. 1). Among them, probe design is crucial for the establishment of MLPA assay, because the PCR product from each DNA sequence must be independently detected and quantified in the capillary electrophoresis based on different length (Kozlowski et al. 2008). Each MLPA probe was composed of a 5′ half-probe and a 3′ half-probe, each half- probe consisted of one primer-specific binding sequence (PBS, red) for PCR amplification, one stuffer sequence (SS, gray) for determination of PCR product size, and one target-specific hybridization sequence (THS, black) for unique binding of probes to targets (Fig. 1). The 5′ and the 3′ THS probes in MLPA were generally adjacent to the predicted mutated regions, namely mutation hot spot region (MHS), and each sister THS sequence should be at least 21 nucleotides in length. The Phage M13 sequences (NCBI/GenBank ID V00604) between 3 and 119 bp were used as SS, which allowed to adjust the length of probes and to result in unique amplicon peak for each probe (Fig. 1a). Notably, the 5′ half-probe had a forward primer PBS, a SS, and a left THS sequence while the 3′ half-probe a right THS sequence, a right SS, and a reverse primer PBS (Fig. 1b). After hybridization, only were the two half-probes that hybridized next to each other on their target sequence under stringent conditions ligated to a single and longer probe (Fig. 1c), which could then be served as the template for the amplification using PBS primer pair (Fig. 1d). One of the PBS primers was labeled with a fluorescence dye, and the SS included in each probe endowed each amplification product a unique characteristic length, thus, the amplification products could be detected by capillary electrophoresis and fluorescence detection based on their sizes (Fig. 1e). Therefore, in the case of CRISPR/Cas9 induced or naturally occurred mutants, even a small mismatch at the ligation point (mutation site) would impair ligation and subsequent probe amplification. As a result, typical chromatogram peak would be utterly absent in the expected position for the affected target-specific probe, while the chromatogram peaks for the non-affected target-specific probes should appear as do for the control probes (Fig. 1e).

Based on this concept, probes for MLPA-based system were designed to identify CRIPSR/Ca9 induced SD1 mutation in rice (Table S1). Those probes included four control probes that were randomly dispersed over rice genome, universal, and likely used in any MLPA assays in rice, and two target-specific probes (THS_SD1.1 and THS_SD1.2) that consisted of both 5′ half-probes and 3′ half-probes.

Specificity and sensitivity of the developed MLPA-based system

The MLPA-based system assay was first optimized and validated using genomic DNA (gDNA) from wild type (WT) and known SD1 deletion mutants. The results showed that control and WT probes generated correct chromatogram peak at the expected positions in both WT and mutant genomes, while probes of CRISPR/Cas9-induced mutants (1–2 bp deletions) (THS_SD1.1 and THS_SD1.2), did not generate corresponding chromatogram peaks in the expected positions (Fig. S1). This result indicated that MLPA-based system can be used for the identification of CRISPR/Cas9 edited mutants in rice. Its specificity was confirmed in different SD1 mutants with a single nucleotide insertion or deletion, or a single nucleotide substitution. In all cases, probes of CRISPR/Cas9-induced SD1 mutants, either 1 bp deletion, 1 bp insertion, or 1 bp replacement, failed to generate correct chromatogram peaks in the expected positions (Fig. S2).

To demonstrate the applicability and accuracy of the developed MLPA-based method for different gene targets, MLPA probes for mutants of three CRISPR/Cas9 targeted genes including Os06g0135460, Os07g0445800, and SD1 were designed (Table S1). MLPA-based analyses were separately performed for each target-specific probe using mixed genomic DNA samples of different mutations targeting the same corresponding sgRNA. Similarly, all MLPA probes generated correct chromatogram peaks in the expected positions when WT DNA sample was used (Fig. 2a). In contrast, no MLPA probes, THS_Os06g0135460.1 for Os06g0135460, THS_Os07g0445800.1 for Os07g0445800, or THS_SD1.1 and THS_SD1.2 for SD1, generated correct chromatogram peaks in the expected positions when mixed DNA samples for corresponding mutants for Os06g0135460 (Fig. 2b), Os07g0445800 (Fig. 2c), and SD1 (Fig. 2d, e) were used, respectively.

Fig. 2
figure 2

Detection of on-target mutations by the developed MLPA-based method. The resulting chromatogram peaks separated by capillary electrophoresis of MLPA-based method are presented in the left panels, while Sanger sequencing results of the PCR products of WT and mutants used for the analysis are presented in the right panels. a Wild-type (WT) genomic DNA. b Mixed mutant genomic DNA samples containing different mutations targeting Os06g0135460.1 (MT1). c Mixed mutant genomic DNA samples containing different mutations targeting Os07g0445800.1 (MT2). d Mixed genomic DNA samples containing different mutations targeting SD1.1 (MT3). e Mixed mutant genomic DNA samples containing different mutations targeting SD1.2 (MT4). Red arrowhead in each left panel represents the expected chromatogram peak site of the corresponding target-specific probe. Red, green, and empty letters in the right panel represent target sequence, inserted and deleted base in the target, respectively. d#, deletion with #bp; i#, insertion with #bp; CTL, control probe; THS, target-specific hybridization sequence; WT, wild-type; MT, mutant (color figure online)

Full size image

The sensitivity of the developed MLPA-based method was tested using mixed genomic DNA samples from both WT with mutant, in which the genomic DNA of the SD1 mutation (1 bp deletion) was mixed with the WT genomic DNA to corresponding mutant/WT ratios of 100% (100% mutant), 50% (50% mutant/50% WT), and 0% (100% WT), respectively. The results showed that, as compared to 100% WT, 50% mixed DNA samples generated significantly reduced chromatogram peak, while 100% mutant failed to generate chromatogram peak for the target-specific probe (THS_SD1.2) (Fig. S3). This result was also confirmed by the quantifying the relative peak of the target-specific probe (THS_SD1.2) using Coffalyser.Net software (MRC-Holland) (Fig. S3). Since the 100% mutant and 50% mixed DNA samples represented the homozygous and heterozygous mutant status, respectively, the developed MLPA-based method could thus be useful for zygosity analysis in CRISPR/Cas9-induced rice mutants.

Analysis on CRISPR/Cas9-induced off-targets in rice

CRISPR/Cas9 may also generate off-target mutation because it tolerates up to three mismatches between the sgRNA and the target sequence (Zhu et al. 2017). To explore if the developed MLPA-based method can be used to screen and detect CRISPR/Cas9-induced off-target mutations, we first applied it to Os06g0135460 and Os07g0445800 mutants using two corresponding probes OTS_Os06g0135460.1 and OTS_Os07g0445800.1, respectively. Our previous screening for the top five potential off-targets of each target identified by CRISPR-P software (https://crispr.hzau.edu.cn/cgi-bin/CRISPR2/CRISPR) using PCR found that there was off-target in mutants targeting Os07g0445800. In MLPA-based method, when WT DNA was used, probe OTS_Os07g0445800.1–1, together with other two probes OTS_SD1.2–1 and OTS_SD1.2–2, generated chromatogram peaks at the expected positions (Fig. 3a). In contrast, when mixed mutant genomic DNA samples were used, probe OTS_Os07g0445800.1–1 did not generate corresponding chromatogram peak, while other two probes OTS_SD1.2–1 and OTS_SD1.2–2 did (Fig. 3b), indicating the occurrence of off-target mutation in Os07g0445800.1.

Fig. 3
figure 3

Detection of off-target mutations by the developed MLPA-based method. The resulting chromatogram peaks separated by capillary electrophoresis of MLPA-based method are presented in the left panels, while Sanger sequencing results of the PCR products of WT and mutants used for the analysis are presented in the right panels. a Wild-type (WT) genomic DNA. b Mixed mutant genomic DNA samples containing different mutations targeting Os07g0445800.1 (MT2). c Genomic DNA sample of SD1.2 mutant (MT4). d gDNA sample of SD1.2 mutant (MT4). Red arrowhead in each left panel represents the expected chromatogram peak site of the corresponding target-specific probe. Red, green, blue and empty letters in the right panel represent target sequence, inserted, replaced, and deleted base in the target, respectively. d#, deletion with #bp; i#, insertion with #bp; CTL, control probe; OTS, off-target-specific; WT, wild-type; MT, mutant (color figure online)

Full size image

The effectiveness and usefulness of this MLPA-based method for screening and detection of CRISPR/Cas9-induced off-targets were validated with other SD1 mutants targeting two different sgRNA (THS_SD1.1 and THS_SD1.2) using corresponding off-target specific probes OTS_ SD1.1 and OTS_SD1.2, respectively. Previously screening for the top five potential off-targets of each target identified by CRISPR-P software using PCR found off-target mutations only in mutants targeting THS_SD1.2. MLPA-based method verified PCR results, in which different from those in WT genomic DNA samples (Fig. 3a), no chromatogram peaks appeared at the expected sites for the off-target-specific probes OTS_ SD1.2–1 and OTS_SD1.2–2 in mixed mutant genomic DNA samples (Fig. 3c–d). Sanger sequencing confirmed that the off-target mutation was caused by a single nucleotide substitution in both off-target sites (Fig. 3c–d). These results indicated that this MLPA-based method is quite effective and highly sensitive for identification of CRISPR/Cas9-induced off-target mutations as well.

Analysis on natural variations in rice

Since the developed MLPA-based method can effectively detect CRISPR/Cas9-induced INDEL mutations, we assume that it might work for the detection of natural variation mutations in rice. To explore this possibility, we designed new set of natural variation (NV) probes for this purpose based on the SNPs of the SD1 gene in the pan-genome identified with SNP-Seek database (https://snp-seek.irri.org/_snp.zul). The designed target specific probes NV_ SD1.1, NV_ SD1.2, NV_ SD1.3, and NV_ SD1.4 were specific for one 1 bp and one 2 bp substitution mutations in rice varieties Minghui63, Kasalath and Xiushui, respectively. A set of controlled probes designed previously were also used. As shown in Fig. 4, no mutant genomic DNA samples generated chromatogram peaks at the expected positions for corresponding probe NV_ SD1.1, NV_ SD1.2, NV_ SD1.3, and NV_ SD1.4, respectively, indicating the applicable of this method for natural variation detection.

Fig. 4
figure 4

Detection of natural occurred mutations by the developed MLPA-based method. The resulting chromatogram peaks separated by capillary electrophoresis of MLPA-based method are presented in the left panels, while Sanger sequencing results of the PCR products of corresponding lines with different natural variations used for the analysis are presented in the right panels. a Wild-type (WT) genomic DNA. b Minghui63 genomic DNA. c Minghui63 genomic DNA. d Kasalath genomic DNA. e Xiushui genomic DNA. Red arrowhead in each left panel represents the expected chromatogram peak site of the corresponding target-specific probe. Red and blue letters in the right panel represent target sequence and replaced base in the target, respectively. CTL, control probe; NV, natural variation (color figure online)

Full size image

Discussion

Although the CRISPR/Cas9 system is still being improved for better targeted genome editing, it has been widely utilized both in basic and applied sciences. The predominant DSB repair pathway of CRISPR/Cas9 system in plants tends to generate small INDELs down to 1 bp or single nucleotide substitutions, and such small INDELs might occur naturally in plants (Grohmann et al. 2019). Therefore, a practical, accurate, and economic screening method for INDELs or single nucleotide substitutions caused by targeted genome editing or natural mutation is essential not only for applied breeding but also for fundamental functional research in plants. Previously developed methods can identify CRISPR/Cas9-induced on-target INDELs (Nekrasov et al. 2013; Shan et al. 2014; Denbow et al. 2017; Hua et al. 2017; Guo et al. 2018; Kohata et al. 2018; Biswas et al. 2019), off-target INDELs (Zischewski et al. 2017), or natural variation (Rigola et al. 2009); separately, very few of them can have multiplex capacity to detect all above-mentioned mutations and variations simultaneously. In this study, we modified an established MLPA method, developed a new approach to detect CRISPR/Cas9-induced on-target and off-target mutants and natural variations, and proved its sensitivity and applicability in different lines in rice.

Various molecular approaches have been developed previously to detect CRISPR/Cas9-induced mutations. These methods included HRMA (Denbow et al. 2017), EMC (Nekrasov et al. 2013; Shan et al. 2014), ACT-PCR (Hua et al. 2017), MSBSP-PCR (Guo et al. 2018), CAPS (Kohata et al. 2018), WGS (Zhang et al. 2014) and others (Biswas et al. 2019). All except WGS proved not to be effective in the identification of them simultaneously. In addition, each method’s intrinsic limitations restrain its fully application (Biswas et al. 2019). Compared with HRMA, EMC, ACT-PCR, MSBSP-PCR, and CAPS, the developed MLPA-based method is sensitive (down to 1 bp INDELs and single nucleotide substitutions) (Fig. 2), consistent (suitable for different targets) (Figs. 2, 3), accurate (consistent with Sanger sequencing results) (Figs. 2, 3, 4), and quantitative (also applicable in zygosity analysis) (Fig. S3). Compared with WGS, one of the most potential approaches to identify CRISPR/Cas9-induced mutations including both on and off-target ones and natural variations in the genomes of interest (Zhang et al. 2014), MLPA-based method is cheap (< 8$ per sample, including probe synthesis cost), time-saving (within maximum 2 days), and no bioinformatics knowledge demanding (no assembly). Furthermore, MLPA-based method is very useful in the identification of CRISPR/Cas9-induced off-targets using non-WGS approaches, particularly at the early screening stage, due mainly to its multiplicity. Amplification of in silico predicted putative off-target sites and subsequent Sanger sequencing are the easiest ways to identify off-target mutation in plants (Zischewski et al. 2017). However, costs of Sanger sequencing and difficulties in management limit its application for a high number of off-target sites and samples. MLPA-based method becomes the good alternative in such cases because of its ability to detect about 60 target (off-target) sites in a single assay (www.mlpa.com). Sequencing of MLPA verified positive mutants will save time and money at this stage. However, for off-target detection, MLPA-based method can only be used for simultaneously detection of previously identified off-target, but not unknown off-targets. In this case, NGS is super advantageous over MLPA. As for the multiplicity of established MLPA for simultaneous detection of on- and off-targets, so far, only at most four target sites plus four controls were simultaneously assessed in this study, more targets need to be included for simultaneous detection of both on- and off-target mutations in the future studies.

What’s more, compared with high-throughput sequence-based EcoTILLING technology, which identifies naturally induced mutations in plants (Rigola et al. 2009), MLPA-based method is labor/time-saving (no need for library preparation), cheap (no library preparation and high-throughput sequence), and sensitive (particularly for the detection of SNPs from pooled of samples). In sum, this MLPA-based method is effective in the simultaneous identification of CRISPR/Cas9 or naturally induced INDELs in plants (Fig. S4).

Different from other previously developed methods such as PCR and Amplicon labeling-based method (Biswas et al. 2019), MLPA-based method designed target-specific paired half-probes to directly adjacent to the predicted mutated-sites overlapping the target sequence. Therefore, any mismatches in the ligation point avert ligation and subsequent PCR amplification, which renders MLPA-based method more sensitive to detect any mutations in the targeted sites. In rice, small (1–2 bp) nucleotide mismatches caused mutations are often observed in CRISPR/Cas9 targets, potential off-targets, and natural variation regions (Grohmann et al. 2019). These molecular features made MLPA-based method the most suitable tool in rice to identify CRISPR/Cas9 or natural induced INDELs in a single-tube assay (Kozlowski et al. 2008), as evidenced by the observation in this study, in which 1 bp INDELs and single nucleotide substitutions successfully prevented ligation and subsequent probe amplification (Figs. S2b–d). In addition, MLPA-based method uses mixed mutant genomic DNA, which endows it an added value for its application in rice research community, specifically for functional analysis of the targeted gene to identify mutations in a pool of mutants. Because genome edited mutants in many other cereal crops have similar characteristics (Zhu et al. 2017), this MLPA-based method developed in rice is plausible to be applicable in other cereal crops for similar purpose.

Similar to other methods, this MLPA-based method also has some drawbacks. For example, it cannot tell the exact genotype of the tested sample, which can only be resolved by Sanger sequencing; it cannot cover the whole-genome analysis and does not allow to detect mutations outside of the targeted regions. Regarding probe design, factors, such as GC contents, Tm value and probe length, can adversely affect the efficiency of the MPLA-based assay, which need careful design following proposed formula (Samelak-Czajka et al. 2017). For target sequences, those are unspecific or those with repeat elements or various SNPs cannot be readily detected by this method (Kozlowski et al. 2008). Nevertheless, combined with Sanger sequencing, the MLPA-based method is one of the most economic alternatives for effective detection or screening for CRISPR/Cas9 or natural induced INDELs in plants. To make it more efficient, it is strongly recommended to sequence the target regions in the intended germ plasm before designing CRISPR/Cas9 guide RNAs.

In summary, an effective, accurate, economic and multiplex capable MLPA-based method was developed and utilized to simultaneously detect CRISPR/Cas9-induced and naturally occurred INDELs in rice (Fig. S4), which would facilitate both rice breeding and rice functional analysis using tools such as genome editing and allelic diversification. In the future, the applicability of this MLPA-based method would be tested for more targets in rice and other plant species.