Hacker News Reader

Comment by zigzigzag | Hacker News Reader

I seem to recall that in prior HN stories about CRISPR some poster was saying he believed CRISPR didn't work the way people thought it worked. He said it was simply killing cells that didn't have the desired mutation and biologists weren't realising that because of design errors in the experiments (or rather, sometimes mass cell dieoffs were being reported but not dwelled upon).

If CRISPR isn't actually editing the DNA but rather just selecting natural mutants that happen to have the desired edit, would that cause what's seen here?

View on HN

Don't think so. Originally the CRISPR mechanism was discovered in bacteria, which only have one cell. It doesn't make sense for bacteria's anti-virus system to kill it's only cell.

This is a really interesting point.

I don't know if this experiment has been done, but I actually think that there's a good chance that a bacterial cell might acquire a sequence that it itself contains - CRISPR is known to work at the population level but my understanding is that the mechanism for acquisition of new sequences is unclear.

I'll take a look for any papers on the topic and repost should I find anything.

Thats assuming that its selection mechanism works at an individual level, and not a population level.

After all a colony of bacteria is (usually) a set of identical clones. As long as a few survive, the DNA lives on.

See my other post here. They don't report the initial number of zygotes injected (why not?) so we cannot assess the "selection" explanation.

They got 11 surviving mice in the end, but only 7 were "edited", and usually we see that ~1%-.1% cells are mutants at any given site I would expect they needed ~ 1000 zygotes.

This only really explains the NHEJ results though, not the HDR (when the repaired DNA includes an exogenous template). They report that 2 (out of the 7) mice had the sequence matching the template. However this was only in some of their cells (36% and 19%).

Two other mice had a sequence that was similar but contained mutations...

Anyway, maybe someone can email them and ask how many zygotes were used originally.

It looks like this strain of mice can produce ~10 pups (a lower bound since not all zygotes will make it to birth) every 5 days for 26 cycles. So, one breeding pair could produce at least ~260 zygotes in 4 months. From this, I'd say it is plausible (economically) that the number of zygotes injected for this study was in the thousands:

>"The fecundity of the FVB/N strain was assessed by data from nine breeding pairs, which produced 43 litters. Litter size ranged from 7 to 13, with a mean value of 9.5. (First litters were generally smaller.) This is superior to other commonly used inbred strains; for example 6.7 for C57BL/6J, 6.6 for SJL/J, 5.4 for 129/J, or 5.0 for DBA/2J (15). A typical breeding pair mated at every postpartum estrous cycle and continued breeding for at least half a year, usually longer." http://www.pnas.org/content/88/6/2065.full.pdf

>"Mice have a 4-5 day estrous cycle and ovulate on the third day. Placing the females with a male on the third day of their cycle will result in the maximum number of pregnancies." Also from the same reference (table 1), number of fertile cycles is ~26: https://www.jax.org/strain/001800

What are the chances that a specific mutation exists once you modify more than 4 or 5 bits of DNA? It's like 1 / (number of base pairs ^ 4) or something. The chances are almost nil.

That isn't how NHEJ works. All they see is some random indel at the site. So all you have to explain is the presence of any mutation at the target site.

Reply on news.ycombinator.com