Thread by @wanderer_jasnah

afternoon session starts here, b/c twitter gets *really* slow this many tweets deep. starting off w/Bette Korber giving a talk on... i missed the title, but it seems to have involved D614G increased infectivity & now is talking about world variant patterns.

https://twitter.com/wanderer_jasnah/status/1448731997757599744

South America is interesting, b/c while Delta is taking over, a set of P.1 variants that have additional +ve charge near the S1/S2 FCS seem to have a competitive edge: Q675H/R, N679K, P681H all on a P.1 background.

while it seems P.1+basic is quite competitive, it's still being replaced by Delta, any Delta declines seem to be sampling artifacts b/c the P.1 variant landscape is surprisingly elaborate.

Peru: Lambda -> P.1 -> P.1+basic, some Mu -> Delta (newly arrived).

P681RH stands out as *THE MOST* variable site in the *entire* @GISAID dataset: it's common to B.1.1.7/Alpha, B.1.617.2/Delta, B.1.621/Mu *AND* within B.1.1.28.1/P.1/Gamma.

interestingly, histidine mutations near/at FCS are proposed to be neutrally-charged when unbound, but become +vely charged once in the active site of kex2/furin.

example of a common Delta variant that's likely actually just founder effect: V1264L.

example of a common Delta variant that seems to actually be increasing relative to background: Q677H

some interesting Delta developments include Q677H, Q675H, H655Y & even *N501Y* on a Delta backbone.

now let's take a look at Mu: it made some inroads in Mexico & Florida & Alaska, was most common in Chile & Columbia, but both are now transitioning to Delta while Mu is now rare in both USA & Mexico to the point where it hasn't been sampled at all in the most recent samples.

C.1.2: it briefly seemed to make inroads in South Africa but now seems completely dead.

so how do we tell which variants are worth keeping an eye on? well, one approach is what's implemented in SHIVER (cov.lanl.gov/content/index).

we *REALLY* need better sampling from Africa!!!

of datapoints from the last 60 days, we have one P.1+N679K, one B.1.1.7+E484K & the rest is literally *all* Delta.

what the hell Cambodia.

currently we have co-circulation of B.1.1.7/Alpha+E484K & B.1.617.2/Delta *MINUS ALL NTD MUTATIONS EXCEPT T19R*.

what the heck will happen? full Delta expansion? this weird partial Delta holding? Alpha+E484K holding? def. worth keeping an eye on.

of 34 PANGO lineages we've been tracking in 2021, 25 have both NTD+RBD mutations, 9 have RBD mutations only & 3 have only NTD mutations.

19 have FCS-related mutations.

Bioinformatic issues:
- using a PANGO lineage to define a consensus & inadvertently developing one that doesn't exist in nature for lab work
- indel issues (such as insertions in Mu being lost)
- common assembly/sequencing errors going unrecognized

an example of a highly prevalent issue: toggling of G142D in Delta Spike that turns out to be an ARTIC 3 primer issue (primers have been updated, switch if you haven't yet!!!): doi.org/10.1101/2021.0…

Indels are common & can be tricky: del69/70, del156/157.

del156/157, R158G seems to be what is happening at the nucleotide level but sometimes people instead treat it as E156G, del157/158 -- they're nearly equivalent but the former is preferred.

finally, looking at transitions in selective pressure: enhanced transmissibility is dominant, possibly also various types of antigenic pressures that can be active (esp. recently).

conclusions:
- once a new variant has a start, a more infectious variant can transition rapidly. both D614G & Delta went from rare to dominant in months.
- 74% of PANGO lineages tracked in 2021 contained both NTD&RBD mutations.
- indels are common in both NTD & near FCS.

"Was it possible to forecast/predict that the mutations seen in the Delta variant (or other variants), would arise before the actual variant emerged?"
A: not that particular combination no, but that those were relevant mutations? yes. they're common mutations that recur often.

"Is it possible to use this large amount of sequence information to predict what mutations (and consequently variants) are likely to emerge in the future (next 6 months - 1 year) in low and middle-income countries where vaccination rates are still low?"

A: i think that the same mutations are likely to be recurrent in both but one of the problems is resource distribution leading to undersampling in LMICs; we need to have more sequences in order to know what's going on. again, we have a very bad portrait of Africa.

A (cont): we can predict recurrence of certain mutations likely to recur, but the combination likely to be present in the next world-dominant variant? hard to predict.

"Any hypothesis why in Perú the Delta Variant take months to be the most common variant? And now that Delta Variant how to explain we didn't see a new wave?"
A: i think that... in Peru, P.1+basic was quite fit, which made it harder for Delta to take over, but otherwise, not sure.

next up, Joel Wertheim talking about... once again, missed the talk title thanks to dodgy internet, but i seem to have landed into tMRCA calculations in early Chinese sequences which extend to ~early Dec.

(posterior distribution landing as early as mid-November).

but we have a high suspicion that there was *some* circulation before this tMRCA. how could this happen? it's possible that if your diversification happens later past the index case & early chains are unsampled.

how can we bridge that gap?

FAVITES: a simulation tool for entire phylogenies: ncbi.nlm.nih.gov/pmc/articles/P…

on average: a simulated #SARSCoV2 epidemic using parameters from early Hubei on average takes no more than 1 week until stable coalescence, *however* periods of up to 2 months are possible. let's now do rejection sampling on this...

from this we get that most likely the index case was infected mid-to-late Nov., & *no earlier than Oct. 2019*.

amusingly, this matches virological.org/t/clock-and-tm… quite well!!!

reconditioning on the WHO report's date of Dec. 8th for earliest confirmed case pushes the posterior distribution forward to a maximum likelyhood for index case ~mid-Nov. & index cases even mid-Oct. are *extremely* rare.

(i think the WHO report erroneously eliminates a case on Dec. 1st, but that would really only move things back by a week or so. regardless, the conclusions that any date before October is impossible is consistent from many angles).

rofl, a picture of a raccoon dog sold at the Huanan market, which was denied to be sold there by the authorities. those animals were never found by the time investigators got on the ground, which is... a bit suspicious.

now, looking at simulations of early establishment: running a simulation of 5000 outbreaks, 70.3% of them *went extinct on their own w/no mitigations*. doing this using rural area parameters, 94.5% of them went extinct on their own.

-> we likely just got very unlucky in 2019.

"Do we know the nationality of the index case in Milan?" [i think he means that 4-year-old boy claimed to be a purported early case in Nov. 2019]
A: unknown, but reported no travel history.

"Joel, do you think that there might be transitions in infectivity in the initial spread that could impact the modeling?"
A: yes, & this was taken into account in the paper.

A (cont): in fact, one hypothesis from Kristian's proximal origins paper was cryptic circulation of a less-transmissible variant. however, if such a virus existed, we know one thing for certain: it went extinct, & so it wouldn't've affected tMRCA much, & you couldn't push...

...the ancestral virus back to even Summer 2019. it's just not possible.

"but was it possible this was a lab strain?"
A: lots of reasons why this is unlikely, but most importantly, it's quite baffling why you'd get a lab strain that then would make it to the one place where CoVs tend to emerge/amplify.

"but we can't discount this b/c people working at WIV go to markets!"
A: none of the workers were Ab-positive, & for example w/Marburg in Eastern Europe you could trace those outbreaks back to the lab using standard epidemiology... which was not the case here.

i feel your pain, Joel. i feel your pain.

next up, Sujan Shresta talking about The Impact of the Pre-Existing Cross-Reactive Immune Response in Protecting Against #SARSCoV2!

i was commenting earlier to @nailbomb3 about how there were *no* technical difficulties whatsoever during this entire symposium & apparently it seems i cursed it LMFAO.

we instead are skipping to Ian Wilson's talk on Structural Insights into Antibody Responses to #SARSCoV2 & VOCs!

primarily going to be talking about RBD-targeting Abs.

brief overview on up vs. down conformations of RBDs (which need to be up to bind hACE2 & also some classes of RBDs).

once again, looking at the 4 major classes of RBD-binding nAbs, some of which are extremely germline-restricted which chains they use.

K417, E484, L452, N501 are differentially involved in Ab binding.

K417N reliably takes out RBS-A/class-1 Ab binding, whereas E484K takes out RBS-B/class-2 Ab binding.

wide variation in buried surface area on Spike in Ab binding, some bind at a very narrow area, some in a much larger contact area than the hACE2 binding surface (which is ~800Å^2).

RBS-A/class-1 nAbs are *almost exclusively* from VH3-53/VH3-66 family & are often barely somatically hypermutated!

VH3-53 Abs interact w/*either K417 or E484*, the combo of K417N & E484K takes out both classes of binding modes for VH3-53 Abs (which, again, are by far the most highly enriched in a #SARSCoV2 immune response).

RBS-C Abs also interact w/E484 in most forms, generally forming salt bridges w/it, explaining why E484K abrogates binding there.

even on unusual, highly potent nAbs E484K, E484Q often can completely abolish binding.

RBS-A sensitive to K417N, RBS-B/C sensitive to E484K/Q & some to L452R, RBS-D tend to be resistant.

this is for WT tho, what about convalescent sera from B.1.351-recovered individuals?

VH1-58 most enriched, also VH4-39 & VH3-30. VH3-53 STILL ENRICHED, DESPITE K417N SENSITIVITY. VH1-2 almost nonexistent.

fascinatingly, VH3-53 canonical mode can accommodate K417N: biorxiv.org/content/10.110

VH1-58 clonotype is also elicited in B.1.351-recovered sera & seems to be quite VOC-resistant & even are cross-neutralizing.

let's now look at broadly-neutralizing sarbecovirus Abs: these do not interact w/VOC sites at all! (for instance, CV38-142, which neutralizes both SARS Classic & #SARSCoV2!).

these primarily bind to the S3022 & S309 sites.

finally, let's look at CC40.8, that binds to the stem-helix region of S2 & it seems to bind efficiently to most β-CoVs using all six CDRs; heavy/light chains form a hydrophobic groove in this Ab to accomodate the S2 stem helix.

@profshanecrotty: "For 3-53 ABs generated in response to Beta infection: How do they accomodate 484? Does the nature of that interaction give indications as to whether 3-53 Abs elicited to non-484 SARS2 would be able to accomodate Beta-like 484 by SHM upon exposure?"

A: we don't really have any answer to that yet b/c we don't have any structures of them binding at E484K in the other orientation; we are currently working on that right now.

"why do you think that germline heavy chains w/o much SHM have such high neutralization capability? do you think this is for some reason or just luck?"
A: unknown, but this isn't unprecedented, for instance VH1-59 gives a good fit for flu A.

we will now have a 5-minute break. christ this is *not* easy LOL.

next up, Sujan Shresta (second attempt)! topic is The Impact of the Pre-Existing Cross-Reactive Immune Response in Protecting Against #SARSCoV2!

brief background on T cell cross-reactivity in flaviviruses.

one important question: how does prior immunity to dengue impact subsequent Zika virus infection & vice-versa given that cross-reactivity in epitopes, co-circulation & the same mosquitoes transmitting both.

turns out prior dengue infection *is* cross-protective against Zika & this effect is dominated by CD8+es!

HLA-transgenic mice recapitulate many of the T cell response in humans; this mouse model is now being used for the same in #SARSCoV2: does prior exposure to seasonal HCoVs provide protection against #SARSCoV2? (or does it contribute to pathogenesis?).

studies hint towards a protective effect: less severe disease in people w/prior exposure to seasonal HCoVs, increased CD4+ & Ab responses in individuals w/pre-existing cross-reactive CD4+ T cells; but correspondingly increased low avidity CD4+ T cells enriched in severe disease.

mapping epitopes in vaccinated HLA-DRB1*0101 mice & using @Baric_Lab's MA10 mouse-adapted virus in HLA-DRB1*0101 mice, we map #SARSCoV2 CD4+ epitopes.

several of these have already been reported in the literature.

we now infect mice w/OC43 & indeed looking at D8, D16 & D30 post-infection compared to naive mice, we actually do indeed find extensive cross-reactivity over a lot of the genome!

correspondingly, looking at the Ab level, we see minimal-to-no cross-reactivity.

now, let's see if prior OC43 infection is protective at all against challenge by #SARSCoV2!

it appears the answer is yes! furthermore, this appears to extend to protection against disease!

finally, we deplete CD4+es via antibody & see that in the animals depleted of CD4+es & in depleted animals, viral RNA titer was higher for both genomic & subgenomic RNA, but in OC43-immune animals disease was minimal in both depleted & non-depleted animals, suggesting that...

...it was not primarily cross-reactive CD4+es that were protective against disease (other components seem to be more important here, possibly CD8+es?).

"How likely do you think it will be that vaccinating against SARS CoV-2 will provide protection against common cold Corona viruses in the future?"
A: likely, there will be some protection via cross-reactivity, especially in the future when developing pan-CoV vaccines.

"Has there been any cases of survivors of SARS getting Covid?"
A: out of Singapore, there was excellent work showing cross-protective T cells for #SARSCoV2. beyond this, so few cases means that it's hard to tell for certain, we don't know.

"Is it possible to make a comprenhensive analysis of the immune infiltrate in the lung after infection. is it a possibility that other cells not only cd8 t resident cells (NK like memory) are providing cross protection?"

A: good question, absolutely agree this is hard to tell & a lot of mouse models don't map to humans nicely... we're currently looking into that.

"have you planned to deplete animals of CD8es after OC43 exposure & see if that affects disease? furthermore, combined depletion of CD4+CD8s?"
A: experiments ongoing, we don't have that data yet.

"what about cross-reactive IgA?"
A: good question -- have not looked into that.

"why did you use HLA knockout mice?"
A: we were trying to generate T cell responses that were comparable to human ones, & so used transgenic HLA mice & indeed this seems to recapitulate most human cellular response dynamics.

next up, @SetteLab on Understanding the Targets of Adaptive Responses to #SARSCoV2 & VOCs in "Natural" Immunity & Vaccination!

first question: what assay to use? ELIspot is simple but doesn't provide detailed information. ICS generates more detailed information but is more labour-intensive. we use Activation-Induced Marker Assays, targeting OX40/CD137 or CD40L for CD4+es & CD69/CD137 for CD8+es.

we first looked at Ab, CD4+ & CD8+ responses in uncomplicated #SARSCoV2 infection: cell.com/cell/pdf/S0092… -- it's convenient that S was immunodominant across all 3 axes (not guaranteed: flavi membrane is a good humoral target but not a good cellular target for instance).

this paper was very important -- to the point where Dr.Fauci waved it in a congressional hearing LMAO.

this was followed up by the excellent peptide pool mapping of immunodominance & breadth of both CD4+ & CD8+ HLA-restricted responses: ncbi.nlm.nih.gov/pmc/articles/P… -- hundreds of epitopes identified, each individual recognizes 15-20 epitopes (which vary based on HLA genotype).

now: we are developing immune history classifiers to detect exposure and/or breakthru infections in field studies where past exposure history cannot be determined.

switching gears for a second: age is a major #COVID19 risk factor. part of that risk is being able to mobilize a naive T cell compartment. study of further risk factors is in progress, to see what the relative contribution of each variable is.

one variable is HLA alelle: both breadth & magnitude vary, for instance some alelles have only a few targets that are very strong, whereas others don't result in strong targeting even w/many alelles.

finally, looking at variants, ~85% of a T cell response (no matter which you looked at, CD4+ or CD8+) was preserved no matter what the variant, given the number of epitopes targeted, it seems variant effects are minimal: ncbi.nlm.nih.gov/pmc/articles/P…

-> T cell reactivity largely preserved across *all* known VOCs.

-> Vaccine Efficacy retained against severe disease & death!!!

Reactivity is also detected in non-exposed individuals!

-> could influence both disease severity & vaccine efficacy due to getting a head start!

this indeed seems to be the case to some extent: ncbi.nlm.nih.gov/pmc/articles/P…

interestingly, response to low-dose Moderna vaccination (quarter dose, 25µg) is stronger in those exposed to seasonal HCoVs, which adds to a growing body of evidence showing that cross-reactive T cell responses have some benefit in prevention of #COVID19 severity.

"In the CD8 T cell data looking at number of epitope and magnitude of CD8 response, in general, the greater number of epitopes the less CD8 magnitude. Have you done a similar analysis witih CD4 T cells?"

A: that is possible, but we don't really know b/c the HLA restriction of CD4+es is a lot more difficult to determine.

& now it's time for the final speaker of the afternoon talk of this symposium, @profshanecrotty talking about Immune Memory to #SARSCoV2 Infection & Vaccines!

trying to determine strength & duration of immune memory to a #SARSCoV2 infection across memory B cells, antibodies & both CD4+ & CD8+ T cells (across 188 individuals): ncbi.nlm.nih.gov/pmc/articles/P…

looking at memory B cells via flow cytometry... this was the biggest surprise, in that memory B cells were *more* expansive at 3-6 months than 1 month post infection, looking at both Spike RBD & nucleocapsid responses.

(this doesn't seem to be unique, a similar effect is known to occur in yellow fever).

for IgG, most people maintained IgG reactivity robustly for S RBD out to 8 months.

virus-specific CD4+ & CD8+ T cell responses appear to be maintained quite well, especially w/CD4+ responses occuring in 100% of people at 1 month post-infection & maintained in >95% of people at 8-months post-infection.

so, how long does #SARSCoV2 immune memory last?

well, it turns out that immune memory is complex. various compartments have different kinetics, also there is heavy inter-personal variation for each.

however, what we *can* say is that ~95% of people maintain out at 6 months at least 3 memory compartments out of: memory B cells, IgG, IgA, CD4+es, CD8+es.

indeed, it turns out that even in people that have fairly low levels of circulating Abs a long period out past infection, people appear to be robustly protected against reinfection.

in summary: it is reasonable to consider that hospitalization-level #COVID19 is prevented by any decent combination of Abs, memory B cells & CD4+es/CD8+es!!!

looking at anatomy of adaptive immunity to #SARSCoV2: different tissues have different kinetics: infection in URT is *fast* (esp. w/Delta!) but lung infiltration is slow: ncbi.nlm.nih.gov/pmc/articles/P…

vaccines give you a head start on the virus!

onto vaccines, we took a look at low-dose Moderna (25µg) & what memory responses it induced: science.org/doi/10.1126/sc…

we were pleasantly surprised that there was only a 2-fold decline btwn peak & 6-month post 2nd immunization; almost 100% of people had CD4+ T cell memory 6mo out!

the CD8+ response is not quite as robust, only 90% of people generated a response at this dose w/~67% maintaining this at 6 months out.

-> most decline in this period; reasonable to expect memory T cell responses to last for *at least* years.

finally, CD8+ responses if anything tend to improve in older pts (conversely slight declines in Ab & CD4+ response, but very slight).

"for CD4+ or CD8+ protective responses, are those largely targeting Spike or elsewhere?"
A: for T-cell intrinsic responses, it should be just as protective no matter where the peptide targeted is, be it in Spike or elsewhere.

"Any early look into tissue specific immune memory responses (TRM, IgA at mucosal sites) related to infection and vaccine?"
A: we just published a paper on this last week! science.org/doi/10.1126/sc…

"is the reduced VE against infection due to limited epitope breadth & would it help to add additional viral proteins?"
A: Spike is quite immunodominant even for T cells due to being a large protein, so it has a good response.

A (cont): would adding another antigen help? possibly, but from a regulatory perspective it's easier to just tweak our current vaccines than run a whole new clinical trial if you added another protein. would it be valuable? probably, but likely not by that much.

and that concludes the afternoon talks (which were likewise excellent). one more short break then the closing keynote at 17:00 (8AM for me lmao).

HOLY SHIT THIS IS HARD OKAY.

& now for the closing keynote by Stanley Perlman, talking about Animal Models of #COVID19!

6 known HCoVs (pre-#SARSCoV2): SARS Classic, MERS-CoV, HCoV-OC43, HCoV-HKU1, HCoV-229E, HCoV-NL63.

immunopathology in MHV: why is it that RAG-/- mice w/no T cells that survive can handle massive amounts of CNS viral load fine but when adding back T cells you get both viral clearance *and* demyelination!

recent spillovers: 8 children in Malaysia developed pneumonia, causative agent was CCoV-II w/S1 from FCoV-I & S2 from FCoV-II.

interestingly, when "common cold" HCoVs reach LRT they can cause severe disease, particularly HCoV-OC43.

#SARSCoV2 combines transmissibility of "common cold" HCoVs w/LRT behaviour of SARS Classic, leading to it being such a huge public health burden compared to SARS Classic.

looking at #COVID19 CNS/PNS manifestations, sequelae common, range from brain fog to seizures to anosmia. these may last for months & are disabling.

pathogenesis is not understood; extremely little evidence if any for direct viral invasion.

anosmia is a well-established complication of #COVID19; generally resolves after infection but not always, may occur w/parosmia (miswired smells, i.e. coffee smells not like coffee, but, say, sewage now) & phantosmia (smelling things that aren't there).

recent review of #COVID19 effects in the olfactory system: ncbi.nlm.nih.gov/pmc/articles/P…

olfactory receptors not infected, infection seems to occur in sustentacular cells along w/myeloid cell infiltrate.

LMAO THE AMAZON SCENTED CANDLE REVIEWS.

recounting this anecdote:

https://twitter.com/wanderer_jasnah/status/1402278948722163717

(hilariously, apparently her idea of "cheap" wine is still more expensive than Stan's idea of expensive wine LOL).

animal models:
- hamsters, macaques, ferrets, cats are all susceptible, but all develop generally mild disease.
- hamsters are *really* good at transmitting virus to other hamsters (TELL ME ABOUT IT LOL).
- mice are resistant to the ancestral strain.

talking about outbreaks on mink farms, mink -> human -> mink transmission chains (science.org/doi/10.1126/sc…) & the possibility of animal reservoirs becoming established.

also talking about recent evidence of deer being infected.

various ways of making mouse models: expression via viral vector, knock-in of hACE2 under various promoters, mouse adaptation of the virus.

talking about K18-hACE2 mice & how they're insanely susceptible to death from neural infiltration of SARS Classic, 3 PFU can be lethal.

interestingly, #SARSCoV2 is less absurdly neuroinvasive in K18-hACE2 mice, making them not completely useless. you get lung inflammation & occasional thrombosis in them, so there *is* some utility, just need to be careful.

can we use these mice to study anosmia? turns out yes, & indeed you get similar effects where sustentacular cells are reliably infected but olfactory neurons are not.

indeed, scent preferences changing in both male & female mice showing equalization of odour preference (for male mice: male vs. female, for female: novel vs. familiar) as the infection progressed.

similarly, control mice find buried cookies in seconds, infected mice could not even after days.

(important note: the mice were not given a glass of milk afterwards).

fascinatingly, convalescent plasma prevents death reliably but does not affect anosmia in these mice.

olfactory cilia are severely damaged post-infection! how this is occuring is unknown.

"Since there are a number of ACE-2 receptors on olfactory cells in the nose, has anyone looked at the the effect of vaccination with the spike sequence/protein and the effect on sense of smell?"
A: haven't looked at this yet, but effect doesn't seem to be prominent.

"<obligatory question about ADE>"
A: not observed in vivo, not observed in vaccines *at all*. in macaques macrophage phenotype was affected, but not outcome severity. still questions exist, but it does not appear to be a risk.

"In the K18-hACE2 human mouse study, it seems there is a gender difference by anosmia in that female mice seem to be more affected?"
A: there actually doesn't seem to be ultimately any differences in histological changes be they olfactory downregulation or cilia damage, but...

...we haven't looked quite that closely at this level.
"is there gut pathology in K18-hACE2 mice?"
A: we have looked but there doesn't seem to be much occurrence.

fascinatingly, major differences in neuroinvasiveness of SARS Classic vs. #SARSCoV2 in K18-hACE2 mice despite this being inherently specific to these mouse models.

"Is anosmia associated with olfactaory neurons dying or damage? Have you looked at the neurons for death and/or altered function during infection?"
A: WE DO *NOT* SEE ANY OLFACTORY NEURON DEATH OR ALTERED FUNCTION.

& with that, this symposium is OVER!!! bloody hell that was fucking difficult but *SO COOL*. thank you to all the speakers for *WONDERFUL* talks!!!

follow @ljiresearch on twitter for more updates! hope to see you next year in-person! (indeed, would *love* to drop by in-person myself someday).