Why SARS-CoV-2 appears to have a natural origin

After my previous blog post about lab leaks, there was some debate, as expected. I learn from that, but while the previous blog post focused on the arguments from the Langsikt memo, here I will focus more on the scientific arguments for why SARS-CoV-2 was most likely not created at the Wuhan Institute of Virology (WIV).

This is quite demanding material, and I am no virologist, molecular biologist or anything close to it. But I have read a lot and tried to find out what made leading virologists believe quite early on that a lab leak as the origin of SARS-CoV-2 was unlikely, and why that conclusion has been strengthened over time.

There is a lot of material to cover, but I have tried to pick out the points I think are most important, and I have included links to further reading at the bottom of the blog post for those who want to delve deeper. I also recommend watching the videos I have included in the blog post.

Should any of what I have written be incorrect, in the form of “misunderstood” or poorly explained biology/genetics, I hope that competent people will correct me so that I can improve the text.

Furin Cleavage Site

One of the important factors that makes SARS-CoV-2 (SARS2) look “man-made” is that it has a Furin Cleavage Site (FCS) that has not been seen in other coronaviruses in the SARS family (the sarbecovirus subgroup of the betacoronavirus genus, of which there are around 1500 variants).

It seems mysterious. How could this have appeared suddenly in 2019?

Most people have seen images of the SARS2 virus as a ball with spikes on it. These spikes are proteins necessary for infecting human cells and consist of two parts, called S1 and S2. Once the virus has attached itself to an ACE2 receptor on the surface of a human cell using S1, S1 and S2 must be split apart so that S2 can perform the actual “connection” to the cell, allowing the virus to enter the cell and replicate there.

This splitting occurs with the help of the enzyme furin (which acts as a kind of biological scissors), and the place where this “cut” occurs is called a Furin Cleavage Site.

Could an FCS have been created at the Wuhan Institute of Virology (WIV)?

We know that research has been done on inserting an FCS into coronaviruses through genetic modification in the past, as described in a handful of studies. And we know that the DEFUSE project described this as one of its objectives.

Here, they wanted to take an existing virus, a so-called “backbone,” and make minor genetic changes to it to see what effects it would have on the virus's transmissibility and pathogenicity in animals and humans.

However, DEFUSE did not receive funding from DARPA, partly due to fears of “gain-of-function” research. This is research that makes an existing virus more dangerous to humans. To the extent that such research was nevertheless carried out with other financial resources from the NIH, the project description states that the role of the WIV was only to collect coronaviruses, while the actual FCS work was to be done at the University of North Carolina (UNC) in the US (under the leadership of Ralph Baric).

WIV does not have the specialist expertise in this field, which UNC does, so it seems strange that this would have happened at WIV. At the same time, we know that research was conducted on hybrid viruses based on a coronavirus called WIV1 in mouse studies there, so related research was probably carried out in Wuhan. However, there is no evidence that WIV ever worked on manipulating specific FCS into a coronavirus that was similar enough to SARS2 to have created this.

To manipulate an FCS and create SARS2, you need a virus that is very similar to start with. This was not available at WIV or anywhere else. The closest known virus in 2019 was the bat virus RaTG13, but it is too different from SARS2 to have been used for this purpose, as it differs by more than 1,100 base pairs from SARS2.

It was only after the outbreak of the pandemic that other coronaviruses more similar to SARS2 were found, such as RmYN02, RpYN06, and PrC31. However, the closest match so far is BANAL-20-52, which was found in bats in Laos a couple of years after the start of the pandemic. This virus is so similar (it differs from SARS2 by about 10-20 years of natural evolution), especially in the areas of the virus that are central to infecting humans, that it could potentially have been the starting point for SARS2 by adding an FCS.

But WIV did not have BANAL-20-52 – or any other similar virus – at the time, so they could not have done so.

Did WIV keep virus data hidden?

Or had they found BANAL-20-52 or a similar virus but kept it hidden?

That makes no sense, because BANAL-20-52 was not a special virus before SARS2 was discovered, so why would you keep what was then a completely insignificant virus hidden from the public?

On the other hand, it doesn't take many mutations to turn BANAL-20-52 into SARS-CoV-2, so that increases the likelihood of natural recombination and the formation of an FCS in nature, which I will come back to later.

The WIV collected viruses from bats several times between 2011 and 2015 from a single habitat near Kunming City, Yunnan Province, China. All information about this was published in a database that contained an overview of all 22,000 viruses and genetic sequences from various virus collections from previous years, last updated in the summer of 2019.

Eddie Holmes has also repeatedly mentioned and described in a now archived Twitter thread (as he later left Twitter) a manuscript that researchers at WIV submitted for publication in 2018 but which was rejected for technical reasons. It was later discovered that they had mentioned the latest viruses sequenced at the laboratory. None of these were genetically close to SARS2 and could not have been used as the backbone to create SARS2.

Another article was published in 2020 listing all the SARS viruses the laboratory had. Since this article was submitted for publication as early as 2019, long before there would be any reason to keep a SARS2-like virus secret, this is further evidence that the WIV did not have any SARS2-like virus that could have been used as a backbone to create SARS2.

US intelligence services have also concluded that the WIV did not have any virus that could have been used to create SARS2.

Then, on September 12, 2019, this database went offline, which has given rise to conspiracy theories that this was done to hide something. This theory fails to mention that the database was constantly unavailable both before and after this date:

Access was very unstable long before September 12, 2019, then stable for a short period before going offline. It then came back towards the end of the year, was unstable, before disappearing again at the end of February 2020.

The official explanation was that the database was subjected to hacking attempts and was therefore taken down, but we do not know if this is the whole truth. Perhaps it was taken down because a laboratory accident had infected a researcher there with SARS2?

Were researchers at WIV infected with SARS2 in the fall of 2019?

It doesn't really make sense, because there is no evidence that any researchers at the WIV were infected with COVID-19 that early (the rumors about hospitalized laboratory employees don't hold water), as described in the previous blog post. And if they had been, it would not be consistent with the infection figures in December.

SARS2 infections doubled every 3.5 days in the beginning. With such exponential growth, an outbreak in August/September would have meant that millions of Chinese would have been infected by December, with overburdened hospitals and a very visible epidemic throughout the fall. However, the first known case of infection was in early December, when very few people were infected. Ergo, the outbreak cannot have started before early November at the earliest, otherwise the math does not add up.

A so-called superspreader event at the seafood market could also be an explanation. Perhaps some people were already infected earlier in the fall, perhaps someone at the laboratory, and it was only when they later visited the seafood market that they started an outbreak? However, the serological and epidemiological findings do not support this hypothesis.

This now infamous wet market in Wuhan is not very densely populated. There are more than 1,500 areas in the metropolis, which has around 12 million inhabitants, where there would be a much greater chance of a virus outbreak than at the wet market, such as train stations and shopping malls. What sets the Huanan wet market apart from other much more densely populated places in the city is that it had animals in captivity that we know can carry SARS2.

If we look at the geography, there are also many such densely populated hubs closer to the WIV than the Huanan wet market. So if the infection started with a laboratory worker who later infected others, it is statistically unlikely that it would have happened at a wet market located half an hour's drive away from the WIV.

So the fact that we see the infection spreading from the market, and that the market was proven to have live animals of species (civets and raccoon dogs1) that we know can carry SARS2, is a strong indication that the virus also originated there. If not, a “superspreader event” from an already infected person would have been much more likely to occur somewhere else entirely.

WIV, wet markets and geography

We also have clear signs that the infection started with two different lineages of SARS2, namely lineage A and B. According to the molecular clock, these two lineages split in late November or very early December 2019. The lineage B variant was found in about 2/3 of all infected people at the start of the pandemic, and the lineage A variant in about 1/3.

Mathematically, this means that with a doubling of infections every 3.5 days, variant B should have infected a human at least 3-4 days before variant A. This is the most likely reason why twice as many people were found to be infected with variant B afterwards. It also means that there must have been several cases of transmission from animals to humans inside the market, both with lineage A and lineage B (and perhaps others that never spread further and which we have therefore not seen), over several days.

Why is this important? Because serological tests have found traces of both lineage A and B inside the wet market, which also makes it highly unlikely that the infection started there via a human from outside. That would require a laboratory employee, who could only have been infected with lineage A, to have infected enough other people (or animals) for lineage B to develop, and then both of these would have had to visit the wet market and infect people (or animals) there. Without infecting people anywhere else other than the wet market first.

It makes no sense. The statistical chance of finding both lines in and right next to the seafood market if they did not originate there is very low.

If it was a laboratory employee who was first infected, that person would not have been able to infect anyone at the seafood market (or elsewhere) with two different variants of the virus several days apart. However, animals in cages at the wet market with the different variants could have exposed humans to infection repeatedly over several days, thereby creating the kind of spread that the data subsequently corroborates.

In terms of timing, this hypothesis does not hold water either. If the WIV had BANAL-20-52 or a similar virus and wanted to keep it hidden, this virus would have had to be discovered after August 2019. But that means that in just three months, they would have had to find the virus, identify and catalogue it, sequence the genome, manipulate it into an FCS (without particular expertise in this area), cultivate enough virus in laboratory mice (without leaving any visible traces in the genome afterwards), and then accidentally infect someone in such a short period of time.

This is not practically possible.

Two of the first three known cases could be linked directly to the wet market, and 28% of all cases detected in December 2019 were also linked to the wet market. A full 55% of all cases detected in December 2019 had a connection to the wet market in Huanan or similar markets. Analyses of infected cases show that they clearly originate from the wet market, and to the extent that this is due to a testing bias, where more people from that area were tested for COVID-19, the same pattern was seen with deaths from “pneumonia” in early January. Furthermore, no such clusters of infection have been found in other parts of Wuhan, including the WIV.

It makes no sense that all the different types of analysis, whether it is mapping of early infections, data on excess mortality, data from mobile apps, etc., all point to the Huanan wet market as the epicenter, while none point to any other place, such as the WIV, unless the wet market was actually the epicenter.

Genetic traces show natural evolution

If the infection came from a researcher at the WIV, then SARS2 would also have shown signs of adaptations to laboratory mice, which has not been found. If the infection came from a researcher at the WIV, SARS2 would also have shown signs of adaptation to laboratory mice, which no traces of have been found. In addition, it would be difficult or impossible to cultivate SARS coronaviruses with FCS in mice, as FCS does not provide any evolutionary advantage there and would therefore have been naturally selected away.

To cultivate such viruses, it would have had to occur in human lung cells or animals genetically modified to have ACE2 receptors, e.g. humanized mice, but then the virus would also have been adapted to infection in mice, including a specific mutation in the spike protein called N501Y. However, the earliest variants of SARS2 could not infect wild mice and lacked mutations similar to N501Y. This also points to the fact that the virus could not have come from infected laboratory mice.

In addition, the way FCS is found in the SARS2 genome is completely different from the way FCS has been “pushed into” coronaviruses in all previous attempts.

DNA is made up of the following four nucleotides: Adenine (A), Guanine (G), Cytosine (C) and Thymine (T), while RNA is made up of Adenine (A), Guanine (G), Cytosine (C) and Uracil (U). These form long chains that we know as DNA and RNA molecules.

A set of three nucleotides, e.g. CGG or ACG, is called a codon, and a series of three codons codes for an amino acid.

As previously explained, the spike protein on the SARS2 virus must be cut using the enzyme furin in FCS, which consists of 12 codons, or four amino acids, in the RNA of the virus.

Furin reacts with the protein where it sees the amino acid sequence RxxR, where R is the amino acid arginine, and x is one of the other 19 amino acids that all our proteins are made up of.

Furin works best if it sees RRxR or RxRR, and best of all at RRKR.

In SARS2's FCS, the amino acid sequence is RRAR.

Of the seven different coronaviruses that infect humans, four of them have FCS, including some that cause the common cold, as well as the more deadly MERS. (SARS1 does not have FCS.) FCS is also found in many other coronaviruses, although (as yet) not in the specific subgroup to which SARS2 belongs (other than in SARS2 itself).

Comparing the SARS2 virus with RaTG13, one finds many similarities, apart from this PRRA sequence, which is new and which many found suspicious in early 2020.

But as mentioned earlier, each amino acid consists of three codons. Different combinations of codons can code for the same amino acid. And when you look at SARS2, you see something strange:

The inserted sequence PRRA is not in the “right place”. The amino acid sequence PRRA is shifted by one letter/nucleotide. It is “out-of-frame”, i.e. not placed in a natural position in the sequence of nucleotides. It starts in the wrong place.

It is unlikely that a human being who wanted to splice the amino acid sequence PRRA into a genetic backbone (existing virus) would have done so in this way, because it has no specific utility and only increases the risk that the virus will not “work” afterwards. This clearly points to a natural mutation.

In addition, I mentioned that PRRA is not FCS in itself. It consists of RRAR, i.e. the last three amino acids in (P)RRA plus a subsequent R that was there before. The sequence is therefore (P)RRAR, as you can see from the image above. So why would researchers want to insert a P at the beginning when it is unnecessary? The virus already had an R at the end, so all they needed to do was add RRA in front, which would have given us RRAR – an FCS.

Peter Miller, the author of this blog post, from which I have drawn much of my information and which you should definitely read in its entirety, has looked at previous studies where an FCS has been created in viruses and summarizes the findings as follows:

A 2006 US study inserted RRSRR into the S1-S2 site in a SARS-CoV-1 pseudovirus.
A 2009 US study inserted RRSRR into the S1-S2 and S2’ site in a SARS-CoV-1 pseudovirus.
A 2008 Japanese study inserted KRRKR into S2’ site in a SARS-CoV-1 pseudovirus.
A 2014 Dutch study inserted RRRRR into S2’ in a mouse hepatitis coronavirus (pseudovirus).
A 2019 Chinese study inserted RRKR into S2’ in a chicken virus (gamma-CoV infectious bronchitis virus).

There are no published data or evidence that WIV has ever attempted this, and in general it has not been attempted very often at all. We also know that such research normally takes place with “safe pseudoviruses,” e.g., WIV1 and similar, not viruses that can infect humans. In this context, it would have been natural to research the SARS1 virus, which does not have FCS, and add an FCS there to see the effect. But SARS1 is far too different from SARS2 to have been the backbone they used.

Traditionally, they have also always used very effective variants of FCS such as RRKR or RRSRR. Variants of RxxR where x is P and A have never been tried, simply because they would be less effective FCS.

PRRAR, as we see in SARS2, gives a significantly poorer FCS than already known sequences that have been used successfully before. Why then would researchers at WIV use a weaker/poorer variant? It makes little sense other than if it was a natural mutation where such considerations are not taken into account.

And since 2019, several viruses have been found with natural mutations in the same place as SARS2 has PRRAR in its genome:

As you can see, all of these have a P first and an A at the end. There is therefore a natural precedent for this to happen, but no precedent for humans doing it this way.

More about codons and coincidences

As mentioned, each amino acid is made up of three codons, but different combinations of these can produce the same amino acid. P (the amino acid proline) can, for example, be coded as CCT, CCC, CCA, or CCG.

What we also see is that P in both SARS2 and the other four natural viruses above is coded as CCT. In other words, we see that SARS2 has coded for proline using the same nucleotide sequence (codon) as other natural coronaviruses, even though there are several other possibilities. This also points to FCS being the result of a natural recombination between two natural coronaviruses.

Evolution has also improved the efficiency of the virus and gradually replaced P with other amino acids that work better. While the original Wuhan variant of SARS2 had the amino acid sequence PRRAR in its FCS, the Alpha variant has HRRAR, and the Delta variant has RRRAR, which made them more contagious and pathogenic.

If one were to optimize a virus for human infection in a laboratory, it would be strange to splice in an “out-of-frame” FCS with an amino acid sequence that has never been tested before and that we know is less effective than others that have been tested before.

Let me quote from another article on this subject, in which Miller participated in a debate on lab leaks vs. natural origin, where he says:

COVID’s furin cleavage site is a mess. When humans are inserting furin cleavage sites into viruses for gain-of-function, the standard practice is RRKR, a very nice and simple furin cleavage site which works well.

COVID uses PRRAR, a bizarre furin cleavage site which no human has ever used before, and which virologists expected to work poorly.

They later found that an adjacent part of COVID’s genome twisted the protein in an unusual way that allowed PRRAR to be a viable furin cleavage site, but this discovery took a lot of computer power, and was only made after COVID became important.

The Wuhan virologists supposedly doing gain-of-function research on COVID shouldn’t have known this would work. Why didn’t they just use the standard RRKR site, which would have worked better? Everyone thinks it works better! Even the virus eventually decided it worked better - sometime during the course of the pandemic, it mutated away from its weird PRRAR furin cleavage site towards a more normal form.

Or from this article written by some of the leading virologists in the field explaining the same problem:

The SARS-CoV-2 furin cleavage site (containing the amino acid motif RRAR) does not match its canonical form (R-X-R/K-R), is suboptimal compared to those of HCoV-HKU1 and HCoV-OC43, lacks either a P1 or P2 arginine (depending on the alignment), and was caused by an out-of-frame insertion (Figure 2). The RRAR and RRSR S1/S2 cleavage sites in feline coronaviruses (FCoV) and cell-culture adapted HCoV-OC43, respectively, are not cleaved by furin (de Haan et al., 2008). There is no logical reason why an engineered virus would utilize such a suboptimal furin cleavage site, which would entail such an unusual and needlessly complex feat of genetic engineering. The only previous studies of artificial insertion of a furin cleavage site at the S1/S2 boundary in the SARS-CoV spike protein utilized an optimal “RRSRR” sequence in pseudotype systems (Belouzard et al., 2009;Follis et al., 2006). Further, there is no evidence of prior research at the WIV involving the artificial insertion of complete furin cleavage sites into coronaviruses.

What about ENaC α?

Some would point out that this may not be so mysterious after all, because a similar amino acid sequence is found in a protein on human cells called ENaC α, which is “a critical component of the epithelial sodium channel” and “is critical for the channel's ability to transport sodium ions across epithelial cell membranes.” This is found in cells in our kidneys, colon, and lungs, among other places.

ENaC α also has a functioning FCS, and since this is well known, it would be obvious to “graft” this into a coronavirus if one wanted to make it more infectious to humans. Both ENaC α and SARS2's FCS have the amino acid sequence RRAR`SVAS, which may seem suspicious. Is it purely coincidental that SARS2 has an FCS identical to that found in humans? Doesn't this suggest that someone at WIV may have taken the FCS from ENaC α and “spliced” it into a secret SARS2-like virus to create SARS2?

Not really. The amino acid arginine, R in the sequence PRRAR, can be coded as CGT, CGC, CGA, CGG, AGA, or AGG. In SARS2, the amino acid sequence RR is coded as CGG CGG, while in ENaC α it is not. There are therefore major differences in the coding of FCS between SARS2 and ENaC α, making it unlikely that FCS from human ENaC α was used to modify the SARS2 precursor, especially since this slightly unusual CGG coding for arginine also occurs naturally in other coronaviruses.

R (arginine) is spelled CGG in 20% of human DNA. And coincidentally also in 20% of bat DNA. Therefore, it may be a natural evolution away from RR spelled CGG CGG, as our immune system can more easily recognize it as viral DNA. Nevertheless, CGG CGG has not changed in SARS2 even after a long period of evolution in humans. It appears to be very useful in SARS2, and when attempts have been made to replace it with, for example, AGG, the virus becomes less effective.

Researchers could not have known this before the pandemic, so if they chose to create an FCS by grafting PRRA out-of-frame in a way where they spelled both R's as CGG, it would be an incredible coincidence and stroke of luck. And spelling arginine as CGG is not a common way of doing it in previous attempts to create an FCS in viruses. All of this therefore argues against human intervention, something even lab leak proponents such as Alina Chang admit. This reduces the likelihood that ENaC α's FCS would have “inspired” a similar FCS in SARS2, especially since the sequence R´SVAS is found in, for example, BANAL-20-52 and other SARS coronaviruses.

But what about DEFUSE?

It may still seem suspicious that finding a sarbeco coronavirus with an FCS “by chance” would be exactly what the DEFUSE project described as its goal. But that argument only holds if you overlook the details.

In the description of the DEFUSE project, they wanted to try to create an FCS in S2, not in the S1/S2 region, as we see in SARS2. They also wanted to grow the modified virus in Vero cells, which are a cell line from monkey kidneys. However, later experiments with this have shown that FCS is selected away through natural evolution and thus disappears from the virus over time.

If they therefore worked at WIV in line with the DEFUSE project description, the virus would look different – and would not exist with an FCS today.

The virus was also not particularly well adapted to humans. It can infect a variety of species and even exist as reservoirs in species without contact with humans. Since December 2019, it has also rapidly adapted to humans, as new mutations were found in January 2020 that made the virus more contagious in humans.

If the virus had been infecting humans for many months already, as some lab leak proponents argue, it would probably have been better adapted to humans, with visible mutations from earlier stages, when it finally became a major outbreak.

Research and time travel

On top of this, the cleavage site in SARS2 has O-linked glycans (sugar molecules attached to the amino acids serine or threonine). These affect how furin and other proteases can cleave the spike protein and are central to the infectivity of SARS2.

However, this was not discovered until long after the start of the pandemic, so no researchers at WIV could have taken this into account if they were to construct an FCS synthetically.

Another problem with FCS in SARS2 is that it is found in the genome in a completely different way than one would expect if it were artificially produced. In laboratories, existing sequences are usually altered through point mutations. It is rare to add completely new nucleotide sequences to create FCS, as seen in SARS2. Adding new nucleotide sequences only massively increases the risk that something will go wrong and that the experiment will fail. Mutating existing genetic code is a much safer and more common way of doing this. However, this is not what we see in SARS2.

However, in nature, we have seen several times that viruses can get unexpected insertions of 12 or more nucleotides. And in HKU1, for example, another Chinese coronavirus discovered in 2004 that caused a small outbreak of pneumonia/colds, an insertion of 15 nucleotides was found right next to the virus's FCS.

In later variants of SARS2, as well as some influenza viruses, several such random insertions of 12-15 nucleotides have also been seen through random mutations. In 2023, it was even found that natural recombination of a coronavirus in shellfish led to the natural development of an FCS in the virus. Once again, we see that everything needed to create SARS2 already exists and occurs in nature. And since there are millions upon millions of opportunities for viruses to recombine and mutate in different animals and humans over many years, it is much more likely that this happened there than that it happened in a handful of people who were infected in a laboratory over a few weeks.

On top of all this, we now know that the RBD (receptor-binding domain) in the spike protein of SARS-CoV-2 has a sequence that is particularly well suited to binding to the ACE2 receptor on human cells. This was not known before the pandemic, and this property could therefore hardly have been constructed synthetically in a laboratory. On the other hand, we have seen optimizations here through natural mutations in newer variants such as Omicron, suggesting that this is something that can happen through evolution.

So again, if researchers at WIV had the goal of inserting an FCS into an existing coronavirus, they would have had to be able to see into the future and know things that no one knew at the time, use methods that are completely unusual and impractical, with viruses they did not have, and on top of that do a very sloppy job that in practice would most likely have led to a failed experiment.

Conclusion

Finally, it should be mentioned that SARS2 has no genetic signatures that violate the laws of nature, such as barcoding. There are therefore no signs in the virus that indicate that it has been genetically manipulated in a laboratory. On the contrary, the genome bears the hallmarks of natural selection, which is very well supported by numerous genetic analyses, and all of its characteristics are found in closely related coronaviruses in nature.

This is, as I understand it, the main argument for virologists generally leaning towards SARS2 having a natural origin rather than being laboratory-created. So regardless of what one might think about the handling of the case, with secrecy, lack of transparency, and suspicious private email communication between researchers back in early 2020, it means little for what we actually have in terms of virological and genetic evidence now in 2024.

And if you base your opinion on that, the conclusion that SARS-CoV-2 arose through natural evolution is the most likely one.

Two thorough articles that go through much of the evidence for natural origin and against lab leak that should be read:

Other interesting reading material, videos and sources:

Pangolins and civets were banned from animal markets after the SARS1 outbreak in 2002/2003, as the infection at that time came from these species. But tragically, the animals were still sold at the Huanan wet market. ↩