Engineering Success

Introduction

Our journey began aiming to apply the engineering principles to every aspect of our research project. The engineering steps of Design, Build, Test and Learn helped us to evaluate, solve and finally optimize every arising problem we faced. These principles helped us to find new ways to approach troubleshooting for the occurring issues. Generally, we understood that the engineering process can be applied to many aspects of a multidisciplinary research project.

Cycle 1

Cloning GG to 3A for final cr-cas plasmid

It is well established that in vivo crispR/cas systems provide immunity against viruses and plasmids in bacteria and archaea (Rath et al., 2015). The silencing of foreign nucleic acids is achieved through a ribonucleoprotein complex of Cas molecule preloaded with crRNA molecules that act as guides for targeting the degradation of the invading nucleic acid (Rath et al., 2015). Once this duplex is necessary for the system to be active, we decided to compose a Cas13a-crRNA plasmid for each crRNA molecule in order to produce efficiently and promptly the ribonucleoprotein complex. It is worth noting that RNA molecules in vitro easily misfold and are usually unstable tending to be degraded in presence of RNases or get trapped in inactive conformations if not preserved in the proper way (Flores and Ataide, 2018). Thus, ribonucleoprotein duplexes protect and stabilize the RNA molecules reducing the risk of degradation. Considering the advantages of the ribonucleoprotein complex we decided to synthesize the cloned Cas13a-crRNA plasmid.

Design stage

In order to synthesize the final Cas13a-crRNA plasmid some factors had to be considered. More specifically, after the sequences of each crRNA as well as Cas13a were amplified and their amplification was successfully confirmed, the most efficient method had to be chosen to integrate the inserts into the pSB1C3 backbone vector. According to bibliography one of the most commonly used assembly methods is Golden Gate assembly. This method is based on an all-in-one mix reaction consisting of a type IIS restriction enzyme and T4 DNA ligase for efficient assembling of the sequences, reducing handling time and increasing cloning efficiency. The DNA sequences get digested via the restriction enzyme and are ligated by the DNA ligase. This process is repeated in cycles, and the elements are forming the desired products since the properly assembled parts lack the restriction sites that exist at the unincorporated DNA parts. Taking into account the advantages of this method, we chose Golden Gate assembly not only for the composition of this specific plasmid but generally as a cloning method for our project in order to achieve the least possible waste of time and the maximum cloning efficiency.

Build stage

We followed a modified protocol of Golden gate called Seva Brick based Golden Gate assembly developed by Damalas Stamatios et. al. 2020 as described thoroughly in the experiments More significantly, the reaction consists of the SevaBrick Assembly mix (BsaI, T4 ligase, T4 ligase. buffer, DpnI, BSA) and the entry parts as well as the vector backbone. The DNA amounts of the final reaction were calculated via an excel file, introduced by Damalas Stamatios et. al. 2020. The final mix underwent the following thermocycling conditions: 37 °C, 20 min; (16 °C, 4 min; 37 °C, 3 min) x 30; 50 °C, 10 min; 80 °C, 10 min.

After the end of the reaction, the samples were transformed into chemically competent E. coli DH5a cells using heat shock transformation. Afterwards, the cells were cultured in LB antibiotic plates at 37 °C overnight.

To confirm if the cloning was successful, samples from the colonies underwent colony PCR and the results were analyzed through an agarose gel.

Test stage

Several tests were performed to achieve the cloning of the two entry parts with the backbone vector. However, its synthesis was inefficient. In order to understand the reasons our protocol was not effective, we tried to improve several factors. More specifically, we firstly checked the initial design of the plasmid in the SnapGene software. Consequently, and as long as we confirmed that the design was correct, we resynthesized each part of the reaction to confirm that the proper sequences were used. Then, we checked the reaction conditions and the enzymes' functionality to confirm that they are working normally. However, the synthesis of the cloned Cas13a-crRNA plasmid was not achieved (Figure 2).

Learning stage

Since Golden Gate assembly was inefficient, we sought in the literature to determine the possible reasons why the final plasmid was formed incorrectly and to find solutions. Based on the results, we observed that the high size difference of the parts was probably the main cause of the experimental failure. While searching for a more functional cloning method for this particular combination of inserts, we found that iGEM also suggests 3A assembly as a cloning method, which has been extensively used to ligate molecules with a difference in sequence size and was firstly described by Shetty et al., 2011. More significantly, 3A assembly stands for 3 Antibiotic assembly and is based on the integration of two parts in a backbone vector. The selection of the correct assembly products is achieved through antibiotics. 3A assembly exploits the negative and positive selection in order to minimize the number of possible incorrect assemblies. The inserts of the reaction are perceived in a) upstream part plasmid, b) downstream part plasmid and c) the destination vector. Each plasmid has a prefix and a suffix region where the restriction sites are. The upstream part is digested with EcoRI and SpeI, the downstream part is digested with XbaI and PstI and the vector with EcoRI and PstI. The plasmids are then mixed to synthesize the final plasmid (Figure 3). The final sample is then transformed into competent cells. The cells are cultured in LB plates in 37 °C overnight. According to Shetty et al., 2011 this method can be used to assemble parts ranging in length from 12bp to 3-4kb. Thus, it could be a more effective method for the synthesis of cloned Cas13a-crRNA.

Verily, we applied 3A assembly method to create cloned Cas13a-crRNA plasmid and we observed that our experiment was crowned with success (Figure 4). In our experiment the upstream part was Cas13a (5422bp) and the downstream part was each crRNA molecule (102-107bp). The destination vector was the pSB1C3 backbone vector (2029bp). According to the 3A assembly the vector must have a different antibiotic resistance than the two input plasmids. However, in our case, the vector had the same antibiotic as the input parts but the result was still successful. Thus, we can safely support that 3A assembly is an effective method for the integration of inserts with high difference in size.

Cycle 2

CRISPR/Cas13a detection assay

After successive engineering cycles and troubleshooting processes we finally managed to produce the system's basic biological components in quantities sufficient for the final DIAS detection assay. The LbuCas13a was expressed from the cloned final T7-SUMO-LbuCas13 (BBa_K4170022) plasmid and purified utilizing a standardized methodology developed by our lab. In addition, the crRNA DNA sequences were successfully cloned into pSB1C3 plasmid (BBa_K4170023 - BBa_K4170023) employing the methodology described at our custom crPrep crRNA preparation kit and produced in satisfactory quantities via in vitro transcription. The next important step towards the implementation of our project was to verify the functionality of the LbuCas13a/crRNA duplex in recognizing the target miRNA and subsequently activating the trans-cleavage activity of the enzyme providing a green fluorescent signal.

Design stage

To develop a detection system with increased sensitivity and broad clinical applicability a lot of factors affecting the CRISPR/Cas13a enzyme reaction need to be investigated. Thus multiple experiments were conducted to estimate the suitable LbuCas13a, crRNA and reporter concentrations to achieve the optimal conditions for the detection assay and better understand the kinetics of the enzyme reactions. The system's detection method is based on a detection standard curve; a bioanalytical method which represents a linear relationship between concentration of an analyte (independent variable-miRNA added) and response (dependent variable-fluorescence intensity). Finally, this relationship is used to predict the unknown concentration of the analyte in a complex matrix such as the blood or serum. To achieve the detection system's maximum effectiveness we had to investigate factors affecting the enzyme reaction such as the appropriate reporter concentration and the specific time point since reaction initiation that the fluorescence signal should be obtained.

Build stage

For the investigation of factors that affect the kinetics of the reaction such as the optimal reporter concentration we designed a simplified detection protocol. The basic steps of the protocol are the preparation of the CC reaction mixture along with the different RS-P and RS-N reaction mixtures as described in detail below (figure 5).

RS reaction mixtures preparation

The RS-P positive reaction mixtures contained a predefined concentration of target miRNA (1nM) in 1X reaction buffer. However, different concentrations of Reporter RNA (1μM, 0.75 μΜ, 0.5μΜ, 0.25μΜ, 01μM) were added in each reaction mixture to investigate the optimal reporter concentration for the enzyme reaction. In addition, the RS-N negative reaction mixtures contained the same RNA reporter concentration as for the RS positive reaction mixtures without the addition of the target miRNA.

• After the addition of the necessary reagents the RS-P and RS-N reaction mixtures remained in ice until use.

CC reaction mixture preparation

The CC final reaction mixture contains 20nM LbuCas13a, 20nM crRNA in 1X reaction buffer.

• The CC reaction mixture after its preparation, remained on ice until further use.
• Before its addition in the microplate, the mixture was left at 37 °C for 10 min without agitation to enable the efficient association of Cas13a enzymes with the crRNAs in the reaction mixture.

Microplate preparation for fluorescence experiments

The CC final reaction mixture contains 20nM LbuCas13a, 20nM crRNA in 1X reaction buffer. The CC reaction mix contains the reagents that are listed in table X on the experiments page in a final volume of 20μl.

• After CC reaction mixture incubation at 37 °C and RS reaction mixtures preparation, a specific volume of the CC mixture (20μl) was added to each RS mixture generating the final detection master mixtures which were gently flicked. Afterwards, 50μl of the final detection master mixtures were added to each corresponding well of the microplate.
• The microplate was incubated for 60 min in a fluorescence plate reader with fluorescence measurements (FAM channel, \( \lambda_{ex} \) 494 nm, \( \lambda_{em} \) 518 nm) taken every 1 min.

Test stage

From the real time kinetic measurement of Cas13a reactions incubated with different reporter concentrations and a predefined miRNA concentration (1nM), we observed a conservative pattern on how the fluorescence intensity changed over time. During the first minutes of the reaction the emitted fluorescence from the RNA reporter increases linearly until the reaction reaches the plateau phase when the fluorescence intensity starts to decrease. On the contrary, the negative control samples exhibit constant background fluorescence at all time points. But, are these the only conclusions that can be drawn from the fluorescence curves?

To construct the following standard curves which plot the miRNA concentration (standards) versus time, the fluorescence capture point should be in the linear part of the curve before the plateau phase. Inspecting thoroughly the graphs (Figure 6), we can easily understand that the samples loaded first at the microplate (1μM and 0.75μΜ) show a limited linear section in the curve which corresponds to the first 1-5 minutes since reaction initiation. On the other hand, the samples loaded last at the microplate (0.25μΜ and 0.1 μΜ) show a wider linear section on the curve that corresponds to the first 1-12 minutes since reaction initiation. This limited linear section of the curve may hinder the practical implementation of the detection system, due to variations in the detected fluorescence caused by the fast enzyme reaction speed and fluctuations in researcher's handling time required for reaction preparation.

1.00μM

0.75μM

0.50μM

0.25μM

0.10μM

Learning stage

Taking into account the above, we decided to investigate solutions to limit the effect of researcher's handling time on assay results. To achieve this, we conducted all the following enzyme reactions on ice, keeping the microplate's temperature at 4°C until its insertion on the fluorescence plate reader. The low temperature prevents reaction initiation until the microplate is inserted in the plate reader which keeps a 37 °C constant temperature. In addition, we removed the 10 min pre-incubation step of the CC reaction mixture at 37 °C. We concluded that the LbuCas13a enzyme can be efficiently associated with the crRNA directly inside the microplate well without the pre-incubation step.

In summary on the following experiments we made the following experimental modifications:

• All reagents, the reaction tubes and the microplate should be kept on ice until the fluorescence measurement.
• The pre-incubation of LbuCas13a enzyme with the crRNA was skipped, to limit the effect of handling time on assay results.

Cycle 3

Microfluidic device manufacturing

One of the directions of our engineering cycles was the process of manufacturing our microfluidic device. This aspect of the project is very crucial since utilizing the technology of microfluidics, we can achieve droplet production resulting into increase of LOD (Level of Detection), (Kaminski et al., 2021).

Design stage

The Design stage of this engineering cycle involves the general design of the microfluidic device and the process of device manufacturing. Since we aim to generate droplets, the microfluidic device should exhibit droplet generation behavior. The chosen design was a flow-focusing T-junction device because it is reported to generate droplet with high mono-dispersity and the diameter of the droplets are a function of the flow conditions and channel dimensions (Ibrahim et al., 2021). The channel geometry and the general device are demonstrated in Figure 8 and 9 respectively. The design is selected to have 3 inlets and 1 outlet with 2 oil inlets and 1 water inlet since the produced droplets must Water in Oil emulsions. The channel depth is designed to be 500μm and the width to be 400,500 and 700 for the oil, water inlets channels and outlet channels respectively. The selected geometry was based on Computational Fluid Dynamics since it exhibited appropriate segmented flow.

The manufacturing process chosen was the Fused Deposition Modelling since it is reported to be able to produce functional microfluidic devices (Nelson et al., 2019). Fused Deposition Modelling is a 3D printing technique based on the deposition of melted thermoplastic material layer after layer until the 3D device is finished. The material used in the FDM process is a filament of polymeric material that is softened and melted with the help of heat and then extruded. The polymer chosen for this type of device was a regular Polylactic Acid (PLA) with a black color filament and a transparent Polylactic Acid filament. The regular PLA exhibits more suitable rheological behavior for FDM technique in comparison with the transparent one. Some of the advantages of the FDM process are the low cost of the materials, the wide range of slicing software and the simple machinery design that makes it suitable for beginners (Kristiawan et al., 2021).

Build stage

After the general design of the microfluidic device and the selection of the 3D printing technique, the next step is to select the appropriate parameters for the manufacturing of the device. The geometry of the design was selected to be suitable for printing. One of the main disadvantages of the FDM technique is the accuracy limitation.

Having that in mind, we decided to use 2 types of nozzles for the material extrusion. The first was 0.4 mm and the second 0.2 mm. The smaller the nozzle diameter, the higher the accuracy and also higher the total time of printing will be. A general rule for selecting the appropriate nozzle is to understand that the nozzle diameter is affecting the XY resolution, and the layer height is affecting the resolution on the Z axis. Since the smaller channel width was 400 μm, hypothetically the 0.4 mm = 400μm nozzle would be fine.

The next step was the selection of the Slicing software for the device and the parameters of the printing. The slicing software used was the Ultimaker Cura. The parameters we used were chosen based on the literature (Nelson et al., 2019 and Zhang et al., 2019).

Test stage

The final stage of the process is the manufacturing of the device and the test of the microfluidic behavior. The manufacturing stage was a challenging aspect. After printing several devices with slightly different parameters to achieve the desired result, the overall outcome was the inability of general liquid flow. The printing process can overall divide into 3 main parts. The first part includes the printing of all the layers beneath the channels that were printed with accuracy. The next part was the layers of the channel geometry. The channel geometry, during the printing with the 0.4 mm nozzle, resulted with the unwanted material inside since the extruder did not retract the material in time. The outcome was that not even one device could perform fluidic behavior. The final part of the manufacturing was all the layers above the channels. The problem in this stage is that if the extruded material is not solidified right after extrusion, unwanted material can again result in the channels. This problem could not be solved by just changing the nozzle.

A second approach to the problem was to split the model and print the first and the second part together and the third alone. In this way, we could evade the unwanted material inside the channels during the printing process. After printing the two components, we tried to assemble them together to test the whole device. The method involved heating the first layer of the upper piece and then putting it on the first piece. Unfortunately, upon testing it with a pumping system for the droplet generation, it showed major leakage issues resulting in failure.

Learning Stage

Facing all those problems with the FDM technique, we tried different approaches to manufacturing a functional device that will exhibit appropriate properties for our project. We understood that even changing the printing parameters and the initial 3D models, we had to face the FDM limitations such as surface roughness and leakage issues (Zhang et al., 2019).

So, we decided to move in another direction. The direction we followed was to change the printing technique to Digital Light Projection (DLP). The process is based on the conversion of liquid to solid, using computer spatially controlled photopolymerization to create solid objects from a vat of liquid resins under light irradiation. DLP offers lower printing times and far better surface roughness, resulting in more accurate manufacturing of the channels (Quan et al., 2020). The materials used for this printing technique are called photopolymers or resins and the materials that we used were a standard silver resin and a transparent yellow resin.

The process of manufacturing the device was like the FDM approach. We printed 2 different parts of the microfluidic device, first to check that the channels would be open and second for better control of the unwanted material inside in the channel geometry.

The surface of the printing device with the DLP technique is far greater. For the FDM device, the channels are rough and can affect the droplet generation. The DLP channels, on the other hand, exhibit a smoother surface and the channels are appropriate for liquid flow. Having a lot of issues trying to assemble them correctly to perform the ideal behavior we tried to print the whole model directly to save time and effort. Unfortunately, not all the channels were printed clear, without undesired resin, meaning that the post-processing was a major drawback.

Finally, after printing the same device under different conditions with undesired results, we observed that nearly all the channels were cleared of the resin, but having 3 inlets and 1 outlet, always one of them would be blocked. Having that in mind, we tried to modify the model of the microfluidic device to involve only 2 inlets, 1 water and 1 oil inlet. Having only 2 inlets made the post-processing successful, and we achieved droplet generation behavior. The selected flow conditions based on the model were 40 μL/min for oil (continuous phase) and 8 μL/min for water (dispersed phase).