2023 Theses Doctoral

Genetic and environmental determinants of recombination in coronaviruses

Diuk-Wasser, Maria

Recombination is increasingly recognized as an important mechanism of evolution among coronaviruses and as a potential driver of cross-species emergence. Here, we show that recombination best explains patterns of ACE2 usage in sarbecoviruses and that recombination of SARS-CoV-1 with a SARS-CoV-2-like virus was likely critical for its emergence in humans in 2003. Despite the demonstrated importance of recombination for SARS-like viruses and for coronaviruses in general, not much is known about the molecular mechanism by which it occurs.

In order to make inferences about possible molecular mechanisms, we designed experiments and analytical methods to study patterns of recombination between two coronaviruses in vitro. We generated mixed infections with Canine coronavirus (CaCoV) and Feline infectious peritonitis virus (FIPV), which have been demonstrated to have naturally undergone recombination with one another, to demonstrate the feasibility and utility of generating and detecting experimental CoV recombination. To facilitate detection of recombinant reads, Oxford Nanopore long-read sequencing was utilized; however, because of the high error rate of this technology, an algorithm was needed to accurately classify reads to either parental sequence and to identify any recombinant reads in the sample. A machine learning likelihood classifier was developed that is able to identify homologous recombinants, structural variants that represent non-homologous recombinants, and sub-genomic RNAs (sgRNA).

The algorithm was applied to a single coinfection between CaCoV and FIPV and showed that homologous recombination is present. Further, when replicated, we observed homologous recombination in every single experiment we performed with these two viruses, even though homologous recombinant reads represented a small proportion of the total reads in the sample. Next, we pooled data from 12 CaCoV/FIPV coinfection replicates together in order to analyze broad patterns of homologous recombination between these two viruses. To analyze the positions of breakpoints on a fine scale, a hidden Markov model was implemented, which increased the accuracy of breakpoint estimation.

Recombination patterns observed across all 12 experimental replicates of CaCoV/FIPV coinfection support a ‘hotspot’ model of recombination, with visible peaks of recombination in specific positions of the genome. Importantly, we find that while some recombination hotspots observed in natural recombinants may be attributable to environmental effects such as natural selection, other hotspots such as within the S gene and in orf1ab appear to be driven by a molecular mechanism and cannot be explained by natural selection alone. Our findings improve our understanding of coronavirus recombination as a mechanistically non-random phenomenon and highlight the need for continued experimental investigation using more distantly related viruses.