CRISPR/Cas9 Demonstration Kit

Welcome to development hell

In 2017 I thought it would be fun to try develop my own demonstration experiment that anybody could use in order to show CRISPR/Cas9 genetic modification in action. The concept was based on this kit from The Odin, with the major difference being that the modification would make the bacteria glow under UV light. This is the story of a project unfinished, lost in the depths of development hell. May you learn from my mistakes.

The concept

CRISPR/Cas9 is the flashiest new genetic modification tool in town and every year it is upgraded to be more efficient and versatile. Unlike older in-vitro tools such as restriction enzyme digest/ligation or site-directed mutagenesis, the CRISPR/Cas9 system is able to operate within a living cell without killing it… well at least without killing all of the cells you target. It isn’t the first in-vivo tool, we’ve had ZFNs and TALENs, but both of these have been prohibitively expensive due to the bespoke manufacturing requirements. The CRISPR/Cas9 system is easily recoded for new purposes and can even be constructed inside the cell you wish to modify, through the delivery of a plasmid containing the genetic information for the system and the target.

How does CRISPR/Cas9 work?

The gene for a protein named “Cas9” is encoded alongside the genetic information to form a two part RNA complex made up of "tracrRNA” and “crRNA” (a). When combined this is known as the “gRNA” or guide-RNA. Cas9 is an enzyme that induces double stranded breaks in DNA, it has no repair function. Cas9 cuts DNA that is perfectly matched to a 20 base pair region contained within the gRNA (b). By modifying this specific gRNA region, a researcher can target this cutting action of CAS9 to any genetic sequence they wish.

Once the cuts are made however, repair is up to the target cell (c). If a matching strand of DNA is in the local area, the cell will use it as a template for repair, also known as “Homology Directed Repair” or “HDR”. If not, it will fill the break with random base pairs, resulting in “Non-Homologous End-Joining” or “NHEJ”.

This distinction between HDR and NHEJ ended up being incredibly important, since successfully pulling off a reliable gain-of-function experiment (such as making a bacteria glow) is far more difficult than a loss-of-function experiment, (such as stopping it from glowing). However I thought - making something glow, that’s fun science! We definitely need to have a kit that makes E. coli glow! Noone wants to be handed such a powerful tool only to use it to intentionally break stuff. Thus the road to development hell was paved with my hubris.

How would the kit work?

This kit would use the pCAS9/pCRISPR system, which was the only available bacterial CRISPR system available on AddGene. The publication that provided these plasmids warned that simply coding the pCAS9 plasmid would lead to cell toxicity, so they’d created the second plasmid to host the CRISPR array and gRNA, which would then be added at a seperate time to the CAS9. To get this kit to work, I believed I would need to construct at least two, but more likely three plasmids, using the following workflow and then stably express them alongside the pCAS9.

However I made two major mistakes in this design process;

1. I didn’t yet fully comprehend the difficulties of getting three plasmids to stably express themselves inside a single cell line. By choosing a plasmid target, rather than a genomic one, I was going to need to find three non-competitive origins of replication and use three different antibiotic resistance genes for selection.

   • This led to a complexity cascade, as the only possible plasmid that offered both of these together was pRS52. While I had high hopes for this plasmid, it’s incredibly low copy number made working with it a nightmare.

2. I assumed that I could easily transfect a single stranded DNA template alongside my plasmids. Instead I would learn that this was extremely difficult, so I would later decide to try add restriction enzyme sites to the template I paid for, in order to then ligate it into the pCAS9-pRS52 plasmid I was working on, subjecting myself to more work with this low-copy-number origin of replication.

early victories

Before I expand on how I fell flat on my face, let’s talk about how I get up to speed. I wanted to make a single-base pair change in a plasmid, in order to make something glow? That would require the plasmid to be almost perfectly coded to produce the “superfolder Green Fluorescent Protein” except for a single error. Thus before I could design my experiment to modify my target, I’d have to build it!

noGFP/dGFP: The sfGFP-pSB1C3 plasmid, modified via Site-Directed Mutagenesis to have a broken start codon ATG > ATT. Changing the T back to a G would re-enable the glowing gene, however it would require a homology template to enable HDR rather than NHEJ.

Using a Round-the-world PCR and primers that annealed back-to-back near the start codon, I successfully managed to mutate the third base in the sfGFP sequence, disabling the start codon and thus the entire downstream gene. This image shows a pellet of E. coli cells containing the noGFP plasmid (left) and the sfGFP plasmid (right) which are clearly distinguishable even without the UV torch. I sent samples to AGRF for sequencing and confirmed that no other changes had been made other than the G > T.

So far, so good…

mired in complexity

Unfortunately, that’s about the high point of my experience. If you want to believe this story has a happy ending, you can stop reading now. If instead you’d like to learn from my mistakes, carry on!

The problem really stemmed from my choice to use the pCAS9/pCRISPR system on a plasmid target. If I’d chosen a genomic target or integrated noGFP, the experiment would have been far easier - but less likely to pass OGTR approval for sale. If I’d chosen a NHEJ knockout experiment I may have succeeded, but I went for the long shot simply because making things glow is awesome! Both of these early design decisions created a complexity cascade that I would not recover from. Whoops!

One of the most important parts of a plasmid is the ‘origin of replication’. This region is necessary to plasmid replication during reproduction (binary fission) of the bacteria, and doctrine states that certain origins are incompatible with one another. I would learn later that this rule of incompatibility is more flexible than I initially feared, nonetheless I tried to adjust my design to compensate.

Another critical plasmid region that became an issue during design was the antibiotic resistance region. Since this gene is used for selection as well as maintenance of the plasmid, I believed it necessary to use a different resistance gene for each plasmid.

This meant I would need three plasmids, each with a unique and compatible Origin of Replication and Antibiotic Resistance Gene. Enter pRS52.

At first pRS52 seemed like the answer to all my problems. It added an Ampicillin resistance gene into the mix, where I already had Kanamycin and Chloramphenicol resistance in the other two plasmids. It also has the pSC101 origin of replication, which solves the compatibility issue. Unfortunately the low-copy number (only about 10 plasmids per cell, as opposed to 200+) made this a bit of an out of the frying pan, into the fire moment.

So not only was the experiment now 20x more difficult, I now had a workflow that looked something like this…

I would spend about 6 months working with pRS52 and trying to get the damn CAS9 gene to ligate into the tiny fraction of plasmid I could purify, even going to far as to Midiprep 200 ml of cells just to get sufficient yield to work with. Eventually the frustration would outweigh my drive and I’d accept that I’d finally reached true development hell.

kill your darlings

So I gave up on pRS52, refusing to invest any more time simply due to sunk cost fallacy.

But all is not lost! Some projects are not meant to succeed, others just need the right technology to drive them forward…

The newest iterations of CRISPR/CAS9 plasmids actually take some responsibility for the cuts they make and can be used to perform single base pair swaps. I’ve still got my noGFP plasmids waiting to be edited, I just need one that swamps T > G in bacteria, then this project will be alive once more. Stay tuned my friends!

- Alex Kelly