Conditional and tissue-specific approaches to dissect essential mechanisms in plant development

Abstract

Reverse genetics approaches are routinely used to investigate gene function. However, mutations, especially in critical genes, can lead to pleiotropic effects as severe as lethality, thus limiting functional studies in specific contexts. Approaches that allow for modifications of genes or gene products in a specific spatial or temporal setting can overcome these limitations. The advent of CRISPR technologies has not only revolutionized targeted genome modification in plants, but also enabled new possibilities for inducible and tissue-specific manipulation of gene functions at the DNA and RNA levels. Additionally, novel approaches for direct manipulation of target proteins have been introduced in plant systems. Here, we review the current development in tissue-specific and conditional manipulation approaches at the DNA, RNA and protein levels.

Introduction

Modulating genes by loss- and gain-of-function approaches are a main pillar of plant functional genomics. However, knockout or overexpression of many critical genes leads to pleiotropic effects. In the most extreme cases, plants can be too severely affected to allow reliable conclusions on primary gene function or even die at early stages of development. Embryonic or seedling lethality has been predicted for up to 10 percent of Arabidopsis genes [1]. Researchers have attempted to circumvent these issues by conditionally modulating genes or proteins at specific times in development by inducible (conditional) approaches, and/or by restricting gene modification or transgene expression to particular cell types, tissues, or organs [2,3]. A classic example of conditional gene modification includes temperature sensitive (Ts) alleles which allow temperature-dependent manipulation of gene function. Though less common than in yeast systems, Ts alleles in Arabidopsis were isolated in early mutagenesis screens [4,5], and have more recently been produced in a targeted fashion [6,7]. Other classic examples of conditional modification of gene function include the expression of RNA interference (RNAi) constructs [8], artificial microRNAs [9] under tissue-specific or inducible promoters, or Cre-Lox recombination systems to generate loss-of-function sectors of complementing transgenes [10]. In this review, we will give an overview of the recent developments in conditional and/or tissue-specific techniques to modify gene or protein function. We will divide the approaches into the three levels at which the modification takes place –DNA, RNA, and protein, highlight advantages and disadvantages of these new systems, and reflect on current limitations and possible future strategies to address them.

Main body

Modifications at the DNA level

Recent developments in DNA modification by CRISPR technology offer the possibility to manipulate a gene – or even several genes at once - tissue-specifically and conditionally. In contrast to traditional knockdown techniques such as RNAi or amiRNA, which act by reducing transcript levels [11], CRISPR-based DNA modifications can generate full knockouts (KO) of the target gene(s). The most commonly used CRISPR system in plants is based on the CRISPR-associated protein 9 (Cas9) DNA endonuclease. Guide RNAs (gRNAs) direct Cas9 to target sequences to create sequence-specific double-strand DNA breaks [12], which are often non-perfectly repaired generating small insertions and deletions (indels) that can affect open reading frames resulting in KO alleles. Cas9-mediated targeted mutagenesis has become the standard for creating single or higher-order KO mutations in plants. While KO of essential genes can have severe effects up to lethality, tissue-specific and/or conditional CRIPSR-Cas9 tools overcome this limitation by creating ad-hoc somatic mutations in a spatially or temporally controllable manner.

In Arabidopsis (Arabidopsis thaliana), the CRISPR-based tissue-specific KO system CRISPR-TSKO is able to generate mutations causing well-defined, localized phenotypes with high efficiencies in specific cell types, tissues, or organs [13**]. Phenotypes were already evident in the first transgenic generation and enabled the growth, development and reproduction of lines with strictly somatic mutations in essential genes such as PHYTOENE DESATURASE 3 (PDS3), YODA (YDA), or cyclin-dependent kinases (CDKs). Importantly, the production of somatic mutations was heritable together with the Cas9 transgene, enabling the scale-up of transgenic lines for replicated experiments. CRISPR-based tissue-specific mutagenesis has been used for a fiber-specific lignin decrease in Arabidopsis to avoid the KO of an essential gene [14] and fruit-specific gene KO in tomato avoiding strong pleiotropic effects known from constitutive RNAi experiments [15].

The initial CRISPR-TSKO reports only employed one or two gRNAs to target up to three genes at a time. However, it was unclear how efficient the system would be using greater numbers of gRNAs to achieve higher-order levels of mutagenesis. A follow-up study demonstrated that mutagenesis of six genes by six different gRNAs was as efficient as single-gRNA CRISPR-TSKO. Moreover, mutagenesis efficiencies across different loci were linked; if one gene was efficiently mutated, the remaining five had a strong tendency to be mutated as well [16*]. These results are promising regarding tissue-specific higher-order mutagenesis and also open the possibility to exploit an easy-to-detect proxy phenotype to select transgenic plants with high CRISR-TSKO efficiency. Whether or not there is an upper limit on the number of simultaneous mutations remains to be determined.

Conditional expression of Cas9 controlled by inducible expression systems has also been used to induce mutagenesis in particular temporal or spatial contexts (Figure 1). An estradiol-inducible gene KO system was reported in Arabidopsis to deplete transgene-reporter proteins ubiquitously or tissue-specifically upon Cas9 induction [17*], though its efficiency for multiplex mutagenesis remains to be demonstrated. A recent preprint reports an example of how the system can be used in a cell-type-specific and conditional manner for gene function discovery. Phloem sieve element lineage-specific KO of PLETHORA2 (PLT2) led to early phloem differentiation in Arabidopsis (https://doi.org/10.1101/2021.01.18.427084).

Figure 1. Schematic representation of current and future applications of temporally and spatially controlled CRISPR/Cas tools in plants.

(A) Conditional and tissue-specific CRISPR-Cas activity in plants has been obtained via restricted temporal and/or spatial expression of a Cas nuclease. Anti-CRISPR proteins inactivate Cas nuclease activity at the post-translational level. Conditionally destabilized Cas proteins allow for activation and deactivation at the post-translational level. (B) Conventional mutagenesis using WT CRISPR/Cas9 enables gene knockout via DNA double-strand breaks adjacent to the PAM generated by the two Cas9 nuclease domains. Subsequent non-perfect repair leads to indel formation. CRISPR/dead Cas9 (dCas9)-based transcriptional regulation is obtained via fusion of effector domains and allow gene activation or repression. CRISPR/Cas9 nickase (nCas9)-based base editors are composed of a cytidine or adenine deaminase fused to nCas9 and enable the base pair conversion of C·G to T·A or A·T to G·C, respectively. Base editing has not been used tissue-specifically or conditionally in plants yet. CRISPR/Cas13a targets and cleaves single-stranded RNA and allows for knockdown of transcripts which has not been used tissue-specifically or conditionally in plants yet. Applications established tissue-specifically or conditionally in plants in black font; applications not established tissue-specifically or conditionally in plants yet in grey font. GOI: gene of interest; PAM: protospacer adjacent motif.

A paramount requisite for tissue-specific and conditional approaches are promoters that are inactive outside the target tissue or in the absence of induction. For example, the stomatal-lineage promoter pTMM caused non-intended KO in mesophyll cells in about 10% of events [13**]. Likewise, a heat-shock inducible approach in rice showed a rate of mutagenesis of around 16% in the absence of heat shock [18]. While these rates are relatively low and do not present a significant burden, newly established promoters or inducible expression systems should be scrutinized for efficiency and specificity. Highly efficient gRNAs targeting a ubiquitously expressed reporter gene, or other genes producing easily detectable phenotypes, are essential tools to assess these parameters [13**,16*].

Another possible limitation of tissue-specific KO may be the appearance of phenotypes depending on RNA and protein turnover rates. The half-lives of mRNAs can range from minutes for unstable messages, to days for stable transcripts, with the average estimated to be several hours in Arabidopsis [19] and the average protein half-life in Arabidopsis leaves is about 3.5 days [20]. Gene products with slow turnover rates might fail to be depleted by tissue-specific or conditional mutagenesis if the onset of Cas9 activity occurs too close to the time of phenotypic analysis. For example, the late stomatal lineage-specific pFAMA promoter resulted in clear DNA mutations in the absence of phenotypes possibly caused by lingering wild-type mRNA or protein [13**].

In contrast to conventional inheritable mutations that can be easily characterized by DNA sequencing and used to create true-breeding lines, it is challenging to accurately assess DNA mutagenesis in tissue-specific approaches – particularly if no mutant phenotype is detected or the mutation results in growth arrest or even elimination of the target cell type or tissue. Overall mutagenesis efficiencies can be determined by fluorescence-activated cell- or nuclear sorting to isolate the DNA of the target cell populations, followed by quantitative sequence analyses to estimate mutagenesis efficiencies [13**,16*]. However, the measured indel rates are only averages and the genotypes of individual cells are unknown which is problematic when multiple KOs are required to observe a phenotype. The use of fluorescently-tagged target genes can help confirm the elimination of a gene product [17*], however such lines are not available for all genes, the number of fluophores is limited and it is not compatible for screens of uncharacterized genes. Therefore, it is advisable when multiplexing to first establish fixed mutations in as many genes as possible without affecting plant development and only in a second step target the remaining gene(s) by tissue-specific approaches.

Modifications at the RNA level

One inherent feature of DNA-based approaches is that once the mutation is made, that cell and its descendants have fixed genotypes. In contrast, targeting at the RNA level allows for a greater level of control as the process is reversible and tuneable. Transgene expression and RNAi-based approaches have been commonly used for inducible and/or tissue specific regulation of transcript abundance in plants [21]. CRISPR tools controlling RNA abundance are typically based on nuclease-dead Cas9 (dCas9). While dCas9 cannot cleave DNA, it is still capable of binding DNA and can be used as a delivery vehicle for a variety of approaches to control gene transcription [22]. Fusing or recruiting transcriptional activators or repressors to dCas9 and/or the gRNA scaffold can increase transcript levels via CRISPR-activation (CRISPRa) or repress them via CRISPR-interference (CRISPRi) [23–25] (Figure 1B). These approaches can generate knock-down or overexpression phenotypes, and by multiplexing, several loci can be targeted simultaneously.

The dCas9-TV system is a potent transcriptional activator in plant cells, which is based on the distinct combination of several transcriptional activation domains fused to dCas9 [23]. Constitutive co-expression of dCas9-TV with pre-screened gRNAs individually targeting the promoters of six Arabidopsis genes led to overexpression levels ranging from 1.6- to 92-fold in Arabidopsis protoplast-based promoter-luciferase assays. The system is ideally suited for genes with modest basal expression levels as those tended to be better induced than those with high expression and allowed for efficient multiplexing [23]. Recently, CRISPR–Act3.0, a novel dCas9 based CRISPRa system, was developed by recruiting multiple transcription activator-like effector (TAL) activation domains (TAD) via an MS2-MCP interaction on the gRNA scaffold. This novel system was reported to achieve four to six times higher transactivation levels when compared to dCas9-TV acting on two genes (OsGW7 and OsER1) targeted by the same gRNA in rice protoplasts [26**]. Furthermore, multiplexed gene activation by CRISPR–Act3.0 targeting seven genes simultaneously resulted in gene activation ranging from less than 5-fold up to 20-fold for all target genes. However, there were substantial differences in efficiencies in rice protoplasts and transgenic seedlings. In protoplasts, four out of six target genes were activated by 5- to 140-fold while the same gRNAs in transgenic seedlings only activated one gene ~20-fold and the other five genes were activated 2- to 8-fold. In Arabidopsis, CRISPR–Act3.0 was also used for the generation of stable Arabidopsis lines simultaneously targeting two genes, FLOWERING LOCUS T and TRICHOMELESS1 which led to activation levels of 130- to 240-fold and 3- to 8-fold, respectively. Transgenic plants showed the characteristic phenotypes which were transmitted to the T2 and T3 generation [26**]. As CRISPRa efficiency appears to be sensitive to the gRNA target position within the promoter, gene activation efficiency can vary highly among different gRNAs for the same target gene [23–26**]. Pre-screening of gRNAs in cell-based assays (e.g. protoplasts) is therefore recommended.

CRISPRi and CRISPRa systems can also be used in tissue-specific and/or inducible settings. For instance, optogenetic control of a CRISPRa system has been achieved via plant usable light-switch elements (PULSE) that combines a blue-light-regulated repressor with a red-light-inducible switch [27*]. In Arabidopsis protoplasts, PULSE controlled expression of dCas9-TV led to reporter gene inductions of up to 24.5- and 40-fold over those in blue-light illumination and darkness, respectively [27*]. One advantage of optogenetics systems is that they allow not only for fast activation, but also for a deactivation of the system in the absence of the stimulus. This could allow for transient gene activation or repression to dissect the temporal order of events during plant development. However, the time it takes for dCas9 to be deactivated once expressed in a plant cell remains to be determined.

At the protein level

Modulating gene function at the protein level can provide a fast and reversible conditional modulation independent of transcript and protein turnover rates. Conditional protein inactivation has so far been attempted in plants via targeted degradation or ectopic sequestration of transgenes in complemented null mutants. Proteins exposing a destabilizing N-terminal amino acid residue are recognized by E3 ubiquitin ligases and degraded by the 26S proteasome via the N-end rule pathway [28]. Essentially any protein of interest (POI) can be targeted for degradation by engineering a destabilizing N-terminal residue to generate a degron. Conditional N-terminal degrons inducibly expose the destabilizing residue in a spatially or temporally controlled manner [29]. Classical approaches like TIPI (tobacco etch virus protease-mediated induction of protein instability) expose an N-degron tag attached to a POI via proteolytic cleavage [30] (Figure 2B). However, this approach is comparatively slow because it requires expression of the inducer protease. Recently, low temperature-sensitive degrons were used as a faster alternative for inducible, tissue specific cell ablation via stabilization of a degron-modified cytotoxic bacterial RNase (barnase) in Arabidopsis trichomes. Restrictive temperatures (27-29 °C) allowed trichome formation, whereas permissive temperatures (13-16 °C) caused tissue specific trichome death [31*] (Figure 2A). Limitations of N-degron systems are the N-terminal 26 kDa degron tag, and that the POI has to be localized to the nucleus or cytosol, compartments with high activity of the N-end rule pathway [32]. A potential disadvantage of temperature-controlled systems in general are the associated physiological changes in the plant as was recently shown for endocytosis and microtubule dynamics [33]. A possible improvement of the N-degron system is to combine it with an engineered destabilizing domain of the 107AA human FKBP12 protein, thereby conferring instability to proteins fused to it while a small synthetic molecule ligand, shield1, protects against degradation. Such a system allows for a fast and tuneable control of protein amounts [34].

Figure 2. Schematic representation of targeted protein degradation and delocalization approaches in plants.

(A) temperature-dependent degron approach. The degron cassette consists of an N-terminal sequence encoding a single ubiquitin (Ub), the thermo-labile mouse dihydrofolate reductase (DHFR^ts) containing 16 Lys (K) and destabilizing (R) residues, followed by C-terminal located protein of interest (POI). At the permissive temperature (13 to 16 °C, green) the fusion protein is stable. To make the cassette liable for the proteasomal degradation pathway, it is subjected to co-translational deubiquitylation by deubiquitylating enzymes and Ub-specific processing proteases to reveal the N-degron. The restrictive temperature (27 to 29 °C, red) triggers DHFR^ts flexibility to expose internal K residues, which are recognized by Ub ligases, and targets the POI for the 26S proteasome degradation. (B) TIPI-degron. The N-degron in the fusion protein is covered by green fluorescent protein (GFP). Upon TEV protease cleavage, GFP is removed to expose the N-degron for polyubiquitination and consequent degradation. (C) deGradFP. This system utilizes the SCF complex that consists of endogenous components: the adapter elements for the F-box protein (Skp1), the cullin scaffold (Cul1) and the E2 ligase-recruiting RING protein (Rbx1). The specificity of the substrate is given by the FBP subunit target recognition (F-box) fused to the vhh-GFP4, an anti-GFP nanobody. The GFP-POI is recognized by the F-box/vhh-GFP4, and undergoes polyubiquitination mediated by the E2 ubiquitin protein ligase to finally be degraded in the proteasomal pathway. (D) GFP-Nb. This system relies on the specificity of an anti-GFP nanobody (GFP-Nb) fused to the import signal of the yeast mitochondrial outer membrane protein Tom70p. In the poi complemented mutant background GFP-Nb targets the POI-GFP and re-localizes it to the mitochondria, therefore sequestering the functional POI at an ectopic subcellular localization.

Nanobodies (Nb) allow for alternative approaches for POI degradation or delocalization. A Nb is a single antigen-binding domain of heavy chain-only antibodies (HCAbs) isolated from camelids [35]. Nanobodies are around a tenth the size of conventional antibodies and bind their epitopes with high specificity and strong affinity. They can be expressed within plant cells as monomeric and stable proteins and are therefore more convenient than two-component conventional antibodies [36,37]. An inducibly expressed anti-GFP Nb fused to the F-box domain of the E3 ligase complex was used to selectively degrade GFP-tagged POIs via the 26S proteasome [38] (Figure 2C). This deGradFP method was also used in transgenic tobacco plants to deplete the C-terminal centromeric histone H3 variant CENH3 of Arabidopsis, demonstrating the capacity of nucleus-specific protein degradation [39]. In Arabidopsis, anti-GFP Nbs were expressed under the control of the ethanol-inducible AlcR/AlcA system to achieve switchable degradation of GFP-tagged WUSCHEL, showing that WUSCHEL controls the auxin response in the stem cells of the shoot apical meristem [40*].

An advantage of deGradFP approaches is its compatibility with any functional POI-GFP fusion that allows easily visualization via fluorescence-based imaging. Of course, Nb-based systems can be designed to operate independent of tags if Nbs targeting endogenous plant genes are available. Similar to the TIPI degron-based approach, a disadvantage of deGradFP is that degradation induction is coupled to the de novo expression of the nanobody-F-box fusion, creating a lag time in temporal control.

Nbs can also be used as ‘nanotraps’, whereby POI inactivation is achieved not by degradation, but by sequestration into a location that precludes its function [29,41] (Figure 2D). Mitochondrial targeting of anti-GFP nanobodies has been shown to delocalize a GFP-tagged functional TML subunit of the endocytic TPLATE complex in Arabidopsis epidermal and cortical root cells. Delocalizing TML to the mitochondria also delocalized subunits of the endocytic AP-2 complex, indicating that this technique is capable of moving interacting multi-subunit complexes. The tissue-specific delocalization reduced endocytic flux while avoiding previously reported seedling lethality using amiRNA approaches [42*,43]. Possible future adaptations to this “anchor-away” system lie in the use of inducible promotors driving nanobody expression. A general drawback however is that the capacity to sequester the POI has to exceed the amount present in the cell and that feedback loops controlling available protein amounts might hamper its efficiency.

Conclusion and Future perspectives

There exist a variety of off-the-shelf tools allowing one to specifically and conditionally control gene expression at the DNA, RNA and protein levels. Each tool has inherent advantages and disadvantages that should be carefully considered before venturing into a long-term project. Inducing DNA modifications using conditional or tissue-specific CRISPR systems presents versatile approaches to generate mutant alleles in one or several genes simultaneously. A potential disadvantage of DNA mutagenesis of course is that the modification is irreversible, and the appearance of a phenotype is dependent on RNA and POI turnover. Depleting gene products by targeting the RNA level can be reversible, but it is often incomplete (CRISPRi) and the phenotype still is dependent on POI turnover. Targeting at the protein level allows for reversible and faster modulation of the ultimate gene product, while multiplexing and targeting of endogenous non-tagged proteins is not yet straightforward. To conclude, we would like to highlight some of the future capabilities that we think will enhance many of these tools and enable researchers to ask, and answer, new kinds of questions.

There are a variety of CRISPR technologies that have not yet been applied to tissue-specific or conditional approaches at the DNA and RNA levels but can be readily tested by swapping their promoter elements, including base editing, prime editing, or targeted DNA methylation/demethylation [44–47] (Figure 1). Given that sufficient levels of efficiency remain to be reached in some cases, these technologies will allow for more nuanced investigations of gene function under specific contexts without using the KO sledgehammer. CRISPR screens are also being actively developed in plants and, combined with conditional and tissue-specific systems, will allow for screening essential gene function while avoiding lethal-gene effects.

The Cas13a enzyme is a CRISPR system that can specifically cleave single-stranded RNA. Comparable levels of knockdown, and improved specificity when compared to RNAi, were reported in mammalian cells and Cas13a was capable of reducing three endogenous RNA transcripts by ~50–80% in rice (Oryza sativa) protoplasts [48]. In plants, Cas13a has been so far primarily used to combat RNA viruses [49–51]. Targeted RNA editing via a Cas13a was also recently reported in human cells and enabled both adenosine to inosine and cytosine to uracil edits on a variety of endogenous gene targets [52]. While tissue-specific or conditional Cas13a expression has not yet been reported in plants, Cas13a tissue-specifically expressed in Drosophila circumvented embryo-lethality by systemic gene knockdown [53]. Hence, Cas13a might be an interesting alternative for tissue-specific or inducible RNAi or Cas9-based CRISPRi approaches in plants (Figure 1).

There are also a variety of Cas9-controlling technologies that remain to be explored. CRE-lox controlled Cas9 expression was recently established in zebrafish [54]. While not yet applied to plants, it could offer the advantage that the CRE-controlled promoter can initiate the constitutive expression of a fluorescently labelled Cas9. This allows for lineage tracing of potentially mutated cells while avoiding the labour consuming floxing of target genes [54]. Alternatively, Cas activity can also be regulated at the post-translational level. Anti-CRISPR (Acr) proteins are naturally evolved CRISPR/Cas antagonists derived from mobile genetic elements such as plasmids and phages [55]. Anti-CRISPR activity of the bacterial AcrIIA4 protein was recently reported in Nicotiana benthamiana to block Cas9-induced indels and CRISPRa in Agroinfiltration experiments (https://doi.org/10.1101/2021.01.08.425920). It remains to be seen how efficient AcrIIA4 and other Acr proteins will function in other plant systems, but these tools might offer another mode for precise control of CRISPR systems in plants. Several novel switch systems are being developed that allow for temporal control over dCas9 at the protein level such as split-dCas9 and destabilizing dCas9 domains [56]. While conditionally destabilized dCas9 has recently been applied in Arabidopsis [57], other switch systems such as split-dCas9 have not been used in plants yet (Figure 1 A).

At the protein level, delocalization of a POI to inhibit its function would benefit largely from a system that does not require a translational step and therefore has a very fast response time. A recent proof-of-concept paper showed that rapamycin-dependent dimerization between the FKBP domain of HsFKBP12 (FKBP) and the FKBP12 rapamycin-binding domain of mTOR (FRB) can be efficiently used to target POIs to a subcellular location of interest within minutes in N. benthamiana. This tool, termed Knocksideways in plants, by analogy to the Knocksideways system developed for animal cells [58], has the potential to be further developed into a conditional in planta KO tool that could be used to study for example otherwise lethal mutants [59**].