Joseph “Joe” Kreitz wants to find ways to make sure that molecular tools go to the right tissues and cell types. And now, the MIT graduate student and his colleagues in Feng Zhang’s lab at the Broad Institute have unveiled a new technology that could be a game changer for the therapeutic delivery of biomolecules: a bacterial “syringe.”

Across the biosphere, there is incredible diversity in extracellular contractile injection systems (eCIS) used to mediate symbiosis between bacteria and their eukaryotic hosts. This seems to be a general strategy among symbionts—bacteria that live within larger organisms.

Kreitz and colleagues focused on an eCIS called the Photorhabdus virulence cassette (PVC), which is produced by a bacteria of insects to target their host, deliver a toxin to kill that insect, and use the carcass of that insect to drive its own reproduction. “We thought it might be possible to actually modify this PVC to target human cells and then deliver a therapeutic payload,” Kreitz told GEN. “It’s a very beautiful strategy for delivering payloads into cells.”

In a new report published in Nature, Kreitz and colleagues from Zhang’s lab reveal how they took a naturally occurring eCIS and turned it into a highly efficient and specific biomolecular delivery tool. “We really need new approaches in the delivery space,” Zhang told GEN. “So much emphasis has been put on viral vectors or LNPs, and new approaches are really important.” Just last month, Zhang launched a new company called Aera Therapeutics that focuses on moving the cargo of genetic medicines—RNAi, antisense RNA, mRNA, or a genetic editing payload—based on protein nanoparticles (PNPs).

“Mechanistically, they are different, as they use proteinaceous effectors to inject through the membrane, as opposed to lipids (e.g., for lipid nanoparticles, or LNPs) that merge with and deliver into the cell,” said Rodolphe Barrangou, distinguished professor and CRISPR lab lead at North Carolina State University, and chief editor of The CRISPR Journal. “The sheath is very much reminiscent of phage tails and is more akin to a prokaryotic viral vector, though rather than package nucleic acid, it delivers proteins.”

AlphaFold adaptations

In the new report, Zhang’s team started by cloning the PVC from nature and expressing it in Escherichia coli, then taking this purified protein and showing that it is active in insect cells. Kreitz and colleagues then demonstrated that PVCs could be modified to target human cells.

“These PVCs naturally target insect cells, so they don’t target human cells at all,” Zhang said. “This is beneficial for us because that’s how Joe is able to make it very specific—you can remove the part that binds to an insect cell and replace it with something that binds to a specific thing on a human cell surface and get it to go into a human cell.”

While it’s not really known what the PVC targets on the insect cells, Zhang’s team used the AI-based protein structure prediction system AlphaFold to identify the region of the tail fiber mediating that interaction that binds to something on insect cells.

“[AlphaFold] gave us the information we needed to make a new delivery strategy that can be changed to target different cells,” said Kreitz. “All it takes is the addition of a very simple modification to its tail fiber protein that extends from the base to the part near the spike that actually gets driven through the membrane. We can add a novel binding domain to this tail fiber that would trick the syringe into binding a human cell instead of an insect.”

Kreitz and his colleagues were able to modify the binding domain, showing different strategies to reprogram these tail fibers. As examples, they retargeted these tail fibers against the human epidermal growth factor receptor (EGFR) and then showed that the whole system produced activity in EGFR-positive cell lines. They could also attach binding domains from human viruses to repurpose these PVCs to bind to human cells.

This system is really effective at killing cancer cells in a very specific way. The PVCs can target cancer epitopes and have the system inject toxins to have the cell die in a very programmable manner that doesn’t appear to show a lot of off-target effects. “It doesn’t seem to kill cells that don’t contain the receptor that you’re targeting,” said Kreitz. “So, you could imagine an application where you’re giving patients a PVC that targets a cancer epitope and then having the system go and deactivate cells that express that cancer epitope.”

Primed for protein delivery

Additionally, the Broad team showed that this re-engineered eCIS could deliver payloads beyond toxins. “One of the things we discovered pretty quickly is that, in terms of protein delivery, the system is actually quite powerful,” said Kreitz. “It can load and deliver a really versatile set of protein payloads. The system naturally delivers toxins because, of course, it’s trying to kill the insect. Those are around 300 amino acids [in size].”

In terms of protein delivery, the system is highly versatile in the types of payloads that it can load and deliver. Zhang’s team reprogrammed the system to deliver a diverse array of proteins—from some very small proteins up to Cas9, which is many times larger than the typical PVC toxin

As to how this works, Kreitz thinks that there may be some protein unfolding at play. “This system likely unfolds these proteins to some extent because it needs to fit the protein into this tube,” said Kreitz. “It is quite remarkable that the system can load such large payloads and then actually deliver them and retain the function of those payloads in target cells.”

A major sticking point, however, is that this eCIS, at least in its current form, is strictly a protein delivery system. That is useful in some circumstances, but if you wanted to deliver a different biomolecule, it would probably require additional engineering.

While Zhang’s group tested to see if they could rewire the PVC to load nucleic acids, the system initially did not cooperate. “Right now, it’s a protein delivery system only, so it’ll be great to get it to deliver RNA or DNA,” Zhang told GEN. “There are a lot of different phage or secretion system mechanisms that deliver things like RNA or DNA, so we’ll be trying to engineer that.”

But Barrangou told GEN that the protein delivery aspect is the major appeal. “Being able to deliver Cas9 and expand this to other genome effectors is exciting and useful at a time when delivery is the challenge for genome editing’s next step into the clinic,” said Barrangou. “Being able to program these molecular machines for cell- and tissue-specific delivery is very intriguing and potentially very useful, so editing strategies can be targeted to specific cell types and tissues.”

The University of Pennsylvania assistant professor Michael J. Mitchell, PhD, who played an integral part in developing the LNP platform during a postdoc in Bob Langer’s lab, agrees that this is an exciting advancement for the field of protein delivery. “LNPs are great for administering RNAs, but protein delivery using LNPs for applications such as gene editing remains challenging,” Mitchell told GEN. “It is difficult to encapsulate proteins into LNPs and deliver them efficiently into cells; this PVC platform can potentially overcome those challenges.”

Barrangou says he’d like to see future research looking into expanding the protein payloads (to other genome editing effectors beyond Cas9) and determining payload capacity (how large can payloads be and how can this be multiplexed to deliver more than one protein, either within the same payload or by delivering cocktails).

Delivery beyond the dish

The rest of the Nature report is devoted to characterizing the efficiency and specificity of this delivery system at targeting cells without creating off-target effects. Using intracranial injections, the team introduced the re-engineered eCIS into the brain of a live mouse; it is not yet known where the PVC goes, whether it has good tissue penetration, or if it just gets cleared.

“Because this is quite a large complex, it’d be interesting to study exactly whether it’s able to move through tissues and reach cells that are farther away from where you’re actually injecting,” said Kreitz. “In our paper, we’re injecting it right into a very defined, small region of the brain that we’re interested in. But for some diseases, you might want it to go in a more systemic way throughout the body. The PVC’s behavior in vivo is a place that will require more work in the future.”

Both Barrangou and Mitchell would like to see the system tested more broadly. Mitchell is keen to see is if the PVCs can be administered systemically (IV) and deliver proteins to therapeutically relevant target cells and tissues. “They show an initial in vivo proof of concept as a direct injection into the brain, but demonstrating IV delivery could open up many therapeutic opportunities in organs such as the liver and lungs,” said Mitchell. Barrangou also wants to see the system tested in many cell types and organisms, including plants, given the challenges of deploying genome editing in plants.

According to Zhang, there’s a lot of diversity in eCIS systems. “You can also find them in the human gut, so it would be very interesting to see what they may do inherently to target human cells,” said Zhang. “There may also be invading properties that they may have because they’ve been living in the human gut. Those are just some examples, but there’s a lot that we can do.”

Zhang thinks that there are probably many things in nature that cells have coopted to be able to transfer molecules back and forth.

“There are also many other versions of this in other bacteria, so you can either engineer this one or explore all the others, which will probably have different properties, to achieve a set of capabilities, and I think we’ll have to do both,” said Zhang.

“We’re just looking at the tip of the iceberg,” he continued. “It’s pretty trivial to re-engineer and is similar to producing any recombinant protein, but in terms of developing this into a biomedical tool, there’s still a lot of work to be done across the board, from discovery to manufacturing. You can grow a lot of bacteria using fermentation processes, and we’ll have to make sure that they’re pure because we need to get really endotoxins and other bacterial contaminants, which could be immunogenic,” said Zhang. “So, that could be part of the challenge, but that’s more about the manufacturing. It’s still early days for this as a technology.”