2.1 Discovery and Mechanism of CRISPR/Cas9

Figure 2: The CRISPR/Cas System A and B, schematic structure of Cas9 and Cas12a with their sgRNA/crRNA, sitting on a DNA strand with their respective PAMs. The sgRNA/crRNA spacer sequence binds the DNA target strand via complementary base pairing. In case of Cas9 the spacer is located at the 5' prime end, for Cas12a at the 3' end of the gRNA. The scaffold of the gRNA forms a specific secondary structure enabling it to be bound by the Cas protein. DNA cleavage sites are indicated by the scissors.

In 2012, Jinek et al. discovered the use of the Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR)/Cas system to induce double-strand breaks in DNA in a programmable manner. Since then, the system has been well established as a tool for genome editing. The CRISPR/Cas system, which originates from the bacterial immune system, is constituted by a ribonucleoprotein complex. For class 1 CRISPR systems, an RNA guide is complexed by multiple Cas proteins, whereas class 2 systems consist of a singular protein binding RNA. The class 2 type II system describes all ribonucleoprotein complexes with Cas9 (Pacesa et al., 2024). They include a CRISPR RNA (crRNA), which specifies the target sequence with a ~20 nucleotide (nt) spacer sequence, and a transactivating CRISPR RNA (tracrRNA), which induces the processing by the Cas protein (Jinek et al., 2012) (Fig. 2 A). Furthermore, a specific three nucleotide sequence (NGG) at the 3' end in the targeted DNA is needed for Cas9 DNA binding and cleavage. This is referred to as the protospacer adjacent motif (PAM) (Sternberg et al., 2014). The most commonly used Cas9 protein is SpCas9 or SpyCas9, which originates from Streptococcus pyogenes (Pacesa et al., 2024).

A significant enhancement of the CRISPR/Cas9 system was the introduction of single guide RNAs (sgRNA[s]), which combine the functions of a tracrRNA and crRNA (Jinek et al., 2012; Mali et al., 2013). Moreover, Cong et al. (2013) established precise targeting of human endogenous loci by designing the 20 nt spacer sequence accordingly.

2.2 Differences between Cas9 and Cas12a

Over the following years, several additional class 2 CRISPR/Cas systems have been discovered, including the Cpf1 system, which has been classified as Cas12a since then (Zetsche et al., 2015). Cas12a forms a class 2 type V system. In contrast to the type II systems, the Cas12a RNA guide only requires a crRNA to mediate Cas12a DNA targeting. Moreover, Cas12a is capable of processing long precursor crRNA transcripts into several, single/independent crRNAs, whereas Cas9 requires the RNase III enzyme and tracrRNA for this process (Paul and Montoya, 2020). This crRNA is often also referred to as a guide RNA (gRNA). However, the stem loop that is formed when binding the Cas protein is structurally distinct to the Cas9 gRNA and positioned on the 5' side of the crRNA (Fig. 2 B). Similarly, the PAM (TTTV) is also on the 5' side (Pacesa et al., 2024). Cas9 possesses RuvC and HNH domains that are catalytically active, each of which cleaves one of the DNA strands at the same site, resulting in the formation of blunt end cuts (Nishimasu et al., 2014). Cas12a possesses one RuvC-like domain that creates staggered cuts with overhangs that are about 5nt long (Paul and Montoya, 2020).

2.3 Dead Cas Proteins and their Application

Specific mutations of these domains result in catalytic inactivity and therefore allow for the creation of nickases that only cut one of the DNA strands, or Cas protein mutants that retain their DNA binding capability, but have no catalytic activity (Koonin et al., 2023) (Kleinstiver et al., 2019). The latter are referred to as dead Cas proteins or dCas9 and dCas12a. These Cas proteins can be used to activate (CRISPRa) or inhibit (CRISPRi) the expression of genes by fusing them to effector domains and targeting the respective genes via complementary spacer sequences (Kampmann, 2017). A common approach for CRISPRa involves fusing Cas9 with the transcriptional activator, such as VP64 or VPR (Kampmann, 2017).

4.1 Editing Endogenous Loci With Fusion Guide RNAs

To show that our fusion gRNA design results in an active CRISPR/Cas ribonucleoprotein complex, a series of different fgRNAs were cloned, each carrying spacer sequences specific to the VEGFA and FANCF target genes. HEK293T cells were then co-transfected with the Cas protein and (f)gRNA encoding constructs. The editing rate was tested 72h after transfection via a T7 endonuclease I assay.
Here, AsCas12a and SpCas9 were used. The AsCas12a spacer, in this case, targets VEGFA, while the SpCas9 spacer targets FANCF. Samples included either (i) conventional single gRNAs co-expressed with the corresponding Cas proteins (positive controls), (ii) the fgRNA co-expressed with only one of the two Cas proteins (as control for Cas ortholog dependency) and (iii) the fgRNA with both Cas proteins simultaneously co-expressed (Fig. 5). The conventional sgRNAs resulted in potent editing ("editing" refers to the observed indel frequency) for both target genes (45% for VEGFA and 15% for FANCF). Note that editing rates for FANCF were consistently lower in all experiments, which likely is due to the specific properties of the FANCF-targeted locus. Importantly, as hoped, targeting FANCF with fgRNAs resulted in noticeable editing of about 10%, when we added the SpCas9 alone or both Cas proteins into the sample. For VEGFA, the AsCas12a only sample resulted in approximately 20% editing in combination with the fgRNA, while adding both Cas proteins led to approximately 40%. These initial results confirmed that our fgRNA design indeed functions, enabling simultaneous recuitment of two different Cas proteins to separate loci in human cells.

Figure 5: fgRNAs Enable Efficient Editing of Endogenous Loci. Indel rates were assessed 72h post transfection via T7EI assay. Editing % was determined by measuring band intensities; Editing % = 100 x (1 - (1- cleaved band/uncleaved band))^1/2. The schematic at the top shows the composition of each fgRNA. The symbols indicate, which Cas variants and which (f)gRNA were present in each sample.

4.2 Efficient Fusion Guide RNA-Mediated Editing With Various Cas Orthologs

After showing efficient editing, the next step was to evaluate the capabilities of the fgRNAs. To this end, we tested fgRNAs in combination with different Cas12a orthologs. Following some initial testing, we decided to use MbCas12a together with SpCas9 in subsequent experiments, since MbCas12a turned out to be more effective than AsCas12a in a dual luciferase assay when co-transfected with SpCas9 (Fig. 6). Of note, SpCas9 reproducibly showed high potency.

Figure 6: Comparison of AsCas12a and MbCas12a with a Dual Luciferase Reporter Assay. HEK293T cells were co-transfected with the indicated constructs and firefly and Renilla luminescence intensity was measured 48 h after transfection. Relative luciferase units correspond to firefly luciferase photon counts normalized to Renilla luciferase photon counts (Renilla luciferase serves as transfection control). Potent activity of CRISPR-Cas effectors result in knockdown of the firefly luciferase and hence smaller values. Samples correspond to cells co-transfected with a firefly luciferase targeting fgRNA and the constructs encoding the indicated CRISPR-Cas effector (see color legend). Data represent the mean +/- SD from n = 3 independent experiments. Statistical analysis was performed using 1way ANOVA with Tukey's multiple comparisons test. ***p<0.001,

Additionally, to determine whether the differences in editing rates observed in the preliminary assays were due to the specific properties of the targeted loci or the distinct characteristics of the Cas orthologs used, the spacers were tested in two configurations. In one setup, the fgRNA-encoded spacer for Cas12a targeted FANCF, while the spacer for SpCas9 targeted VEGFA; in the other, the targets were reversed. To more precisely assess the impact that the utilization of a fgRNA design has on the editing rates, conventional crRNAs/sgRNAs were tested separately or, alternatively, combined in one sample.
Combining a conventional crRNA/sgRNAs with the individual Cas proteins in the same sample showed no significant difference in editing rates (Fig. 7) as compared to using them in separate samples. While the fgRNA led to an overall lower editing rate as compared to the conventional guide RNAs, editing was still clearly noticeable for both targeted loci, indicating that the fgRNA works in principle. While editing efficiency for VEGFA remained consistently around 20%, the editing rate for FANCF dropped significantly, but was still detecable. Under identical conditions, MbCas12a consistently exhibited lower editing rates compared to SpCas9 when targeting the same gene, which is expected based on experience with this Cas orthologs in the CRISPR field.

Figure 7: FgRNA-mediated dual genome editing with SpCas9 and MbCas12a. In A and B the editing rates were determined 72h post transfection via T7EI assay. Editing % was determined by measuring band T7 cleavage band intensities as explained above. The schematic at the top shows the composition of the fgRNA. Below each spacer, the target gene is indicated. The symbols below indicate which parts are included in each sample. The fgRNA in A and B differs by the order of used spacer sequences (FANCF-VEGFA vs VEGFA-FANCF).

4.3 Fusion Guide RNAs are Compatible with Linkers of Various lengths

To further assess the effect of the genomic locus on the editing rate, we included CCR5 as an additional gene target. For this assay, a fgRNA with a 20 nt long linker sequence between the two spacers was included. As above, fgRNA-mediated editing was assessed via T7E1 assay in HEK293T cells. The editing rate for VEGFA was again relatively consistent throughout the samples (Fig. 8). For CCR5, the editing rate with the conventional sgRNAs was in a similar range as that of VEGFA (at about 30%), but dropped to below 10% in the fgRNA sample. The addition of the 20 nt linker had no effect on the editing rates compared to no linker.

Figure 8: Fusion gRNAs with A 20 Nucleotide Spacer Still Mediate Editing For CCR5 and VEGFA HEK293T cells were co-transfected with the indicated components followed by assessing editing rates 72h post transfection via T7EI assay. Editing % was determined as described above. The schematic at the top shows the composition of the fgRNA. Below each spacer, the targeted gene is indicated. The symbols below indicate which parts are included in each sample. Cas12a targets VEGFA and Cas9 targets CCR5 in this case.

4.4 Fusion Guide RNAs Enable Efficient Activation of Gene Expression via CRISPRa

The data above provide manifold evidence that fgRNAs can indeed be employed for simultaneous genome editing at two defined loci with catalytically active Cas9 and Cas12a. Next, to establish the foundation for the use of fgRNAs as scaffold for Cas-staple protein assembly and hence 3D genome enineering, we next combined fgRNAs with catalytically dead Cas9 and -12a in an CRISPRa (CRISPR activation of gene expression) experiment. For this, we intended to recruit the transcriptional activator VP64 to a firefly luciferase reporter gene to induce expression. The VP64 protein is genetically fused to the catalytically inactive Cas9 protein, which is then guided by gRNAs to the luciferase-driving minimal promoter. More specifically, the gRNAs here target a TetO repeat sequence, which is positioned 5' of a minimal tata site. The luciferase reporter activity was then quantified and photon counts for Firefly luciferase (to be activated by CRISPRa) were normalized to the photon counts of Renilla luciferase, which is expressed on a separate plasmid under an ubiquitous promoter (transfection control). In two biological replicates we saw similar relative luciferase activity with fgRNA as a guide compared to a sgRNA (Fig. 9).

Figure 9: The Efficacy of CRISPRa is Comparable Between Our fgRNAs and Conventional sgRNAs. The schematic at the top shows the composition of the fgRNA. Symbols show which parts were co-expressed in each sample. The empty spacer control is a negative control with a gRNA lacking the TetO-targeting spacer, thus preventing binding of dCas9-VP64 to the Firefly-luciferase driving promoter. Firefly luciferase activity was measured 48h post transfection and, as above, normalized to an ubiquitously expressed Renilla luciferase (transfection control). The tetO repeats were targeted by Cas9-VP64, once with a sgRNA (center) and once with a fgRNA (right) that carried a non-targeting spacer sequence for Cas12a.

4.5 A Proof-Of-Concept for 3D Genome Engineering: Stapling Two DNA Strands Together In Human Cells Using fgRNAs

Having demonstrated the excellent compatibility of the fgRNA design with simultaneous genome editing at two distinct loci and CRISPR activation, we set out to provide a proof-of-concept for engineering 3D genome conformation in living cells. To achieve this, we designed an experiment to "staple together" two regulatory elements—a synthetic enhancer and a minimal promoter—located on different strands of DNA. We developed a two-plasmid system consisting of an enhancer plasmid and a reporter plasmid. The reporter plasmid encodes firefly luciferase driven by a minimal promoter containing several repeats of a Cas9-targeted sequence, while the enhancer plasmid includes a Gal4 binding site (UAS) positioned next to several repeats of a Cas12a-targeted sequence. We co-expressed these two plasmids with dCas9, dCas12, an fgRNA simultaneously co-targeting (i.e., tethering) the reporter and enhancer plasmids, and Gal4-VP64. This resulted in the expression of luciferase (Fig. 10, Panel A). Different linker lengths (as mentioned above) were tested for the fgRNA. Firefly luciferase values were normalized against ubiquitous renilla expression as a transfection control. Using no linker between the two spacers produced luciferase activity values similar to the baseline control (i.e., no reporter activation) (Fig. 10, Panel B). Excitingly, however, the stepwise extension of the fgRNA linker (i.e., the linker separating the Cas12a and Cas9 spacer elements) from 20 nt to 40 nt led to progressively increasing, and ultimately potent, luciferase reporter activation. This experiment provided the first demonstration that an fgRNA with an optimized linker can effectively "staple" two otherwise separate pieces of DNA. As this experimental readout—mediated by the spatial proximity of two separate regulatory elements—mimics naturally occurring enhancer hijacking events, we refer to this setup as "synthetic enhancer hijacking."

Figure 10: Applying Fusion Guide RNAs for Cas staples: Proof-Of-Concept 3D Genome Engineering Via Synthetic Enhancer Hijacking. A, schematic overview of the assay is shown. An enhancer plasmid and a reporter plasmid are brought into proximity by an fgRNA-Cas staple complex, which co-targets and tethers both plasmids. Target sequences were included in multiple repeats upstream of the functional elements. Firefly luciferase serves as the reporter gene, while the enhancer is composed of multiple Gal4 repeats bound by a Gal4-VP64 fusion protein. B, HEK293T cells were co-transfected with the reporter plasmid, the enhancer plasmid, a Gal4-VP64-encoding plasmid as well dCas9 and dCas12. Firefly luciferase activity was measured 48 hours post transfection and normalized to ubiquitously expressed Renilla luciferase. Observed differences were tested for statistical significance using ordinary one-way ANOVA with Dunn's method for multiple comparisons. (*p < 0.05; **p < 0.01; ***p < 0.001; mean +/- SD from n = 3 independent experiments). The assay included sgRNAs and fgRNAs with linker lengths varying between 0 nt to 40 nt as indicated.

5.1. Enhancer Hijacking is Successfully Studied In Silico

Figure 11: Cas stapled plasmids.

With the Cas staple, we aimed to simulate the principles of enhancer hijacking experiments we conducted in the lab. For these experiments, we modeled the two plasmids also used in the wet lab (BBa_K5237023 and BBa_K5237024). On top of the two plasmids, we modeled the Cas staple by applying a “DNA-DNA tethering force" throughout our simulation, selectively on the regions targeted by the fgRNA. This force was based on simulation data acquired in earlier phases of DaVinci. As there is no suitable model available that also simulates proteins, this proved to be the most effective modeling strategy.

Our simulation showed the expected behavior, holding the target sequences of the Cas staple (Fig. 12). Overall, these exciting results demonstrate that we can successfully model the core principles of enhancer hijacking with a total of 20 thousand simulated nucleotides in silico.

5.2. Cas Staple Forces do not Disturb DNA Strand Integrity

Figure 13: Cas stapled plasmids.

Next, we aimed to stress test our system to determine the amount of force required to induce DNA double-strand breaks. To achieve this, we used an identical setup to the previous experiment but instead of experimentally determined forces, we used artificial forces of varying strength. It is important to know that our in silico model responds to forces that cause double-strand breaks by scattering the nucleotides across the simulation box. As the specified bonds cannot actually break within the simulation, the phosphate backbone bonds become severely distorted and exhibit erratic behavior in the simulation videos. From our extensive simulations, we concluded that the forces required to damage the DNA are approximately 320 times (Fig. 15) and 680 times greater (Fig. 16) than the forces exerted by the Cas staple, respectively.

This provides important evidence regarding the safety of our staple approach for 3D genome engineering: According to our model, DNA-DNA tethering with our Cas staples is not expected to have a negative effect on the DNA stability.

5.3. DaVinci Helps to Design Multi-Staple Arrangements

Figure 18: Double Cas stapling within 40 nucleotide distance induces double-strand breaks.

Finally, we simulated the behavior of multiple staples at once. Therefore, we expanded our previously introduced experimental setup by a second Cas staple.
In a first approach, we targeted an additional region next to the original chosen one. This region is 40 nucleotides away from the first target region on the plasmid displayed in blue connecting it to the opposite site of the orange plasmid, therefore bringing both ends of the orange plasmid together (Fig. 18).
Our simulation predicted that with these staple points, DNA double-strand breaks would occur as clearly visible by the scattered nucleotides (Fig. 19).

Figure 20: Stable multiplexing with 2 Cas staples. On the blue plasmid, the Cas binding sequences are 980 nucleotides apart.

To simulate a setup where we expect no double-strand breaks, we increased the distance between the stapling sites on the blue plasmid (BBa_K5237023) from 40 to 980 nucleotides (Fig. 20).
With this increased distance between stapling sites, we observed a stabilized system. Most interestingly, the non-stapled regions showed maximum distances close to 500 nm, indicating that the two staples led to more compact plasmid structures.

In conclusion, we show that applying multiple staples on the same structures can lead to double-strand breaks if the staples are positioned closely to one another. However, increasing the separation of staples leads to a stable system without DNA damage. Thus, it is definitely possible to multiplex Cas staples, thereby increasing the compactness of DNA structures (Fig. 21). This shows the potential of creating complex regulatory networks by bringing genetic elements into spatial proximity with a multitude of Cas protein staples.