A recent paper has found spike protein antigens in the blood of people who received the mRNA-1273 vaccine from Moderna, and this is naturally being spun to mean the vaccines are dangerous. But actually there are several points about this paper that are very reassuring. Firstly, the reason some are concerned has to do with the discoveries that the spike protein alone may play a role in some of the disease manifestations of COVID (such as those explored here, and here). I have a detailed post on spike that I am working on to explain basically everything anyone ever wanted to know about the protein but in brief when spike protein interacts with the endothelial cells that line the blood vessels, it has been shown to reduce expression of the ACE2 protein (this makes some sense as ACE2 is how the virus enters cells, so naturally the cells are driven to evade infection) and this contributes to inflammation of the endothelial cells. Additionally, in Buzhdygan et al, it’s shown that spike protein can affect the permeability of a model of the blood-brain barrier, which may lend clues as to some of the neurological diseases that seem to result from COVID-19. Spike is also shown to activate the NLRP3 inflammasome which can induce inflammation that alerts the immune system to a problem (but can cause damage if allowed to go on uncontrolled); this can cause cells to undergo an inflammatory type of cell death called pyroptosis.

Anyway with that context, let’s discuss the paper. The paper uses a technique called SIMOA to identify the S1 fragment of the spike protein and the intact spike protein in the plasma of vaccinees. I’ll focus on those data which are shown below:

This finding is initially a bit surprising. A key mechanistic point regarding the safety of the vaccines has been that the spike protein does not get to freely float around in the blood where it might be able to cause these deleterious effects we discussed earlier. We know that the spike protein does not get secreted because it lacks the signal sequence for that, and the protein as specified by the mRNA is membrane bound. However we need to be aware of a few things. Firstly, this assay is measuring picograms per milliliter quantities of the proteins. A picogram is a trillionth of a gram, so this is a very small quantity indeed. Additionally, we see quite clearly that the S1 and spike proteins both disappear from the blood despite their initial tiny concentration (note the day scale). And intact spike is scarcely detectable in any of the participants (though there weren’t many). In other words, a test that doesn’t get to this level of sensitivity- picograms per milliliter- will likely show that no spike antigens are detectable in the vaccinees. The question now, however, is how small this really is- how does this compare to those studies that did find a concerning effect from spike protein alone?

Ogata et al summarize their results:

S1 antigen was detected as early as day one post vaccination and peak levels were detected on average five days after the first injection (Figure 1A). The mean S1 peak levels was 68 pg/mL ±21 pg/mL. S1 in all participants declined and became undetectable by day 14. No antigen was detected at day zero for 12 of 13 participants, as expected. However, one individual presented detectable S1 on day zero, possibly due to assay cross reactivity with other human coronaviruses or asymptomatic infection at the time of vaccination. Spike protein was detectable in three of 13 participants an average of 15 days after the first injection. The mean spike peak level was 62 pg/mL ± 13 pg/mL. After the second vaccine dose, no S1 or spike was detectable, and both antigens remained undetectable through day 56. For one individual (Participant #8), spike was detected at day 29, one day after the second injection and was undetectable two days later.

Let’s be cautious and round up to 100 pg/mL of S1 subunit and 100 pg/mL of spike protein in the plasma of these vaccinated individuals. Before we begin though, a caveat: the spike protein in the Pfizer and Moderna vaccines is not quite the same as the wild-type spike protein found on the virus. This protein has been prefusion stabilized which means it lacks the ability to change conformation into its postfusion state (via a double proline substitution). This change is thought to significantly enhance the ability of the spike protein to elicit neutralizing antibodies from the immune system, but it also has another functional consequence: the spike protein has drastically less ability to cause syncytium formation. Syncytia form when cells fuse together, as may happen in this case if the spike protein binds a receptor (e.g ACE2) on an adjacent cell and form a single cell, potentially leading to the formation of giant cell structures. It’s worth noting that it is thought that Merck’s rVSV-vectored COVID-19 vaccine was ineffective because there was not sufficient viral receptor at the injection site for it to adequately stimulate the immune system and elicit a productive response. However, syncytium formation may indeed play a direct role in the disease process of COVID-19, as demonstrated by papers here (where it’s suggested that this process results in killing of cells of the immune system and interferes with the ability to initiate a productive response) and here (where it’s discussed how syncytium formation may play an important role in the pulmonary disease inherent to severe COVID-19). For this reason, the spike protein of the mRNA vaccines, as well as the Johnson and Johnson/Janssen vaccine, likely has important differences in its properties compared with the wild type spike proteins of SARS-CoV-2. But for the sake of argument, let’s be cautious and assume that the biological, pathologic properties of spike protein as produced by vaccinees is the same as that from virus (though again, this is likely not entirely true).

Using Lei et al as a reference, the toxic effects of spike resulted from a concentration of 4 mcg/mL on the endothelial cells. A microgram is one-millionth of a gram. If we assume that the plasma concentration of spike and S1 were 100 pg/mL we can conservatively estimate that this concentration is 40,000 times higher than that which was detected in the patient’s plasma.

Using Buzhdygan et al, a concentration of 10 nM was used- a nanomolar is 1 billionth of one mole per liter of solution. Spike protein has a mass of about 146.1 kDa (divide the mass of the structure by 3 because that’s the trimer) and the S1 subunit has a mass of about 76.5 kDa. A 10 nM solution of these would equate to 14,610,000 pg/mL and 7,650,000 pg/mL respectively which are respectively 146100 times and 76500 times more spike protein than is found in plasma of vaccinated people. Here’s the math for the calculation as it’s a bit more involved due to the many conversions:

Ah but I hear you protesting- the experts lied! They said no spike circulating- clearly there’s spike circulating. Not exactly. For one thing, the data available until this point didn’t show evidence of spike circulating, and we have a tendency in shorthand to say that that means there is no spike because we can’t prove a negative. All assays have limits of detection (in this case it’s labelled). A 10 nM concentration is very small- and yet this is still about 100,000 times more spike than what we find in plasma. This assay is pretty special to be able to find anything reliably at this concentration and I would be skeptical of its accuracy at this level if not for the time points that these things are appearing. Also note that this isn’t evidence of spike protein being secreted by the cells that receive the mRNA, which was the key consideration behind such claims and indeed based on the tiny quantities noted, that doesn’t appear to be happening. The appearance of intact spike in the plasma of this admittedly small sample is very rare and transient. The authors attribute it to T cell killing of infected cells, which seems plausible (though I would imagine that this would generally occur by an apoptotic pathway which should preserve the contents of the cell as being membrane-bound, although nothing is absolute and given these quantities it could reflect small leaks). My initial guess was that this was from pyroptosis triggered by spike but that doesn’t fit with the timing of the appearance of spike protein (it happens too late). Roy Heesbeen suggested that the spike protein may be spontaneously forming virus like particles on the surface of the cells which get released when they concentrate at high levels (I don’t know that spike has been shown to do this- but it has been seen with other proteins as he points out). The authors are uncertain about the source of the S1 subunit- I think there’s probably a protease on the surface of transfected cells that cleaves spike and that’s accounting for this.

So in short, this study is interesting- but it in no way impugns the safety of mRNA vaccines for COVID-19, or mRNA vaccines as a whole. In fact I would take it a step further and say we don’t need to know anything about the biology of a pharmaceutical to make judgments about whether or not it’s safe. That determination is based on epidemiologic surveillance. To date, the epidemiologic data on mRNA vaccines is exceptional and reassuring: anaphylaxis can occur but is rare and very treatable. Otherwise no events have produced safety signals in these vaccines to date, and outcomes in pregnant patients are reassuring. We have given out hundreds of millions of doses of these vaccines and despite a pharmacovigilance system sensitive enough to detect an adverse event reported in fewer than 1 per million doses, we are seeing no such problems with mRNA vaccines. Papers analyzing the mechanisms of how these vaccines work are valuable because they can be used to guide smarter vaccine design. But they cannot tell us anything definitively about safety. All of this data must be held in context. COVID-19 has killed nearly 600,000 Americans, and the death toll globally is staggering. People are additionally experiencing disabling complications after getting over even seemingly mild cases. Vaccines are the way out.