Egg cell-specific promoter-controlled CRISPR/Cas9 efficiently generates homozygous mutants for multiple target genes in Arabidopsis in a single generation

Targeted mutations of multiple Arabidopsis genes in the T1 generation

Two reports have demonstrated that DD45/EC1.2 (At2g21740) is an egg cell-specific gene 49], 50]. In situ hybridization of tissue sections revealed that EC1.2 transcripts are specifically present in the egg cell, whereas GUS activity and GFP
signals were observed in EC1.2p:GUS and EC1.2p:GFP transgenic zygotes and early embryos;
the carryover of the signal into later stages of embryogenesis likely resulted from
higher stability of the reporter mRNA and/or protein 49], 50]. We reasoned that Cas9 driven by the EC1.2 promoter would be specifically transcribed in the egg cell; Cas9 mRNA would reside in one-cell stage embryos due to mRNA stability and continue to
translate Cas9 protein. Also, newly translated Cas9, together with residual Cas9 that
remained due to Cas9 protein stability, would function in one-cell stage embryos,
thus allowing creation of Arabidopsis T1 homozygous or biallelic mutants, rather than mosaic plants.

Since combinations of the same promoter with different terminators might result in
significantly different amounts of protein 51], we made the two constructs to examine the effects of two terminators, the Pisum sativum rbcS E9 terminator, in the pHEE2A-TRI construct, and the Agrobacterium nos gene terminator, in the pHEN2A-TRI construct, on the expression of Cas9 driven by the EC1.2 promoter (Fig. 1). We used two single guide RNAs (sgRNAs) to target three genes, ETC2, TRY, and CPC (Fig. 1a, b), since the try cpc double and etc2 try cpc triple mutants have easily observed phenotypes (clustered leaf trichomes) and the
triple mutant has a different phenotype from the double mutant 26].

Fig. 1. Arabidopsis T1 homozygous triple mutants obtained via EPC CRISPR/Cas9. a Physical maps of the T-DNAs of two CRISPR/Cas9 binary vectors, each harboring Cas9 driven by the egg-cell specific promoter EC1.2p and two sgRNA genes driven by Pol-III
promoters U6-26p and U6-29p, respectively. RB/LB, T-DNA right/left border; EC1.2p,
EC1.2 promoter; rbcS-E9t, rbcS E9 terminator; Nost, nos gene terminator; sgR, sgRNA; 2-sgRs, two sgRNA expression cassettes; zCas9, Zea mays codon-optimized Cas9; U6-26p and U6-29p, two Arabidopsis U6 gene promoter; U6-26t, U6-26 terminator with downstream sequence; Hyg, hygromycin-resistance gene. For the sgRNAs, the yellow part represents 20-bp target
and the green part represents 76-bp sgRNA scaffold. b The alignment of the sgRNA with its target genes and potential off-targets. Only
aligned regions of interest are displayed. rc, reverse complement. c Phenotypes of two triple mutants segregated from T1 transgenic lines. The other plants
in the same pot are from the same batch of T1 transgenic lines with normal phenotypes.
Seeds from the T0 plants were sown on MS medium containing 25 mg/L hygromycin, vernalized
at 4 °C for 3 days, and grown under long-day conditions (16 h light/8 h dark) at 22 °C
for 9 days. Hygromycin-resistant seedlings (T1) were transplanted to soil and allowed
to grow for 33 days before photographing

In our first attempt, we obtained 24 T1 EC1.2p:zCas9-rbcS_E9t lines using pHEE2A-TRI
(Fig. 1a) and 54 T1 EC1.2p:zCas9-Nost lines using pHEN2A-TRI (Fig. 1a). Among the 24 T1 EC1.2p:zCas9-rbcS_E9t lines, two lines (#1 and #3) were likely
triple mutants (Fig. 1c), and one line (#21) was a mosaic with two branches that displayed the double-mutant
and wild-type phenotypes, respectively. We sequenced the regions surrounding the target
sites of the three genes from the two putative triple mutant lines, and confirmed
that they were indeed triple mutants (Table 1). In this instance, all observed mutations were single base pair insertions or deletions.
Sequencing analysis and examination of the phenotypes of T2 plants derived from these
two T1 lines further confirmed the identity of the two mutant lines (Tables 1 and 2). Unexpectedly, we failed to identify a likely triple mutant, double mutant, or mosaic
among 54 T1 EC1.2p:zCas9-Nost lines, suggesting that the combination of the EC1.2 promoter and rbcS E9 terminator performed much better than the combination of the EC1.2 promoter and nos terminator.

Table 1. Mutation analysis of three T1 likely triple mutants and their T2 progeny

Table 2. Phenotypic segregation analysis of T2 transgenic lines

To examine the specificity of the mutagenesis, we searched the Arabidopsis genome for potential off-targets with fewer than four mismatches with the targets
of the sgRNAs. This identified three potential off-targets of the sgRNA targeting
ETC252]. We sequenced these regions in the two triple mutants and found no mutations, demonstrating
the high specificity of the EPC CRISPR/Cas9 system.

To confirm the repeatability of the results from EC1.2p:zCas9-rbcS_E9t transgenic
lines, we performed two additional, independent Arabidopsis transformation experiments with the construct pHEE2A-TRI. In the second transformation,
we obtained 41 T1 lines, among which three were likely triple mutants (Additional
file 1: Figure S1). In the third transformation, we obtained 43 T1 lines, including four
that were likely triple mutants (Additional file 1: Figure S2). Therefore, approximately 8.3 % (9/108) of the T1 plants were likely
homozygous triple mutants.

We also demonstrated the usefulness of the EPC CRISPR/Cas9 system by performing targeted
mutation of two Arabidopsis genes, CHLI1 and CHLI2, in T1 plants. Simultaneous disruption of CHLI1 and CHLI2 leads to an albino phenotype. We obtained 99 T1 lines, including 18 putative chli1 chli2 double mutants (albino plants, see Additional file 1: Figure S3). We sequenced the regions surrounding the target sites, and found that
10 lines were double mutants and five were mosaic plants (Additional file 2: Table S1). These results indicate that the EPC CRISPR/Cas9 was functional, not only
in one-cell stage embryos, but also in some early embryos, likely due to Cas9 mRNA and/or protein stability, and/or reduced egg cell-specificity. Among the 18
albino lines, three grew poorly, and we were unable to obtain sequence data from these
lines (Additional file 1: Figure S3, Additional file 2: Table S1). These three albinos were most likely double mutants rather than mosaics,
based on their poor growth. Thus, the ratio of homozygous T1 double mutants to T1
plants was approximately 13 % (13/99). Together, these results demonstrate that our
EPC CRISPR/Cas9 system could be used to efficiently produce confirmed T1 homozygous
or biallelic mutants in less than 3 months (Additional file 3: Figure S4). In practical applications, users might have no visible phenotypes that
they could use to screen for T1 homozygous mutants. However, this obstacle can be
easily overcome by screening 25–50 T1 lines by restriction enzyme digestion analysis,
T7E1 assay, or Surveyor assay (Additional file 3: Figure S5). After these primary screens, users will be able to quickly obtain a
few candidate lines for sequence analysis (including direct sequencing of PCR fragments,
sequencing individual clones of PCR fragments, and deep sequencing of PCR fragments)
(Additional file 3: Figure S5).

To confirm that the T1 mutations are germline transmissible, we sowed 20 T2 seeds
per T1 line derived from the two T1 triple mutant lines (#1 and #3) on MS plates.
We observed no phenotypic segregation of these T2 plants (Table 2). Moreover, sequencing analysis of four T2 plants per T1 line showed no novel mutation
types (Table 1). These results strongly suggested that germline transmission of T1 mutations occurred.
To further confirm the germline transmission of the T1 mutations, we screened for
non-transgenic T2 lines and analyzed their mutations. Since we harvested 30 T2 seeds
per T1 line from the two triple mutant T1 lines (#1 and #3) due to their poor growth,
no additional T2 seeds were available for screening of non-transgenic lines. We then
turned to screening for non-transgenic T2 plants derived from the T1 triple mutant
(#C1) produced in the third transformation (Additional file 1: Figure S2). We sowed 36 T2 seeds on MS plates, and transplanted the seedlings to
soil. All 36 T2 plants were phenotypically triple mutants. We screened all 36 T2 plants
for non-transgenic plants and obtained only one such plant, much fewer than the nine
or so plants we expected, which may reflect insertions of two or more copies of T-DNAs
into the genome of the T1 plant. We analyzed the mutations of the T1 mutant (#C1)
and the non-transgenic T2 mutant (#C1-17) by sequencing (Table 1), which demonstrated that the T2 mutations are derived from the originally confirmed,
rather than newly produced, T1 mutations through germline transmission (Table 1).

Analysis of mutations in the phenotypically wild-type T1 plants and their T2 progeny

Since CRISPR/Cas9 should continue to function in T1 egg cells, T2 one-cell stage embryos,
and T2 early embryos, and since T1 plants with normal phenotypes might be heterozygotes
or mosaics rather than wild type, T1 plants with no clear phenotypes should be able
to give rise to homozygous or bi-allelic triple mutant T2 plants. To ensure that triple
mutants could be differentiated from double mutants, we re-examined the phenotypes
of the triple/double mutants, finding no differences from our previous observations
(Additional file 3: Figure S6). Then, we examined T2 plants derived from the 24 T1 EC1.2p:zCas9-rbcS_E9t
lines, revealing that approximately 50 % (12/24) produced likely triple mutant T2
progeny (Fig. 2 and Table 2). The segregation ratio of the likely triple mutants to total T2 plants examined
was higher than 20 % for each of the 12 T1 lines and averaged 24.8 % for all 24 T1
lines (Table 2). Of the 54 T1 EC1.2p:zCas9-Nost lines (100 T2 plants per line examined), only two
lines, equivalent to 3.7 % (2/54), produced likely triple mutants in their T2 progeny.
These results further demonstrate that the combination of EC1.2 promoter and rbcS E9 terminator performed much better than the EC1.2 promoter and nos terminator combination, suggesting that in egg cells, the terminator is a key factor
in stabilizing the Cas9 mRNA and thus enhancing its translation.

Fig. 2. Phenotypic segregation of T2 transgenic lines. Phenotypic segregation of T2 transgenic
lines derived from two representative T1 lines with normal phenotypes. Seeds from
T0 plants were sown on MS medium containing 25 mg/L hygromycin, vernalized at 4 °C
for 3 days, and grown under long-day conditions (16 h light/8 h dark) at 22 °C for
7 days. Hygromycin-resistant seedlings (T1) were transplanted to soil and allowed
to grow for 20 days before photographing

We sequenced the three target genes of two representative T1 lines (#4 T1 and #6 T1),
which had normal phenotypes (Table 3). We also sequenced their likely triple mutant T2 progeny (10 T2 plants per line;
Table 3). The sequencing results revealed that the two T1 lines were mosaic with different
degrees of mutation in the three target genes, demonstrating that the mutation frequency
of a single gene in the T1 population was much higher than the frequency of simultaneous
mutations of all three target genes. The formation of mosaic plants could be attributed
to Cas9 mRNA and protein stability (Additional file 2: Table S2). For example, for a two- or four-celled embryo derived from a zygote that
had undergone two or three rounds of mitosis, each of the two or four cells would
contain three-quarters or half the amount of Cas9 protein of that in the egg cell
(if Cas9 mRNA and protein were sufficiently stable; Additional file 2: Table S2). Two types of mosaic plants resulted from EC1.2p:Cas9 transformation:
mosaics with a wild-type allele of a target gene and mosaics without wild-type alleles,
which could be regarded as homozygous mutants. Analysis of the mutations present in
the T2 progeny of the T1 mosaic plants demonstrated that most of the triple-mutant-like
T2 plants were homozygous or biallelic triple mutants (Table 3).

Table 3. Mutation analysis of likely triple mutants segregated from two representative T1 lines
with normal phenotypes

Functional comparisons of 12 combinations of eight promoters and two terminators

In an attempt to improve the efficiency of generating T1 homozygous mutants, we first
tested another egg-cell specific promoter, using the promoter from EC1.1, and then we tested EC1.2 or EC1.1 promoters fused with enhancers (Fig. 3). Similar to our tests of the EC1.2 promoter, we also tested two combinations of the EC1.1 promoter with the rbcS E9 terminator (pHEE2B-TRI) or nos terminator (pHEN2B-TRI) to further examine the effects of terminators on mutation
efficiencies (Fig. 3a). We obtained 32 plants with observable mutations out of 224 T1 EC1.1p:zCas9-rbcS_E9t
transgenic lines (Additional file 4: Figure S7). However, most plants with observable mutations seemed to be likely double
mutants or mosaics, and only four plants seemed to be likely triple mutants (Additional
file 4: Figure S7), suggesting that the EC1.1 promoter is less egg cell-specific than the EC1.2 promoter. The existence of a high ratio of mosaics means that the likely triple mutants
(1.8 %) from EC1.1p:zCas9-rbcS_E9t transgenic lines are more likely to be phenotypically
severe mosaics. We obtained only three mosaic plant out of 102 T1 EC1.1p:zCas9-Nost
transgenic plants (Fig. 3a), demonstrating for the third time that the rbcS E9 terminator performed much better than the nos terminator. To exclude the possibility that the pGreen backbone of pHEN2A-TRI and
pHEN2B-TRI was the reason for the low mutation efficiencies, we constructed pHEN2C-TRI
by replacing the rbcS E9 terminator of pCambia1300-derived pHEE2A-TRI with the nos terminator (Fig. 3a). We obtained only four likely double mutants out of 134 T1 EC1.2p:zCas9-Nost (pCambia)
lines, demonstrating for the fourth time that the rbcS E9 terminator performed much better than the nos terminator, and the effects of the terminators were independent of the backbones
of the binary vectors.

Fig. 3. Structural and functional comparisons of twelve combinations of eight promoters and
two terminators. a Seven combinations of EC1.1, EC1.2, or 2x35S promoters and rbcS E9 terminator (rbcS-E9t) or nos terminator (Nost). The pHEN2A-TRI and pHEN2C-TRI constructs have the same combination
but different vector backbones: pGreen for the former and pCambia for the latter.
The data for pHSN2A-TRI come from the publication and p2gR-TRI-A is renamed pHSN2A-TRI
in this paper 26]. b Five combinations of five fusion promoters and the rbcS E9 terminator. Physical maps of the T-DNAs of seven (a) or five (b) CRISPR/Cas9 binary
vectors are indicated. For each binary vector, the vector name, the promoter, the
terminator, and the mutation frequencies of T1 transgenic plants are indicated at
the same row under the maps. See Fig. 1 for RB/LB, zCas9, 2-sgRs, and Hyg. EC1p, EC1.1p or EC1.2p; 35Sen, CaMV 35S enhancer; EC1.2en, enhancer
from EC1.2 promoter; LTM, likely triple mutant; Total, total number of T1 plants;
Mosaics-I, type I mosaic plants with strong phenotypes indistinguishable from the
double mutants; Mosaics-II, type II mosaic plants with the phenotypes appearing only
in some parts of the whole plants. The ratios of T1 plants with the mutations (LTMs,
Mosaics-I, or Mosaics-II) to total number of T1 plants are indicated

In our previous work, we demonstrated that constitutive overexpression of zCas9 driven by the double 35S promoter in T1 2x35Sp:zCas9-Nost transgenic lines (using
construct p2gR-TRI-A, renamed pHSN2A-TRI in this paper) efficiently produced mutations
for TRY, CPC, and ETC2, but all the mutants were mosaics 26]. Since EC1p/rbcS-E9t combinations (pHEE2A-TRI and pHEE2B-TRI) performed much better
than EC1p/Nost combinations (pHEN2A-TRI, pHEN2B-TRI, and pHEN2C-TRI), we reasoned
that the 2x35Sp/rbcS-E9t combination (pHSE2A-TRI) would perform much better than the
2x35Sp/Nost combination (pHSN2A-TRI). We constructed pHSE2A-TRI (Fig. 3a), and obtained 109 T1 2x35Sp:zCas9-rbcS_E9t transgenic lines. None of the T1 lines
are likely triple mutants (Fig. 3a), demonstrating again that almost all mutants produced from the T1 2x35S:Cas9 transgenic
lines are mosaics. The ratios of mutants from T1 2x35Sp:zCas9-rbcS_E9t lines with
strong (30.3 %) or observable (68.8 %) phenotypes to total number of T1 lines are
much lower than those from T1 2x35Sp:zCas9-Nost lines (78.8 % and 97.0 %, respectively).
These results demonstrated that 2x35Sp/rbcS-E9t combination did not perform much better
than 2x35Sp/Nost combination, suggesting that in vegetative cells, the nos terminator seemed to work better than the rbcS-E9 terminator. Considering statistical errors (for example, due to insufficient sample
population for 2x35Sp:zCas9-Nost transgenic lines), another possibility is that zCas9 mRNA stability is not as important for strong constitutive promoters as it is for
egg cell-specific promoters.

To determine whether the 35S enhancer could increase the expression driven by the
egg cell-specific promoters, we constructed three fusion promoters by fusing the 35S
enhancer with the egg cell-specific promoters and then generated transgenic lines
for the analysis of these fusion promoters’ activities (Fig. 3b). The ratio (26/67, 38.8 %) of 35Sen-EC1.1p:zCas9 plants with observable mutations
to the total number of T1 transgenic lines was much higher than that (32/224, 14.3 %)
of EC1.1p:zCas9-rbcS_E9t plants with observable mutations (Fig. 3b). In comparison with the ratio for T1 EC1.2p:zCas9-rbcS_E9t transgenic lines, the
ratios of plants with observable mutations to total numbers of T1 35Sen-EC1.2p(900-bp):zCas9
or 35Sen-EC1.2p(565-bp):zCas9 transgenic lines greatly increased – 11.1 % (12/108),
25.2 % (29/115), and 29.0 % (20/69) for the three transgenic lines, respectively –
whereas the ratios of likely triple mutants decreased (8.3 %, 3.5 %, and 1.4 % for
the three transgenic lines, respectively) (Fig. 3b). These results demonstrated that the CaMV 35S enhancer increased the expression
of EC1.1 or EC1.2 promoters but not in an egg cell-specific way. Thus, the CaMV 35S enhancer is not
suitable for improving the EPC CRISPR/Cas9 system, which is consistent with our notion
that the CaMV 35S promoter has weak activity in egg cells and one-cell stage embryos.

To determine whether the enhancer from the EC1.2 promoter or EASE 53] could improve the performance of the EPC CRISPR/Cas9 system, we constructed another
two fusion promoters by fusing the enhancer from the EC1.2 promoter (EC1.2en), or EC1.2en plus double EASE enhancers (EC1.2en-2xEASE), with
the EC1.1 promoter (Fig. 3b). The ratio (17.0 %) of EC1.2en-EC1.1p:zCas9-rbcS_E9t plant-derived likely triple
mutants to total number of T1 transgenic lines greatly increased (Fig. 3b, Additional file 4: Figure S8) in comparison with those for the EC1.2p/EC1.1p:zCas9-rbcS_E9t plant-derived
mutants (8.3 % and 1.8 %, respectively, Fig. 3a). The ratio (28.3 %) of EC1.2en-EC1.1p:zCas9-rbcS_E9t plants with observable mutations
to total number of T1 transgenic lines also greatly increased (Fig. 3b, Additional file 4: Figure S8) in comparison with those for the EC1.2p/EC1.1p:zCas9-rbcS_E9t plants
with observable mutations (11.1 % and 14.3 %, respectively, Fig. 3a). These results demonstrated that the EC1.2en-EC1.1p fusion promoter performed much
better than the single EC1.2 or EC1.1 promoters, and the enhancer from the EC1.2 promoter significantly improved egg cell-specificity and expression strength of EC1.1 promoter. Unexpectedly, when we added double EASE enhancers into the EC1.2en-EC1.1p
fusion promoter, the resultant fusion promoter caused lower mutation efficiency: only
8.3 % (10/120) EC1.2en-2xEASE-EC1.1p:zCas9-rbcS_E9t plants harbor the observable mutations
(Fig. 3b, Additional file 4: Figure S9). These results suggested that EC1 and EASE have different mechanisms for egg cell-specific expression, and the two
mechanisms seem to be antagonistic.