For TU Delft
We started collaborating early after the European iGEM Meet-Up with TU Delft in the wetlab.
Since we were both employing Cas13a as the molecular heart of our project, it made sense to
exchange data and experiences from wet lab experiments. At some point during a skype meeting,
we realised that we were doing similar things in the dry lab and thus, decided
to collaborate and exchange ideas in that field as well.
Secondary Structure Prediction
In our drylab collaboration with Team Delft, we performed a Cas13a design verification
of crRNAs they used in their project with our developed software. Since the secondary
structure of the crRNA is essential for its affinity to Cas13a, one can make assumptions
on the post-binding RNase activity of the protein based on secondary structure predictions.
We then compared these predicted structures to secondary structures of crRNAs that have
been shown to work experimentally. For the five crRNA sequences Team Delft provided,
the predicted secondary structure matched one of the secondary structures in our databank.
The only one that was not recognized by neither the NUPACK nor the mFOLD databank was
crRNA 2. It's predicted secondary structure in Vienna notation is depicted below:
It looks like the backbone structure needed in the first 35 bases is not constructed completely.
When looking at the graphical output of NUPACK depicted in Figure 1, however, it is visible that the binding probability
seems to be rather low regarding the structures that interact with the terminal 28 bp target sequence.
Thus, crRNA 2 would most probably still be active when tested in experiment, but might show decreased activity.
Figure 1: Graphical output of the secondary structure prediction of crRNA 2 in NUPACK.
Figure 2: Cleavage experiments of crRNA 2, crRNA 3 and crRNA 4 as provided by TU Delft.
Indeed, decreased activity using crRNA 2 in comparison to other crRNA is observed when performing experiments using
these crRNA designs. Though this has to be treated with caution since crRNA 4's secondary structure was predicted
to be correct but the Cas13a-crRNA complex of that crRNA does not show higher activity. We conclude that further testing
of the software has to be done to show whether it can predict activity of Cas13a.
Furthermore, in order to determine off-target effects, we constructed a database from Transcriptome data obtained
from the Ensembl and Ensembl Bacteria databank. We tried to make the search as tailor-made to the Delft project as
possible and thus considered species that were relevant to their project of detecting bacterial resistances related
to cattle and milk production. The included species were:
- Bos Taurus (Cattle)
- Clostridium perfringens
- Corynebacterium diphtheria
- Fusobacterium necrophorum
- Lactobacillus casei
- Lactococcus lactis
- Providencia stuartii
- Staphylococcus aureus
- Streptococcus pneumoniae
Three of the five crRNAs showed no off-targets in the constructed database. For crRNA 3 and crRNA 4,
possible off-targets were detected but just as a result of the identity parameter being rather low
in the BLAST-short run. When looking at the results, it is clearly visible that the sequences are not
identical enough to show any RNase activity of Cas13a since the 28 bases target is specific to up to 2 mutations,
but all the hits that have been found are with a match of 17 bases far from the 28 bases long.
Finally, we can conclude that our software did not find any problems in the crRNA design of Team Delft.
It seems that their constructs will most probably show activity and, at least for the mentioned species,
it will not show any off-target effects.
From TU Delft
The TU Delft team developed a simulation code for modeling the on and off-target
activities for sequences on all possible frames on the genome. In this, they employ a kinetic model
by Depken et al. in 2017 to determine whether off-target activities are high enough to give a false positive
collateral RNase activity of the Cas13a protein. They deducted from these model cleavage probabilities
for different crRNA sequences and could thus make statements of the possibility to distinguish
between our two bacterial targets, the 16S rRNA of E. Coli and B. subtillis. By running the crRNA
against both genomes of these species, they determined the off-target probability to be very low for all our crRNA
design as shown in Figure 3. Thus, they deduced from these that our crRNA design would be able to differentiate between E. Coli and B. subtilis.
Figure 3: Off-target probability of activation of the Cas13a protein for our crRNAs sensing E.Coli (1.3) and B. subtillis
crRNA 1-3 are taken into one since the spacer sequence of the crRNA was identical for these constructs.
We would like to thank Team Delft for the fun collaboration and wish you all the best for your project.
We are looking forward to seeing you in Boston.
- Klein, M., Eslami-Mossallam, B., Gonzalez Arroyo, D., Depken, M. (2017).
"The kinetic basis of CRISPR-Cas off-targeting rules." doi: 10.1101/143602.