ABOVE: Scientists miniaturized the Cas13 enzyme with the help of AI. © iStock, ArtemisDiana

CRISPR-Cas holds great promise for treating and curing human disease. One of the few ways that scientists can usher CRISPR elements into a large number of mammalian cells is via an adeno-associated viral (AAV) vector. These vectors are small, so squeezing the large Cas enzyme gene plus its promoters, the polyadenylation signal, guide RNA genes, and other needed components into one vector is difficult; splitting the components into two vectors can create problematic doses or reduce the efficiency.

One way to fit it all into one AAV vector is to shrink the large Cas enzyme gene. In a new paper published in Nature Communications, geneticist Zhanjun Li and his team at Jilin University miniaturized a Cas13d enzyme by carving out unnecessary bits and then using an artificial intelligence platform called AlphaFold2 to visualize the structure of other Cas13d and Cas13b proteins and miniaturize them.1 They found that these mini-enzymes worked just as well as the wild-type ones.

“There are not many examples of proteins that can be shrunk down to smaller sizes and still have them work,” said Jeffrey Chamberlain, a geneticist at the University of Washington who was not involved in the study. “That could be huge in terms of simplifying gene editing through the use of these AAV delivery vehicles.”

In the past, researchers have searched for and found naturally smaller Cas proteins, but these didn’t always perform as well as the full-sized Cas.2 They’ve also developed other methods of miniaturization, but these alter function. No efficient, generalizable method of making Cas proteins smaller has been developed.

See also: “Predicting the Next Level of CRISPR Control

Li and his colleagues used a strategy that they call Interaction, Dynamics, and Conservation (IDC) to find sections of the Cas13d enzyme that were not necessary to its function. (Cas13 is one of the many Cas enzymes used in CRISPR, but it targets RNA, instead of DNA like Cas9.)

They first looked at one example of a Cas13d protein called EsCas13d. Previous studies had determined the crystal structure of this protein using Cryo-EM, so the team used their IDC strategy to examine the interactions between the functional sites of the EsCas13d enzyme, the conformational changes that occur as the enzyme performs its functions, and the structural units that are conserved and generalized.3 The researchers chose eight sections of the amino acid chain that they hypothesized were not necessary to its function, and deleted them, creating mini-EsCas13d.

Next, they used AlphaFold to predict the structure of two other Cas13d members, RfxCas13d and RspCas13d, based on their amino acid chains. They found their structures were similar (with some key differences) to EsCas13d, so they set out to miniaturize them in the same way.

Li found that each mini-Cas13d knocked down RNA for various genes in a line of HEK293 cells with approximately the same level of efficiency and similar optimal spacer length, mismatch tolerance, and off-target effects as the wild-type enzymes. Western blot analysis showed that the wild-type and mini-13ds also showed similar levels of protein in the cells. The team then tried miniaturizing Cas13b using AlphaFold2 and their IDC strategy, and created two mini variants with 12 deleted sections, each with the same amount of targeted RNA degradation as their wild-type counterparts.

See also: “CRISPR-like Abilities in Eukaryotic Proteins

To see if the mini-enzymes worked in vivo, Li and his team targeted the murine proprotein convertase subtilisin/kexin type 9 (Pcsk9), which raises blood cholesterol. They injected a mini-RfxCas13d AAV plasmid containing crRNAs targeting the coding sequence of Pcsk9 mRNA into the tail vein of 8-week-old mice. mRNA levels of Pcsk9 decreased significantly in the injected mice, and protein levels also decreased. “These small proteins exhibited full activity similar to that of the wild-type enzyme, and the total serum cholesterol level in the AAV-injected mice was reduced to 61%+8.3% that of the normal level,” confirmed Li in an email. Indicators of liver damage were normal, and the liver looked healthy.

“The next plan is to design more reasonable proteins that contain not only miniaturization but also addition, replacement, and mutations to get high-fidelity, hyper-accurate, and high-efficiency Cas proteins,” said Li.

“The nice thing about that is they incorporate artificial intelligence and all sorts of learned features to really analyze protein structure in enormous detail,” said Chamberlain. “I hope that similar work can be applied to some of the other Cases, such as Cas9, which is the more commonly used gene editing enzyme that is focused more on DNA rather than RNA.”


  1. Zhao F, et al. A strategy for Cas13 miniaturization based on the structure and AlphaFold. Nat Commun. 2023;14:5545.
  2. Kannan S, et al. Compact RNA editors with small Cas13 proteins. Nat Biotechnol. 2022;40(2):194-7.
  3. Zhang C, et al. Structural basis for the RNA-guided ribonuclease activity of CRISPR-Cas13d. Cell. 2018;175(1):212-223.e17.