mRNA vaccines… still don’t reverse transcribe and integrate into your genome
The gist: Not too long ago, there was a paper published which was seized upon by anti-vaccine circles purporting to show that there is production of DNA encoding the spike protein from the mRNA vaccines- something that was long stated to be impossible. In truth, however, the paper does not show this. With the most charitable interpretation I can offer, the paper shows that a small fragment of the mRNA vaccine can be used as a template to make a DNA copy in a cell line derived from a liver cancer at doses far larger than any that are experienced by any actual liver cell. It does not at any point establish or attempt to establish that the DNA in question is integrated into the genome of the liver cancer cells though, which would be critical to establish because DNA that is not integrated into our chromosomes is lost as cells divide. Beyond this, there is an absolutely critical flaw here in that liver cancer cells make a protein called LINE-1 which has the ability to carry out reverse transcription (taking an RNA copy and making DNA) and integration into the chromosomes- which normal, healthy cells do not do because they do not make LINE-1. Additionally, the fate of a cell that expresses the spike protein is ultimately death- the immune system will treat it like an infected cell and kill it. In the setting of cancer, that could very well be a good thing, and in fact, mRNA vaccines have a long history predating COVID-19 for cancer immunotherapy. In short, this study is so flawed that I would argue it does not merit publication and has no relevance to vaccination and I would urge that it be retracted given both these flaws and how it is being misused in the discourse.
The latest study that some anti-vaccine activists have latched onto is one which claims that the mRNA vaccines are reverse transcribing into the genome, a claim I’ve addressed before, which some are extending to indicate the vaccines are unsafe. In reality, this study doesn’t show much of anything beyond that cancer cell lines behave… like cancer cell lines. I explain below.
A summary of the Central Dogma of Molecular Biology as taken from my old molecular biology notes (I am not positive what the original source of this figure is and if anyone knows I would be more than happy to provide credit).
Background
Much of this is redundant with my prior posts on the subject but here’s a condensed version so you don’t have to dig through them. A good starting point is the central dogma of molecular biology which states (in simplified terms):
DNA serves as a template for RNA which results from a process called transcription which is facilitated by an enzyme called a DNA-dependent RNA polymerase (or RNA polymerase for short).
There in turn mRNA (messenger RNA) serves as a template for the production of proteins in a process called translation mediated by a complex molecular machine (think 3D printer) called a ribosome.
DNA is also capable of serving as a template for DNA in DNA replication, mediated by an enzyme called a DNA-dependent DNA polymerase (but commonly just called a DNA polymerase).
RNA can also serve as a template for RNA but this is mainly done by viruses with RNA genomes via an enzyme called an RNA-dependent RNA polymerase (commonly abbreviated RdRP).
But, importantly, in general, RNA cannot reverse transcribe back into DNA.
In (healthy) human cells one of the major reasons is a geographical one. Mature RNA is in an entirely separate compartment of the cell from the DNA with no simple means to get to it.
The second reason is that in general, human cells do not have any significant reverse transcriptase activity. But there are exceptions.
The entity that is probably most famous for reverse transcription is likely HIV, a retrovirus. The reverse transcriptase makes a DNA copy of the virus's genome and the virus's RNA genome is destroyed in the process. But the fact that it makes a DNA copy of its genome is not in fact what makes HIV so persistent. That requires a second enzyme that is encoded by HIV: integrase. Integrases cut and attach DNA into our genome. In this way, the HIV genome can persist inside cells and any daughter cells they divide into for as long as those cells live. If not for integrase, over time, HIV's genome would be lost as the cell divided. Thanks to integrase, HIV can persist in the genome of the cells it infects and any daughter cells that result from the divisions of that infected cell. In general, DNA that is not associated with a chromosome will progressively be lost during cell division (herpesviruses are an exception to this but even so rely on being physically associated with chromosomes to persist). Most probably, this is related to the fact that during cell division, a mitotic spindle pulls apart sister chromatids during anaphase, attaching at the centromere; without a centromere, such as if the DNA is not associated with a chromosome, that cannot occur, and so the DNA does not continue to propagate with the daughter cells.
But we have other reverse transcriptases in nature too. For instance, stem cells (as well as cancer cells) express high levels of an enzyme called telomerase, which maintains structures at the end of chromosomes called telomeres. Before a cell divides, the chromosomes replicate and in so doing, the chromosomes shorten a bit. Telomeres are caps at the end of the chromosomes that prevent the replication from eventually erasing something important, and the telomerase enzyme generates them with an RNA template. The reason that we don’t maintain high expression of telomerase indefinitely is that cells accumulate damage to their DNA as they age and that can lead to problems down the line- so having them continue to divide in spite of that could readily result in things like cancer.
Another genetic element (formally known as a retroelement) capable of reverse transcription is LINE-1 (L1). LINEs make up a sizable chunk of our genome (about 17%), and can be thought of more or less as evidence of ancient encounters with retroviruses which eventually lost the ability to make functional retroviruses and are now essentially genetic parasites (although the exact role and importance of LINEs is debated). Most LINEs, including L1s, are fossils- nothing productive is generated from their sequence. A very small fraction however do produce a product capable of a phenomenon called retrotransposition. This is essentially a copy-paste mechanism in our genome. L1s encode 2 proteins which form a structure that recognizes RNA sequences (through an RNA recognition motif), a structure that has reverse transcriptase activity, and a structure that has endonuclease activity, together giving it the ability to reverse transcribe and integrate the DNA it makes from reverse transcription into the genome (L1s, however, are thought to do this mainly during cell division, when the nuclear envelope dissolves rather than the conventional means of nuclear trafficking through the nuclear pore complex). Like telomerase, the L1 proteins do not grab RNA at random (though they are not that picky). In particular, L1 shows a strong preference for its own mRNA for reverse transcription, which is known as cis preference (it likely results from a proximity effect of the RNA being right by the protein once it is synthesized on the ribosome and from requirements of the mRNA sequence that is retrotransposed). Another key consideration is that L1 is transcriptionally repressed in most healthy cells.
The Paper’s Findings
Figure 1 demonstrating the sequence of Bnt162b2 and the PCR amplicon.
Figure 3 showing the relative expression of LINE-1 across different concentrations of Bnt162b2 and over time.
Aldén et al use Huh7 cells to interrogate whether or not the Pfizer/BioNTech mRNA vaccine (Bnt162b2) can be reverse transcribed into the cell. The basic design concept is simple enough and relies on qPCR, an extremely powerful, sensitive, and specific analytical technique (provided you have well-designed primers) that has been the subject of much ire from some during this pandemic. The schematic for their qPCR examining this question however already betrays a big limitation. The amplicon (the region of DNA that’s amplified- it is a DNA copy of the mRNA vaccine because you cannot do PCR on RNA) does not capture the entire gene sequence for spike. RNA and DNA can readily break inside a cell through any number of processes. Even if you do find evidence of reverse transcription through this work, you will not be able to show with this set of primers that the entire spike gene is being reverse transcribed by sequencing your PCR product to confirm it matches the sequence in the vaccine (which to their credit, the do).
Figure 5 showing the presence of a DNA sequence consistent with the PCR amplicon in size when the cells are treated with Bnt162b2 but not when they are not.
The paper then tries to loop in the L1 hypothesis by showing that introduction of the mRNA into Huh7 cells induces an increase in the expression of L1 both at the level of the mRNA and the product. This builds on prior work that I have addressed previously; in short, a paper claimed to show evidence that SARS-CoV-2, the virus, was able to integrate into the genome (or rather fragments of its genome) and proposed this as a possible explanation for long COVID. The methods used in that paper were prone to artifactual results, and since that time a letter responding to that publication has been published arguing that it does not show what it claims to show, another paper has been published demonstrating that the results from this publication could not be reproduced and were artifactual, and a third publication performing detailed sequencing of the genome of cells that had been infected with SARS-CoV-2 found no evidence of reverse transcription or integration either. The claim a priori was on very shaky ground and the fact that the findings can’t be reproduced is a poor sign about their reliability- yet the authors of this paper are undeterred by such considerations. This is especially surprising when their own experiment demonstrates that 48 hours after the mRNA vaccine is introduced, it is actually the negative control that did not receive the mRNA that has the highest L1 gene expression levels with otherwise relatively similar levels across the various concentrations of RNA tested through all timepoints. Given how wildly the L1 content can fluctuate in these cells at baseline, it’s hard to compel me from these data that the mRNA vaccines are inducing increased L1 expression (and the immunofluorescence data they present aren’t particularly supportive therein either). Furthermore, there are 5 replicates for each sample and yet the statistical test used is parametric which isn’t justified in the paper (parametric tests like the Student’s t test performed here assume normality in the distributions and no outliers which is not really something that can be reliably assumed with a sample of 5 replicates; the same issue applies to their quantifications of the immunofluorescence in Figure 4). Basically, from this figure I can’t reliably conclude anything about the ability of the vaccine to induce or repress L1 beyond that if I wait 48 hours and do nothing, L1 levels will rise precipitously in the culture. That’s really not a strong case for suggesting L1 induces integration of an mRNA sequence of interest. Another statistical issue is that despite doing multiple comparisons, they do not make any attempt to correct for a type 1 error e.g. Bonferroni. This is essential because with every test run, there is a probability of a false positive result, and thus the more tests run, the more likely that at least 1 of the comparisons is a false positive.
The key piece of data from this paper comes from Figure 5 which shows bands on a DNA gel electrophoresis (first treated with RNA) demonstrating the presence of a band that is about the size of their PCR amplicon for the vaccine sequence. Table 2 then also shows the Sanger sequencing result for the PCR product in the band which the authors report is identical to the sequence in the vaccine (and I did verify and confirm that indeed this does correspond to the vaccine sequence with a BLAST search). So… aha! That’s it. The Pfizer vaccine reverse transcribes into my DNA via L1 and so it must also integrate in because that’s how L1 works right? No, not quite. While this does support that there is reverse transcription occurring in this experimental model, no experiments are ever done to show integration; this is not an automatic process- transport into the nucleus wherein integration could occur is highly regulated so it cannot be inferred simply from these results that the integration in question is occurring. Furthermore, no actual experiments are even done to confirm that this is the result of L1 retrotransposition. To demonstrate this, you could perform a repression of L1 in the cell and see whether or not the reverse transcription occurs such as through Rhox, as discussed here. To the authors’ credit, they at no point argue that the DNA integrates into the genome (though some actors less concerned about the actual meaning of their findings and more concerned about advancing an agenda of vaccine hysteria have made this claim from their data).
Even 2 major clones of the Huh7 cell line have profound genetic disparities when compared with one another and both are quite different from normal human cells.
Another point worth discussing is the title of the paper. Calling Huh7 a “liver cell line” is a bit like calling a spoiled, rotten egg “food.” Maybe it was at one point, but you’ll have a hard time convincing me to eat it. This actually is, I think, the most critical problem of this study that alone should have made it unpublishable (and maybe if it weren’t MDPI, it would not have been). Huh7 is a cancer cell line, which has properties that are quite different from that of any healthy liver cell. If you ever examine the genome of a cancer cell in a karyotype to visualize the chromosomes, it’s been described as looking like a bomb went off in the nucleus. For example, cancer cell genomes show a phenomenon called chromothripsis where the chromosomes rip apart and kataegis where there is a drastic acceleration in the rate of base changes in the DNA which is highly abnormal in most healthy cells. With Huh7 though this is harder to understand because Huh7 is derived from hepatocytes and they exhibit a unique quirk of cell biology. Normally, a cell duplicates it genome and then divides, splitting the chromosomes between the two daughter cells. Hepatocytes, however, divide without a key step in this process called cytokinesis, which allows the to accumulate many extra copies of each chromosome. Hepatocytes with as many as 16 copies of each chromosome (our cells normally have 2) are entirely normal, and it is believed that these extra copies slow hepatocyte replication precisely so that you don’t get tumors. Nonetheless, Huh7 cells are so genomically unstable that the karyotype of even 2 clones from the cell line have dramatic differences. Using a cancer cell line to attempt to claim something happens in normal human cells is absolutely unacceptable and without delving any deeper, that would be sufficient to disqualify this study as written from publication. But in fact, there is another layer to this that further makes this choice egregious.
As I said before, in any typically healthy cell, retrotransposition by L1 is an incredibly rare phenomenon that, owing to cis preference, would generally just result in the insertion of the L1 mRNA into some site in the genome. Choosing a cancer cell line however engineers particular results in the same manner that prior papers have claimed to show that there is reverse transcription and integration of SARS-CoV-2’s genome into the cell when in reality the findings were artifactual (as discussed here, here, and here). However, L1 is substantially overexpressed in cancer cells. L1 expression is furthermore directly related to the cell’s own DNA damage response, as the integration process induces double-stranded breaks in the DNA that should trigger repair pathways and if they should fail, cause the cell to commit suicide (something that cancer cells do not do). L1 in fact is such an important aspect in the development of cancer that it is being considered as a biomarker for cancer. This study in fact manages to support that these observations extend to L1 from the fact that 48 hours after doing nothing, the cells have higher expression of L1 than any cell treated with the mRNA vaccine. Another point that argues against the external validity of this study (i.e. how it applies to the real world) is the fact that L1 is known to induce type 1 IFN responses, which in turn contributes to the aging of cells by promoting a low grade state of chronic inflammation. This paper attempts to argue that there is derepression of L1 in the genome caused by the mRNA vaccines (as stated before, I do not think the case for this is strong), which in principle should result in detectable elevation of type 1 interferon if it is occurring to a significant extent. However, the mRNA vaccines are not associated with a robust type 1 interferon response (there is robust activation of interferon-γ, which is the only type 2 interferon). This is a good thing because induction of type 1 interferon prematurely in response to the mRNA vaccination would prevent translation of the spike protein and accelerate breakdown of the mRNA vaccine (this likely contributes to the poorer responses seen with mRNA vaccines that are not modified like Curevac and vaccine responses in elderly individuals with mRNA vaccines, although there are a number of other factors as well).
Another factor worth discussing here is the dose of Bnt162b2 given to the cells. Firstly, the authors misrepresent pharmacokinetic data with this remark:
The Pfizer EMA assessment report also showed that BNT162b2 distributes in the spleen (<1.1%), adrenal glands (<0.1%), as well as low and measurable radioactivity in the ovaries and testes (<0.1%) [26].
In actuality, this part is about the lipids in the nanoparticle, and not the mRNA. It cannot be inferred that these lipids are still part of the nanoparticle and contain the mRNA from the experiments; mRNA escapes the endosome once it is inside cells and lipids are known to redistribute throughout the body (this is for instance the mechanism by which some anesthetics like thiopental lose their effect- initially the lipids enter the brain but then redistribute out of the brain). That limitation should be explicitly noted in the discussion- it is not. Also contained within that EMA document is this:
Moreover, according to the Applicant given the large difference in dose between the toxicity studies and the clinically efficacious dose (300-1000x), it is unlikely that the administration of a booster dose will lead to significant accumulation.
Doses matter. Here, Aldén et al report that they gave a dose of vaccine corresponding to 0.5% of the local injected dose to 200,000 Huh7 cells. As pointed out by Terrence Bartlett, this quantity of mRNA is likely not a reasonable approximation of the RNA quantity delivered to the liver; per the EMA document 21.5% of the mRNA-LNP is distributed to the liver, which has more than 170 billion cells in an adult. This amounts to 6.45 mcg of the dose reaching the liver (using a 30 mcg initial dose). If you compare this quantity of the mRNA-LNP to that used in the experiment, the experiment in effect exposed these cells to 20,000 times higher concentrations of mRNA than that any liver cell would experience (and again, Huh7 is not really a reasonable model for the genetic behavior of normal liver cells as is being explored here). So not only does this not reasonably model the L1 expression of healthy cells, but it doesn’t even reasonably approximate the clinically relevant dose. It may be worthwhile to point out though that though this does represent a significantly exaggerated quantity of mRNA for a liver cell, it may be relevant for antigen-presenting cells or muscle cells at the injection site; however neither one would express L1 at detectable levels.
There are however aspects of this paper that raise questions of bad faith rather than mere incompetence. One aspect is the fact that this journal is published by MDPI, which was at one point on Beall’s list, a catalogue of journals and publisher’s with predatory practices, and subsequently removed after an appeal from the publisher. Since that period however, MDPI has faced a number of problems and scandals, most recently with their journal Vaccines after an egregious failure of peer review wherein a study (now retracted) claimed to show that COVID-19 vaccines killed more people than they saved which resulted in much of their editorial board resigning in protest. It is for this reason that I actively avoid reading anything from MDPI journals unless I have no other choice or I know that the authors’ work is trustworthy based on other factors- a choice that this work very much validates as being correct. Another matter is the fact that this paper was reviewed extremely quickly (about 1 month). In isolation, this would not be incriminating as there have been papers during the pandemic that have been reviewed at more or less light speed (particularly work on Omicron and the antibody responses therein), but the manuscript has considerable errors and limitations that make it more or less meaningless as a study.
Implications
Truthfully, it is very hard to conclude anything meaningful from this study. It essentially demonstrates that cancer cells behave like cancer cells. If you’re not a cancer cell, nothing done in this paper is meaningfully interpretable as far as the implications for your health. In real life, a cancer cell made to express the spike protein which the immune system was strongly primed against may have potential to be an effective immunotherapy one day (with some caveats e.g. potential for tumor lysis syndrome). I also think it’s critical to point out that all the mechanistic data in the world is meaningless if you cannot connect it to a real-world effect. As it happens, a recent study of more than 298 million doses of mRNA vaccines administered in the US over a 6-month period found that the vast majority of adverse events were mild and transient, serious adverse events were very rare, and no major safety signals were detected. When it comes to looking at vaccine safety, the pharmacovigilance data is the be-all end-all. Mechanistic studies can help to clue in what pharmacovigilance data should focus on, but they cannot be used to argue against what is clear as day- the mRNA vaccines are safe. Genotoxic effects from mRNA vaccines are not biologically plausible any more than they are for any RNA in our body at any given moment.
