The main goal of the wet lab team is to test and validate, in vitro, the AI-generated plasmid parts the dry lab team creates. Validating each individual plasmid part allows us to determine with more granularity which specific AI-plasmid components are functional.

To achieve this goal, we utilize a pipeline in which we assemble the AI-parts into a known plasmid backbone using Golden Gate Assembly (GGA – a high-throughput transgene insertion method), and test the viability of the AI-parts using bacterial growth assays. Finally, we sequence the new working plasmid parts to confirm successful propagation of the AI-parts, such as the origin of vegetative replication (oriV) and antibiotic resistance genes (ARGs). These developments provide a proof-of-concept for the tractability of synthesizing plasmids de novo using dry lab’s machine learning framework, which sets the path forward to ultimately create fine-tuned plasmids to tackle antibiotic resistance.

The wet lab pipeline for validating various AI-plasmid parts, in particular the oriV and ARGs, is described in detail in the following sections. Thus far, the wet lab team has synthesized and tested novel oriVs generated by dry lab, of which samples 741, 9371, 5276, 5727, and 4089 have been determined as functional.

The main AI-plasmid part the wet lab team tested was the oriV. In general, the AI-generated oriVs consisted of various control elements (such as RNAI-, RNAII-, rep- and rop-coding regions) and non-coding regions. As such, to test this part, the wet lab team cloned the dry lab oriVs into suitable backbones of which have similar control elements for their oriVs. This ensured compatibility between the dry lab oriV and backbone.

A backbone conferring resistance to a specific antibiotic serves as an effective positive and negative control for growth assays that are essential for ascertaining dry lab oriV functionality. Such a backbone eliminates confounding variables that may skew the interpretation of experimental results by observing E. coli growth as well as number of colonies to determine if the backbone was successfully annealed to the ori, and successfully uptaken into the E. coli.

Initially, the plasmid pSB1C3 was used as a backbone. pSB1C3 is a common plasmid backbone from E. coli which confers chloramphenicol resistance. Thus, pSB1C3 was to be used as a positive control for the experiments, since it was predicted that it would successfully grow on the plates. pSB1C3 was miniprepped and underwent PCR to increase the amount of this plasmid, transformed into E. coli, which was then plated. However, no growth was observed from the transformed E. coli on the chloramphenicol plates.

After we repeated this experiment a number of times, the team decided to have the plasmid sequenced to check if the structure was correct. Sequencing results revealed that the backbone contained additional BsaI sites than initially described, meaning that it was being cut in additional locations, meaning the ori and the backbone would fail to be annealed together.

This meant that afterward, a different positive control had to be used for the experiments. Interlab_NegC was used, however, there it exhibited a similar problem as pSB1C3. After no growth was observed again, Interlab_NegC was taken for sequencing and found to have an extra BsaI site, which caused the backbone to be cleaved during GGA as our GGA workflow utilized the exact same restriction sites.

Finally, we found that eforRed/eforCP Biobrick K592012, an ampicillin (Amp) backbone was successful as a positive control backbone, indicated by numerous colonies on Amp agar plates. This led us to use this backbone for future experiments. eforRed/eforCP Biobrick K592012 had a length of 2807 bp, and contained a 861 bp long bla gene, coding for beta-lactamase and thus conferring Amp antibiotic resistance.

Positive Control
The positive control served to validate the wet lab team’s experimental approach of cloning dry lab oriVs into functional backbones to test their functionality. Here, this control is carried out by cloning a known working oriV (pJump27) into eforRed/eforCP Biobrick K592012 using GGA. This would theoretically create a construct that confers Amp resistance provided that the cloned oriV is functional.

As such, this construct is then transformed into competent E. coli K12 thawed on ice. The culture was then plated on Amp plates, incubated at 37°C overnight, and then observed for growth of colonies the following morning (which would indicate successful oriV functionality).

Once we verified that this assembled plasmid was functional by plating transformed bacteria, we began our assemblies using the Dry Lab-generated sequences.

Negative Control
The negative controls are made by transforming competent E. coli cells with a functional plasmid lacking an oriV sequence. The negative control contains only a linear plasmid backbone without an ori, since it has not been cloned and annealed to anything. When assembled and plated onto Amp plates, these cells will theoretically die as they have no oriV to propagate any Amp-resistance cassettes.

The negative control is used to ensure the assembly’s procedure was correct, as E. coli without the ori to replicate the plasmid would not survive on the antibiotic-plated media. For our GGA using the Interlab plasmid, eforRed/eforCP Biobrick K592012 with water was used as a negative control.

LB Amp agar plates
Before beginning transformations, it is important to prepare nutrient rich plates with the correct antibiotic conditions to ensure only E. coli DH5ɑ that have successfully taken up the designed plasmid with antibiotic resistance (i.e. successfully transformed) can grow. DH5ɑ competent E. coli cells were used for the transformation. LB agar miller was dissolved in milliQ water, then autoclaved using the wet cycle. Once cooled, Amp was added to the agar at a concentration of 100 µg/mL. The agar was then poured into sterile petri dishes and allowed to solidify at room temperature, with aseptic technique applied throughout. The plates were then stored at 4°C and warmed up to 37°C before the transformed E. coli cells were plated. Prior to plating of transformed E. colicoli, the plates were warmed up to room temperature to facilitate their growth.

Competent Cells
In our experiment, we used E coli DH5α cells due to their high efficiency in transformation and stable amplification [1]. These cells were stored at -80°C and subsequently thawed on ice before use. To initiate the transformation, we added the assembled plasmid, containing the oriV drylab generated sequence with the eforCP backbone, to the thawed cells and incubated the mixture. We then performed a heat shock by incubating the cells at 42°C for exactly 30 seconds to facilitate the uptake of the plasmid DNA. Following heat shock, we added SOC medium to the cells and then incubated the mixture at 37°C to ensure sufficient recovery. Finally, we plated the transformed cells on agar plates containing Amp to select for successful transformants.

GGA is a one-pot molecular cloning technique that allows for the combination of multiple inserts into a backbone without the need for multiple restriction enzymes [2](Figure 1). We used GGA to assemble all our plasmids for testing of oriV sequences and ARGs.

Figure 1. Golden Gate Assembly uses type IIS restriction enzymes, which have separate recognition (yellow) and cut sites (green, purple, light pink). This allows for the creation of many possible unique overhangs without the need for multiple restriction enzymes. GGA can then proceed in a one-pot reaction, allowing for a backbone and multiple inserts to be combined simultaneously. Figure created with BioRender.com.

We developed a molecular cloning (MoClo) system for our oriV and ARG assemblies (Figure 2). Inserts for both experiments were assembled into backbones with a selection marker, the ARG backbone also containing an oriV sequence (Figures 4-5). Dry lab generated oriV sequences were either well aligned to the oriV control element RNAII or the noncoding and Rop sequence, so 5 different assemblies were used to test their functionality (Figure 3).

Figure 2. OriV and ARG MoClo system overhangs and their assigned letters. Overhang locations in assemblies are indicated in Figures 3 and 4. Figure created with BioRender.com.

Figure 3. Type B oriV and sequences used to test generated oriV alignments. Dry lab has been generating type B oriV sequences, which contain sequences for RNAI, RNAII, Rop, and a noncoding region. Generated oriV sequences have been aligning best to either RNAII or the noncoding region and Rop sequence, so we made assemblies to test these generated sequences by themselves and with the sequence that they are best aligned to from a wild type (WT) oriV sequence. Figure created with BioRender.com.

Figure 4. The assemblies used to test generated oriV functionality, letters indicating the locations of specific MoClo overhangs. Each generated oriV sequence was tested by itself and with a portion of a WT oriV containing the region of the oriV sequence that the generated sequence wasn’t best aligned to. The entire WT oriV was also used as a control. Amp was chosen as the selection marker, and bla was used for the backbone’s ARG. Figure created with BioRender.com.

Figure 5. The assembly used to test generated ARG functionality, letters indicating the locations of specific MoClo overhangs. A selection marker was included to allow for selection if the generated ARG was nonfunctional. Figure created with BioRender.com.

After finishing the assemblies, we transformed our backbone eforRed/eforCP Biobrick K592012 with different dry lab generated oriV sequences. We use competent E. coli DH5a cells to transform the plasmids in. We add 2 uL (1 pg – 100 ng) of our vectors to thawed competent cells. We then incubate on ice and heat shock the cells, which is followed by cell recovery using SOC media for 1 hour. Afterwards, we plate the liquid cell culture on Agar plates with corresponding antibiotics.

To test the function of the ARGs, we will also transform the assembled plasmids into competent E. coli DH5a cells. We will be using double antibiotic plates, which includes CmR, our antibiotic selection marker, and Amp.

We began with validating our positive control experiment using the pSB1C3 device and the oriV from the pJUMP27-1A plasmid from the 2024 iGEM Distribution Kit Plate. However, this vector was found to have unreported BsaI cut sites. Therefore, we switched to the Interlab_NegC device, but that was also discovered to also have unreported BsaI cut sites. Finally, we used the eforRed_eforCP K592012 device and obtained visible colonies and a valid sequencing result, thus verifying our positive control.

Next, we began assembling the AI-generated sequences generated by our dry lab team into the eforRed/eforCP K592012 device (BBa_J435099). The first batch of sequences generated by dry lab were not found to be functional as the transformed bacteria did not grow into visible colonies. 6 generated oriVs from the second batch of sequences generated visible colonies, and 5 of them were confirmed to be functional via whole plasmid sequencing. Testing for the third batch of sequences, which includes sequences encoding for ARGs and oriVs, is currently in progress.

Controls

We first began by assembling and validating our positive and negative controls. After switching over to the eforRed_eforCP vector, we amplified the positive control pJUMP27 insert and verified via agarose gel electrophoresis its correct size of 1.6 kb before purifying, transforming, and plating.

Batch 1

Unfortunately, numerous attempts at transforming the first batch of AI-generated sequences did not produce viable colonies on the plates. Since our positive control produced colonies, we concluded that this first batch was nonfunctional.

The wet lab pipeline for validating various AI-plasmid parts, in particular the oriV and antibiotic resistance genes (ARGs), is described in detail in the following sections. Thus far, the wet lab team has synthesized and tested novel oriVs generated by dry lab, of which samples 741, 9371, 5276, 5727, and 4089 have been determined as functional.

Batch 2

After completing the required workflow to prepare the inserts, we introduced our second batch of plasmids into E. coli for uptake. This batch included eight new AI-generated oriV sequences, six of which produced viable colonies on the plates. This result aligned with our expectations, as Batch 2 showed improved alignment with the oriV control element RNAII.

We also tested the wild-type oriV which the sequences aligned to, pColE, and saw viable colonies, which provides a reliable point of comparison for the AI-generated oriVs. The positive control, pJUMP27, also produced viable colonies, indicating that oriVs are indeed modular. Upon cloning this batch, we found that sequences with the IDs 741, 4089, 5276, 5727, 6769, and 9371 produced visible colonies. Upon sequencing, we confirmed that the sequences with the ID 741, 4089, 5276, 5727, 9371, and the wild-type pColE1 oriV were indeed functional. Unexpectedly, sample 6769 from this batch lacked a promoter but colonies were still observed. However, the sequencing data for 6769 was unreadable, so we cannot draw any conclusions about its functionality. The observed colonies might very well be due to contamination or high background levels. Upon reviewing our designs, we realized that samples lacking a promoter (like 6769) could theoretically produce functional plasmids if driven by the promoter of the ARG. We had assembled the resistance gene and oriV in the same orientation, which could allow the ARG's promoter to inadvertently drive the expression of oriV. To address this issue, we plan to assemble the parts in our next batch such that they run in opposite directions, which is currently ongoing.

Positive Control

The oriV from the pJUMP27-1A (BBa_J428326) vector was cloned into the eforRed/eforCP (BBa_J435099) backbone and transformed into E. coli DH5a, which generated visible colonies. This functioned as a positive control and confirmed that oriVs are modular.

5 Functional Sequences

After testing batch 1 of the AI-generated oriVs, we found no functional sequences as no visible colonies were seen. Upon testing batch 2 sequences, we found 5 functional sequences, summarized below:

Batch 2 Sequence Testing Result Summary

The pColE1 wild-type oriV was also confirmed to be functional. The following is a whole plasmid sequencing result showing 100% alignment between the eforRed/eforCP + 9371 sequence in silico assembly and the isolated plasmid:

These results demonstrate our proof of concept, indicating that the AI model generated functional oriVs which allow the plasmids to replicate in vivo. Currently, we are testing batch 3 of the AI-generated sequences, which includes both oriVs and ARGs.

All experiments were carried out using the Invitrogen One Shot™ MAX Efficiency™ DH5α-T1R Competent Cells.

Table 1. Vectors and synthesized DNA used for the cloning process

Table 2. Primers used for cloning process

For the safety of our team members, we ensure all experiments are performed in functional, appropriate, isolated and properly organized facilities.

All our experiments are performed in a biosafety level 1 facility, with appropriate Personal Protective Equipment open to access. All our laboratory facilities are labelled and documented to the drawer, with proper labeling and segregation mechanisms in place.

All wetlab members are required to undergo rigorous safety training. This includes:

1. A 6 hour online module specifying all operational risks, safety, emergency situations, and proper lab attire during lab work in Biozone.
2. A lab tour with the BIozone lab manager to ensure knowledge about emergency equipment.
3. Equipment training with the lab technician to ensure operational knowledge on standard operating equipment.

Human Health/Safety Factors - Lab Members :
iGEM Toronto prides itself in involving both experienced and new researchers into collaborating to create impactful and meaningful synthetic biology projects. As a group of undergraduates, some of our members do not have prior wet lab experience. Therefore, proper training, lab safety, abiding to EHS regulations and sufficient understanding of experimental procedures must be implemented before experiments are allowed to proceed. Furthermore, biologicals and chemicals are invaluable resources. Proper coordination and inventorization measures must take place for a large group of students to seamlessly carry on from previous experiments without generating extra waste.

We have been in touch with Endang Susilawati – the lab manager for our lab space and a member of the biosafety council from the University of Toronto’s Department of Chemical Engineering. WHMIS guidelines and the University of Toronto’s Health and Safety committee govern our work.

The list of safety measures taken to ensure lab member’s individual safety as well as proper training are attached in the table below:

In this project, wetlab validates the dry lab sequences generated. Novel oriVs should not confer hazardous functionality to E. coli DH5a cells.

The main risk of our technology is potential misuse of our language model and the uncontrolled development of our language model generating unforeseen and potentially hazardous sequences. Currently in our experimental procedures, We define risk as a combination of vulnerability, magnitude of hazard and exposure, and our hardware team has been working on methods to reduce vulnerability and exposure.

Vulnerability
Wetlab ensures all members are dressed in appropriate personal protection equipment (PPE), and all reagents are contained in the lab. Handwashing procedures are mandated, and sick members are not allowed into the lab to reduce vulnerability.

Exposure
Our hardware team is working on a colony counter and RFID chip to automate colony counting and allow better inventorization of chemicals. In our workflows, validation of plasmids are performed through observing and counting colonies. An automated method of validation reduces the need of human exposure to bacteria containing novel sequences. The RFID chip also reduces human exposure inventorization of chemicals and provides a thorough tracking mechanism - allowing easier root cause identification and chemical location should errors occur. These processes reduce exposure, preventing potential containment breaches and consequently, risk.