Variability ≠ “We don’t know” how COVID vaccines work

It’s been a while since I’ve written about a basic science paper misinterpreted and/or misused by antivaxxers. It’s also been a while since I’ve encountered an antivax influencer or blogger with whom I was unfamiliar when first seeing their blather. So when I came across a post by someone going by the ‘nym of Joomi entitled, New study shows how little we know about how mRNA vaccines “work”—along with a blurb claiming that “Vastly different levels of spike protein depending on cell type”—I knew I had my topic for today. I have no idea who “Joomi” is other than this description from the Substack Let’s Be Clear:

I’m a biologist. My PhD research was in how the cell cycle connects to metabolism. I love science, but abhor Institutionalized Science. You can follow me on Twitter here.

So apparently this blogger’s name is Joomi Kim, if her Twitter profile is any indication and the photo is actually of “Joomi.” (These days, I take nothing for granted.) It does appear that Joomi Kim is whom we’re dealing with, as a quick Google search found an interview with her on a podcast about a Substack post by her comparing the disinformation leading up to the invasion of Iraq 20 years ago to the “information climate” around COVID-19 and the invasion or Ukraine as well as a list of publications over at ResearchGate. Unfortunately, Kim apparently got her PhD at Rutgers University, where I was faculty with The Cancer Institute of New Jersey before I switched jobs 15 years ago, which made me sad to learn, even though she was apparently in the Environmental Biophysics and Molecular Ecology Program, Department of Marine and Coastal Sciences at Rutgers. It didn’t take me long to find her article, published about a year ago, and, indeed, she did liken the disinformation campaign used by the Bush Administration about weapons of mass destruction in Iraq in the wake of 9/11 to justify the invasion to “nonstop COVID propaganda” during the pandemic, while asking;

How much can we trust the videos and images coming from Ukrainian officials? Are they any better than the anonymous sources, aka “US officials,” in the lead up to the Iraq War? What really happened in Bucha (see here and here)?

So, right away, perusing her Substack, listening to part of an interview with her, and seeing her painting messaging on COVID-19 and the Russian invasion of Ukraine as being very much like the Bush/Cheney disinformation campaign about weapons of mass destruction in Iraq, I knew what I was dealing with.

Still, Kim states that she’s a biologist—although in a different post in which she lamented having been “deceived” about COVID vaccine safety and referencing antivax cranks Steve Kirsch, Jessica Rose, and Mathew Crawford, she says she “used to be a biologist”—and had done her PhD work “manipulating algae to make lipids for biodiesel,” after which she worked in tech for a while before joining a startup called Goodloops, which turns out to be a fly fishing company, at least if her ZoomInfo profile is any indication. I suppose it’s possible that this isn’t the same Joomi Kim, but she did mention having recently joined a startup called Goodloops in her interview, and the podcast is dated August 2022. Given that her profile now says she’s been with Goodloops for year, this information all seems to indicate that we’ve found our Joomi Kim, but, again, I take nothing for granted.

Whoever Kim is, I was actually rather surprised at her take on the study in question, an in vitro study in which the researchers tested the level of expression of spike protein (how much spike protein is made) in different cell lines in cell culture when they exposed to the Pfizer and Moderna COVID-19 vaccines, which are mRNA-based. I’ll start with her take as posted on her Substack, and then I’ll look at the study itself, which in which “two different cell types (Jurkat or K562) were given either the Comirnaty (Pfizer) or Spikevax (Moderna) vaccines.” Much to Kim’s amazement, the study found major variability in the expression of spike protein depending on the vaccine, noting:

We also see that cells that were given the Comirnaty vaccine showed a lot less spike expression compared to Spikevax.

And including this graph, showing that variability:

Here’s how the experiment was done. The cell lines were exposed to vaccine, incubated for 24 hours in the presence of the lipid nanoparticle/mRNA vaccines, after which the cells were harvested and subjected to flow cytometry to determine what percentage of the cells were expressing (making) spike protein. Where did the authors get the vaccines for their experiments? Simple. They used leftover discarded vaccine, basically what’s left in the multidose vial that doesn’t get used:

Three vaccination centers in Perugia (Italy) provided us with residual vaccines present in vials after administration. The vials were collected between September and December 2021, during the first cycle of vaccination (first/second dose). The content of four vials of each vaccine was pooled under sterile conditions and used within 1 h after administration of the last dose.

My first question was whether the variability was due to that, but then I remembered something, which, to her credit, Kim mentions, namely that the Moderna SpikeVax has way more mRNA in it than the Pfizer Comirnaty vaccine does:

This may not be that surprising given that Spikevax has a higher concentration of mRNA. In this study, 1 and 10 µL of both vaccines corresponded to 0.1 and 1 µg mRNA of the Comirnaty vaccine, and 0.2 and 2 µg mRNA of the Spikevax vaccine.

However, what is surprising is that in the Jurkat cells, there isn’t that much difference in spike expression between the cells that got 1µL and 10µL Comirnaty. Strange.

Being a biologist, Kim should also have been aware of lots of other factors that could account for this supposedly astonishing result. I thought of a few right off the bat. For instance, the lipid nanoparticle carriers used by Pfizer and Moderna might also have had different levels of efficiency in transfusing cells (binding to the cells and helping the mRNA get into them). The authors even mentioned these possibilities:

Differences in S-protein expression levels following vaccine treatment may be attributed to variations in the efficacy of lipid nanoparticles, differences in mRNA translation rates and/or loss of some lipid nanoparticles’ properties and mRNA integrity during transport, storage, or dilution, and may contribute to explaining the slight differences in the efficacy and safety observed between the Comirnaty and Spikevax vaccines.

Note: Slight differences in efficacy and safety between the two vaccines.

Nor is it particularly surprising that in one cell line there wasn’t much increase in expression at a ten-fold higher dose of vaccine. What easily could have accounted for would be a 1.0 µL dose already being close to the saturation point for the cells, so that adding more doesn’t result in a lot more mRNA getting into the cells. It is not at all uncommon for a dose-response curve for a drug to reach a plateau, and likely that was the case for the Jurkat cells. Moreover, the only thing measured was the percentage of cells positive for spike protein. They did not do any Western blots to examine overall protein levels after treatment. It is, of course, possible that the K562 cells might have been making a lot more spike protein in a lower percentage of cells.

Indeed, look at the higher dose, and you see a lot less variability in expression compared to the lower dose. While this could be due to variability in the vaccine batches used, more likely it is due to approaching that plateau at a lower dose of mRNA. The way to work that out, of course, would have been to do a more detailed dose-response curve that went down to lower doses, say 0.10 µL, or one-tenth the lowest dose used in this experiment. Seriously, this experiment is lacking in data points in its dose-response curves. Perhaps the authors couldn’t get enough vaccine to use, but, even so, the data presented here are pretty sketchy. Had I been reviewing the paper, I would have asked for at least a couple of more doses on the dose-response curves. I rather suspect that if the investigators had included a dose between 1 and 10 µL and maybe a dose lower than µL and higher than 10 µL, what we would have seen is a pretty conventional sigmoidal dose-response curve looking something like this representative curve that I got from Wikipedia:

Indeed, I facepalmed when I read this part of the discussion in the paper:

Notably, dose escalation of the Spikevax vaccine increases S-protein expression (Figure 1B (right panel) and Figure 2), whereas dose escalation of the Comirnaty vaccine does not (Figure 1B (left panel) and Figure 2).

Note that the authors have not demonstrated that at all, as they have done an utterly inadequate dose-response curve, with only two data points. Again, how do they know they didn’t start near the plateau of the dose-response curve. Answer: They don’t, because they didn’t test enough doses.

Kim seems to draw a much stronger conclusion than is warranted about the differences in expression between the two cell lines as. well, with treatment of the Jurkat cells resulting in about two- to three-fold more cells expressing spike protein than the K562 cell line. Again, it is not at all uncommon for different cell lines to have different levels of expression after introduction of DNA plasmid or mRNA sequence designed to drive expression of a gene.

She also makes a lot of the observation that two forms of spike are made, both the full-length spike, which is, as designed, embedded in the cell membrane, and truncated spike protein, which floats loose from the cell membrane in the supernatant media in which the cells are submerged. Surprise! There’s variability in how much soluble spike protein fragments each cell line makes after treatment with each vaccine:

If we focus on the bars that received 10µL of Spikevax (circled in red); this time, we see more spike associated with the K562 cells compared to the Jurkat cells (notice the difference in scales). This is the opposite of what we saw with spike that was attached to the surface of cells.

Remember what I said about the Western blot and how K562 might be making more spike in fewer cells? This rather suggests that I might have been correct. Still, this is a pretty bare-bones set of experiments, even as observational experiments, to justify all the handwaving in the discussion about what might be causing the differences, which include reasons alluded to above and others including differences and variability in:

• Lipid nanoparticle composition
• Cell line characteristics
• mRNA dose
• mRNA sequences surrounding the gene sequence for spike protein
• Storage conditions under which the vaccines were stored
• Dilution necessary to achieve the desired dose
• Codon optimization strategies in the sequences of the two vaccines. Codon optimization involves which of the redundant three-nucleotide codes, or codons, are chosen to code for each amino acid in the protein, with different codons resulting in different efficiencies with which the protein is translated (made).

This paper claimed to “evaluate whether the S-protein expressed following treatment with the two vaccines differs in the real-world context.” However, testing just two cell lines, both of which are highly artificial compared to normal human cells, is hardly “real world context. Jurkat cells, for instance are an immortalized cell line—artificially altered so that they can divide indefinitely and never become senescent, or unable to divide anymore—human T lymphocyte cells originally established in the mid-1970s from the peripheral blood of a 14-year-old boy with T cell leukemia commonly used to study acute T cell leukemia, T cell signaling, and the expression of various chemokine receptors susceptible to viral entry, particularly HIV. In contrast, K562 cells were the first human immortalized myelogenous leukemia cell line to be established and were originally derived from a 53-year-old female chronic myelogenous leukemia patient in blast crisis, which is the phase of chronic myelogenous leukemia in which more than 30% of the cells in the blood or bone marrow are blast cells (immature blood cells). In other words, neither of these cell types are much like the normal muscle and lymphoid cells exposed to the vaccine that take up the mRNA for spike protein, and they are grown on Petri dishes in growth media in incubators, rather than existing in organs in living animals or humans. As such, using them to test how much spike protein the mRNA vaccines can induce cells to make is not how I would choose to test how the vaccines work to produce spike protein in the “real world.”

Far be it from me to disparage testing how well mRNA vaccines work to induce different cells to produce spike protein, but this is a shockingly bare bones effort: Only two cell lines, chosen apparently because they can be purchased from ATCC and represent T cell-derived cell lines, is wholly inadequate, as were the range of doses of vaccine chosen. I realize that perhaps the investigators were very limited in how much “real world” vaccine they could get their hands on, although by the time the study was carried out in late 2021 vaccine supplies were abundant in Europe. Even so, just two cell lines that don’t include any muscle cell lines? Only two doses per cell line? This paper reeks of leftover data that the authors tried to get published as what we in the biz call a “minimal publishable unit” (MPU), which is not a compliment. Also, don’t get me wrong. I’ve been very tempted to do some MPU papers myself, particularly now given that progress in the laboratory has been slow. I also hope, though, that if I do succumb to the temptation I won’t also be tempted to oversell my results as more important than they are, as these investigators did when they tried to suggest that they might explain differences in the likelihoods of each vaccines to cause myocarditis:

Notably, myocarditis following vaccination with mRNA-based vaccines affects young males much more frequently than other demographics [13,16,17,18]. In these subjects, Spikevax shows a higher frequency of myocarditis than Comirnaty, with increased risk ranging from 2.5 to 8 folds in different studies [16,17,18]. The higher mRNA dose of Spikevax compared to Comirnaty is believed to be the reason for the increased incidence of myocarditis. If the different in vitro S-protein expressions by Spikevax and Comirnaty vaccines reflect in vivo conditions, our results could contribute to explaining the disparity in myocarditis frequency.

That’s a huge “if” there, or, as I like to say, “if” is doing some very heavy lifting there.

But back to Joomi Kim, who also notes that a higher fraction of K562 cells die 24 hours after transduction with Spikevax (Moderna), specifically 3.7% of the Jurkat cells and 11.4% of the K562 cells, leading her to conclude:

So K562 cells seem to be more susceptible to vaccine toxicity compared to Jurkat cells.

Unfortunately, the controls used for this were not adequate. Specifically, the controls were untreated cells, which makes it impossible to tell if the differences were due to the lipid nanoparticles, the mRNA, or something else. At the very least, a lipid nanoparticle-only control without the mRNA in it should have been included if at all possible, although I realize that if you’re using commercially available vaccines you get the whole package. Also unmentioned is that even the untreated K562 cells had 2.2% dead cells after 24 hours compared to 0.8% of untreated Jurkat cells, which should have been mentioned. If you think of it this way, you’ll be less impressed. Spikevax increased cell death by 4.8-fold in Jurkat cells and by 5.3-fold in K562 cells, which is not a very impressive difference at all.

I could go on and on pointing about examples in which Kim is unduly impressed by these results, but let’s go on to see the conclusions she makes about the variability in this study:

The main point is this: we do not understand fundamental things about what happens after mRNA vaccines are injected into people.

For example, we saw that the Pfizer vaccine led to less spike protein expression, and yet there are studies that seem to show that the efficacy between Pfizer and Moderna are similar (example here). How could that be?

And what accounts for the differing levels of spike protein expression in the different cell lines?

Jurkat cells are T lymphocyte cells that were originally derived from a 14-year old boy with leukemia. K562 cells are bone marrow cells that were originally derived from a 53-year old female with leukemia. 

Do the levels of spike expression differ between these cell lines because they are from two different people, or because they are different cell types, or both? What would happen with heart or brain cells? How much spike would they express if they accidentally took up vaccine LNPs?

And why is it that we see more surface spike in one cell line, but more loose spike in the other? Is it because one of the cell lines express more enzymes that cleave the spike off the surface of cells?

“Fucking magnets, how do they work?“

Not to say that the questions above are not worthy of investigation, but that’s not what Kim seems to be doing here. Rather, she is JAQing off, handwaving, and incorrectly concluding that the variability in spike production between the two vaccines in two different cell lines has way more to say than it actually does about how the vaccines work and might produce different risks of myocarditis than it, in fact, does, which is actually very little to say.

Not that that stops Kim from concluding:

As usual, the more we learn about these vaccines, the more questions arise.

Maybe, but we also have copious data showing that the vaccines are very effective at preventing severe disease and are very safe. What Kim is doing is what antivaxxers always do: Take scientific observations (particularly rather uninteresting ones from lower quality publications reporting in vitro studies) and then using variability in the results to cast doubt on vaccine safety. That’s what science deniers do. They misunderstand, either unintentionally or sometimes intentionally, the role and significance of uncertainty and variability in scientific findings. Because they want certainty, variability and uncertainty frighten them and they turn that fear into doubt and do their best to make that doubt contagious by harping on what is not known rather than what is known.