|
What my MBE paper is about
- A viral pathogen adapted to one mammalian species (e.g., bat) may sometimes switch to another (e.g., human), or from one organ/tissue to another (e.g., from intestine to gall bladder). The new host or new organ/tissue is typically not prepared for the viral infection, and often gets sick.
- It is important to detect (and ideally predict) host-switching or tissue-switching events. However, it is difficult to identify host-switching or tissue-switching events unless you catch the virus in a different host or in a different tissue, which is often too late. In the case of SARS-CoV-2, human populations did not have it before. Now we do. So a host-switching event is detected, but it is too late. Furthermore, we do not know the source species from which SARS-CoV-2 switched to human.
- It would be nice if a particular species or a particular tissue would imprint some genomic signature on the virus, so that we can check and say "Oh, the virus is from intestine of Rhinolophus bats because it bears that specific signature." Analogously, if you see a Caucasian boy in Ottawa who speaks fluent Chinese, you might roughly infer that he had spent significant amount of time in a Chinese environment, but had switched country lately.
- What would imprint a genomic signature on a viral genome so that we can infer its past dwelling? If there is an enzyme X in kidney and another enzyme Y in lung that modify the viral genome in different ways and leave different genome signatures (say Sx and Sy), then we only need to check for the presence of Sx and Sy to determine if the virus has always been in kidney or lung or shuttled in between in their evolutionary trajectory.
- DNA methyltransferases leave signatures on DNA viruses and retroviruses with a DNA stage. For coronaviruses with an RNA genome, we need enzymes to "write" on RNA. However, few RNA modification enzymes are known. In fact, scientists originally thought only DNA methyltransferases exist. When researchers characterized the first RNA methyltransferase, they named it Dnmt2 (standing for DNA methyltransferase 2). Only many years later did they find that Dnmt2 modifies RNA instead of DNA, but the name stays. Thus, if kidney has a lot of Dnmt2 and lung has little, then we would see different genomic signature in RNA viruses in kidney and lung. Unfortunately, Dnmt2 seems to act only on small RNA molecules, not on the large RNA genomes of coronaviruses. So this thread of the story reached an early end.
- Vertebrate DNA methyltransferases increase mutation of CpG (shorthand for 5'-CG-3') to TpG, so our DNA genome has strong CpG deficiency (and TpG surplus). Consequently, our mRNAs also exhibit CpG deficiency. DNA in different tissues exhibits differential DNA methylation patterns. By looking at this DNA methylation patterns, one can often predict which tissue the DNA is from.
- One thing peculiar is that many RNA viruses also exhibit strong CpG deficiency. Given that no known CpG-targeting enzyme that is modifying the viral RNA genome leading to biased CpG to UpG mutation, what is driving down their CpG dinucleotide frequencies?
- Nature has only two important genetic sculptors. One is mutation and the other is selection. If we can't find a clue in biased mutation mediated by methylation, then let's look at selection. In other words, viral RNA genome can increase or decrease their CpG, but selection may favor those low-CpG variants against those high-CpG variants.
- If selection really favor low-CpG variants, then those low-CpG variants should be able to replicate faster (fitter) than those high-CpG variants. Researchers checked these variants that arose either naturally or through experimental manipulation in those RNA viruses. They consistently found that low-CpG ones are fitter than those high-CpG ones. Thus, there is indeed strong natural selection against viral genomic variants with increased CpG.
- What is the proximate mechanism for this selection? There must be something targeting against CpG-rich viruses in the cell, but what is it?
- It is not until 2017 when a zinc finger antiviral protein (ZAP) was found to specifically target CpG dinucleotides in viral RNA genome to mediate its degradation. That explains why increasing CpG in the viral genome is bad for those RNA viruses.
- The distribution of ZAP is quite pervasive, but especially abundant in more exposed areas of our body. ZAP is high in all parts of our digestive system. It is also high in urinary bladder, testis, epididymis, seminal vesicle, vagina, fallopian tube, endometrium cervix/uterine, placenta, breast, gallbladder, pancreas, kidney, cerebellum and caudate. It is expectedly high in lymph system (lymph nodes, spleen, appendix, and tonsil). ZAP seems to be deployed to protect all vital organs. However, ZAP abundance is only moderate in human lung and nasopharynx (I do not know why ZAP is not deployed more in the lung). Given differential distribution of ZAP in different tissues, selection pressure exerted by ZAP against CpG dinucleotides would differ among tissues. Besides, ZAP gene sequences differ substantially among different species. All these suggests the possibility that differences in ZAP function and abundance among species and tissues may imprint different RNA genomic signatures on RNA viruses infecting different species and different tissues.
- It is in this context that I decided to address two questions. First, has SARS-CoV-2 evolved to have low CpG to evade ZAP-mediated host defense? Second, do coronaviruses from different mammalian species and mammalian tissues exhibit different CpG signature (typically characterized genomic GC% and an index of CpG designated as ICpG)? I will not go into details (which takes up most of the space of the paper), but the results from genomic analysis is consistent with a 'Yes' answer to both questions. In particular, SARS-CoV-2 (together bat-derived TG13 and pangolin derived pangolinCoV) exhibit lowest CpG among all betacoronaviruses.
- I was hoping to find some Betacoronavirus lineages (say X, Y, Z) that also have extremely low CpG. If X, Y, Z are all kidney pathogens, then we can infer that the SARS-CoV-2 progenitor is likely to have evolved in kidney. However, there is no such X, Y, Z among betacoronaviruses, so I went beyond to look at alphacoronaviruses to see if I can find X, Y, Z in alphacoronaviruses.
- I found only one clade of canine coronaviruses (CCoV) to have extremely low CpG comparable to that of SARS-CoV-2. These are digestive pathogens, including a most deadly CCoV that can go from digestive system to infect other organs. In particular, when one of the CCoVs was propagated in cell culture (which should neither mount an immune response to kill the virus nor get eliminated by the virus), genomic CpG quickly increases. This is consistent with the interpretation that selection is pushing down genomic CpG which rebounds when the selection is removed.
- This led me to check publications documenting SARS-CoV-2 in human digestive system. Indeed SARS-CoV-2 is frequently found in human digestive system. The bat-derived TG13 is also from a fecal swab. I therefore speculated, close to the end of my MBE paper, that mammalian digestive system, especially canine digestive system, appears to favor the origin of low-CpG coronaviruses. The viral progenitor could infect the digestive system of a canid that chewed a bat, and then gain a low-CpG genome better able to evade host defense. I was careful not to use 'dog' but used 'canids' (which include wolves, jackals, foxes, coyotes, etc.). I did suggested that feral dogs should be included in viral monitoring. If we know that feral dogs carry certain coronaviruses that are never found in pet dogs, then we should aim to know where feral dogs get those coronaviruses from and if they have potential for zoonotic transmission.
A bit of information not present in the paper: Because SARS-CoV-2, TG13 and pangolinCoV all have low CpG, the simplest hypothesis is that their common ancestor gained a low-CpG genome that has not changed since their divergence. The common ancestor might have lived quite a long time ago. It is for this reason that, when I received the press release from MBE, I quickly did a rough dating of the common ancestor to about 1880s, so that we know approximately when the reduction in genomic CpG has occurred in a SARS-CoV-2 progenitor, presumably in the digestive system of a mammalian species (likely a canid). However, I am not certain if the simplest hypothesis is correct. A virus could gain a low-CpG genome only recently and went on to infect bats and pangolins.
|