Microfluidic tool for rapid functional characterization of CRISPR complexes.

RNA guided nucleases are regarded as the future genome editing technologies. As such, they need to meet strong safety margins. Two major challenges in incorporating CRISPR technologies into the clinical world are off-target activity and editing efficiency. The common way to tackle such issues is to measure the binding and cleavage kinetics of the CRISPR enzyme. This can be challenging since, for example, DNA is not released from the CAS9 protein post cleavage. Here a promising new microfluidic approach to characterizing Enzymatic Interaction and Function of CRISPR complexes on a microfluidic platform (EnzyMIF) is presented. The method can rapidly detect the kd, koff, km and kcat for various RNA guided nucleases. In this work, two single guide RNAs with significantly different in-cell cleavage efficiency, RAG2 and RAG1, are used as proof-of-concept. The EnzyMIF assay results provide biochemical characterization of these guide RNAs that can explain the difference in cleavage using both wild type (WT) CAS9 and HiFi CAS9. Notably, it is shown that EnzyMIF characterization correlates with cell culture genomic editing efficiency results. It is suggested that EnzyMIF can predict the quality of cleavage rapidly and quantitatively.

Introduction

CRISPR-CAS9 is the state-of-the-art gene editing technology that promises tools for specific DNA editing, for both the research and medical communities [1,2]. The CAS9 protein of the CRISPR system is complexed together with a designed guide-RNA to create a Ribonucleoprotein (RNP) complex. The RNP is guided to the target DNA for specific cleavage near a protospacer adjacent motif (PAM) site [[3], [4], [5]], enabling guide RNAs to be engineered for various applications. Nevertheless, several challenges still hinder effective usage of this technology for medical applications.

Characterization of enzyme binding kinetics is important for determination of their degree of specificity and reaction rate. The common perspective is that there is rapid equilibrium of the enzyme-substrate binding prior to enzyme activity. Thus, the characterization includes measurement of both the binding and the activity, as described by the Michaelis Menten model [5,6]. Regarding CRISPR-CAS9 enzymes and other RNA guided nucleases, specificity can be defined as the ratio of cleavage efficiencies of the target to its competing substrates [[7], [8], [9], [10]]. Current in vitro methods of cleavage quantification either fail to fit the Michalis Menten model or require lengthy gel-based assays and/or modified proteins [[11], [12], [13]]. The lack of in-depth enzymatic kinetic analysis impedes the understanding of the mechanistic basis for CAS9 enzyme specificity [13].

While major studies have shown that CAS9 RNP is a single turnover enzyme and therefore does not fit the Michaelis Menten model [14], others have theorized that it is still the best model to explain the’ excess energy’, i.e. the energy above a certain threshold that is needed for the stabilization of both on-target and off-target site binding. If the binding energy of two molecules or complexes is much higher than the minimal threshold, the stability of this binding is lower than binding that is maintained with lower excess energy.

It has been shown that models based on kcat/km can assist in engineering RNPs with less off-target activity [[15], [16], [17]]. Off-target cleavage, leading to undesirable mutations, is a major concern in CRISPR technology [18], occurring in 0.1%–60% of cells, depending on cell type and target sequence [4]. Many efforts have been made to reduce the off-target effects at the protein level [9,10] and several online tools to predict off-target cleavage due to’ mistargeting’ of the guide RNA have been published [19,20]. However, predicting guide RNA binding efficiency is still a major challenge, since it has been shown this can vary significantly between guide-RNA sequences [3,[21], [22], [23]]. Also, while bioinformatic tools benefit the scientific community, providing large amounts of data in short amounts of time does not always accurately reflect experimental data, and thus cannot completely replace experimental data [24].

Regarding prediction of on-target cleavage, several online bioinformatic tools also exist [19], but there are few experimental platforms that enable the characterization of correct targeting of a guide RNA towards a specific DNA sequence. The kinetics of CAS9 RNP DNA binding and cleavage have been tested in various ways both in vitro and in cellula [11,13,14,25] however, these methods require either Total Internal Reflection Fluorescence (TIRF) microscopy, gel electrophoresis, protein engineering or deep sequencing, thus making them low throughput, expensive and time consuming. As for gel electrophoresis, the separation itself takes the interaction out of equilibrium and precise quantification of results is challenging, thus raising questions regarding its resolution abilities [[26], [27], [28], [29]]. Furthermore, existing methods also do not differentiate between the rate limiting and the specificity determining reaction pathways, needed for complete characterization of a single enzyme reaction cycle [13,30,31]. These major disadvantages mean that there is currently no method that can fully automate the characterization process while giving an accurate comparison of both binding and cleavage for several CRISPR complexes simultaneously. Thus, these drawbacks highlight the need for a more economical, efficient and thorough technique for analysis of CAS9 kinetics.

Detection and characterization of kinetics of DNA binding proteins has been very successful using integrated microfluidics assays [32]. Microfluidics is a broad and growing field that enables quick, accurate and low-volume experiments. Microfluidic devices can be designed and tailored towards specific assays, such as protein-DNA or protein-RNA interactions [[32], [33], [34], [35], [36], [37], [38]]. Furthermore, the use of a mechanically induced trapping of molecular interactions (MITOMI) valve allows the detection of interactions at equilibrium [38] thus enabling the detection and quantification of weak interactions.

In this work, an assay has been engineered based on a novel microfluidic device that can address the kinetics of a single enzyme reaction cycle, and automated microfluidics is used to characterize CAS9 RNP DNA binding and cleavage efficiency. The binding assays determined the kd and koff, while a cleavage assay was used to determine the km. To measure the km, the challenge of the “single turnover” enzyme was overcome by using a protease to release the DNA, successfully’ forcing’ the enzyme into a Michaelis Menten model. By timing the cleavage using MITOMI valves and degrading the protein, we were able to quantify the velocity of the reaction at different time points. Moreover, we were able to observe two steps of the reaction, one rate-limiting step and the other specificity-determining.

Materials and methods

Chip fabrication

Chip design was performed using AutoCAD, followed by laser lithography using the MLA-150 maskless aligner (Heidelberg Instruments, Heidelberg, Germany). For the control layer, a 4” wafer was coated with SU-8 negative photoresist while the flow wafer was coated with SPR 220 7.0 positive photoresist, as previously described [[32], [33], [34], [35], [36], [37], [38], [39]]. Soft lithography was then performed by polydimethylsiloxane (PDMS) casting on the wafers and baking until PDMS was cured. Alignment of the two layers was performed using the μDAS system [40], and the full device was bound to an epoxy coated glass slide.

Surface chemistry

The preparation of the inner surface of the EnzyMIF (epoxy coated glass) for the experiments, included the following steps. Biotinylated-BSA (1 mg/mL, Pierce, Waltham, Massachusetts, USA) was flowed for 20 min through the device, binding BSA to the epoxy surface. Thereafter, NeutrAvidin (0.5 mg/mL, Pierce, Waltham, Massachusetts, USA) was flowed for 20 min. The ‘MITOMI’ valve was then closed and biotinylated-BSA was flowed again through the chip for passivation of all areas except the button where reaction happens. For the association and dissociation experiments, the ‘MITOMI’ valve was opened and biotinylated anti-His antibody (QIAGEN, Hilden, Germany #34440) was flowed and bound to the exposed NeutrAvidin for further immobilization of the various His6-tagged CAS9-RNP complexes. Various concentrations of Cy5-labeled target DNA were then separately but simultaneously introduced to provide accumulation of DNA around the closed MITOMI valve. The valve was then elevated to allow DNA binding to the CAS9 RNP. The reaction was captured at equilibrium and scanned using the Tecan's LS Reloaded Scanner (Tecan Trading AG, Männedorf, Switzerland) (Fig. 1D).

Fig. 1

Association and dissociation assays preformed on the EnzMIF device.

A) A photograph and diagram of the EnzMIF device which has 32 outputs/inputs and can test up to 32 different sgRNAs simultaneously. B) Schematic illustration of the binding assay to determine kd: His-tagged RNP is immobilized to the surface under the MITOMI valve, and then exposed to a dose response of on-target Cy5 labeled DNA. Interaction is captured at equilibrium and the fraction bound is calculated using anti-CAS9 488 fluorescence. C) Schematic illustration of off-rate determining assay; once Cy5 labeled is captured at equilibrium to immobilized RNP, the MITOMI button is released, and the decay of the reaction detected after incubation in PBS for 1/2/5/10 min. D) Image of a section taken from an experiment showing anti-CAS9 488 labeled RNP and Cy5-labeled on-target DNA. E) Association and dissociation curves of three CAS9 proteins using two types of sgRNA: RAG1 (red) and RAG2 (blue) (Representative experiment, each experiment was repeated at least 3 times).

For the cleavage assay biotinylated DNA–Cy5 was flowed and bound to the NeutrAvidin below the MITOMI valve. The various CAS9-RNP complexes were then flowed through the chip and allowed to react with the DNA. All washing steps were performed using phosphate buffered saline (PBS) buffer pH 7.4.

Cy5 labeled dsDNA synthesis for dose response assay

Template DNA was ordered from IDT (Coralville, Iowa, USA) and the complementary strand was synthesized using Klenow fragment (NEB, Ipswich, Massachusetts, USA #MO2102) and a Cy-5 labeled primer. Primer and template were mixed at equimolar concentrations in NEBuffer2NEB # B7002S. With dNTPs, and the mixture was heated to 95 °C and slowly (0.1c/sec) cooled to 37 °C using a PCR machine. Following the hybridization of the primer and template, the Klenow fragment was added to complete the dsDNA. The final product was diluted in NEBuffer2 and 100 ng/ul Salmon sperm DNA (Sigma Aldrich, St. Louis, Missouri, USA, #D1626-1 G) for the dose response assay.

CAS9 RNP preparation

Alt-R.® S.p CAS9 (IDT, #1081059), Alt-R® HiFi S.p CAS9 (IDT, #1081060) and dCAS9 (Sigma- Aldrich St. Louis, Missouri, USA, # D110010) were mixed with Alt-R® RAG1 (single guide (sg) RNA sequence: 5’ – AACUGAGUCCCAAGGUGGGUGGG-3’) and RAG2 (sgRNA sequence: 5’ – UGAGAAGCCUGGCUGAAUUAAGG -3’) sgRNA (IDT) in a 1:1.5 M ratio (protein: sgRNA) in NEBuffer2. The mixture was incubated for 20 min at room temperature until full formation of the RNP complex. For the TIRFm assay, RAG2 Alt-R® CRISPR-CAS9 CRISPR (cr) RNA (IDT, #1072532) and Alt-R® tracrRNA, ATTO™ 550 (IDT, #1075927) were hybridized before incubation with the CAS9 protein.

Biotinylated and Cy5 labeled dsDNA preparation

RAG1 and RAG2 on-target DNA templates were amplified using 5' biotinylated primer and 3' Cy5 primer using Real time PCR SYBR Mix (Quantabio, Beverly, Massachusetts, USA) and a real time PCR machine (CFX-96, Bio-Rad, Hercules, California, USA). The first run was 40 cycles in order to determine the optimal cycle number (OCN), at which the beginning of the exponential phase can be determined. The next run was then stopped at the OCN in order to prevent unwanted PCR byproducts. The reactions were then cleaned from excess primers using a 30,000 MWCO spin-column (SARTORIUS Göttingen, Germany, VIVACON 500, #VN01H21). Concentration and quality of the cleaned products were determined by Qubit Fluorometric Quantification (Thermo-Scientific, Waltham, Massachusetts, USA) and automated electrophoresis TapeStation system (Agilent, Santa Clara, California, USA).

Immobilized DNA concentration detection and CAS9 degradation via Proteinase K

Biotinylated and Cy5-labeled dsDNA products were diluted to 50 nM with 10 nM of free biotinylated primer to compete with the matching dsDNA product during the surface binding process. This mixture was then introduced to the NeutrAvidin-coated surface below the MITOMI valve, together with the biotinylated primers. The gradient signal intensity was determined by the microarray fluorescent scanner (LS Reloaded, TECAN) before the exposure to CAS9. The concentration of each button was determined by a Cy5-primer calibration curve of 100, 50, 25 and 0 nM diluted in PBS, which was introduced to the side chambers and scanned for their fluorescent signal. The cleavage activity of CAS9 was determined using Proteinase K (Thermo-Scientific, Waltham, Massachusetts, USA, #EO0491, 1:40 in PBS) which was incubated with open MITOMI valves for 10 min at 42 °C. The difference in the signals before and after the incubation with Proteinase K and degradation of CAS9 were calculated, subtracting the background fluorescence level (which changed due to bleaching and Proteinase K degradation of the surface). The normalized calculated delta of signals was translated to nM using the calibration curve described above.

Data analysis and non-linear fitting

All scanned images were analyzed using GenePix7.0 (Molecular Devices, San Jose, California, USA) and the raw data was normalized to background levels and in some cases to DNA concentration based on the calibration curve described above. Data was then fitted by non-linear least squares regression available on: https://statnfo/nonlin.html. The equations used to evaluate the various constants (kd, koff, T1/2) were as follows:

For the association assay:

For the dissociation assay (no normalization to protein concentration):

T1/2 was then determined by calculating: ln2/K.

For the CAS9 cleavage assay- determining km and Vmax:

kcat was then determined by calculating:

Results and discussion

The microfluidic device for measurement of enzymatic interaction and function (EnzyMIF)

The EnzyMIF device contains 32-channels with 16 MITOMI valves in each channel. Besides the common manifold for addressing all 32 channels, each channel has two in/out ports at its end, allowing direct flow into individual channels or elution out of the channels in an isolated manner (Fig. 1A). This design allows for 32 simultaneous experiments, e.g., the interaction of 32 DNA concentrations with the same CAS9-RNP complex that is bound to the surface. This is analogous to a gel with 32 lanes that can be loaded in parallel. Moreover, each set of 8 channels is controlled by a separate set of valves, enabling testing of interactions between eight concentrations of DNA for four protein complexes simultaneously for kd determination (Fig. 1B). This provides an important advantage over previous MITOMI devices, as it allows the application of the same surface chemistry for protein immobilization while enabling the screen of a vast range of ligands, avoiding experiment specific biases. Moreover, various DNA concentrations can then be introduced without the need to print a microarray. In this experimental setup, surface chemistry was applied using the common manifold, as described in the Methods section and previously [36].

Association and off-rate determination using the EnzyMIF device – comparison of Single Guide RNAs (sgRNA)

A set of experiments was performed for the determination of off rate values of two sgRNAs; RAG1 and RAG2 (Fig. 1C, Suppl. Fig. S1). Each sgRNA was complexed with three types of CAS9: Streptococcus pyogenes (S.p.) CAS9 nuclease, S.p. High Fidelity (HiFi) CAS9 Nuclease [41] and S.p. dead (inactive) CAS9 [42] (dCAS9, Fig. 1E). The resulting RNPs were then immobilized onto the surface. The evaluation of bonded protein complex level was then determined using fluorescently labeled anti-CAS9. The protein complexes then were exposed to a range of concentrations of target DNA-Cy5 sequences. The MITOMI was activated to capture the interaction while the chip was washed to remove unbound DNA. The dose response binding levels were detected by their fluorescent signals and fitted to the curve to evaluate the kd (Fig. 1E). To measure the off rate, the complex was then washed with PBS (by MITOMI release) for 1, 2, 5 or 10 min, followed by data acquisition by a fluorescent scanner. The bound DNA, evaluated by Cy5 fluorescence was normalized to the plateau level and the data was fitted to determine the off-rate value.

As presented in Fig. 1E, the binding efficiency as presented by kd of the WT CAS9-RAG2 complex is higher than that of the WT CAS9-RAG1 complex and the binding is stronger as indicated by the slower release and higher plateau level of the dissociation curve (off-rate assay). However, when dead CAS9 was used, both complexes had similar on- and off-rates. These differences can be attributed to the cleavage process, which occurs only with live CAS9, and is the only difference between the two WT protein RNPs (Fig. 1E, Suppl. Table S1). For negative controls, the experiment with miss-matching sgRNA for each RNP type was performed (Fig. S2). The association and dissociation of WT CAS9 to both DNA sequences was also tested in the absence of sgRNA. As expected, in most of these experimental set-ups, a linear correlation between DNA concentration and the binding curve was observed, with almost identical kd levels, indicating nonspecific binding. The dissociation curves were also similar and yielded similar koff values. CAS9 without an sgRNA showed strong non-specific binding, as expected. The bound DNA was not fully released, plateauing at about 40 %. The control data are presented in Suppl. Table 2.

Fluorescence and TIRF microscopy analysis demonstrate a two-step dissociation process

TIRF microscopy (TIRFm) was used to confirm that the system behaves as previously observed in the literature [14] and further characterize the CAS9 binding kinetics. While it is known that CAS9 has a very long off-rate of up to 6 h [43], the results show that this process comprises two steps. The first step takes less than one minute, whereas the second step is very gradual, reaching a plateau. After one hour of TIRF monitoring (Fig. 2), using two part fluorescently labeled guide RNA, the RAG2 T1/2 was found to be 5.55 h, similar to that reported elsewhere [43]. As with fluorescent scanner monitoring (Fig. 1), dCAS9 presented similar T1/2 levels for both sgRNAs, implying that the binding of DNA to CAS9 also depends on the activity capabilities of the complex. T1/2 levels of the WT CAS9-RAG2 complex were found to be higher than that of RAG1, indicating stronger interactions between the enzyme and DNA. The HiFi CAS9, which has, by definition, reduced affinity to DNA, indeed presented lower T1/2 plateau values for both RAG2 and RAG1. Technical noise from the MITOMI valve was only observed for the first sec and cannot account for the rapid decline observed in the first minute.

Fig. 2

Off-Rate assay evaluated by TIRF microscopy. Following the binding assay at equilibrium, the MITOMI valve was open and the complex WT CAS9- RAG2 was exposed to PBS for the off-rate assay. A rapid decline lasting for ∼1 min was first seen, followed by a long decline that lasted for the entire monitoring period (up to 60 min). Full line - real data, dashed line - fitting curve.

WT CAS9 and dCAS9 show similar kd values at equilibrium, yet plateau with different off-rates

Although WT CAS9 and dCAS9 respond similarly to DNA binding at equilibrium, presenting nearly the same kd values for both RNP complexes (RAG1 and RAG2), the off-rate of the DNA was found to differ. When measuring the off rate for dCAS9, both RAG1 and RAG2 RNP complexes reached ∼56−60% of the initial signal. However, the WT CAS9-RAG2 complex appeared to bind DNA more efficiently and with stronger affinity, reaching a plateau level of 70 %, while RAG1 reached a plateau level of 43 % (Suppl. Table S1). This interesting observation may again imply the effect of enzymatic activity on the off-rate levels, suggesting that this system can differentiate between the efficiency of the hybridization process of different sgRNAs without requiring additional complicated assays. It is important to emphasize that the off-rate evaluation is possible only due to the ability to capture the DNA-CAS9 complex and to take a snapshot of the reaction at the desired time points within the dissociation phase. This is not possible in techniques such as chip-hybridized association-mapping platform (CHAMP) [44], providing an advantage to this platform.

sgRNA type affects DNA binding profile differently for HiFi CAS9 compared to WT CAS9

As described above, the DNA binding affinity comparison between dCAS9 and WT CAS9 showed that the DNA binding was affected by CAS9 ability to cleave DNA. When cleavage abilities were absent – dCAS9, the sgRNA type had no effect on the association and dissociation profiles of the enzyme, whereas when activity capabilities were intact, the DNA binding character was affected by the sgRNA sequence. Based on these insights, it was of interest to test the effect of sgRNA type on the HiFi CAS9 enzyme, which by definition uses lower DNA binding affinity to enhance its specificity [41].

As presented in Fig. 1E, the application of HiFi CAS9 switched the binding patterns of both RNPs (RAG1 vs. RAG2) compared to the WT CAS9. Additionally, the DNA off-rate of both RNPs changed significantly. While the final plateau level for RAG1 (43 %) and the rapid phase of dissociation were similar for both WT and HiFi CAS9, the second phase of dissociation was slower with HiFi CAS9, yielding a higher T1/2 value (2.7 min vs. 1.55 min). It is concluded that RAG1 sgRNA allows for an RNP-DNA interaction of longer duration (with HiFi CAS9 vs. WT CAS9), increasing cleavage activity. However, when HiFi CAS9 with RAG2 sgRNA was tested, final plateau levels were significantly lower than when using WT CAS9 was used (35 % Vs. 70 % respectively) (Suppl. Table S1A) and the plateau stage was achieved 4.5 times faster (T1/2 0.39 min vs. 1.8 min).

A rapid off-rate is one of the manipulations used to increase CAS9 specificity: as the binding between the enzyme complex and the DNA is weaker, the enzyme has a shorter time window for cleavage activity. Thus, a mismatching DNA-protein complex, which is less stable than a matching one, remains at its binding state for a shorter period, reducing the number of off-target cleavage events, increasing its specificity. This hypothesis is also explained by [18], where it is claimed that one way to increase specificity is to increase off-rate levels. Others also showed enhanced specificity of RAG2 without losing cleavage efficiency, unlike RAG1 [45], as also indicated in the present study. Another way to increase specificity according to this hypothesis is to decrease kcat, an additional challenge tackled with the cleavage assay as presented here.

Experimental setup for measuring cleavage

The experiments conducted for the evaluation of DNA binding state of various CAS9 proteins and RNP complexes included the immobilization of the protein complex, while DNA sequences were in the solution. To evaluate the cleavage activity, the experimental setup was reversed (Fig. 3B). ‘On target’ DNA sequences were amplified with biotinylated, and Cy-5 labeled primers, with the PAM sequence being closer to the biotin moiety (Fig. 3A). In order to induce a gradient of DNA concentrations under the MITOMI valves in each row, the purified PCR product was mixed with biotinylated primer of decreasing concentrations. The primer concentration in the sample was higher than the amplified DNA, hence decreasing the DNA signal below the MITOMI valve, due to the higher propensity of shorter DNA sequences to bind to the surface. The evaluation of DNA concentration in each chamber was determined using a calibration curve of free Cy5 primer trapped in the DNA chamber (Fig. 3C). As previously indicated in the literature, the cleaved DNA is not released from the CAS9-RNP complex following CAS9 cleavage [15,46]. In order to observe a change in fluorescence level before and after cleavage, Proteinase K was used to elute cleaved DNA via CAS9 degradation (Fig. 3B, D).

Fig. 3

Cleavage assay using Proteinase K for rapid release and detection.

A) Design of target dsDNA probe used in the km assay. DNA was immobilized using Biotin and NeutrAvidin, cleavage product was detected as a decrease in Cy5 signal from the surface. B) Schematic illustration of the cleavage assay on-chip (i) immobilized DNA is detected; (ii) CAS9 is introduced for 1 min and then trapped using the MITOMI valve; (iii) CAS9 is then degraded by Proteinase K and cleavage is measured by levels of signal loss. C) Heat map presentation of the immobilized DNA gradient under the MITOMI valve, before and after cleavage and CAS9 degradation via Proteinase K. D) TapeStation analysis of washing and elution solutions presenting cleaved and non-cleaved DNA. Positive Wash (PW) shows no signs of cleaved product. Positive elution (PE) following incubation with Proteinase K shows the cleaved product (blue arrow). Unrelated sgRNA was used for Negative binding, Wash (NW) vs Elution (NE) showing no cleaved DNA. CAS9 alone (no sgRNA) shows no cleavage product as well as release of the DNA with 5 min washing (CW) only after elution (CE).

Microfluidics evaluation of the enzymatic efficiency of various CAS9-RNP complexes

The cleavage process was timed and the reaction velocity determined by using the gradient of DNA concentrations, applying equal concentration of CAS9-RNP, and using Proteinase K to stop the cleavage reaction, simultaneously in all reaction chambers. Following introduction of the RNP around the immobilized DNA, the MITOMI valve was opened to allow cleavage activity and then closed again to trap the reaction at the desired time point. Proteinase K then was incubated together with the RNP for degradation of CAS9 complex, thus releasing cleaved DNA. Non-cleaved DNA at said time was trapped beneath the MITOMI and detected via fluorescence. Even if some of the DNA molecules were cut while under the MITOMI button, the cleaved products remained trapped under the valve, ensuring that the fluorescence signal under the button accurately represented the activity at the specific time point. The DNA signal was then calculated for each concentration before and after CAS9 binding and elution, with the delta being the y axis for the Michaelis Menten model. Two experiments were performed for WT CAS9 and for HiFi CAS9 (Fig. 4A). Results showed that using WT CAS9, both RAG1 and RAG2 shared a similar pattern, unlike HiFi CAS9 where RAG1 and RAG2 differed in their cleavage velocity profiles.

Fig. 4

Cleavage evaluation of in vitro and in cellula experiments of different experimental setups (CAS9 and SgRNA’s).

A) Real data and fitted Michaelis Menten curve of the velocity (nM/min) of the CAS9-RNP activity at different DNA concentrations is presented for WT CAS9 and HiFi CAS9 using RAG1 sgRNA (red) and RAG2 sgRNA (blue). Calculated enzymatic efficiency (kcat/km) and Vmax (nM/min) values are presented for both setups. B) TIRF microscopy analysis of CAS9-RAG2 (0.18 nM) 1 min incubation cleavage assay. Left: Bound DNA signal to immobilized RNP over time in equilibrium. Right: DNA-Cy5, with and without 1 min incubation with RNP, and RNP-ATTO550 signal loss after 5 min with Proteinase K. *P < 0.004 (n = 7) C). In cellula analysis of the editing efficiency of WT CAS9 and HiFi-CAS9 RNP’s complexes: two different sgRNA’s. While WT CAS9 showed similar editing abilities with both sgRNA’s, HiFi CAS9 showed significantly higher editing efficiency with RAG2. * p < 0.004 (n = 3).

kcat and km were calculated for all 4 combinations (Suppl. Table S1) by fitting the results to the Michaelis Menten model. Both constants were successfully determined using this assay. The enzymatic efficiency (E.E.) could be calculated and the cleavage efficiency of the CAS9-RNP was expressed by the value of kcat/km. As presented in Fig. 4A, the cleavage efficiency of both sgRNAs was very similar to that of WT CAS9. However, for HiFi CAS9, the efficiency was higher for RAG2 compared to RAG1. This may be related to the lower DNA binding affinity of HiFi CAS9 that in turn highlights the differences between sgRNA effects on the CAS9 activity, as considered above. No marked difference was found between the velocity curves of both sgRNAs with the WT CAS9-RNP complexes.

The results were further supported by the TIRF microscopy analysis of the cleavage assay (Fig. 4B), showing that the assay is sensitive enough to detect CAS9 cleavage and that the Proteinase K is necessary to detect CAS9 specific signal loss. The results were further supported by the experiment shown by [45], in CD34+ hematopoietic stem and progenitor cells (HSPCs), additional comparative analysis of previously published results [45] (Fig. 4C) showing a similar editing efficiency for WT CAS9 with both sgRNAs and significantly higher efficiency (p < 0.004) for the CAS9-RAG2 complex vs. CAS9-RAG1 complex.

Interestingly, the maximal velocity of DNA cleavage by CAS9 WT (Vmax) was more rapid for RAG2 RNP than for RAG1. This is since the Vmax value alone indicates the maximal velocity of the enzyme for a specific concentration of the enzyme at its saturated state, not taking into account the substrate concentration at that point. Thus, if an enzyme has a high Vmax and also a high km value, its efficiency is lower than if its km is low. A high Vmax with a low km indicates that the enzyme reaches its maximal turnover velocity at low substrate concentrations, making it highly efficient. When using HiFi CAS9, the enzymatic efficiency as well as Vmax levels favor RAG2 over RAG1.

TIRFm analysis of microfluidic cleavage assay

The TIRFm cleavage assay was used to show that the experimental setup is on a par with previous studies that have used TIRFm for kinetics evaluation of CAS9 binding and activity. First, ATTO 550 labeled two-part system CAS9 RAG2 RNP was immobilized to the surface under the MITOMI valve, followed by introduction of Cy5 labeled DNA around the valve. Subsequently, the levels of Cy5 that accumulated beneath the MITOMI valve were measured. The results (Fig. 4B, left) showed that, as previously described [14], the DNA bound and was not released after cleavage. However, following incubation with Proteinase K it was possible to detect and measure the DNA cleavage by signal loss after removal of the RNP complex by protease degradation (Fig. 4B, right). The results showed almost full degradation of the RNP and over 20 % loss of the DNA signal. A substantial amount of background signal loss was also observed, due to the Proteinase K degradation of the surface, as well as Cy5 photobleaching. This was accounted for and did not interfere with detection of the DNA cleavage by the RNP complex.

The higher efficiency of RAG2 over RAG1 was observed in a recent in vitro model [45] (Fig. 4C) showing that HiFi CAS9-RAG2 has more efficient editing abilities in CD34+ HSPCs than HiFi CAS9-RAG1 (88 % vs. 60 %). Moreover, this complex also showed reduced off-target editing frequencies. However, this effect was not observed when WT CAS9 was applied. When comparing WT with HiFi CAS9 for RAG1 sgRNA, cleavage abilities were reduced (83 % vs. 60 %) while off target events were only partially reduced [45]. The comparison between RAG2 and RAG1 in both enzymatic efficiency and cleavage kinetics (Vmax, kcat) presented in Suppl. Table S1B shows that the enzymatic efficiency of RAG2 was less affected by the HiFi CAS9 (decreased by 7.5 fold vs. 13 fold for RAG1).

Conclusions

A unique microfluidic based assay for the rapid, comprehensive characterization of CRISPR CAS9-RNP enzyme kinetics has been introduced. The DNA binding was characterized of four CAS9 complexes: CAS9 WT with RAG1 or RAG2 sgRNA’s and HiFi CAS9 with both sgRNAs. The measurements included kd, off-rate and enzymatic efficiency of the four complexes.

The microfluidic platform - EnzyMIF - allowed simultaneous running of all four experimental setups for the kd and off rate experiments while the cleavage assay was tested on a separate second device as it demanded a different experimental setup. The results were in correlation with the in vitro experiments, showing that the sgRNA has a significant impact on the enzymatic efficiency of CAS9 DNA editing abilities. The assay focused on testing the RNP effect in the case of full DNA hybridization (matching DNA sequences), eliminating other mismatch factors.

A remaining open question is whether the monitoring of RNP kinetics of the on-target sgRNA-DNA hybridization and cleavage (as herein) can serve as a predictor of the enzyme efficiency and specificity. To answer this question, it is proposed to screen and analyze a large set of validated highly efficient vs. low efficiency sgRNAs. EnzyMIF holds great potential for biochemical profiling of guide-RNAs or RNA guided nucleases as a tool to improve genomic editing efficiency.

Author contributions

DP-C, experimental design, performed experiments, data analysis, manuscript writing; GS, performed TIRFm experiments; LB, Microfluidic device optimization; AI, performed TIRFm experiments; OI, in cellula data, manuscript editing; EBM, experimental design, data analysis, manuscript writing; AH, experimental design, data analysis, manuscript writing; DG, experimental design, data analysis, manuscript writing.

Consent for publication

All authors provide their consent for publication.

Funding

Kamin (grant no 65359 and 61962) by Israel innovation authorities (2018) and European Research Council (ERC) under the European Union Horizon 2020 research and innovation program (Grant no 755758).

Data Availability

No data was used for the research described in the article.

Data will be made available on request.

Declaration of Competing Interest

The authors report no declarations of interest.