Introduction

Engineering loss- and gain-of-function alleles by reverse genetic studies is a powerful approach for dissecting the biochemical function of a gene product in vivo. Several different tools and resources are available for reverse genetic approaches in plants, including RNA-mediated interference (RNAi)1, virus-induced gene silencing (VIGS)2, targeted induced local lesions in genomes (TILLING)3, clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein 9 (Cas9) (CRISPR/Cas9)4 and gene targeting (GT)5. Gene targeting is a technique for introducing site-specific genome modifications at an endogenous locus via homologous recombination (HR) with exogenous DNA. Because of its site-specific precision and the almost unlimited types of DNA modifications that can be introduced, e.g. deletions, inversions, insertions and point mutations, GT remains one of the most powerful methods available to analyze gene function. However, in angiosperm plants as well as in most higher eukaryotes, GT is inefficient due to the fact that the frequency of random DNA integration by illegitimate recombination (IR) exceeds that of targeted integration by HR by several orders of magnitude6, 7. Mosses, particularly Physcomitrella patens, where GT is as efficient as in Saccharomyces cerevisiae 5, 8, and much more efficient than in mouse ES cells9 and in the green alga Chlamydomonas reinhardtii 10, represent an exception to this rule in plants. Recently, it was demonstrated that GT-efficiency could be significantly improved in moss using the CRISPR/Cas9 system11, 12. Highly efficient GT in P. patens also permits complex in vivo assembly of multiple DNA fragments with overlapping sequences of as short as 12 bp13. This also expands the potential of this model organism for biotechnological applications similar to that of yeast14, 15.

The intrinsic efficiency of HR in P. patens has made GT the predominant method for functional genomics in this species. Knock-out of a gene in P. patens is routinely achieved using a construct harboring a positive selection marker gene, such as those conferring neomycin (neo) or hygromycin (hpt) resistance, flanked by targeting sequences (TS) homologous to the 5′ and 3′ flanking regions of the sequence to be modified16. Upon transformation, homologous exchange between the TSs in the vector and their genome counterparts results in a targeted gene deletion and insertion of the marker cassette. Constructs for homology-dependent gene insertion (knock-in) are similar in design to knock-out vectors, except that the sequence to be introduced is fused in-frame with one of the TSs in the vector. The selection marker gene used in GT vectors can be flanked by loxP sequences that later can be exploited to obtain marker-free mutant lines by Cre/lox-mediated recombination17,18,19,20,21,22. The site-specific Cre recombinase, which is introduced by transformation of a Cre-expressing vector23, specifically recognizes the recombining loxP sites and mediates deletion of the marker gene, leaving a singular loxP site at the recombined locus23.

The majority of HR-promoted gene disruptions in P. patens represent single gene knock-out mutants generated to study gene-function relationships, in addition to gene tagging experiments allowing in vivo monitoring of target gene expression and fusion protein localization. However, since most eukaryotic proteins are composed of multiple domains, often with specific functions, mutants harboring specific partial gene deletions are often desirable for functional studies. Design of the vector for targeted gene manipulations of individual domains is a critical step, and requires knowledge of the intron-exon borders in order to precisely modify the nucleotide sequence corresponding to the domain of interest while preserving the reading frame of the gene. In addition, deletion of individual domains located internally in a protein and/or for which the coding sequence spans two or more exons are further complicated by the insertion of the selection marker into intronic sequences, which potentially can interfere with splicing. P. patens mutants harboring partial gene deletions have been reported, the deleted domains being either located at the N- or C-terminus of the protein24. Although common practice in yeast25, 26 and mice27, 28, reports of targeted partial deletion of sequences corresponding to internal protein domains are rare in plants.

DEK1 is a large multi-domain membrane-anchored cysteine protease of approximately 240 kDa composed of an N-terminal membrane domain (MEM) harboring 23 transmembrane segments linked to the C-terminal calpain domain (CysPc-C2L) by a Linker segment, containing a Laminin-G3 (LG3) domain (Fig. 1)29. Previously, we created the P. patens DEK1 knockout mutant (Δdek1), in which the entire 14,569 bp DEK1 gene of 30 coding exons was deleted using GT. The Δdek1 mutant develops protonemata, filamentous cells exhibiting polar tip growth; however, leafy gametophores fail to form due to an early arrest of bud development30. The same phenotype is obtained in a mutant (dek1 0)29 in which the active site cysteine residue of the DEK1 protease core domain is substituted with serine, a mutation that also inactivates DEK1 calpain activity in vitro 31. Thus, mutant plants with calpain-null alleles display a clearly detectable phenotype with filamentous growth without gametophores. A much milder phenotype is observed in the P. patens dek1Δlg3 mutant, generated by in-frame removal of the sequence corresponding to the DEK1-LG3 domain. This mutant forms reduced gametophores with narrow phyllids lacking several features of the WT phyllids, including the midrib29.

Figure 1
figure 1

Intron-exon structure of the Physcomitrella patens DEK1 gene. DEK1 ORF consists of 30 exons (grey rectangles) encoding the MEM domain (blue line), the LINKER segment (red line), the LG3 domain (blue rectangle) and the CALPAIN domain (green line). The numbers identify the exons of the DEK1 ORF.

Full size image

In this report, we evaluate the performance of different DEK1 GT genome editing constructs for the production of mutants harboring specific internal gene deletions and insertions in P. patens. Generation of mutagenized DEK1 transcripts with an intact open-reading frame was used as a criterion for successful gene modification. Our results show that intragenic mutagenesis in P. patens is critically dependent on a vector design that conserves intron splicing signals. We propose useful guidelines and discuss important considerations in the design of GT vectors for targeting internal sequences of multi-exon genes in P. patens.

Results

Design of gene targeting vectors for internal exon deletions

Two different GT vector designs G1 and G2 were tested to delete the Linker encoded by DEK1 exons 17–24 (Fig. 1) in order to create a P. patens (Pp) mutant expressing a truncated DEK1 protein lacking the internal Linker segment. The G1 vector was designed to insert the 5′ and 3′ ends of the loxP-flanked Hygromycin Resistance Cassette (HRC) into DEK1 introns 17 and 23, respectively. With this vector, we expected to produce a mutant with targeted deletion of a 3198 bp DEK1 fragment (nucleotides (nt) 8347–11544; XP_001774206.1) of the Linker sequence. In addition, transformation was expected to yield a DEK1 hybrid intron composed of the HRC flanked by the 5′ and 3′ ends of introns 17 and 23, respectively (Fig. 2a). The G2 vector was designed to insert the loxP-flanked HRC into DEK1 intron 17 without affecting the intron nucleotide composition (Fig. 2b). To this end, the G2 5′ targeting sequence (TS) contained ~1.3 kb genomic sequence (nt 7249–8404) ending in DEK1 intron 17. The ~1.2 kb 3′ TS was a mutagenized genomic sequence (nt 8402–12893, Δ8462–11718) composed of DEK1 nucleotides from intron 17 followed by the first 12 nucleotides of exon 18 fused in frame with the 3′ end of exon 24, to intron 26. Using the G2 vector for transformation, we expected targeted deletion of the DEK1 sequence from within exon 18 to within exon 24, in addition to insertion of the HRC into the full-length intron 17 (Fig. 2b).

Figure 2
figure 2

DEK1-Linker knock-out. Schematic representation of the DEK1-Linker knock-out strategies using the (a) G1 and (b) G2 vectors and elimination of the hygromycin resistance cassette (HRC) (light blue box) by Cre/loxP-mediated recombination. The grey boxes correspond to exons in the P. patens DEK1 gene. Introns are colored in orange, with the exception of introns 17 (blue) and 23 (red). The 5′ and 3′ targeting sequences (TS) are boxed. The loxP sites are shown as vertical black lines. Numbers identify the exons of the DEK1 gene. (c) cDNA sequencing electropherogram of the dek1∆linkerG1/cre DEK1 transcript showing deletion of exon 18–23 and retention of the loxP-harboring hybrid intron 17/23, creating a premature termination code (red box). (d) cDNA sequencing electropherogram of the dek1∆linkerG2/cre DEK1 transcript showing in-frame fusion of exons 17, 18 and 24. (e) Phenotype of 3-week-old dek1∆linkerG1/cre and dek1∆linkerG2/cre plants. For comparison, wild-type (WT) and ∆dek1 are also shown. WT plants with developing gametophores. dek1∆linkerG1/cre and dek1∆linkerG2/cre mutant plants are showing filamentous growth and aborted buds (arrows). dek1∆linkerG1/cre bud development stops at an early stage leading to small arrested buds, similar to ∆dek1; the buds of dek1∆linkerG2/cre mutant continue to divide, giving rise to larger disorganized bud structures. Scale bars: grey = 2 mm, black = 500 μm and white = 25 μm.

Full size image

Following transformation of wild-type P. patens protoplasts with the G1 and G2 vectors, six dek1ΔlinkerG1 and one dek1ΔlinkerG2 hygromycin-resistant (HygR) on-locus targeted mutant lines, respectively, were identified by PCR (data not shown). These plants all displayed a phenotype similar to the dek1 0 mutant29 with filamentous growth and lacking gametophores, as shown for the dek1ΔlinkerG1#20 and dek1ΔlinkerG2#1 lines in Supplementary Fig. S1. Insertion of the 1.8 kb HRC in an intron may potentially cause a null mutant phenotype by interfering with posttranscriptional modification of the transcript, preventing expression of the active protein. This was subsequently confirmed by reverse-transcription (RT)-PCR analysis using primers annealing outside TS and within the resistance marker cassette, with positive amplification signals for the presence of the HRC in the cDNA amplified from the two mutant lines (data not shown). Thus, in order to eliminate the resistant marker from the DEK1 locus, dek1ΔlinkerG1#20 and dek1ΔlinkerG2#1 were subjected to Cre/loxP-mediated recombination, resulting in the hygromycin-sensitive (HygS) lines named dek1ΔlinkerG1/cre and dek1ΔlinkerG2/cre. The mutants were verified by Southern hybridization confirming targeted deletion of the Linker, no sign of secondary non-specific integration, and loss of the HRC from the locus (Supplementary Fig. S2a). To investigate DEK1 transcript processing in dek1ΔlinkerG1/cre and dek1ΔlinkerG2/cre, a cDNA spanning exon 11 to exon 27 was amplified by RT-PCR using primers annealing outside the TSs, resulting in the generation of a single PCR product (Supplementary Fig. S2b) which was subsequently sequenced (Fig. 2c,d). Analysis of the dek1ΔlinkerG1/cre cDNA revealed retention of the hybrid intron 17/23 in the truncated DEK1 transcript (Fig. 2c). Retention of this intron introduced a premature stop codon 5′ to the calpain coding sequence (Fig. 2c), caused by a frame-shift mutation. In contrast, analysis of the dek1ΔlinkerG2/cre cDNA confirmed the expected splicing product of the truncated DEK1 transcript, with the 5′ end of exon 18 fused in frame with the 3′ end of exon 24 (Fig. 2d). Consistent with the molecular data, dek1ΔlinkerG1/cre displayed a typical DEK1 calpain-null mutant phenotype29, 30 with lack of gametophores and aborted buds (Fig. 2e). In contrast, dek1ΔlinkerG2/cre mutant buds divide beyond those of Δdek1, forming larger disorganized buds with abortive growth (Fig. 2e), suggesting the presence of residual levels of DEK1 calpain activity in the mutant29.

Design of gene targeting vectors for internal exon insertions

We next tested two different GT vector designs (In23-5′ and In23-3′) for knock-in of wild-type, heterologous and mutated DEK1-LG3 sequences into the Pp dek1Δlg3 mutant. The LG3 domain is encoded within exon 22 and 23 of the PpDEK1 gene (Fig. 1), and the dek1Δlg3 mutant contains a deletion in the genomic region spanning the LG3 domain29. Both for the In23-5′ and the In23-3′ designs, the complementing LG3 sequence is fused in frame with the 5′ TS, and the loxP flanked HRC are targeted to intron 23, however the position of the HRC insertion within intron 23 differed between the two (Fig. 3a,b).

Figure 3
figure 3

DEK1-LG3 knock-in. Schematic representation of the DEK1-LG3 knock-in strategies using the (a) In23-3′ and (b) In23-5′ vectors, and elimination of the hygromycin resistance cassette (HRC) (light blue box) by Cre/loxP-mediated recombination. The grey, blue and green boxes correspond to the exons in the P. patens DEK1 gene. The blue and green boxes represent the fused exon 22 and 23, respectively, of the dek1Δlg3 mutant. Introns are colored in orange, with the exception of intron 23, which is colored red. The 5′ and 3′ targeting sequences (TS) are boxed. The loxP sites are shown as vertical black lines. The numbers correspond to the exons of the DEK1 gene. For simplicity, the PpLG3-M1, PpLG3-M2 and PpLG3-M3 mutant lines are collectively denoted as Pp*LG3. (c) cDNA sequencing electropherogram of the PpLG3-3′ DEK1 transcript showing retention of the loxP-harboring intron 23, resulting in a premature stop codon (red box). The slanted lines denote omitted sequences due to space limitation. (d) Phenotype of 3-week-old PpLG3-3′ mutant plant showing filamentous growth without gametophores. (e) cDNA sequencing electropherogram of the PpLG3-5′ DEK1 transcript showing in-frame fusion of exon 23 and exon 24. (f) Phenotype of 3-week-old PpLG3-5′ mutant plant showing well-developed gametophores, demonstrating successful complementation of the dek1Δlg3 mutant phenotype. Scale bars: 2 mm.

Full size image

In the In23-3′ design, the HRC was targeted to the end of intron 23, 27 bp upstream of the 3′ end of the intron (Fig. 3a and Supplementary Fig. S3a). In this set of experiments, wild-type genomic LG3 sequences from Pp, Arabidopsis thaliana (At) and Marchantia polymorpha (Mp) were targeted to the Pp DEK1ΔLG3 locus. In the In23-5′ design, the HRC was targeted to the start of intron 23, 28 bp downstream of the 5′ end of the intron (Fig. 3b and Supplementary Fig. S3b), and in this set of experiments, we targeted the wild-type and three mutated Pp LG3 versions (LG3/M1, LG3/M2 and LG3/M3, see Methods) to the DEK1ΔLG3 locus. Following transformation of dek1Δlg3 protoplasts with the vectors built using the In23-3′ design, we obtained 40 PpLG3-3′, 105 MpLG3 and 74 AtLG3 HygR lines in which four, five and one, respectively were on-locus targeted as demonstrated by PCR genotyping (data not shown). The HRC was subsequently removed from one representative of each of the PpLG3-3′, AtLG3 and MpLG3 lines using Cre/loxP-mediated recombination. DEK1 cDNA of marker-free PpLG3-3′/cre, AtLG3/cre and MpLG3/cre lines were amplified and analyzed as described in the preceding section, confirming insertion of the wild-type (Supplementary Fig. S3c(i)) and heterologous LG3 sequences (Supplementary Fig. S3c(ii,iii)) at the DEK1 locus. In all lines, however, intron 23 was retained in the DEK1 transcript, as shown by DNA sequencing of the DEK1 cDNA amplification-product of line PpLG3-3′/cre in Fig. 3c. Retention of the loxP-harboring intron in the DEK1 transcript caused a frame-shift mutation leading to a premature termination codon and a putative truncated DEK1 protein lacking the catalytic calpain domain. Consistently, all mutant lines displayed a calpain-null mutant phenotype similar to Δdek1 and dek1 0 with the lack of gametophores as shown for PpLG3-3′/cre in Fig. 3d, and for AtLG3/cre and MpLG3/cre in Supplementary Fig. S3d. In addition to retention of Pp DEK1 intron 23, the AtLG3/cre DEK1 transcript displayed additional pre-mRNA splicing errors, specifically the loss of 64 exonic nucleotides corresponding to 61 nucleotides at the 3′ end of At DEK1 exon 21 followed by the first three nucleotides of At exon 22 (Supplementary Fig. S3c(ii)). In contrast, sequencing of the DEK1 cDNA from MpLG3/cre confirmed proper splicing of the heterologous Mp LG3 sequence (Supplementary Fig. S3c(iii)).

Transformation of dek1Δlg3 protoplasts with vectors built using the In23-5′ strategy (Fig. 3b and Supplementary Fig. S3b), yielded 3 PpLG3-5′, 3 PpLG3-M1, 5 PpLG3-M2 and 8 PpLG3-M3 HygR on-locus targeted lines (data not shown) with the HRC from one representative line for each having been removed by Cre/lox-mediated excision. Sequence analysis of DEK1 cDNA amplified from marker-free lines showed in-frame re-targeting of the Pp LG3 sequences and proper splicing at the locus, as shown for mutant PpLG3-5′/cre in Fig. 3e. Sequencing also confirmed the presence of the desired LG3 mutations in the DEK1 transcript of PpLG3-M1/cre, PpLG3-M2/cre and PpLG3-M3/cre (Supplementary Fig. S3e). The PpLG3-5′/cre (Fig. 3f) and PpLG3-M1/cre, PpLG3-M2/cre and PpLG3-M3/cre mutants (Supplementary Fig. S3f) all developed gametophores with fully expanded phyllids similar to wild-type plants, and clearly distinguishable from the narrow phyllid phenotype of the dek1Δlg3 mutant29 (Supplementary Fig. S3f), thus confirming successful complementation at the DEK1ΔLG3 locus using the In23-5′ vector design.

Discussion

In vivo genetic engineering of internal exonic sequences is complicated by the targeting of intronic sequences with a selection marker cassette, which may interfere with post-transcriptional modifications, disrupting gene expression. Although the resistance marker cassette can be efficiently eliminated from the targeted intron using e.g. the Cre/lox system, one loxP site will be retained at the locus following recombination. In this study, we show that successful genetic modifications of internal protein domains is dependent on the critical and rational design of the GT construct in order to avoid gene targeting-associated mutagenesis of introns having detrimental effects on splicing of the targeted transcript.

Two different knock-out vectors (G1 and G2) for targeted deletion of the internal DEK1 Linker were constructed and tested in P. patens plants. Transformation with both vectors resulted in mutants with targeted deletion of the Linker. However, while the G2 strategy resulted in a mutant plant (dek1ΔlinkerG2/cre) expressing a DEK1 transcript with in-frame deletion of the Linker, the G1 strategy resulted in a mutant plant (dek1ΔlinkerG1/cre) expressing a DEK1 transcript containing a frame-shift mutation leading to a premature termination code. The frame-shift mutation was caused by retention of the loxP-harboring hybrid intron 17/23, impairing DEK1 pre-mRNA splicing. Thus, the G1, but not the G2 strategy resulted in the introduction of mutations causing splicing defects of the targeted intron. Three structural elements are used to correctly excise introns from primary transcripts. These include the 5′ and 3′ splicing sites spanning the 5′ exon/intron and the 3′ intron/exon borders respectively, and the internal branch point (BP) sequence located 19–50 nt upstream of the 3′ splicing site32. The hybrid DEK1 17/23 intron of dek1ΔlinkerG1/cre contains both the heterologous 5′ and 3′ splicing sites, but lacks the internal sequences of the native introns since only the 5′ and 3′ portions of intron 17 and 23, respectively, are retained after recombination between the G1 donor vector and the DEK1 locus (Fig. 2a). This likely resulted in a loss of internal splicing signals, e.g the BP sequence within the targeted intron, thus abolishing splicing of exon 17 and 24. In contrast, splicing of the mutagenized, loxP-harboring DEK1 intron 17 of the dek1ΔlinkerG2/cre mutant occurred properly. In this intron, no nucleotides were deleted during targeted insertion of the HRC. These results demonstrate that when designing GT knock-out vectors aimed at deleting internal exonic sequences, a strategy that preserves the native sequence of the targeted intron(s) and which minimizes or eliminates the loss of intronic nucleotides, should be chosen. This is achieved by using a strategy (G2) as described in this study; the selection marker cassette in the donor vector is placed within the sequence of the native intron (Fig. 2b). The 3′ end of this intron is followed by a sequence of the adjacent (3′) exon, which is fused in frame to the 3′ TS starting with a downstream exon. This design thus creates a heterologous exon at the targeted locus, containing a mutagenized sequence harboring a deletion of the genomic sequence intervening the original exons of the heterologous exon. Using this strategy, intronic sequences are preserved.

We also designed and evaluated two different GT knock-in vectors (termed In23-5′ and In23-3′) for targeted insertion of wild-type, heterologous and mutated versions of DEK1-LG3 sequences into the P. patens dek1Δlg3 mutant. Knock-in of a gene fragment is commonly performed in genetic complementation tests to confirm that a mutant phenotype is caused by a specific gene deletion, or to introduce modified versions of the gene for functional studies. In complementation tests where the outcome of the gene modification cannot be predicted, possibly giving a phenotype indistinguishable from that of the background line used for transformation, a positive selection marker is needed to identify transformed plants. Similar to gene knock-out approaches, knock-in of DNA fragments into internal regions of a gene is complicated by the necessity of targeting the selection marker gene to introns. Transformation of dek1Δlg3 protoplasts with In23-5′-based vectors, where the resistant marker was inserted 28 bp downstream of the start of intron 23, expressed DEK1 transcripts in which the reading-frame was maintained, and the plants reverted to a wild-type gametophore phenotype, confirming successful knock-in of DEK1-LG3 into the DEK1ΔLG3 locus. In contrast, dek1Δlg3 protoplasts transformed with In23-3′-based vectors, where the resistance marker was inserted 27 bp upstream of the end of intron 23, expressed DEK1 transcripts containing the loxP-harboring intron 23. These plants displayed a calpain-null mutant phenotype, as expected for DEK1 plants lacking the catalytic calpain domain. These results demonstrate that HRC insertion into the 3′ end, but not the 5′ end, of DEK1 intron 23 impairs splicing, suggesting that a residing loxP sequence located 3′, but not 5′, corrupts the internal splicing signals of the intron. To explore this possibility further, we used the ERISdb program (http://lemur.amu.edu.pl/share/ERISdb/home.html)33, which predicts splice sites and splicing signals in various plant species, including P. patens. Using this tool, we predicted a putative branch site 36 bp upstream of the 3′ splice site of DEK1 intron 23, close to the position in which the loxP sequence is retained in the mis-spliced mutagenized DEK1 intron 23. The BP sequence is important for splicing in plants34, the sequence between the BP and the 3′ splicing site, tending to be U-rich, strongly influences splicing efficiency and splice site selection in plants35. Thus, insertion of a loxP site into this region may have corrupted the splicing signal of the intron, causing intron 23 to evade spliceosome activity. Our results demonstrate that selection of the marker cassette insertion point within the intron is important when designing GT vectors. BP sequences are normally located 19–50 nucleotides upstream of the 3′ splicing site. In order to avoid interfering with essential BP signals, we recommend to insert the selection marker close to the 5′ end of the intron and, if possible, at least 50 nucleotides upstream of the 3′ end of the intron. Correspondingly, the G2 donor vector that was successfully used to create the mutant plant expressing the DEK1 transcript harboring an in-frame deletion of the Linker, was also designed to insert the resistant marker to the 5′ end of the targeted intron. Thus, our results show that targeting of the resistance marker cassette to the 5′ end of the intron should be preferred. It should be mentioned that choosing an appropriate position to insert the marker cassette may be problematic if the targeted intron is small. In such case, identifying a position where GT-associated mutagenesis (loxP retention) will not affect splicing, may be challenging. Therefore, if the insertion position within the gene is not critical for the experiment, we recommend the targeting of larger sized introns.

Recently it was shown that the piggyBAC transposon could be used for the precise and effective excision of selection marker cassettes from the genome without leaving any residual ectopic sequences both in mammalian cells36 and rice37, 38. The transposon, which derives from the cabbage looper Trichoplusia ni 39, integrates into the host genome at TTAA elements and has been used efficiently for transgenesis in animals40. The piggyBAC transposon thus offers an alternative method to the Cre/loxP system for elimination of a selection marker cassette from genomes. However, with the exception that the piggyBAC system has yet to be tested in P. patens, the main limitation of the system is the requirement for specific insertion into a TTAA sequence, maxing the system less flexible especially for precise site-directed mutagenesis.

An attractive and powerful alternative to classical GT methods in P. patens is the CRISPR/Cas9 system12. In P. patens both transformation frequency and GT efficiency have been found to be significantly enhanced using CRISPR/Cas9 when compared to the classical method using GT donor vectors only11. However, no reports have been published on the use of the CRISPR/Cas9 system for precise partial deletions or insertions of gene sequences at internal positions. Ultimately, the CRISPR/Cas9 approach may be efficient enough to bypass the need of the resistance cassette for such partial deletions. But for now, as with classical GT approaches, insertion of a resistance marker cassette to internal intronic sequences is necessary in order to select for transformants receiving the donor DNA. Thus, the results reported in the present study would also apply to the design of the donor vector using the CRISPR/Cas9 system.

In this study, we also demonstrate that P. patens has the capacity to properly splice a genomic fragment from the liverwort M. polymorpha, suggesting conservation of splicing signals between mosses and liverworts. In contrast, the genomic A. thaliana DEK1 sequence in P. patens is incorrectly processed, suggesting the possibility that the moss spliceosome machinery does not properly recognize the splicing signals from dicot plants. This may stem from differences in the UA content of the introns; P. patens introns display a significantly lower UA content, 60%, compared to 67% in A. thaliana 41. Notably, Perroud and Quatrano42 showed that P. patens correctly splices the BRK1 gene from A. thaliana, indicating that accurate processing of certain A. thaliana introns occurs in the moss.

It is worth noting that another knock-in GT strategy can be used when the expected phenotype of the resulting mutant is morphologically distinguishable from that of the transformation background. In this strategy, the GT vector does not carry a resistance marker and the complementing DNA is flanked by the 5′ and 3′ TSs. After transformation with the resistance marker-free vector, transformed mutants harboring single copy on-locus GT events can easily be identified based on the phenotype. The use of this approach offers two advantages; elimination of the time-consuming Cre/loxP-mediated transformation step to remove the resistance marker and no introduction of loxP-associated mutations affecting splicing integrity of the targeted intron.

Finally, in this study we investigated the functional significance of highly conserved amino acid residues in the DEK1-LG3 domain by performing site-directed mutagenesis using the In23-5′ strategy to knock-in mutated versions of LG3 into the dek1Δlg3 mutant. HRC-free PpLG3-M1, PpLG3-M2 and PpLG3-M3 mutant lines, harboring substitution of amino acids E1477Q1478(M1), E1481(M2) and S1497(M3) to alanine, respectively, all developed wild-type like gametophores indicating that the targeted amino acids are not critically important for DEK1-LG3 function during gametophore development. Also, worth noting is that we have observed that transformation of wild-type protoplasts with constructs designed to introduce site-directed amino acid substitutions may result in selection of marker-resistant plants harboring the wild-type sequence at the targeted position. Therefore, we recommend the use of deletion mutants (lacking the sequence for which the substitution is to be made), rather than wild-type plants, as the transformation background for creating mutants harboring single (or few) amino acid substitutions.

Methods

Plant material, growth conditions and transformation

The Physcomitrella patens Gransden strain and the P. patens dek1Δlg3 mutant29 were used throughout this study. Tissue maintenance and production was performed on BCDA medium as described by Cove and co-workers43. P. patens tissue and protoplasts were grown under long day conditions (16 hours light [70 to 80 µmolm−2 s−1]/8 hours dark) at 25 °C. P. patens protoplast transformation was performed as described in ref. 43. Growth medium was supplemented with 20 µgl−1 of Hygromycin B for selection of transformed plants. Tissue for phenotypic characterizations was grown on BCD medium. Tissue for reverse transcription PCR (RT-PCR) and Southern blot analysis was grown on BCDA medium overlaid with cellophane discs under long day conditions. The tissue was harvested after 7 days, snap-frozen in liquid nitrogen, ground in a mortar and used for RNA and DNA extraction.

Construction of gene targeting vectors

The sequences of the primers used for vector constructions are listed in Supplementary Table S1. Plasmid cloning was performed using the In-Fusion® HD Cloning kit (Clontech Laboratories) according to the manufacturer’s instructions, unless otherwise stated. The PCR-generated fragments used in the In-Fusion reactions were amplified with Thermo Scientific® Phusion High-Fidelity DNA Polymerase. To verify the constructed vectors, restriction digestion and DNA sequencing analyses were performed. All nucleotide (nt) numberings are relative to the A1TG start site of the P. patens (Pp) DEK1 genomic sequence (Pp3c17_17550; www.phytozome.net), unless otherwise stated.

Knock-out vectors: Construction of the pBHRF∆LinkerG1 vector is described by Johansen et al.29. In short, fragments spanning nt 7249–8346 and nt 11545–1254 as the 5′ and 3′ TSs, respectively, flank the hygromycin resistance cassette in pBHRF∆LinkerG1 (Fig. 2a). GenScript Inc. (USA) performed fragment synthesis and cloning of the pBHRFΔLinkerG2 vector. The following fragments were synthesized de novo: the 5′ TS, nt 7249–8404 flanked by HindIII and NruI recognition sequences and the 3′ TS, a mutagenized fragment spanning from nt 8402–12893 with a 3257 bp deletion from nt 8462–11718 and flanked by 5′ SpeI and 3′ NsiI recognition sequences. The 5′ and 3′ TSs were cloned sequentially into pBHRF20 using HindIII/NruI and SpeI/NsiI restriction enzyme pairs, respectively, resulting in vector pBHRF∆LinkerG2 (Fig. 2b). Prior to P. patens protoplast transformation, the pBHRF∆LinkerG1 and pBHRF∆LinkerG2 vectors were digested with the restriction enzymes BmrI and PacI.

Knock-in vectors: The vectors to re-insert WT and heterologous DEK1-LG3 sequences into the DEK1ΔLG3 locus in the dekΔlg3 mutant using the In23-3′ strategy (Fig. 3) were constructed from vector pBHRF∆LinkerG129. To make the vector containing the WT LG3 sequence, nt 8347–11544 was first PCR-amplified using primers SP_Inf_6 N and ASP_Inf_6 N from genomic P. patens DNA. The resulting PCR product was mixed with linearized (at nt position 8346) pBHRFΔLinkerG1 DNA, produced by PCR using primers SP_Inf_5 N and ASP_Inf_5 N. The linearized vector and insert were ligated using an In-Fusion reaction, resulting in the WT LG3 complementation vector pBHRF_PpLG3_In23-3′ (Fig. 3a). To make the construct for DEK1 LG3 cross-species complementation tests, the WT genomic DEK1 LG3 sequences from Arabidopsis thaliana (At) (nt 6741–7450; relative to the A1TG start site in AT1G55350) and Marchantia polymorpha (Mp) (nt 1258573–1259287; scaffold 26, JGI scaffold genomic data v3.1) were first PCR-amplified using primer sets MpLG3_Inf-F/MpLG3_Inf-R and AtLG3_Inf-F/AtLG3_Inf-R, respectively. Vector pBHRF_PpLG3_In23-3′ was PCR-amplified to delete nt 10567–11239, corresponding to the DEK1 LG3 sequence, using primers SP_Inf_2 and ASP_Inf_2X. Finally, the linearized vector and the AtLG3 and MpLG3 amplicons were ligated in two separate In-Fusion reactions creating the cross-species complementation vectors pBHRF_AtLG3_In23-3′ and pBHRF_MpLG3_In23-3′, respectively. The vectors to re-insert P. patens DEK1 LG3 WT and mutated sequences using the In23-5′ strategy (Fig. 3b) were built from pBHRF20. First, the 3′ TS (nt 11334–12511) was PCR-amplified using primers 3TS-F and 3TS-R. The resulting product was cloned into linearized plasmid pBHRF, produced by PCR using primers pBHRF-F and pBHRF-R, resulting in plasmid pBHRF_3′ TS. The 5′ TS (DEK1 nt 9568–11333), which harbors the LG3 coding sequence, was PCR-amplified from genomic P. patens DNA using primers 5TS-F and 5TS-R and subsequently cloned into the pCR-Blunt vector (Invitrogen) to obtain the pCR_5′ TS vector. To make mutated versions of the LG3 sequence, the pCR_5′ TS plasmid was subjected to mutagenesis reactions using the GeneArt® Site-Directed Mutagenesis PLUS kit (Invitrogen) with the mutagenic primer pairs M1-F/M1-R, M2-F/M2-R and M3-F/M3-R, resulting in three mutated 5′ TSs containing the M1 (E1477A;Q1478A), M2 (E1481A) and M3 (S1497A) substitutions, respectively. The WT and mutagenized versions of the 5′ TS were each amplified with primers 5TS_Inf-F and 5TS_Inf-R, and sub-cloned into XhoI-linearized vector pBHRF_3′ TS, resulting in the generation of the complementation vectors pBHRF_PpLG3_In23-5′, pBHRF_PpLG3_M1_In23-5′ pBHRF_PpLG3_M2_In23-5′ and pBHRF_PpLG3_M3_In23-5′. Prior to P. patens protoplast transformation, these vectors were digested with the restriction enzymes BmrI and PacI.

PCR genotyping of transformants

Each mutant was subjected to three rounds of PCR genotyping using the Phire Plant Direct PCR kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. The sequences of the primers used for genotyping each set of transformants are provided in Supplementary Table S1. For primary genotyping, knock-out and knock-in mutants were genotyped for the loss and gain of genomic sequence, respectively. Secondary genotyping PCR was performed to verify proper 5′ and 3′ targeting at the locus using a primer pair annealing within the DNA fragment used for transformation and 5′ or 3′ to the 5′ and 3′ TSs, respectively. Finally, tertiary genotyping PCR, mutants harboring single-copy insertion events at the locus were identified using primers annealing 5′ and 3′ to the 5′ and 3′ TSs, respectively.

For dek1∆linkerG1 and dek1∆linkerG2, primary genotyping was performed using primers Linker-F and Linker-R. Lines without a detectable PCR product were subjected to the second PCR using primer pairs Linker-2F and 35S_rev1 for the 5′ targeting and, Term-fwd and Linker-2R for the 3′ targeting. Lines with both 5′ and 3′ targeting were assessed for single copy insertion at each locus using the primers Linker-2F and Linker 2 R. For the In23-3′ lines, gain of sequence was assessed using primers ARMSeq 6 and ASP_PpARM_inf 1. Positive lines were genotyped for on-locus targeting using primers TM2Seq1 and 35S_rev1 for 5′ targeting, and Term_fwd and Ex30 R for 3′ targeting. Single copy insertion was assessed using TM2seq1 and Ex30 R. For In23-5′ lines, primary genotyping was performed using primers LG3_fwd and LG3_rev; secondary genotyping: ARMSeq3 and 35S_rev1 for 5′ targeting and, Term_fwd and Ex30 R for 3′ targeting, respectively. Tertiary genotyping was performed using primers ARMSeq3 and Ex30 R.

Molecular characterization of transformants

Genomic DNA for Southern Blot analysis was extracted using the Nucleon™ PhytoPure™ Genomic DNA Extraction Kit (GE Healthcare). Southern Blot was performed as described19 using ca. 1 µg DNA per digestion. Probes were DIG-labelled using the DIG Probe PCR synthesis kit (Roche, Indianapolis, USA) according to the manufacturer’s instructions. pBHRF∆LinkerG1 DNA was used as template for PCR amplification of the TS and hygromycin-resistance probes; primers for amplification of the; 5′ TS probe: Δarm5′-fwd and Δarm5′-rev; 3′ TS probe: Δarm3′-fwd and Δarm3′-rev; HRC probe: HRC-fwd and HRC-rev. The sequences of the primers used to generate the probes are provided in Supplementary Table S1.

RT-PCR and DNA sequencing were employed to amplify and analyze the DEK1 cDNA to verify proper deletion/insertion and splicing of the transcript. Total RNA was isolated from protonema tissue using the Plant RNeasy Kit (Qiagen). Five hundred ng DNaseI-treated total RNA was reverse-transcribed using 100 U Superscript™ III Reverse Transcriptase (Invitrogen) primed with 200 ng random hexamers at 55 °C for 60 min. Phusion® High-Fidelity DNA polymerase was used to amplify the DEK1 cDNA target using primers TM2Seq1 and ASP_PpCALP, annealing in exons 11 and 27, respectively. After Exo-SAP treatment to remove unincorporated primers, the PCR product was sequenced with appropriate primers using BigDye® v3.1 chemistry (ABI) according to the SteP method44. Extension products were precipitated using sodium-acetate/ethanol, dissolved in Hi-Di™ Formamide (ABI) and finally sequenced by Capillary Electrophoresis using the 3130xL Genetic Analyzer. DNA sequences were analyzed using Genomic Workbench v7.

Microscopy

The mutants were observed by light microscopy.