Abstract
Detailed functional analyses of many fundamentally important genes via conventional loss-of-function approaches are impeded by severe pleiotropic phenotypes. In particular, mutations in genes that are required for basic cellular functions or plant reproduction often interfere with the generation of homozygous mutant plants precluding further functional studies. To overcome this limitation, we devised a CRISPR-based tissue-specific knockout system, CRISPR-TSKO, enabling us to generate mutations in particular plant cell types, tissues, and organs. In Arabidopsis, resulting somatic mutations even in essential genes caused well-defined, localized phenotypes with high efficiency in the root cap, stomatal lineage, or in entire lateral roots. The underlying modular cloning system allows for efficient selection, identification, and functional analysis of mutant lines directly in the first transgenic generation. The efficacy of CRISPR-TSKO opens new avenues to analyze gene functions in virtually any given spatial and temporal context of plant life while avoiding pleiotropic effects of system-wide loss of gene function.
Introduction
The generation of stable, inheritable loss-of-function mutant alleles has been indispensable for functional genomics studies in plants. Such knockout or knockdown lines have been generated with various techniques such as ionizing radiation, ethyl methanesulfonate, T-DNA or transposon insertions, RNA interference (RNAi), or artificial microRNAs. In recent years there has been an explosion in the generation of plant lines mutated by clustered regularly interspaced short palindromic repeat (CRISPR) gene knockout.
CRISPR technology contains two components, the CRISPR-associated (Cas) nuclease and CRISPR RNAs (crRNA) that direct the nuclease to the target nucleic acid. The most widely used CRISPR system in plants is based on the DNA endonuclease Cas9 and its artificial crRNA, the guide RNA (gRNA)1. In plants, Cas9 is very efficient at inducing double-stranded breaks in DNA. The DNA breaks repaired by the error-prone non-homologous end joining pathway ultimately result in the formation of short insertions and/or deletions (indels) at the break site2. These indels most often lead to frame shifts and early stop codons, facilitating an effective knock-out of the targeted gene(s).
Most, if not all, CRISPR efforts to date have focused on generating stable and inheritable DNA mutations in plants for reverse genetics approaches. Yet the ability to study essential genes is limited using this approach as loss of many fundamentally-important genes convey severe pleiotropic phenotypes up to lethality. Of the approximately 25,000 protein-coding genes in the Arabidopsis thaliana genome, 10% are estimated to be indispensable3, presenting a considerable challenge for functional analyses especially of genes with essential functions.
An approach to overcome these problems is the use of tissue-specific RNAi4. However, gene silencing is often incomplete, interfering with the interpretation of the observed phenotypes. Furthermore, small RNAs can be mobile, limiting tissue specificity5. Therefore, the results obtained using RNAi are often not comparable with the use of stably transmitted DNA-based mutants. For certain genes, transgenic vectors generating dominant-negative protein versions have been developed. Expressing these mutant versions in a tissue-specific context can locally interfere with endogenous gene functions6, 7. Other methods include the conditional knockout of genes in specific cell types or tissues by using a CRE-recombinase8. These approaches, however, can be cumbersome and difficult to scale up9.
Outside of the plant field, researchers have recently overcome such limitations with the development of conditional knockouts using CRISPR technology. In zebrafish, the gata1 promoter driving Cas9 expression has been used to knockout genes specifically in the erythrocytic lineage10. In Drosophila, targeted knockout mutations in two essential genes Wingless and Wntless only in germ cells permitted the generation of adult flies, whereas constitutive knockout individuals did not survive past the pupal stage11. Additionally, cardiomyocyte-specific expression of Cas9 led to organ-specific knockout in a mouse model12. In plant research, however, analogous approaches have not been reported thus far.
Here, we describe the development of a CRISPR tissue-specific knock-out (CRISPR-TSKO) system in Arabidopsis that allows for the specific generation of somatic DNA mutations in plants. The CRISPR-TSKO toolset is simple to use, highly efficient and allows for multiplexing and large-scale screening approaches. We show the potential CRISPR-TSKO for somatic gene knockouts of essential genes in diverse plant cell types, tissues, and even entire organs. Our approach opens new opportunities to study the function of fundamentally important genes in specific contexts of plant biology.
Results
Tissue specific GFP knockout in the lateral root cap
We reasoned that by utilizing tissue-specific promoters to drive Cas9 expression, CRISPR could be used to generate cell type-, tissue-, and organ-specific DNA mutations with high precision. To test this hypothesis, T-DNA vectors were constructed with Cas9 expression controlled by the promoter region of SOMBRERO/ANAC033 (SMB; AT1G79580). The SMB promoter (pSMB) is highly root cap-specific and is activated directly after the formative division of root cap stem cells13, 14. The pSMB:Cas9 expression cassette was combined with one of two individual U6-expressing gRNAs targeting the GFP coding sequence, GFP-1 and GFP-2, and transformed into a homozygous Arabidopsis line with ubiquitous expression of a nuclear-localized GFP and β-glucuronidase (GUS) fusion protein (pHTR5:NLS-GUS-GFP15, henceforth, NLS-GFP). Primary transgenic plants (T1 seedlings) were selected via resistance to the herbicide glufosinate and investigated for loss of GFP signal in the root tips of five-day-old seedlings. Strikingly, six out of eleven pSMB:Cas9;GFP-1 events and three out of ten pSMB:Cas9;GFP-2 events showed an almost complete loss of GFP specifically in the root cap, suggesting high efficiency CRISPR-mediated knockout soon after the formative division of the root cap stem cells (Supplementary Fig. 1a). All other root tissues maintained GFP expression, indicating that specific Cas9 activity in the root cap cells led to cell-autonomous GFP knockout. The tissue-specific effect is heritable, as three lines with pSMB:Cas9;GFP-1 and three lines with pSMB:Cas9;GFP-2, had T2 progeny without GFP fluorescence in root cap cells while having normal NLS-GFP expression in all other tissues examined (Supplementary Fig. 1b). These results clearly indicated that the use of a tissue-specific promoter driving Cas9 can result in somatic, tissue-specific knockout phenotypes.
Design of the CRISPR-TSKO gene knockout toolset
While we were encouraged by these initial results, two limitations were identified: (1) the specific expression of Cas9 in each line could not be easily monitored; and (2) herbicide selection on the T1 seedlings caused stress and limited the ability to directly phenotype primary transgenic T1 seedlings. To overcome these limitations, and to facilitate a wide range of future gene-modification approaches, we devised CRISPR-TSKO, a modular and versatile vector toolset based on Golden Gate technology using modified GreenGate vectors16, 17. CRISPR-TSKO is thus inexpensive and immediately compatible with modules already in use by other laboratories that have adopted the GreenGate vector system. The modularity allows for the combination of Cas9, or any nuclease, with virtually any promoter sequence of choice. Furthermore, Cas9 protein fusions are possible on the N- and C-termini allowing for the wide range of CRISPR technologies such as base editors18 or transcriptional regulators19. The promoter, Cas9, N- and C-tags, and terminator modules can be combined with an “unarmed” gRNA cassette to generate an unarmed destination vector (Fig. 1). One or two gRNAs can be directly cloned into the destination vector with a single Golden Gate reaction (see methods).
Alternatively, when an AarI linker is used instead of the unarmed gRNA cassette, a second round of Golden Gate assembly can be performed for the cloning of up to 12 gRNAs in a single destination vector (Supplementary Fig. 2).
A collection of binary destination vectors that contain different selectable markers and/or a non-destructive fluorescent markers based on the fluorescence-accumulating seed technology (FAST) system20 were generated to take advantage of this general cloning strategy (Supplementary Table 3). The FAST system allows for the antibiotic- or herbicide-free selection of transformed T1 seeds, and permits a time-saving screening for phenotypes directly in T1 seedlings. To facilitate the determination of tissue specificity and expression levels of Cas9, a nuclear-localized fluorescent mCherry tag to was fused to the Cas9 transcript via a P2A ribosomal skipping peptide21. Using this strategy we targeted different tissue types, cell lineages, and organs in Arabidopsis to explore the potential of CRISPR-TSKO in plants.
Root-cap specific gene knockout
Utilizing CRISPR-TSKO, Cas9-mCherry was expressed under pSMB and combined with the gRNA GFP-1. Ten of the 21 T1 seedlings showed a complete loss of GFP fluorescence specifically in the root cap, while six showed partial loss of GFP (chimeras), and five maintained normal GFP expression (Fig. 2a, Table 1). Of note, even in strong knock-out lines, cells of the youngest root cap layers retained some GFP signal (Fig. 2b). This result suggests that mRNA and/or protein turnover is required after the onset of Cas9 expression for the phenotype to become apparent. Interestingly, we could observe a clear correlation between the intensity of mCherry expression and the penetrance of the knockout phenotype. All ten highly-expressing mCherry lines were entirely devoid of root-cap GFP signal, the medium mCherry-expressing lines had chimeric knockout phenotype and the low-to-no mCherry lines had chimeric or full expression of GFP (Table 1). By comparing the intensity of both fluorescent proteins in root cap nuclei, we confirmed our observations that higher Cas9 expressing lines have a significantly higher probability of gene editing success (Fig. 2c).
To test if a root-cap expressed gene, SMB itself, could be successfully targeted by CRISPR-TSKO, the gRNA GFP-1 was combined with one of two different gRNAs targeting SMB (SMB-1 and −2) with Cas9 under pSMB. The loss of SMB activity delays root cap maturation and preparation of programmed cell death in root cap cells, causing larger root caps and a delayed and aberrant root cap cell death with a lack of cell corpse clearance14. In CRISPR-TSKO T1 seedlings, we found smb-like phenotypes for both independent SMB targeting gRNAs coupled with the disappearance of the root cap GFP signal (Supplementary Fig. 3a). Both gRNAs appear equally effective as 13 out of 21 and 9 out of 12 T1 events gave clear smb and GFP knockout phenotypes, respectively (Table 2, Supplementary Fig. 3a). The smb phenotype and loss of GFP in the root cap cells were also observed in four independent next-generation T2 lines, demonstrating the heritability of CRISPR-TSKO (Supplementary Fig. 3b).
To confirm that the observed phenotypes were due to root cap-specific DNA mutations, root tips from four independent lines (two for each SMB target) were protoplasted and used for fluorescence-activated cell sorting (FACS) followed by target sequence analysis. The mCherry-positive cell populations had indel frequencies >95% for the GFP-1, SMB-1 and SMB-2 target sites, (Table 3) as determined by TIDE analysis22. In contrast, the mCherry-negative cell populations had indel frequencies of 1-5% which are equivalent to wild-type using this method. Thus the consistent and heritable knockout phenotypes were due to active Cas9-induced mutagenesis specifically in root cap cells.
Cell-lineage-specific gene knockout in stomata
The cis-regulatory promoter elements of TOO MANY MOUTHS (TMM; AT1G80080) and FAMA (AT3G24140) control gene expression during stomata development, with TMM expressing early in the stomatal lineage23 and FAMA expressing later in the lineage of guard mother cells and young guard cells24. These two promoters were used to produce CRISPR-TSKO constructs targeting both GFP and Phytoene desaturase (PDS; AT4G14210) in the stomatal lineage. PDS is essential for chlorophyll, carotenoid and gibberellin biosynthesis and pds null mutants show a dwarfed and photobleached lethal phenotype25. Consistent with this, pPcUbi:Cas9;GFP-1, PDS vectors targeting PDS in a ubiquitous manner gave rise to severe phenotypic effects ranging from full albino to variegated leaves and stunted plants (Fig. 3a).
Given the expression patterns of FAMA and TMM, we expected the pTMM-CRISPR-TSKO constructs would show a loss of chlorophyll in all stomata-lineage-derived epidermal cells whereas the pFAMA-CRISPR-TSKO constructs should have the knockout phenotype restricted to the late stomatal lineage. At five days after germination, T1 seedlings for both pTMM:Cas9-mCherry;GFP-1, PDS and pFAMA:Cas9-mCherry;GFP-1, PDS were assessed for the presence of chlorophyll autofluorescence and GFP fluorescence by epifluorescence microscopy. Eighteen out of 20 pTMM:Cas9 T1 lines were clearly lacking both GFP and chlorophyll autofluorescence (Fig. 3b), indicating highly efficient gene knockout. Intriguingly, restricting the loss of PDS to the stomatal lineage did not markedly affect plant development, as T1 plants were able develop and set seed. Two out of 23 pTMM:Cas9 lines did exhibit some mild bleaching similar to the pPcUbi:Cas9 events (Supplementary Fig. 4a), suggesting that Cas9-mCherry expression from the TMM promoter can leak into mesophyll cells at a low frequency.
In contrast to pTMM:Cas9, we observed neither a loss of GFP nor a loss of chlorophyll autofluorescence in any of the 21 mCherry-expressing pFAMA:Cas9 T1 seedlings evaluated (Fig. 3b). We hypothesized that due to the later induction of Cas9 by pFAMA, residual PDS and GFP mRNA and/or protein might still be present in the targeted cells and would have to be depleted before a loss of signal could be observed. Therefore we investigated cotyledons at ten days after germination in five pTMM:Cas9 and eight pFAMA:Cas9 T1 events. Despite this extended incubation time, mCherry-positive guard cells still showed GFP and chlorophyll signal in the pFAMA:Cas9 lines (Supplementary Fig. 4b).
To determine if DNA mutations were induced in the pTMM:Cas9 and pFAMA:Cas9 lines, protoplasts from T2 cotyledons of four independent lines (two for each promoter) were sorted for mCherry expression. Sequencing results of the mCherry-positive and - negative protoplast populations showed an indel frequency of ~80% for the GFP and PDS target genes for the pTMM:Cas9 lines and 30-74% for the pFAMA:Cas9 lines (Table 4). The observation of mutations in the pFAMA:Cas9 lines was surprising given that no phenotypic effects were observed in any of the events. Despite the lack of phenotypes in the pFAMA:Cas9 lines, these results indicate mutations were generated in the mCherry-expressing cells for both CRISPR-TSKO constructs.
Knockout mutants in YODA (YDA, AT1G63700) show clustered stomata, severe developmental defects, frequent seedling growth arrest and, if yda mutants do manage to survive until flowering, sterility 26, 27. Given these features, targeting YDA specifically in the stomatal lineage with pTMM-CRISPR-TSKO presents a perfect test case. When targeted by ubiquitously-expressed Cas9 by a dual-gRNA construct, pPcUbi:Cas9-mCherry;YDA-1, YDA-2, 33 out of 35 mCherry-positive T1 seedlings contained clustered stomata on the cotyledonary abaxial epidermis to varying degrees of severity (Fig. 3c). Strikingly, all 40 mCherry positive T1 plants carrying pTMM:Cas9-mCherry;YDA-1, YDA- 2 copied this stomata phenotype.
Remarkably, in contrast to the pPcUbi:Cas9 lines, the pTMM:Cas9 lines developed similar to the wild type. From a separate batch of T1 plants transferred to soil, all 19 pTMM:Cas9 T1 plants were fertile and set seed while 8 out of 19 pPcUbi:Cas9 lines were completely sterile. PCR and DNA sequence analysis confirmed efficient mutagenesis at the YDA target sites in whole seedlings transformed with both constructs. As a pair of gRNAs are used to target YDA, an 813 bp deletion is expected from the excision of the intervening DNA sequence. Such deletion events are observed for both constructs (Supplementary Fig. 4c). Indel frequencies for pPcUbi:Cas9-mCherry;YDA-1, YDA-2 events were higher than pTMM:Cas9-mCherry;YDA-1, YDA-2 events as is expected for ubiquitous versus stomata-specific targeting.
These results illustrate that by utilizing the stomatal lineage-specific TMM promoter, we were able to uncouple the pleiotropic growth defects and sterility in systemic YDA knockouts27, but still retain the characteristic clustered-stomata phenotype and obtain T2 seeds.
Organ-specific gene knockout in lateral roots
Next to specific gene-knockout in particular tissues and cell lineages, we tested the potential of CRISPR-TSKO to generate entire mutant organs on otherwise wild-type plants. To this end we made use of the promoter sequence of GATA23, a gene that marks the initiation of lateral root organogenesis and is expressed in pericycle cells primed to become involved in lateral root formation in Arabidopsis28. GATA23 expression is transient and disappears prior to the emergence of the primordium from the primary root, except for some remaining expression at the base of the primordium28 (Fig. 4a). Strikingly, 20 out of 23 mCherry-positive T1 seedlings carrying pGATA23:Cas9-mCherry;GFP-1 showed a complete or partial loss of GFP fluorescence in lateral roots, while maintaining normal GFP expression in the primary root (Table 5, Fig. 4a, b). Lines with undetectable mCherry expression showed chimeric or normal expression patterns in lateral roots (Table 5). Sequence analysis of lateral roots from six independent events confirmed >94% of the alleles were mutated in these organs (Table 6). These results indicate that organ-specific gene knockout in lateral roots is feasible via the lateral root founder cell-expressed GATA23 promoter.
To test the pGATA23-CRISPR-TSKO construct on genes involved in lateral root formation, the auxin-response factors ARF7 (AT5G20730) and ARF19 (AT1G19220) were targeted. Lateral root organogenesis depends on the partially redundant action of ARF7 and ARF19 as arf7 arf19 double-knockout mutants are not able to develop lateral roots29. As both ARFs are broadly expressed in Arabidopsis seedlings, it is unknown whether this phenotype depends on ARF7 and ARF19 function strictly in pericycle cells30. To test this, we first recapitulated the systemic knockout of ARF7 and ARF19 in T1 plants with two ubiquitously-expressed Cas9 vectors containing two gRNAs targeting both ARFs, pPcUbi:Cas9-mCherry;ARF7-1, ARF19-1 and pPcUbi:Cas9-mCherry;ARF7-2, ARF19-2. Consistent with the published phenotype of arf7 arf19 double mutant30, a significant reduction in lateral root density was observed for pPcUbi:Cas9-mCherry;ARF7-2, ARF19-2 T1 plants compared to pGATA23:Cas9-mCherry;GFP-1 T1s with 18 out of 26 T1 plants completely lacking lateral roots (Fig. 4c). No reduction in lateral root density was observed for the pPcUbi:Cas9-mCherry;ARF7-1, ARF19-1 T1s. In agreement with these phenotyping results, sequencing of the gene target sites in whole roots from four independent events showed the indel frequency was >93% for most events with the pPcUbi:Cas9-mCherry;ARF7-2, ARF19-2 vector (Table 7). The pPcUbi:Cas9-mCherry;ARF7-1, ARF19-1 construct was particularly ineffective as the ARF19-1 target site had a low indel frequency of only 9-13% (Table 7), and explaining the lack of mutant phenotype in those T1s.
By targeting ARF7 and ARF19 in the xylem-pole pericycle via the pGATA23:Cas9-mCherry-ARF7/ARF19-2 construct, a significant reduction in lateral root density was also observed and 3 out of 13 T1 seedlings produced no lateral roots (Fig. 4c). These results show that CRISPR-TSKO can efficiently knock out two genes at once in lateral root founder cells, and that ARF7 and ARF19 action specifically within the GATA23-expressing xylem-pole pericycle is required for lateral root formation.
Discussion
Conditional knockout of essential genes enables their study in specific contexts
Targeted gene knockout experiments in plants typically have the objective of generating stably transmitted, inheritable DNA mutations. The generation of such knockout lines is a powerful tool for the functional analysis of many genes of interest. However, this approach is difficult to apply to genes that are essential for cell survival or reproduction, or those that have severe pleiotropic effects when mutated. In cases of lethality or sterility, such alleles need to be maintained in a heterozygous state31. Moreover, context-specificity of key regulators of developmental processes is often taken for granted without further experimental proof, while being aware of the non-context specific expression. In this report we describe the design and validation of CRISPR-TSKO, a gene knockout approach in plants that can be used to overcome these limitations.
We targeted six genes using four different tissue-specific promoters. Two of the target genes, PDS and YDA, are essential for plant growth and development. Mutations in PDS induced with the pTMM:Cas9-mCherry;GFP-1, PDS vector led to the expected defects in chlorophyll content and chloroplast formation25 specifically in the stomatal lineage. Importantly, the active photosynthetic mesophyll tissue was not markedly affected in non Cas9-expressing cells which allowed these plants to develop as vigorously as the wild type (Fig. 3a, Supplementary Fig. 4a). This stands in contrast to the ubiquitous CRISPR knockout lines that were primarily dwarfed albino plants not viable in soil.
In wild-type Arabidopsis, chlorophyll-containing chloroplasts are formed in epidermal pavement as well as stomatal guard cells, though they are much smaller than mesophyll chloroplasts32. The function of chloroplasts in guard cells has been subject to debate33. Very recently, the discovery of the gles1 (green less stomata 1) mutant suggests that functional chloroplasts in guard cells are important for stomatal responses to CO2 and light, resulting in stomatal opening34. CRISPR-TSKO plants with mutated PDS specifically in the stomatal lineage represent a powerful resource to test these and other hypothesized functions of chloroplasts in guard cells.
The mitogen-activated protein kinase YDA has several roles during plant development including embryogenesis, epidermal patterning and root development35, 36. Accordingly, yda mutants have severe pleiotropic phenotypes. Already directly after fertilization, yda mutants fail to establish the first asymmetric division of the zygote and ensuing embryo development is severely compromised27. Some yda embryos do continue to develop into seedlings, but these rarely survive on soil and the few yda plants that flower are severely dwarfed and completely sterile26, 27.
Using CRISPR-TSKO to target YDA only in the stomatal lineage, all transgenic events expressed a range of clustered stomata yda phenotypes while the other aspects of plant development were not notably compromised. Critically, all lines transferred to soil were fertile. These results demonstrate how CRISPR-TSKO can facilitate the analysis of gene function in specific developmental contexts, and it shows that YDA function in the stomatal lineage can be uncoupled from the pleiotropic defects caused by YDA mutations in other developmental contexts.
The root cap is a particularly interesting plant tissue as it encompasses all stages of plant cellular life – generation by stem cells, proliferation, differentiation, and finally programmed cell death and separation – in a compact spatial and temporal frame of a few hundred microns and a couple of days37. In this light, the pSMB-CRISPR-TSKO vector system could particularly useful to study genes essential for basic cellular functions in an easily accessible root tissue.
SMB is a key transcription factor required for root cap maturation and programmed cell death and its expression starts immediately after the formative stem cell division in the root cap daughter cells13, 14. Our results show that pSMB-CRISPR-TSKO allows gene knockout in new root cap cells from the moment of their inception. Diminishing GFP fluorescence is detectable in the two youngest root cap layers (Fig. 2a, b). This is likely due to remaining pools of mRNA and protein that have to be turned over before the GFP loss-of-function phenotype becomes apparent. The speed of these processes likely depends of the stability of the particular mRNA and protein, or even the cell type in question. The results presented here highlight that these dynamics have to be taken into account on a gene-by-gene and tissue-specific promoter basis when designing and analyzing tissue-specific gene knockout experiments.
Organ-specific knockout
In the case of the stomatal lineage and the root cap, specific cell types and tissues were targeted by CRISPR-TSKO, whereas in our lateral root experiments, we targeted the organ primordial founder cells to generate entirely mutated lateral roots. The generation of completely GFP-negative lateral roots at high rates impressively shows the efficiency of CRISPR-TSKO under pGATA23. Having a promoter at hand that is activated in the founder cells of a new organ thus represents an effective means to generate whole organs devoid of the function of a gene of interest. This experimental approach is an attractive alternative to grafting to study the transmission of mobile signals such as small RNAs38. Moreover, essential genes for primary root development such as MONOPTEROS hinder loss-of-function analysis during lateral root development39. The pGATA23-CRISPR-TSKO will enable us to uncouple the function of such genes involved in primary- and lateral root development.
Plants possess a modular body structure and a high regenerative potential. Thus, mutated lateral roots can be used to generate entire plants via shoot regeneration culture40. In that way, above-ground organs could be investigated for phenotypes generated by lateral-root CRISPR-TSKO. We observed a strong correlation between gene knockout and Cas9-mCherry expression, which can be used to facilitate selection of mutated lateral roots. Furthermore, targeting of GFP alongside a gene of interest in a ubiquitously-expressing NLS-GFP background revealed that knockout of GFP strongly correlated with mutagenesis of the endogenous genes SMB and PDS. Thus, both tagged Cas9 as well as knockout of reporter genes facilitate the selection of successful CRISPR knockout events in tissues and organs.
CRISPR-TSKO; A powerful and versatile tool
To facilitate the construction of CRISPR-TSKO reagents, we developed a modular vector-cloning scheme based on the GreenGate system17. The modularity of the cloning system allows for the rapid assembly of new promoter sequences with Cas9 or any nuclease of choice. Targeting specificity in spatial and temporal contexts can be readily achieved with the recently-developed inducible and tissue-specific plant expression systems that utilize GreenGate technology41. The cloning reagents used here are inexpensive to use and can be readily adopted by any laboratory. The system can be easily expanded to accommodate up to twelve gRNAs by the use of an AarI-SacB linker and later additions of six paired-gRNA entry modules (Supplementary Fig. 2 and Supplementary Methods). For ease of use, a workflow was developed to substitute the AarI restriction sites in the AarI linker for BsaI restriction sites flanking the ccdB and CmR selectable markers (Fig. 1 and Supplementary Fig. 2). This strategy avoids the need for separate AarI digestions of regularly-used destination vectors and provides a negative selection marker for the original destination vector in ccdB sensitive cells (e.g. DH5α). Alternatively, additional expression cassettes can be sequentially inserted into the AarI-SacB or BsaI-ccdB linkers. Hence, the system presented here can be easily used and modified for a variety of applications.
One general characteristic of CRISPR-TSKO is the continuous de-novo generation of mutations in cells that start to express Cas9. In the case of the root cap, every formative stem cell division creates a novel gene knockout event. In the stomatal lineage, every cell that starts expressing TMM or FAMA generate independent mutations, and in the case of the lateral root, every GATA23-expressing founder cell will contribute individual mutations to the lateral root primordium. All cells the lateral root primordium go back to a small number of founder cells, each of which is responsible for the formation of a portion of the lateral root primordium42.
Most gene mutations lead to small indels causing frame shifts and early stop codons, but a few should also lead to in-frame mutations, thereby disturbing the eventual amino-acid sequence of the protein only minimally. The GFP-1 gRNA used here targets the essential Gly67 residue, so that even in-frame mutations are likely to result in a loss of fluorescence43. Hence, gRNAs targeting genes of interest in particularly sensitive sites, as crucial interaction domains or active sites can further increase the effectiveness of CRISPR-TSKO.
Based on the results of this study, we suggest several tissue-specific promoters be tested before more labor-intensive phenotyping is performed for genes of interest. This test can simply be performed by analyzing the ubiquitously expressing NLS-GFP plant line transformed with the gRNA GFP-1 under the control of a new promoter of choice. This approach clearly demonstrated the context-specific knockout with the SMB, TMM, and GATA23 promoters. However, results might not always be straightforward as clearly illustrated with the FAMA promoter. The use of this promoter did not result in an observable phenotype while DNA mutations were detected. Therefore, ideally more than one promoter should be tested when targeting novel cell types with CRISPR-TSKO.
In conclusion, cell type-, tissue-, or organ-specific gene knockout by targeted expression of Cas9 is a powerful means of functional analysis in specific spatial and temporal contexts of plant development. Especially if genes are widely expressed or have fundamental roles for cell survival or plant reproduction, CRISPR-TSKO allows for the generation of stable transgenic lines that target mutagenesis specifically to the cell, tissue or organ of interest. Unlike in system-wide, heritable mutant approaches, no defined mutant alleles are generated. Nevertheless, due to the high efficiency of CRISPR mutations causing loss of gene function, heritable expression of CRISPR-TSKO provides a powerful tool to detect and analyze phenotypes in specific cellular or tissular contexts. As with other CRISPR applications, CRISPR-TSKO is forward-compatible to incorporate upcoming future variations of CRISPR gene modification.
Together with the virtually unlimited possibilities to combine different promoters, reporters, or tags in CRISPR-TSKO, this technology presents a powerful addition to the molecular genetics tool-box in plant biology research.
Distribution of modules, plasmids, and protocols
All cloning modules and plasmids reported here are available via the VIB-UGent Center for Plant Systems Biology Gateway Vector website (Supplementary Table 1; https://gateway.psb.ugent.be/search). See Supplementary methods for detailed cloning protocols.
Seeds for the pHTR5:NLS-GUS-GFP line are available upon request.
Methods
Cloning
All cloning reactions were transformed via heat-shock transformation into ccdB-sensitive DH5α E. coli or One Shot™ ccdB Survival™ 2 T1R Competent Cells (ThermoFisher Scientific). Depending on the selectable marker, the cells were plated on LB medium containing 100 μg/mL carbenicillin, 100 μg/mL spectinomycin, 25 μg/mL kanamycin or 10 μg/mL gentamycin. Colonies were verified via colony-touch PCR, restriction digest and/or Sanger sequencing by Eurofins Scientific using the Mix2Seq or TubeSeq services. All cloning PCR reactions were performed with either Q5® High-Fidelity DNA Polymerase (New England Biolabs) or iProof™ High-Fidelity DNA Polymerase (BioRad Laboratories). Gibson assembly reactions were performed using 2x NEBuilder Hifi DNA Assembly Mix (New England Biolabs). Column and gel purifications were performed with Zymo-SpinTM II columns (Zymo Research).
Golden Gate entry modules
Golden Gate entry modules were made by PCR amplification of gene fragments and inserting the purified PCR product into BsaI-digested GreenGate entry vector17 via restriction-ligation using BsaI (New England Biolabs) or Gibson assembly. See Supplementary Table 3 for all primers used. The entry modules pGG-F-AarI-AarI-G and pGG-F-AtU6-26-AarI-AarI-G were made by annealing oligos 361 and 362 and ligating into the BsaI-digested entry vector pGGF000. The Golden Gate entry linker modules were constructed as previously described44. The AarI-SacB-AarI linker, based on the SacB sequence from pMA7-SacB45, was synthesized on the BioXP3200 DNA synthesis platform (SGI-DNA) and inserted into a BsaI-digested pGGF000 plasmid via Gibson assembly. The correct clone was verified by Sanger sequencing. The coding sequence for mTagBFP2, based on a previously reported mTagBFP246, was PCR amplified with primers RB42 and RB43 from a synthesized fragment (Gen9) and inserted into a BsaI-digested pGGC000 plasmid via restriction ligation.
The unarmed gRNA modules were cloned by amplifying the AtU6-26 promoter and gRNA scaffold from previously-described Golden Gate entry vectors44. The amplification was done using the forward primer 120 and the reverse primers 283, 284, 230, 231, 232 and 233 for the B to G overhangs, respectively. This removed an unwanted attB2 site. PCR products were digested with BsaI and ligated into the respective empty entry vectors. The unarmed gRNA modules were further adapted by adding the ccdB negative selectable marker. The ccdB gene was PCR amplified from pEN-L1-AG-L2 with oligos 391 and 392 and inserted into BbsI-digested unarmed gRNA modules via Gibson assembly. All clones were verified by Sanger sequencing.
GATEWAYTM destination vectors
pGG-A-pOLE1-B, pGG-B-OLE1-C, pGG-D-linker-E17, pGG-E-NOST-F and pGG-F-LinkerII-G were assembled with either pGG-C-mRuby3-D or pGG-C-GFP-D17 into pEN-L1-A-G-L2. The ligation reactions were used as templates for PCR with the primers 195 and 196. The PCR products were cloned via Gibson assembly into pGGK7m24GW linearized with KpnI and XbaI. Clones were verified by Sanger sequencing. The resulting vectors containing the red and green fluorescent FAST markers were named pFASTRK24GW and pFASTGK24GW, respectively.
Proof-of-concept vectors
The Golden Gate entry modules pGG-A-pSMB-B, pGG-B-Linker-C, pGG-C-Cas9PTA*- D, pGG-D-Linker-E, pGG-E-G7T-F and pGG-F-linkerII-G were assembled in pEN-L4-A-G-R144, resulting in the vector pEN-L4-pSMB-Cas9PTA-G7T-R1. The Golden Gate entry module pGG-A-AtU6-26-BbsI-BbsI-B and pGG-B-linkerII-G were assembled in pEN-L1-A-G-L244, resulting in the vector pEN-L1-AtU6-26-BbsI-BbsI-L2. The BbsI restriction sites were swapped with a fragment containing the ccdB and CmR selectable markers flanked with BsaI sites. This fragment was PCR amplified from the plasmid pEN-L4-A-G-R1, using primers 1436 and 1437. The fragment was BsaI-digested and ligated with T4 DNA Ligase in the BbsI-digested vector pEN-L1-AtU6-26-BbsI-BbsI-L2, resulting in the vector pEN-L1-AtU6-26-BsaI-BsaI-L2. pEN-L4-pSMB-Cas9PTA-G7T-R1 and pEN-L1-AtU6-26-BsaI-BsaI-L2 were recombined in pGGB7m24GW47 via a MultiSite Gateway reaction according to manufacturer’s recommendations. This vector was called pB- pSMB-Cas9-G7T-AtU6-BsaI-BsaI-gRNA scaffold.
Oligos 138 and 139 (GFP-1 target) and oligos 134 and 135 (GFP-2 target) were annealed by adding 1 μL of each 100 μM oligonucleotide in 48 μL of MQ and incubating with a slow cooling program on the thermal cycler (5 minutes at 95°C; 95-85°C, - 2°C/second; 85-25°C, −0.1°C/second). These annealed oligonucleotides were cloned via a Golden Gate reaction into pB-pSMB-Cas9-G7T-AtU6-BsaI-BsaI-gRNA scaffold. The Golden Gate reaction conditions are described in Supplementary Methods. The resulting vectors were named pB-pSMB-Cas9-G7T-AtU6-GFP-1 and GFP-2.
Golden Gate destination
The Golden Gate destination vectors were cloned by amplifying the CmR and ccdB selection cassettes, flanked by the Golden Gate cloning sites A-G, from pEN-L1-A-G-L2 using primers 298 & 313. PCR products were column purified and cloned via Gibson assembly in the HindIII and PstI linearized Gateway destination vectors pGGP7m24GW, pGGK7m24GW, pGGB7m24GW, pGGPH7m24GW47, pFASTRK24GW and pFASTGK24GW. The resulting vectors were named respectively pGGP A-G, pGGK A-G, pGGB A-G, pGGH A-G, pFASTRK A-G, pFASTGK A-G. All clones were verified by Sanger sequencing and diagnostic digest with NotI.
To generate the pFASTR-A-G destination vector, the Golden Gate entry modules pGG-A-OLE1P-B, pGG-B-OLE1-CDS-C, pGG-C-mRuby3-D, pGG-D-Linker-E, pGG-E-NOST-F, pGG-F-linkerII-G were assembled in pGGP-A-G. Subsequently the CmR and ccdB selection cassettes, flanked by the Golden Gate cloning sites, were PCR amplified using primers 298 and 430 from pEN-L1-AG-L2 and inserted via Gibson assembly.
One-step CRISPR-TSKO cloning vectors
For cloning two gRNAs in a destination vector, we followed as similar approach as previously described48, with some modifications. A plasmid was generated to serve as a PCR template for 2-gRNA vectors. The Golden Gate entry modules pGG-A-AtU6-26-BbsI-BbsI-B, pGG-B-Linker-C17, pGG-B-AtU6PTA-C44 and pGG-D-linkerII-G were assembled in pEN-L1-A-G-L2 to generate pEN-2xAtU6 template. The clone was verified by Sanger sequencing.
The extended protocol for making one-step, CRISPR-TSKO cloning vectors can be found in Supplementary Methods. In summary, six different entry modules are combined via Golden Gate assembly in a destination vector. The A-B entry module contains the tissue-specific promotor, the C-D module contains the Cas9 endonuclease and can be combined with an N-terminal tag (B-C) or C-terminal tag (D-E), and the E-F entry module contains the terminator. For making a vector that is compatible with cloning one or two gRNAs, the F-G module pGG-F-AtU6-26-AarI-AarI-G is used (Fig. 1). Upon digestion with AarI, this vector can be loaded directly with one or two gRNAs. Alternatively, the AarI sites can be replaced for a fragment containing BsaI sites flanking the ccdB and CmR selectable markers. Two gRNAs can be cloned via a PCR reaction on the pEN-2xAtU6 template using primers that contain gRNA spacer sequences via a Golden Gate reaction. More details can be found in Supplementary Methods.
For making a vector that is compatible with multiple gRNA’s (up to 12) the Golden Gate cloning is slightly modified. The initial Golden Gate reaction is performed with an F-G linker containing AarI restriction sites (Supplementary Fig. 1). Upon AarI digestion, this vector can be directly loaded with six Golden Gate entry modules containing one or two AtU6-26 promotors and gRNAs. Alternatively, a similar strategy to replace the AarI sites by the ccdB and CmR selectable markers flanked with BsaI sites can be followed.
Plant lines used in this study
The smb-3 line is derived from the SALK collection (SALK_143526C). The NLS-GFP line (pHTR5:NLS-GUS-GFP) was previously reported15. The arf7 arf19 double mutant30 is derived from the ABRC collection (CS24629).
Plant transformation
Plant vectors were transformed in Agrobacterium tumefaciens C58C1 by electroporation. Transformation in pHTR5:NLS-GUS-GFP via the floral dip method49. For constructs containing the bar selectable marker or FASTR screenable marker, the T1 seed selection was done on 1/2 MS medium + 10 mgL−1 Glufosinate-ammonium (Sigma-Aldrich) or by selection under a Leica M165FC fluorescence stereomicroscope. Resistant seedlings or pFASTR-positive seeds were transferred to Jiffy-7 pellets® and grown in a greenhouse at 21°C under a 16-hour day regime.
DNA extraction and molecular analysis
Seedling, leaves or roots were frozen and disrupted to powder using a TissueLyser (Retsch MM300). DNA was extracted using a modified version of the protocol from Edwards et al. (1991 Nucleic Acids Research 19:1349). The modifications consisted of an adapted extraction buffer (100 mM Tris HCl pH 8.0, 500 mM NaCl, 50 mM EDTA, 0.7% SDS) and a 70% ethanol washing step before dissolving the pellet. A region around the CRISPR/Cas9 target site was PCR amplified using the ALLin™ Red Taq Mastermix, 2X (highQu GmbH) with the following program on the thermocycler: 95°C for 3 minutes, followed by 33 cycles (30 seconds at 95°C, 30 seconds at the annealing temperature, 1 minute/kb at 72°C), 72°C for 5 minutes. The PCR products were analyzed via agarose gel electrophoresis and the clean-up was either done by bead purification with HighPrep™ PCR (MAGBIO) or column purification with the DNA Clean & Concentrator™ kit (Zymo Research). The purified samples were send for Sanger sequencing (Eurofins Scientific) and analyzed using TIDE22.
All gRNA target sites can be found in Supplementary Table 2.
Confocal microscopy for original proof of concept
T1 seedlings were imaged on a Zeiss LSM710 confocal microscope. GFP was excited at 488nm and acquired between 500-550nm. T2 seedlings were imaged on a Leica SP8X confocal microscope. GFP was excited at 488nm and acquired between 500-530nm.
Confocal microscopy
Seedlings were imaged on a Leica SP8X confocal microscope. For root imaging, GFP was excited at 488nm and acquired between 500-530nm. mCherry was excited at 594nm and acquired between 600-650nm. Samples were either stained with 20ug/mL of DAPI or with 10μg/mL of propidium iodide in 0.43gL−1 Murashige and Skoog salts with 94μM MES.H2O medium. DAPI was excited at 405nm and acquired between 410-480nm in sequential mode.
For stomata imaging, cotyledons were vacuum infiltrated with with 20ug/mL of DAPI in 0.43gL−1 Murashige and Skoog salts with 94μM MES.H2O medium. Samples were imaged in sequential mode. DAPI was excited at 405nm and acquired between 410-450nm. GFP was excited at 488nm and acquired between 500-530nm. mCherry was excited at 594nm and acquired between 600-650nm. Chlorophyl autofluorescence was excited at 488nm and acquired between 680-730nm. Images were analyzed using Fiji50.
To image lateral root primordia, T1 seedlings were cleared using the ClearSee protocol51, 52 in combination with cell wall staining using Calcofluor White M2R (Sigma) on a Leica SP8X confocal microscope. Calcofluor White was excited at 405nm and acquired between 430-470nm. GFP was excited at 488nm and acquired between 500-525nm. mCherry was excited at 594nm and acquired between 600-630nm.
Epifluorescence microscopy
Cotyledons of FASTR positive seedlings were mounted on distilled water and observed on a Zeiss Observer.Z1 using a Plan-Apochromat 20x/0.8 DICII objective. GFP fluorescence was observed with a BP 470/40 filter for excitation, a FT 495 beam splitter, and a BP 525/50 emission filter. mCherry was observed with a BP 545/25 filter for excitation, a FT 570 beam splitter, and a BP605/70 emission filter.
Segmentation and analysis of root cap nuclei
Root tip image stacks were segmented and nuclei intensity measurements performed using the interactive learning and segmentation toolkit ilastik 1.3.053. Intensity of GFP and mCherry were measured for segmented nuclei with a probability equal or higher than 0.95 of belonging to root cap cells. Based on mCherry measurements in the NLS-GFP line (Supplementary Fig. 5), a threshold of 25 was established as a minimum signal for mCherry.
Protoplast preparation and cell sorting
Protoplasting was performed as previously described54. Briefly, for pSMB-CRISPR-TSKO lines, root tips of 5 day old seedlings grown under continuous light on 0.43gL−1 Murashige and Skoog salts with 94μM MES.H2O medium were incubated in protoplasting solution consisting of 1.25% cellulase (Yakult, Japan), 0.3% Macerozyme (Yakult, Japan), 0.4M mannitol, 20mM MES, 20mM KCl, 0.1% BSA and 10mM CaCl2 at pH 5.7 for 3 hours. Samples were then filtered through a 40μm filter and the flow through centrifuged at 150xg for 10min. Supernatant was discarded and protoplasts were recovered in ice-cold resuspension buffer. Resuspension buffer was of same constitution as protoplasting buffer with the omission of both cellulase and macerozyme. For lines targeting stomatal lineages, cotyledons of 5-day old seedlings were processed as above but with a 12 hours incubation time to get proper release of guard cells.
Root tip protoplasts were sorted into 1.5ml Eppendorf tubes containing 500μl of resuspension buffer using a BD FACSAriaII, equipped with 3 lasers (405nm, 488nm and 633nm). To account for the double presence of GFP and mCherry in some samples, the cotyledon protoplasts were sorted into 1.5ml Eppendorf tubes containing 500μl of resuspension buffer using a BD FACSMelody, equipped with 3 lasers (405nm, 488nm and 561nm). The 561nm laser in the BD FACSMelody made a better separation possible due to a better excitation of the mCherry. All FACS sorting reports can be found in Supplementary Fig. 6.
Quantification of lateral root density
FASTR positive seeds were sown on half-strength Murashige and Skoog (MS) medium (Duchefa Biochemie B.V.), supplemented with 1% (w/v) sucrose and 0,8% (w/v) agar, at pH 5,7 and stratified for 2 days in the dark at 4°C. Seedlings were grown vertically for 12 days in continuous light (100 μmol m−2s−1) at 22°C. Presence/absence of Cas9-mCherry signal was scored using a Leica M165FC fluorescence stereomicroscope. The number of emerged lateral roots was determined for every seedling using a stereo microscope and root lengths were measured via Fiji50 using digital images obtained by scanning the petri dishes.
Stomata analysis of cotyledons in YDA targeting lines
The cotyledon epidermis of seedlings 10 days post germination was visualized by clearing cotyledons in 100% ethanol and incubation at 60 degrees in 90% ethanol / 10% acetic acid for 30 minutes and ethanol / 1,25M sodium hydroxide (1:1 v/v) for 2 hours. Next, cotyledons were incubated overnight at room temperature in lactic acid saturated with chloral hydrate, washed in 100% lactic acid and mounted for differential interference contrast microscopy (Olympus BX51). Images (430 μm x 566 μm) from the midline to the margin on abaxial surfaces were generated. Thirty five to 40 cotyledons of individual seedlings were evaluated per genotype.
Statistical analysis
For segmentation and analysis of root cap nuclei, Spearman’s Correlation Coefficient between median root cap signal of GFP and mCherry was calculated using SAS (Version 9.4M5 of the SAS System for windows 7 64bit. Copyright © 2013 SAS Institute Inc. Cary, NC, USA (www.sas.com)). The SAS code is available upon request. For lateral root density analysis, statistical significance was assessed using the Graph-Pad Prism software, version 7.03 (GraphPad Software, Inc.). The comparison of ELR densities for different constructs was done using the Kruskal-Wallis test (non-parametric) with the Dunn’s multiple comparison test.
Author contributions
W.D., R.A.B., M.K.N., and T.B.J. conceived and devised the study. M.K. adapted the GreenGate vectors and provided additional cloning support. W.D. constructed the vectors and performed the genotyping analysis. R.A.B., M.L.P., N.V. and J.J. performed the experiments, imaging and analysis. W.D., and R.A.B. performed statistical analysis. G.V.I. and R.A.B. performed the FACS experiments and analysis. W.D., T.B.J., R.A.B., M.K.N. and T.B. wrote the manuscript with contributions from all other authors.
Additional information
Supplementary information accompanies this paper at
Supplementary Table 1. Overview of plasmids
Supplementary Table 2. gRNA target sequences, with the PAM in bold
Supplementary Table 3. Primers used
Conflicts of interest
The authors declare no conflicts of interest.
Acknowledgements
We thank Dominique Bergmann, Camila Lopez-Anido, Michael Raissig, and the members of the Programmed Cell Death and Plant Genome Editing teams at VIB for constructive discussions. We also thank Veronique Storme for assistance with statistical analysis. The GreenGate plasmid kit used for generation of plant transformation constructs was a gift from Jan Lohmann (Addgene kit # 1000000036). The coding sequence for mRuby3 was derived from pNCS-mRuby3, which was a gift from Michael Lin (Addgene plasmid # 74234). We thank Carina Braeckman (VIB-UGent Center for Plant Systems Biology) for the help with the A. thaliana floral dip transformation. We thank Nico Smet, Miguel Riobello y Barea, Thomas Farla, and Sandra Lefftz for greenhouse support. We thank Debbie Rombaut and Freya De Winter for the help with the preparation of root protoplasts. We thank Eugenia Russinova for a critical review of the manuscript. We gratefully acknowledge funding by the ERC StG PROCELLDEATH (Project 639234) to M.K.N., and by the FWO 1174119N to M.L.P.