Simulations predict stronger CRISPRi transcriptional repression in plants for identical than heterogeneous gRNA target sites

Abstract

Plant synthetic biologists have been working to adapt the CRISPRa and CRISPRi promoter regulation methods for applications such as improving crops or installing other valuable pathways. With other organisms, strong transcriptional control has typically required multiple gRNA target sites, which poses a critical engineering choice between heterogeneous sites, which allow each gRNA to target existing locations in a promoter, and identical sites, which typically require modification of the promoter. Here, we investigate the consequences of this choice for CRISPRi plant promoter regulation via simulation-based analysis, using model parameters based on single gRNA regulation and constitutive promoters in Nicotiana benthamiana and Arabidopsis thaliana. Using models of 2–6 gRNA target sites to compare heterogeneous versus identical sites for tunability, sensitivity to parameter values, and sensitivity to cell-to-cell variation, we find that identical gRNA target sites are predicted to yield far more effective transcriptional repression than heterogeneous sites.

Open in new tabDownload slide

1 Introduction

Genetically engineered plants hold immense promise for helping to address critical global challenges such as food security, environmental sustainability, and renewable resource production. Establishing engineering control over a transcriptional regulation has been an important tool for other organisms and is expected to be so in plants as well, which is inspiring many efforts in this area (e.g. [1–5]). Native plant promoters can be used to control heterologous gene expression, however, they may be limited in their strength and regulatory capacity, and can cause a metabolic burden to the cell [6]. These limitations can be addressed with synthetic promoters and transcription factors [7].

Two of the key properties that synthetic promoters must achieve for effective regulation are orthogonality and high dynamic range. Orthogonality refers to the promoter’s ability to function independent of the plant’s native regulatory elements. This isolation is beneficial to both the host plant and the engineered circuit [1]. By using orthogonal promoters, the circuit’s desired function is protected from perturbations caused by the host plant and the host plant itself is protected from undesired regulatory impacts from the circuit [8]. Dynamic range refers to the range of output expression levels (e.g. RNA or protein production) that is achieved when the promoter is in its binary “on” state (high rate of transcription) or “off” state (low rate of transcription). A promoter with a high dynamic range is capable of producing a significant difference in expression levels between the on and off states, both allowing greater modification of behavior and enabling more complex regulatory networks [9].

Synthetic transcription factors based on the CRISPR/Cas transcriptional regulatory system are both orthogonal in plants and have the potential for a high dynamic range. In these systems, the catalytically inactivated Cas9 protein (dCas9), either alone or fused to a transcriptional activation or repression domain, binds to a specific promoter region determined by the sequence of a customized guide RNA (gRNA) [10], thereby activating (CRISPRa) or inhibiting (CRISPRi) transcription. Both have previously been used in plants, but have often resulted in a low dynamic range in both plants [11–13] and other systems [14–16]. Some degree of range improvement can be achieved through careful engineering of the gRNA sequence and binding location [17]. The efficacy of regulation can also be greatly increased by using multiple gRNA target sites to control a single promoter. For example, previous studies have used 3–4 gRNAs for CRISPRa in mammalian cells [18], two gRNAs for CRISPRi in mammalian cells [19], and four gRNAs for CRISPRi in yeast [20]. Multiple gRNAs have also been shown to have synergistic effects for CRISPRa in plants [13], but have not yet achieved the dynamic range seen in other organisms.

When designing synthetic promoters, a key question when using gRNA to target multiple sites on a promoter is whether to have the targets be heterogeneous (i.e. each gRNA has a unique sequence) or identical (i.e. the same gRNA is used for each target site). Heterogeneous gRNA target sites are often easier to use, since there is no need to find or engineer identical nearby sequences, but each site requires a separate dCas9 binding event to contribute to transcriptional repression (Fig. 1a). On the other hand, using multiple identical binding sites reduces the competition for Cas9 between gRNA species and can allow for a single binding event to result in multiple target sites being occupied due to the dCas9–gRNA complex laterally diffusing along the DNA strand (Fig. 1b) rather than unbinding completely [21]. Heterogeneous target sites have been used in mammalian cells [18] and yeast [20], and identical target sites have been used in mammalian cells [19]. However, in these cases, the choice between heterogeneous and identical target sites was a pragmatic choice based on the ease of engineering. To the best of our knowledge, there has not been a detailed analysis of the efficacy of heterogeneous or identical target sites for CRISPRa or CRISPRi in any system.

To better understand the trade-offs for CRISPRi between heterogeneous and identical gRNA target sites for CRISPRi in plants, we thus perform a simulation-based analysis of the potential for building highly repressible synthetic plant promoters using multiple gRNA targets. Specifically, we built a model of CRISPRi regulation in Nicotiana benthamiana using parameters drawn from constitutive promoters in that organism and previously published values of CRISPRi repression in other plants [17] and evaluate it for up to six gRNA heterogeneous or identical targets. Evaluating these circuits in simulation with respect to a range of biologically plausible parameters, we find a strong benefit for using identical gRNA target sites, both in achievable fold repression and in the scaling of repression per additional gRNA target site.

2 Materials and methods

2.1 mEmerald expression and quantification in N. benthamiana

The data used during model construction and parameter fitting came from an approximate 200-h time course of constitutive mEmerald fluorescent protein expression in N. benthamiana, and the fluorescence measurements were calibrated and reported as molecules of equivalent fluorescein (MEFL). For this purpose, 4- to 6-week-old N. benthamiana plants were agroinfiltrated in a vacuum chamber as described before [3]. A total of three plants were transformed with an Agrobacterium tumefaciens LBA4404 strain carrying a pAGM4723 binary vector backbone [22] with a transgene expression cassette for overexpression of the mEmerald fluorescent reporter gene (CaMV2x35S::TMVΩ::mEmerald::35S-polyA) in plant cells. The transformation vector was assembled by Golden-Gate cloning using regulatory elements (CaMV 2× 35S promoter, TMV Ω leader, and 35S poly A signal) and the mEmerald coding sequence from publicly available databases [22; 23]. As negative control, three N. benthamiana plants infiltrated with wild-type A. tumefaciens LBA4404 were used. The fluorescence signals produced from leaf tissue of mEmerald plants and negative controls were measured by fluorometric analysis of alive leaves using a scanning fluorescence spectroscopy apparatus (Fluorolog®-3, Jobin Yvon and Glen Spectra, Edison, NJ), as described before [24]. The mEmerald signal was analyzed at an excitation wavelength of 475 nm, and maximum fluorescence emission peaks were collected at 509 nm. For time course experiments, fluorometric data (counts per second) were collected from the same plants starting at time 0 to  200 h after infiltration, for a total of 18 time points. For each time point, 27 readings (three plants, three leaves per plants, and three readings per leaf) were collected per mEmerald genotype and wild-type LBA4404 control. At the end of the experiments (time 18,  200 h), leaf protoplasts were isolated from mEmerald plants and wild-type controls, and then, leaf protoplasts were subjected to flow cytometric analysis using an Attune NxT acoustic focusing cytometer (Life Technologies, Carlsbad, CA, USA), as previously described [3]. For this purpose, protoplasts were suspended in Washing Incubation buffer (500 mM mannitol; 20 mM KCl; 4 mM 2-(N-Morpholino) ethanesulfonic acid hydrate, pH 5.7) at a concentration of  |$1\times10^5$| cells per mL and then analyzed by flow cytometry at an acquisition volume of 750 |${\mu}$|L and a flow rate of 500 µL/min. The forward scatter and side-scattered light voltages were set at 50 and 180 V, respectively. The mEmerald fluorescent reporter was excited using a 488-nm laser with a 510/10 bandpass filter, while the voltage was kept at 200 V.

SpheroTech URQP-38-6K calibration beads were used in our experiments to calibrate flow cytometry measurements to MEFL units using the TASBE Flow Analytics software package [25] as has been done with plant protoplasts before [3]. The calibrated flow cytometry measurements and the scanning fluorometer (Fluorolog-3) measurements taken at the final time point were then used to standardize all the scanning fluorometer measurements into MEFL units.

2.2 Model construction

All the models investigated were represented using SBOL3 [26; 27]. Specifically, we constructed SBOL generators for each of the modular elements of the circuit: the expression of dCas9 and gRNA(s), their binding kinetics, and the transcriptional regulation of GFP. Because of the modular creation of the circuits, the number of gRNAs and the number of interactions between each gRNA and the GFP promoter could be varied combinatorially to generate a genetic regulatory network model for each of the possible configurations. Stantardized ontology terms were used to describe each circuit component in order to support the following machine processing steps [28]. Both LaTeX equations and MATLAB code were then programmatically generated for each genetic regulatory network. Complete information about the models used, including full ODE equations for all systems, is provided in the Supplementary Material. Model generators and products are also available on GitHub at https://github.com/TASBE/CRISPRi-promoters.

2.3 Parameter fitting

As detailed in the Supplementary Material, the parameters needed for the models were taken from previously published studies (Table S3), fit to a long time course experiment (Fig. S1), or hypothesized based on observed repression values. For parameters fit to newly generated experimental data, we performed a least-squares fit for parameters in a logarithmic parameter value space in MATLAB using the ODE model for no gRNA transcriptional repression. The complete set of parameter base values and their sources is shown in the Supplementary Material (Table S2).

3 Results

We first detail the different repressible synthetic promoter architectures under investigation and then present an exploration of their behaviors with respect to biologically plausible parameters. Further evaluation of the promoters’ tunability, sensitivity to parameter values, and sensitivity to cell-to-cell variation (presented in the Supplementary Material) shows that the conclusions from this study are not sensitive to the specific values used in our investigation (Figs S2 and S3).

3.1 Generation of possible promoter architectures

A simple CRISPRi system with a single gRNA and single target site is shown in Fig. 2A. The system consists of an engineered genetic vector (V1) for expression of the single guide RNA (gRNA1) and a second vector (V2) for expression of dCas9 and the GFP fluorescent marker controlled by the target promoter sequence.

The utilization of a two-component strategy has several advantages. The control of multiple transgenes from individual expression cassettes integrated in the same transformation vector is often difficult. The level of transgene expression can be affected by both direct acting regulatory elements and proximal regulatory elements located in other transgene cassettes, causing different expression levels from the same transgene cassette based on vector context. Furthermore, the utilization of a two-component strategy allows improved system flexibility by enabling “mix-and-match” combination of different components. The utilization of separate vectors for gRNA and dCas9 expression also enables modulation of the stoichiometry of the two components simply by changing vector dosages during the transformation of plant cells.

The expression of dCas9 is constitutive, and the dCas9 protein can bind to the gRNA to form a dCas9–gRNA complex (in this case dCas9–gRNA1), which, in turn, binds to a gRNA binding site in the GFP promoter that represses transcription. Fig. 2A also shows the mathematical symbols for the concentrations of the molecular species in the model and the parameterized rates that modulate the changes in these concentrations (see Table S1 for a full list of variable symbols and meanings). All the stable molecules, including the genetic constructs and the proteins, dilute at a rate of λ, while the gRNA molecules degrade at a faster rate of δ_g. Each genetic product is associated with its own production rate, α, where α_r is a transcription rate to produce gRNA and α_p is a combined transcription and translation rate to produce protein. The different α rates can be varied independently to reflect the strength of each protein’s individual promoter. The dynamics of the dCas9–gRNA complex formation are described by the binding constant |${k_{C_g}}$|⁠. An ordinary differential equation (ODE) model was constructed for this system and for constitutive expression in the absence of gRNA (the “on” condition). Details for equation and parameter development are described in the Methods section, and the equations and parameter values are provided in the Supplementary Material.

To investigate the potential for stronger repression, we generated models for using 2, 3, 4, 5, or 6 heterogeneous or identical gRNA target sites in a promoter. The generalized forms of the heterogeneous and identical target site systems are shown in Fig. 2b and c. As the focus of this investigation is the regulatory effects of multiple gRNA target sites, we defer any additional details related to construct ordering and specific gRNA site location for future investigation. For the purpose of this discussion, the ordering of functional units and the placement of gRNA target sites are both notional, selected primarily for clarity of illustration. ODE models were generated for these 10 multiple gRNA target circuits using the same methods as for the single gRNA target circuit. These models are also provided in the Supplementary Material and use the same parameter values as the single target circuit.

3.2 Analysis of promoter repressibility

We use simulation to evaluate the strength of repression seen for each of the 11 promoter architectures under consideration. Fig. 3 shows the results of simulating all possible promoter architectures over a 200-h time course using the base parameter values, starting with an initial 10 copies of the genetic construct encoding the gRNA(s) and 3 copies of the genetic construct with dCas9 and GFP. Previous CRISPRi studies have used co-transfections with gRNA constructs on separate plasmids than other parts of the circuit (e.g. Cas9 protein and reporter coding regions) and have tuned the dosage of the plasmids for optimal circuit behavior [19]. While our ratio of 10 gRNA containing constructs to 3 dCas9/GFP constructs is not identical to the previous optimized plasmid dosages, it is within a reasonable range and is a value that can be readily manipulated experimentally. Fig. 3a shows the concentration of the GFP over the time course for each of the possible promoter architectures. The fold repression can be calculated at any point for one of the models using the ratio between the Base Expression (no gRNA) model and the model of interest. Fig. 3b shows the average fold repression for each of the models between 72 and 96 h of the time course. The mean fold repression was shown for the 72- to 96-h window because during this time, GFP expression is no longer rising in the unrepressed system (Base Expression, No Regulation) and vector dilution has not yet significantly decreased levels of dCas9-mediated repression.

Fig. 3 shows higher levels of repression when using identical target sites than using the same number of heterogeneous target sites. This is seen in Fig. 3A, where the GFP concentrations for the identical target site repression systems are lower than almost all GFP concentrations for the heterogeneous target site repression systems, and in Fig. 3b, where the fold repression for identical target sites is always higher than the fold repression for the same number of heterogeneous target sites. Fig. 3B shows that this increased efficacy in repression allows fewer identical target sites to have the same effect as a larger number of heterogeneous target sites. For example, two identical gRNA target sites give roughly equivalent levels of repression as five heterogeneous gRNA target sites.

Fig. 3 also shows that increasing the number of gRNA target sites in a promoter increases the ability to repress that promoter. However, adding identical target sites increases that ability more quickly. For example, when adding a sixth identical target site, mean repression goes from 43.5-fold with five identical target sites to 50-fold repression with the sixth. This difference of 6.5-fold repression is much larger than the equivalent 1.3-fold difference between five and six heterogeneous target sites. For both the heterogeneous and identical target sites, Fig. 3 shows that the largest increase in repressibility comes from adding a second target site to the single gRNA system. Adding a third, fourth, fifth, or sixth binding site increases repression, but with a smaller increase each time.

Finally, while models with low levels of repression (such as the Single gRNA model) result in an approximately log-linear decrease in GFP concentration, other models show inflections in the rate of change in GFP. This inflection is both stronger and begins later for models with higher levels of repression, with the maximum found in the six identical target site model (Fig. 3A).

4 Discussion

Models with identical gRNA target sites show greater repression than models with the same number of heterogeneous target sites, and additional identical target sites caused a greater increase in repression than the addition of heterogeneous target sites. This difference between the heterogeneous target sites and identical target sites is caused by introducing dependence between the effect of the multiple gRNA target sites. Heterogeneous target sites have independent effects; for each target site to have an effect on repression, the dCas9 must bind to the correct site on DNA, which has a relatively low likelihood of occurring. While the initial binding of a dCas9–gRNA complex to the DNA is still relatively low for a promoter with multiple closely spaced identical gRNA sites, there is now a higher likelihood of the dCas9 complex remaining attached, as Cas9 does not remain strongly bound to a single site, but instead scans short distances along DNA for additional target sequences [21]. By shifting from events with all low independent likelihoods of occurring to one low likelihood event of binding and subsequent higher likelihood events of binding, multiple identical target sites are expected to yield higher levels of transcriptional repression. The Supplementary Material includes parameter perturbation studies that show that these conclusions are not dependent on precise parameter values.

Fig. 3A also shows two phenomena that we would expect based on the biophysical constraints on the system. The first is a decrease in GFP for the “Base Expression (No Regulation)” system. While there is no repressor turning off GFP expression, protein levels still go down in transient expression, as the GFP expression construct (V₂) dilutes in actively dividing cells. A second phenomenon seen is a plateaued level of GFP for the very strong repressors (e.g. “6 Identical Target Sites”), as a tight off signal will eventually not be able to go significantly lower.

While this model does not account for the physical design of the promoter, including the precise location of the gRNA target sites, recreating these results experimentally will depend on the physical constrains of dCas9 lateral diffusion. Lateral diffusion of Cas9 is typically limited to a local distance of approximately 20 basepairs [21] and is facilitated by binding to protospacer adjacent motif (PAM) sequences. To account for these limitations, promoter design could insert PAM sites within 20 basepairs of each other, potentially including PAM sites that are not adjacent to a target site, in order to prolong lateral diffusion between targets. These design considerations will need to be balanced with the needs of stability and functionality of the promoter.

For further development of identical target site promoters, next steps include more accurate simulation of repression with models of the biophysical dynamics of gRNA and promoter DNA binding, as well as validation of these circuits in the laboratory. If the expected high levels of repression are verified experimentally, CRISPRi with identical target sites may be a valuable addition to the plant synthetic biology tool kit, with potential applications in food security, environmental monitoring, and biomaterial production.

Supplementary data

Supplementary data is available at SYNBIO Online.

Conflict of interest:

None declared.

Funding

This work was supported by DARPA contract HR0011-18-2-0049. This document does not contain technology or technical data controlled under either U.S. International Traffic in Arms Regulation or US Export Administration Regulations. Views, opinions, and/or findings expressed are those of the author(s) and should not be interpreted as representing the official views or policies of the Department of Defense or the US Government. This study by approved for public release, distribution is unlimited (DISTAR Case 39 302).

Data availability

Data are available at https://github.com/TASBE/CRISPRi-promoters.

References