I wrote here about a new company (Homology Medicine) that claimed to have a viral method for gene editing that did not involve any sort of double-strand DNA breaking enzyme (as you need during the CRISPR, TALENs, or zinc-finger nuclease methods). That’s a pretty interesting claim, because double-strand breaks (DSBs) are powerful but can be hard to control. But as is always the case, interesting claims need interesting evidence to back them up. As I said at the time:
This is a testable hypothesis, of course. The company will be working very hard to produce convincing results, and if they do, they can expect to see a lot more than $127 million. I’d be a little jumpy about investing at this point, though, because it’s certainly true that a lot of neat stuff doesn’t quite pan out. (The current research boom, looked at from that perspective, is due to a slightly higher-than-normal run of neat stuff actually working).
Even at the time (as a link in that post shows) there were skeptics in the field, and now a new manuscript on bioRxiv will give those skeptics more support. It’s from Paula Cannon’s lab at USC, and it makes for pretty grim reading (if you’re an investor in Homology, anyway). The earlier work from City of Hope (the stuff that kicked off Homology’s founding, PNAS or PubMed Central) seemed to involve homology-directed repair (HDR), although the details hadn’t been fully worked out. But this new paper reports that they just cannot make the viral vectors do what they are advertised to do.
Not at all. The original paper reported that AAV genomes packaged into capsids from Clade F viruses (AAV9, for example, and apparently only the capsids from that group) could do the gene-editing trick. Cannon’s lab has been trying to reproduce that, using good old green fluorescent protein (GFP) as a marker. They’ve packaged that into the AAV9 vectors as well as the more traditional AAV6, which has been reported many times to insert such genomic material into human stem cells, and tried editing both a stem-cell line and three others. It didn’t go well. The AAVg experiments altered between 20% and 80% of the stem cells, depending on the viral vector load, but even at the highest load the AAV9 experiment only seemed to transduce about 3% of the cells. Which is basically the opposite of what was reported.
The same went for the other three non-stem cell lines, including one (K562) that the original paper had reported as being efficiently transduced in the absence of any DSB nuclease enzymes. They went on to try these experiments at more than one gene locus, but in every case they couldn’t show that the viral payload had been incorporated into the targeted cell genomes. In general, if you tried the gene editing with AAVs alone, no matter what clade they were from, the amount of GFP seen in the cells dropped off rapidly, consistent with it just being expressed from some exosomes at first and not really getting edited in. No matter what experiment they tried, they could not find anything that contradicted the previous conventional wisdom (that unless you can target some double-strand breaks, that AAVs by themselves can only do less than 1% transduction). And they were able to reproduce none of the earlier findings that caused such interest at the time.
At present we are unable to explain our inability to reproduce the findings of Smith et al. Perhaps there are unidentified features in their AAV constructs, or some aspect of their vector production, purification, or titration methods that contributed to this phenomenon. Nevertheless, the reported ability of clade F AAVs to perform highly efficient nuclease-independent genome editing by homologous recombination is clearly not a universal phenomenon.
Nope, sure doesn’t look like it. This news came out yesterday, highlighted by Anthony Regalado’s Twitter feed at Technology Review, and it’s not hard to see just when that happened. We’ll see what response the company has – this is one of those times where it’s both a scientific question and a business one. . .