Omicron Origins, Part 2
The somewhat fragile lab origin theory.
Continued from Part 1, which disputes other possible origins for the Omicron siblings, including the “early leak” theory by Ethical Skeptic.
And now, the “graphical abstract” - or rather my botched first draft of it, which will hopefully be replaced soon.1 Presenting, the “lab loss + recovery mode theory”:
Please note that even I have first-look objections to certain elements of the above - an example being the semi-dogma that cellular passage leads to loss of the Furin Cleavage Site™, and yet both BA.1 and BA.2 have kept those genes in tact. As I will restate below, this version of the theory is meant to be used as a template, to organize future thinking about the problem of BA.1 and BA.2’s divergent mutation signatures.
Put differently, the purpose of this version of the lab loss + recovery mode theory will be to eliminate all the current reasons BA.1 and BA.2 “don’t make sense,” and put a new set of reasons in that empty space, so we can work from there.
A final caveat: I doubt that these ideas are novel. Once I spent enough time with the mutations for BA.1 and BA.2, this theory seemed semi-obvious. What I can’t say, is if the theory has already been published by anyone who has already developed it before me. If so, apologies for the accidental resemblance.
But before delving into the theory, there is one more potential origin to dismiss.
The “escape from a published mouse experiment” theory.
Test Tube Babies (cont’d!)
3: It Wasn’t Leaked from a Published Mouse Experiment
In Part 1, we toyed with the rules of the “mystery” a bit, to propose, not unreasonably, that viral particles bearing the triplet of mutations which comprise the signature for “B.1” might have already been present in the first infections and isolates of the “Wuhan” version of SARS-CoV-2, even though they won’t show up in sequencing.
Perhaps, then, a mouse-passage experiment using Wuhan isolates could not-so-miraculously “reveal” that triplet, accounting for its appearance in the descendants that arrived onto the scene last November.
And from there, why, we could imagine that perhaps one of the Wuhan-era-isolate-mouse-passage studies published in 2021 was implicated in an accidental leak. Even better, what if the described work continued after publishing, and the leak occurred then? What if it was only a couple months before November? Suddenly, the “implausibility” factor of a Wuhan-mouse-passaged “BA.0” ancestor splitting into BA.1 and BA.2, experiencing different mutation pressure, and yet evading detection, and yet gaining whatever magic mutation makes them stop evading detection at almost the same exact time, etc., becomes less of a reach - as now this magic mutation is something like a “1 per 2 month event”!2
Well, this was a setup. Every mystery needs a good red herring.
If the B.1 triplet is hiding in those ancient isolates, passaging said isolates into mice seems like a good way to kill it off. None of the 7 published studies involving serial passage of pre-B.1-era isolates finds B.1 genes, whether early on or after 30+ rounds of mice infection.3
And why should they? If we are supposing that B.1 is only .01% of the particles in those early isolates, passing the isolates through a narrow sieve - such as by spraying them into a foreign animal host’s nose, to see if any take hold - is not the way to “nurture” B.1 into the dominance ordained by natural selection in real human infection.
So, it turns out to be pretty simple. Either there is a not-yet-published experiment involving a rare, “20B.S.732A”/Alpha ancestor B.1 isolate4 for mouse serial passage, which experienced a lab-leak sometime before November, or we are back to zoonotic double-crossover.
So far, we haven’t even addressed zoonotic double-crossover, since this theory suffers from all the same problems as the “2+4 year theory” in accounting for why BA.2 and BA.1 branched so long ago in their evolutionary path, demonstrated differing non-mouse evolutionary pressures, and yet emerged near-simultaneously (or, why there is no apparent pre-existing BA.2 in the US).
However, I don’t want to be too hard on this one. It does put a lot of the “pieces” in the right place, even if it doesn’t satisfy the rules of the mystery very well when it comes to the co-discovery timing problem; as with the 2+4 year theory, a resort to “Rashocronian Uncertainty” is required to rescue the narrative. So, I invite the reader to keep it on their list of possibilities.5
The grander point of the red herring was to demonstrate that, due to BA.1 and BA.2 both possessing the B.1 signature, right down to the “silent” letter-change at 3037, an “innocent” lab origin for our mysterious siblings becomes fantastically more implausible per the rules of our mystery, as all the published studies flunk out.
There’s one more way in which an innocent lab origin doesn’t fit for BA.1 and BA.2: The published studies using the Wuhan-era isolates, by pushing the envelope all the way to 36 passages in one case, demonstrated that mouse-serial-passage alone cannot produce the scale of spike-remodeling that is evident in the “shared” portion of the siblings’ mutation-set, which either were acquired before their “break” or by coincidental co-mutation afterward.
In Montagutelli, et al., 27 passages through naive mice leads to only two mutations in the spike protein. In Sun, et al., 36 passages through naive mice leads to only three spike mutations.6 Wong, et. al. finally hit it out of the park with a whopping 4 mutations in the RBD via 30 passages.
Although BA.0 (as imagined via the siblings’ shared set of mutations), possesses mutations in all five of the RBD “hotspots” identified in the mouse serial passage experiments, none of those experiments was able to produce more than four (without self-editing the desired change). In other words: Once the virus has one or two of the pro-mouse amino acid revisions, it doesn’t seem to bother making any more, even if you keep passaging it forever.
As it turns out, the evidence for the mouse serial passage origin of BA.1 and BA.2 is less “smoking gun” and more “water-pistol.” We will revisit the question of whether this calls for a reassessment of the evidence for mouse-passage having actually occurred at all - but for now, further analysis of the theory requires actually describing the theory, and so, “On with the show.”
What Could Go Wrong
So, imagine you are trying to mouse-ify SARS-CoV-2 in a lab, because you think this will make it stronger just in time for “Dark Winter,” or a “white hat” ordered you to whip up a “live vaccine,” or who-knows-what, but either way, seriously you should take a look at your life.
Anyway, all at once, on pre-innoculated mouse serial passage 116 or whatever (you’re really “going for the gold” with this baby) the virus goes quiet. You blend up the lungs of the mice it last infected, spray the resulting goo at the next set of mice, just as you’ve done 115 times before, and… nothing happens. “No problem,” you think - you’ll just passage it through a non-immunized set of mice instead. And so you spray the goo, and… nothing happens.
Unbeknownst to you, things were beginning to go wrong a long time ago. Your “virus” has been contaminated with co-infecting variants.
Due to being in such a radical pressure environment for so long, the virus - unbeknownst to you - was making a particular “error” over and over. This error could have been anything - maybe it made the RNA less easy to read, for example. RNA is a “code,” but it is also a long molecule full of self-attracting parts. The “nucleocapsid” protein helps to stabilize this long molecule and cut down on unwanted tangling; but let’s say there was a situation here where a particular mutation, one that improved cellular entry every time the virus made it, also created a “tangle” of this sort, and led to deficient within-cell replication of whatever unlucky mouse you sprayed that virus into.
Let’s say this “defect” is only present in 1 in a million of all viral copies, but, because it leads to better cellular entry, especially as you’re in here pre-immunizing your mice to the “common” version of the spike protein in advance, the rate at which the “defect” ends up successfully entering a cell is 1 per overall 1000 of viral copies.
And this is happening over and over, every time you run a serial passage. Unbeknownst to you. Well, the result of this is that you have created a new species, with a new radical evolutionary pressure. This species doesn’t “want” to stop having the gene that is creating the problem (because then it just becomes “not the species”), and so the pressure that this species is experiencing, unbeknownst to you, is: Fix whatever’s wrong with replication, while keeping the new gene.
Well, there are several ways it could “go” about this, which is to say several second-order or simultaneous mutations that could allow for a slight improvement to our hypothetical “genetic entanglement” conflict. Maybe just randomly changing genes in Orf1 somehow improves things, and so the new species starts to randomly change genes in Orf1 (which is to say, copies of the new virus that contain these mutations are slightly better at replicating). Who can say? Nature = Mysterious. (Unbeknownst to you.)
In fact, these types of Schrödinger's cat species of disparate defects which, together, have the goal of “fixing the defect,” but stop existing if the “fix” is to remove the defect, are likely a constant reality for any given virus. In a genome that is ~30,000 letters long, there’s likely always one or two potential “transmission enhancing” mutations that mess up replication in some manner, leaving all the copies of the virus that have this mutation in the condition of being a new “species.”
But in real life, viruses ebb and flow, which leads to fringe versions of the virus having a short life span. If only 1/10,000,000 copies of measles virus were going to survive to the next season, back when measles had seasons, the most common version (without the defect) was favored; the fringe version had to “re-emerge” from scratch as a defect of the common version, and start the work of evolving out of its replication problem all over again. Who knows, maybe Nature does things like “seasons” for a reason? Unbeknownst to you.
Meanwhile, you’re here spraying mice with the blendered-up lungs of other mice.
Anyway, while you’re doing this, just what’s the environment like for “your” “actual virus”? Well, for one thing, you’ve noticed that mutations to the RBD have slowed down, which is upsetting the bosses quite a bit. Now, you don’t really know why this thing needs even more, considering that it’s already set several records, but you share their frustration that Nature is daring to Not Do What You Tell it To. Unbeknownst to you, things have gotten easier for the common, consensus version of the virus, as now some of them are arriving in cells that have already been co-infected with the transmission-enhanced, replication-deficient copy. The copy has already thrown out a bunch of immune suppression proteins, but seems to be stuck after there.
As far as your consensus virus is concerned, this is the life. The radical pressure environment you were subjecting it to for all these passages has suddenly relaxed, just a bit. It no longer has to look at every passage as a sort of “death,” in which only the copies that have new mutations to the spike protein (defective or not) survive. (Do you think your little viruses like it that only 1/100 of them even have a chance of passing on?)
Meanwhile, the defect species (which is still also being fed by new mutations of the “common” species, remember) benefits from occasionally being passaged into other cells via the common version’s successful employment of the fusion feature, so that’s nice. So in a way, the common virus has become dependent on the defect, and the defect vice-versa. But for the defect, the radically high pressure on “fixing whatever is wrong” remains, and the changes to Orf1 continue to be what hit the right buttons on this front. Gradually, it gets closer and closer to being able to “stand on its own legs.” At the same time, the co-infection environment has become permissive for other off-shoots of the common virus - it’s like there’s a whole ecosystem going on in here. (Unbeknownst etc.)
And that’s where things begin to head toward The Crash.
One day, the “defect” version is 1/2 of the copies that come out of any given infection, and the copies of the “common” version that can outcompete it for transmission are, all at once, now effectively zero.
Now, no matter how many times you spray your lung-goo into a mouse’s nose, the “common” version is going to fail to get a foothold, and the replication-deficient version, however much it has made a recovery, is going to be suppressed by the innate immune system. The virus, as contained in the goo of the sample from 115, is in crisis. There’s no way for you to go in there and pluck out the defective 1/2 of the billion copies. This goo is kaput.
Well, you’ve got save-points.
But… they aren’t working either.
Because Nature is mysterious and “quantum,” (whatever that means), now your back-up copies of passage 110, 100, and 90 don’t work either. This is very frustrating. You feel that Nature is being unreasonable. As if, because you are suddenly aware of the problem, something about time and space has been altered by your knowledge. As if your knowing that Nature is doing stuff has suddenly made it impossible to escape the consequences of that “stuff.”
Anyway, at this point, notebooks are being slapped onto the back of heads. Security tapes are being reviewed. There’s serious questions about sabotage. No one wants to believe you that “It was just working, and then it just wasn’t.” You’re gone.
Replacement-You takes the obvious approach: If the virus can’t be passaged into mice, no big deal. It’s still Alive™, but just a bit weak - as you will verify by effect on cell cultures.
Well, turns out the virus isn’t growing in these cultures either! Certainly not enough to allow for a full sequence. All you can recover is little bits and pieces - just enough to confirm it’s still definitely something that was in the goo. But things are getting a bit dire - sample size is near 0 at this point, and can’t be re-upped, because the thing won’t replicate - and so, drastic measures are called for.
“Drastic,” as in, recombination surgery.
This one’s out of your hands - the “SARS whisperer” has been flown in, and is handling the more delicate steps. Eventually, the operation is complete, the fragmented genes of the virus have been injected into a cell and have kicked out a culture-viable product (though the sequence still flunks the quality benchmarks), and the result is handed back to you for full cell culture recovery, further sequence quality and infectivity confirmation, and mass lot production - all of which must be finished within the month to meet the release schedule. What could go wrong?
Anyhow, after another few rounds of cellular passage, you get the sequence to comfortably pass the benchmark - and it looks, well, it looks OK. Certainly, the spike is moused-up to the wazoo. But there’s an awful lot of letters in the Orf1 region that don’t match the last known sequence of the consensus virus - and what’s more, all the most recent Orf1 changes of the common virus are missing (it was beginning to mutate in that region more as well, your predecessors notes happen to mention - you find this a bit strange). And on one of the reads, there was a giant hole in the NTD, next to what looked like a weird insert. You forward all this to the bosses, simultaneously downplaying the warning signs while half-wishing they will “pull the breaks” for you anyway, so that you can be reassigned to a job where you are 100% sure your predecessor was not fed to a shark.
So it’s down to the pathogenicity check. Oh, boy, do those mice and hamsters keel over. There’s quite a bit of fist-pumping when you show the weight loss charts after that one. This is it: Full recovery! Why, you even half believe the groupthink hype.
Unbeknownst to you, the favorable, immune-system-free conditions of the cell culture recovery were enough to get the “defect” version over the last hill of fixing “whatever needed to be fixed,” by un-mutating the defect that was favored by the mice passage challenge and had caused the problem to begin with.
They also gave a single, unseen “recombination failure” mutant time to slowly grow within the stock.
Now it’s only a few weeks before the release date, and things are more out of control than ever. The scaled-up stocks are failing sequence quality benchmarks routinely. The NTD deletion keeps showing up. Likewise for the nearby EPE insert. If there’s one thing the bosses are clear on, it’s that you do not mess with the NTD. RBD mutations are the goal, not NTD. The NTD in these hard to pin-down error sequences look like a shark mauled it. Did you say “shark”? You meant “bear.” Sharks are on the brain a lot these days, for some reason. Just, sharks sharks sharks.
Measures are taken. Production is ramped up; and lots that fail the (new, somewhat lower) benchmark are tossed. Everything that is a big problem one day suddenly becomes totally normal and acceptable the next day. You’re assured this is how the first two runs went, and things still came out fine; the virus still behaved adequately “like a virus.”
“In a way, Nature fixes whatever we mess up. And if not, the news just lies anyway.” Ah yes, Nature. That thing with the sharks in it.
The lots ship; the news breaks. Your nursed-to-life baby has set off alarms thanks to a sequence in South Africa (where it was released a few days earlier than the other sites); the traders in Wall Street have all gotten the tipoff from your bosses that this is the big one; the world is on fire. The WHO is declaring that it is skipping the two names that don’t grant desired anagrams, because reasons.
A bit queasily, you take your first look at the South Africa sequence.
It’s… not the right virus.
Harder Than it Looks
At this point, the only thing that can rescue an “innocent mouse-passage leak” is B.1 itself. If there is an unpublished study involving an early B.1 isolate, perhaps we could imagine that the “secret sauce” of B.1’s removal of Wuhan’s defective position-4715-Proline in the gene-copying protein allows the virus to achieve a more “kinetic” mouse-passage experience.
Of course, the same thing, if true, would likely be known and exploited by anyone conducting a non-innocent mouse-passage experiment. B.1, in other words, would be the default template for any intentionally-developed variant.
As in Part 1, I want to evoke the eerie echos with the B.1 triplet’s somewhat overly “clean” appearance in our famous late-2020 scariants, and with B.1’s emergence itself. Alpha, Beta, Gamma, and Delta all arrive on the scene with B.1’s mutation triplet, including the “synonymous” (as in not changing which amino acid gets placed in the resulting protein) disagreement with the original Wuhan sequence in letter 3037 of the genome (where B.1 has a U instead of a C). They bear no or very little evidence of the early 2020 “top 80” spike protein mutations. They appeared synchronously, at roughly the same time we can imagine BA.2/BA.1 to have been making its first infections, pre-sequencing, a year later.
If B.1 is the template, however, does that fully account for the scope of BA.0’s “creativity” with the receptor binding domain? Not really. BA.0’s incredible receptor-binding-domain revision may, in fact, be evidence of a super-passage approach, in which successive generations of mice are first immunized against the RBD of the spike protein of the current version of the mouse-passaged virus before they are infected with it, over and over, so that the virus cannot stop at the “one or two” mutations which enabled it to thrive in naive mice. (Note, also, that the zoonotic double-crossover theory does well here again, though also implies that antibodies should be detectible for BA.0 somewhere out in the wild).
And moreover, as long as we’ve transitioned into discussing a theory of mind for a non-innocent origin for our siblings, we’ve simultaneously put anything and everything on the table: Including that the six “mouse mutations” were themselves an edit, rather than the product of mouse-passage.
While Wei, C. et al. are compelling in their analysis of BA.1’s “letter-swap signature” (which BA.2 happens to preserve) in so far as it seems clear that BA.1 did not acquire its mutations in a human host, the use of the signature to implicate mice is a bit hazardous, now that we know that BA.1 and BA.2 are so different from each other, and that none of their unique mutations fits the “mouse receptor binding domain optimization” signature.7 So, in so far as the template “lab loss + recovery mode” theory includes mice passage at all, it’s still only one of many elements that can be modified at will.
I’d like to throw out one more suggestion: That the B.1 triplet, as a “variant template,” may offer other values beyond promoting mutation. We could imagine, for example, that B.1 offers better retention of the Furin Cleavage Site in cell-passage; that would make it a very useful template indeed. If the Wuhan version of SARS-CoV-2, for example, likes to “lose” the Furin Cleavage Site after a few passages in cell culture, then how are you ever supposed to grow an isolate into a great big stockpile without losing the thing that makes it “it”?
Even if your ambitions are more modest, the potential for a virus to lose its essential character whenever you stop and try to measure it presents a problem. If the virus is close enough to being a single thing (an Elvis, as opposed to a lobby full of Elvis impersonators), you can ignore this hazard for the most part, or rely on your quality benchmarks to flunk a sequence that seems unsure of itself. But you’re still operating all the time in a “will the car start” paradox: The only way to prove that a “save-point” sample of the virus can still go from cell culture to sequence without losing expected genes, is to turn the key, at which point you’ve just used a bit of the battery, and now will have no idea wether the car will start the next time you turn the key.
What you really want, is a “battery” you don’t have to wonder so much about to begin with. Is that what B.1’s removal of the “Proline RdRp defect” accomplishes - stability, rather than variation? Who can say? (Obviously, whoever’s hypothetically using B.1 as a variant template can say.)
There’s one more interesting observation that arises from our switch to a non-innocent origin: BA.1 and BA.2 don’t seem to satisfy the same intention. So what, exactly, could that “intention” possibly be? One could imagine that BA.2 is fulfilling the role implied by its “stealth” moniker; but the fact that BA.2 only seems to “follow” BA.1 if BA.1 arrived in a region early suggests that BA.2 isn’t fulfilling this role very well. Additionally, late 2020’s apparent “scheduled release” was more spaghetti-at-the-wall, in that there were four “versions” which were allowed to compete for dominance. This year’s apparent “scheduled release” is more eggs-in-one-basket.8
Lab loss + recovery mode attempts to answer this part of the mystery in a manner worthy of any TV series reveal, as the culprit’s “one mistake.” Because it is more delightful to the imagination, I have placed BA.2’s sequence as the version of “Omicron” that was intended to be released. Really, do imagine the preparations that were in place, and had to be reversed when the “I’m a Cold” headlines broke: In the itineraries of our hypothetical variant-creators, we’d all be at mandated fourth-dosed firemen by now, instead of all this expert contrition and the ongoing attempted consolidation of the ashes of the media/state’s credibility. Judging from how reluctant and ambiguous the “reversal” of the Pandemic has been - as if the reversal itself comes with a pending reversal, for which planning is already underway - Omicron flopped. Additionally, I find it implausible that BA.1 could be selected for an “intended release” in any context, given the troubling disfigurement of the NTD genes. Meanwhile, BA.2, if it is the product of a “recovery,” seems to have come out not looking any worse for wear.
Though here, again, the theory’s parts are exchangeable. Perhaps BA.1 was intended for release as a “Pandemic exit viral vaccine,” and BA.2 was the accident.
In either case, the loss + recovery scenario proposes that “Omicron” was lost during serial passage, and had to be reconstituted via recombination plus cell culture, or cell culture alone. Either of these steps could shift the balance between “Elvis” and “Elvis impersonator,” or create a situation where there are only “Elvis impersonators,” and increase the chance that a post-cell-culture sequence “passes” quality check when the genetic material being sequenced is not actually very sure of itself.
If deadlines are involved (because God forbid the ants that you’re roasting in the sun don’t witness your precious, overly on-the-nose “Media Control” brag), we can obviously imagine a scenario where this initial ambiguity starts to unravel, but the signs are nonetheless ignored. Perhaps as “Omicron” was scaled up for wide release (you don’t actually want to trust that the virus will take off from a Patient 0, especially since you haven’t actually tested this product in the real world yet), sequence errors became more common, and it started to become clear that “something else” was in the original stock. What’s more, this something else looked, well, damaged. There were apparent huge holes punched into the NTD genes, a puzzling insert, etc. Perhaps efforts were made to prune scaled-up “Omicron” back down to only the lots that seemed “clean.”
Given the decades of issues with polio vaccines and monkey-viruses, there should have been no reason for any sane mad scientist to imagine that “pruning” the product grown from a contaminated stock would actually prevent unwanted release of the alternate sibling - but hope springs eternal when you’re trying to hide a program-fatal QA problem.
I am between vector-editors right now, having idiotically decided to learn a yearly-subscription based product in January, 2021. Anyway, I only finally conquered my decades-long total bafflement at Bézier Curve UI yesterday (it is why this post was delayed another day), and should be able to produce decent-looking results soon.
Though, a late leak of a Wuhan-era-isolate-mouse-passage study still leaves the problem of BA.1 and BA.2 no longer evincing a “mouse serial passage” mutation signature! But there was no need to point this out, since the B.1 problem defeats the theory anyway.
See footnote 19 in “Mouse Party.” Note that “7” excludes the UNC experiment (which was not based on an isolate, but rather a recombination-born SARS-CoV-2) and includes the Australia experiment (even though it didn’t involve mice infection passage, but rather mouse-ified cell passage, resulting in loss of the precious Furry Cleveland Sights).
In one of the most extreme of the published mouse-optimization experiments, as well as the one published last (and thus closest to the emergence of Omicron), a January, 2020 isolate from Paris was passed through mice 15, 16, and finally 27 times, and then the most mouse-pathogenic isolates from each timestamp were sequenced. No B.1 signature. And, far less mutations than “BA.0” overall:
Once you become aware of the B.1 signature, it becomes a fun way to compare the less-renowned variants, and to imagine whether B.1 looks geographically more “artificial” than non-B.1 variants. I’m sure I will be looking into this subject for a while, though whether I can hash out any “proof” that all the most hyped variants bear clear signs of artificial introduction is anyone’s guess. It seems like someone else should have already managed this by now.
As for “20B.S.732A” (see here at covariants), a B.1-signed variant that appeared in North America in, when else, October 2020, it is mysteriously apparently related to Alpha (as if both are on the same branch of the B.1 tree) and not as highly-mutated. Because it is not as highly mutated, it is much closer to our imaginary pre-mouse “BA.0” than Wuhan (there are fewer synonymous mutations to require BA.0 to have coincidentally “reversed” than for Alpha, for example).
In particular, double-crossover accounts for another of the problems visible in the graphic above: In mouse serial passage with naive mice, SARS-CoV-2 seems to stop evincing evolutionary pressure on the receptor binding domain pretty quickly (though, other studies were able to show slightly more progress in this regard, such as Sun, S. et al.).
Montagutelli, X. et al. “A mouse-adapted SARS-CoV-2 strain replicating in standard laboratory mice.”
(Again, see “Mouse Party” footnote 19 for study details.)
See Wei, C. et al. “Evidence for a mouse origin of the SARS-CoV-2 Omicron variant.” biorxiv.org, and, again, “Mouse Party.”
The famous Figure 2, which demonstrates that “Omicron” BA.1 instantly reverts to human-like mutations after the initial sequence, while it’s pre-discovery mutations lack the same touch. Again, it turns out that BA.2 possesses the same signature in its unique set of mutations: