The role of indels in evolution and pathogenicity of RNA viruses

The combination of high mutation rates, large population size, and short generation time endow RNA virus populations with high evolvability and adaptability (1). However, high mutation rates are double edged. On the one hand, mutations fuel selection processes that promote adaptation to changing environments. On the other hand, most mutations are deleterious and pose a threat to virus survival (2). The low fidelity of virus-encoded RNA-dependent RNA polymerases (RdRp) results in a wide range of mutations, including single-nucleotide variants (SNVs), deletions, and genomic reorganizations. Mutant genomes with impaired replication ability are known as defective viral genomes (DVGs) (3). DVGs are ubiquitously generated by RNA viruses. While SNVs have traditionally received significant attention, the role of deletions and insertions (indels) has been relatively neglected, despite their potentially stronger fitness effects (such as altering reading frames, disrupting regulatory elements, and affecting RNA and protein structures).

An exception to this lack of interest is the study of defective interfering particles (DIPs), a particular type of DVGs, which has garnered attention mainly due to their therapeutic applications (4). DIPs act as parasites of wild-type (WT) viruses by hijacking essential viral proteins and interfering with their replication and encapsidation processes. Additionally, DVGs are known to trigger interferon responses (5), induce viral persistence in cell cultures (6), and contribute to diversity in long-term infections of immunosuppressed patients (7). The study by Rangel et al. (8) emphasizes the significant phenotypic consequences of indels. Understanding the full spectrum of indel variants and accurately mapping their diversity is crucial for unraveling the mechanisms of their generation, selective pressures shaping their dynamics, and production of viral novelty at protein and RNA levels. By shedding light on indel dynamics, Rangel et al. provide a deeper understanding of viral evolution and potentially open new avenues for antiviral strategies.

Indels have not been extensively studied, partly due to the technical challenges associated with accurately identifying and distinguishing them from sequencing errors. Rangel et al. propose using the CirSeq high-fidelity high-throughput sequencing method (9) in combination with a specifically developed bioinformatic pipeline, MultiMatch, which utilizes CirSeq data to identify, map, and quantify indels. CirSeq relies on rolling-circle amplification of circularized RNAs to generate cDNAs containing an average of three tandem repeats, enabling the differentiation between artifacts and true variability. MultiMatch addresses issues related to read rearrangement, orientation, and the detection of large insertions and deletions. To further validate the performance of MultiMatch, in silico CirSeq datasets containing indels were generated. The algorithm demonstrated good sensitivity, successfully recovering 84.2% of insertions and 93.4% of deletions, including some very low-frequency variants. Unfortunately, no information is provided on false-positive rates to assess specificity and precision.

Having established this hybrid experimental–bioinformatic approach, authors conducted a characterization of indel repertoires in poliovirus (PV) evolving through serial passages in HeLa cells. Several interesting observations emerged. First, indels occurred at low frequencies and were evenly distributed across the PV genome. Deletions occurred at a rate approximately 100 times higher than insertions. Most indels were less than 10 nucleotides in length, although a small fraction covered a significant portion of the genome. The dominance of deletions over insertions has been previously described in several organisms (10) and can be attributed to slippage being a more common mechanism than the random insertion of long nucleotide stretches. Second, indels were predominantly deleterious and imposed a significant fitness burden on viral populations compared to SNVs, with most indels being lethal and subsequently lost and regenerated after each passage. However, a small number of indels were positively selected (Fig. 1).

Fig. 1.

Interaction between WT viruses and DVGs. In cell culture, the WT virus replicates its genome (r > 1) and generates DVGs (indels) at a rate m > 0. DVGs are replicated by WT RdRp with an advantage k > 1 due to their shorter genomes (typically deletions). DVGs interfere with WT virus replication (f > 0) by sequestering its products. At the end of each passage, a random sample is taken to inoculate the next cell culture. This sampling follows a binomial distribution, determining the proportion of WT virus (V₀) and DVGs (D₀) in the inoculum. To maintain a constant multiplicity of infection, the sample is diluted further. It is important to note that experimental uncontrolled errors in viral population sampling can be described as Gaussian-like (e). Hypothetical time dynamics of WT virus (blue) and DVGs (yellow) are shown for within- and between-passages levels.

By shedding light on indel dynamics, Rangel et al. provide a deeper understanding of viral evolution and potentially open new avenues for antiviral strategies.

Third, Rangel et al. investigated how PV RNA structure and function influenced the observed diversity in indels. They discovered a preference for indels in unstructured regions of viral RNA, particularly at junctions between stem-loop structures in the canonical internal ribosome entry site. Indels were more frequent at these junctions between canonical stem-loop structures. Interestingly, a five-nucleotide deletion was identified as beneficial and found at a high frequency. This deletion likely impacts translation initiation rates, providing a fitness advantage in HeLa cells (Fig. 1).

Another potential effect of indels, especially short ones, is their ability to alter coding reading frames. Consequently, resulting proteins may differ significantly in sequence from the WT, imposing a fitness penalty. Consistent with this expectation, the length and periodicity of observed indels were constrained by the maintenance of translation frames. Deletions in coding sequences exhibited a trinucleotide periodicity indicative of purifying selection favoring indels that do not disrupt codons.

Fourth, Rangel et al. explored the relationship between indels and RdRp fidelity using PV RdRp mutants with increased and decreased misincorporation and recombination rates. In experiments involving variants affecting only SNV mutation rates, the researchers observed a negative correlation between indel rates and misincorporation rates. Specifically, the high-fidelity RdRp allele G64S exhibited fewer indels than the WT, whereas the low-fidelity allele H273R showed the opposite trend. In contrast, the recombination-defective allele D79H doubled the number of indels. Double mutants G64S/D79H and H273R/D79H displayed similar indel rates to the SNV fidelity mutants. Based on these findings, the authors concluded that increasing fidelity slows down RdRp elongation and increases the likelihood of template switching, thereby generating indels.

In a second set of experiments, Rangel et al. examined the effect of indels on the evolution of host range. They chose to study the adaptation of Dengue virus (DENV) to human or mosquito cells (11). Throughout experimental evolution in these two cell types, indels were observed across the DENV genome. Indel diversity was present in both human and mosquito populations. However, the rates of indels differed depending on the host, with significantly higher rates in human cells. Similar to PV, deletions were approximately 100 times more frequent than insertions. Another significant difference between the two cell types was the distribution of mutational fitness effects along the DENV genome. While lethal indels were evenly distributed between coding and noncoding sequences in both cell types, the distribution of beneficial indels sharply differed: They were enriched in the 3′ untranslated region (UTR) in mosquito cells but in the 5′ UTR in human cells. This suggests that distinct selective pressures act upon the same RNA elements in the two hosts. Additionally, another disparity between the two cell types was the rise in frequency of ~100-nucleotide long indels at early passages in mosquito cells, which was not observed in human cells. Host-specific selection and accumulation of indels have been previously described, such as the accumulation of a DVG with varying interference effects on tomato black ring virus accumulation in different hosts (12).

Previous studies have already revealed the importance of indels in noncoding regulatory regions in flavivirus pathogenesis and immune selection (13). One notable hotspot of indel variation is the SL-II RNA structure of the 3′ UTR, which is enriched in beneficial indels in mosquito populations. Interestingly, not all beneficial indels are equal: While some destabilize SL-II, others delete part of the structure. Both types of changes result in the removal of a halting site and enhanced replication (14).

It is worth noting the consistent 100:1 proportion of deletions versus insertions rates in both viruses. Although PV and DENV have mutation rates that are within the same order of magnitude (~10⁻⁴ per site per replication) (15), their SNV mutational spectra exhibit notable differences. PV shows a high frequency of CG dinucleotide bias (16), while DENV exhibits an A-to-G substitution bias (17). These differences in SNV spectra are usually considered as evidence of mechanistic differences in RdRps. Given that PV and DENV RdRps likely differ structurally in their interactions with the template and nascent RNAs, as well as incoming nucleotides, the apparent uniformity in indel formation requires further investigation.

An intriguing observation involves a specific indel that has been consistently observed across evolutionary passages in all DENV populations. This indel involves the insertion of an adenine within a stretch of six adenines in the NS3 region, leading to a premature stop codon downstream of the self-proteolytic cleavage site. Consequently, a NS3 protease is produced that does not require autoproteolytic processing. Another suggested advantage of this frameshift mutation is its impact on viral protein stoichiometry. Despite these claimed advantages, the mutation has not risen and remains at only 1%. This may be attributed to the mutation having pleiotropic negative effects that counterbalance the suggested benefits or its linkage with other nearby deleterious mutations, with background selection preventing its rise to higher frequencies. Additionally, the fitness effect may be smaller than the inverse of the inoculum size, rendering it a quasi-neutral mutation, given the dilutions to maintain a low multiplicity of infection at the beginning of each passage (11) (Fig. 1). Utilizing more advanced statistical techniques, such as Wright–Fisher Approximate Bayesian Computation (18), on the indel frequency time series data would enable the simultaneous inference of fitness effects and effective inoculum size.

Finally, it should be noted that the experiments conducted by Rangel et al. were performed in cell cultures. Thus, one might question whether the dynamics of indels inferred in PV and DENV populations are solely a result of the simplified pathosystem or whether they have correlates in in vivo virus population dynamics. As highlighted by the authors, previous studies have reported the presence of large deletions in the serum of DENV-infected patients (19). Furthermore, the authors reanalyzed three RNA-seq datasets generated from PV-infected mice (20) using MultiMatch and identified deletions similar to those found in the HeLa cultures. Notably, DVGs of varying sizes have been described in natural infections with numerous animal and plant viruses (3).