The amino acid sequence of the receptor binding domain (RBD) of the A subgroup of FeLV varies little between isolates, constraining the virus to usage of the thiamine transporter THTR1 for infection. The switch in subgroup from A to C in cats with pure red cell aplasia is marked by amino acid alterations in the RBD that shift receptor usage from THTR1 to the haem transporter FLVCR1. While all anaemogenic strains of FeLV bear such substitutions, little is known about the genesis of the A to C switch. To date, each isolate of FeLV-C studied has displayed a unique RBD sequence, suggesting that recombination with endogenous env sequences is an unlikely source of the mutated RBD. In contrast, the emergence of subgroup B viruses is associated with recombination between exogenous and endogenous FeLV env sequences [13, 36, 37]. A more likely mechanism for the derivation of subgroup C viruses would be the acquisition of mutations or deletions in vivo in response to a selective pressure from the host, either through pressure to escape the adaptive immune response or through receptor availability in the tissue in which the virus replicates. Such a mechanism predicts the presence of variants with an intermediate tropism; subgroup A viruses with point mutations in the RBD that confer an enhanced or expanded receptor usage. Previous studies identified a subgroup C virus, FY981 that had retained the ability to utilise the subgroup A receptor THTR1 for infection , confirming that there are indeed “dual-tropic” or “poly-tropic” viruses amongst primary isolates of anaemogenic strains of virus. FY981 is actually a poly-tropic virus as it is able to utilise a third receptor (FLVCR2) in addition to THTR1 and FLVCR1 . Here, we demonstrated that subtle variations in the RBD of subgroup A viruses may have significant effects on the way the viruses interact with their receptors, potentially predisposing the viruses to in vivo mutagenesis. Accordingly, the presence of the combination of D83 and D91 in the background of A (Glasgow-1) was sufficient to enhance receptor binding, viral entry and viral replication. Residue D91 is particularly intriguing as it is present in the well-characterised Rickard strain of FeLV. In two separate studies examining recombination in FeLV infection, it was noted that inoculation of cats with a molecular clone (pFRA) of the Rickard strain of FeLV resulted in 1 of 3  and 1 of 5  cats developing an FeLV-C associated anaemia. As FeLV-C is thought to arise in <1% of infected cats, the high incidence of FeLV-C emergence following inoculation with FRA (33% and 20% respectively) may suggest an enhanced propensity for the development of FeLV-C. Mechanistically, a scenario can be envisaged whereby some subgroup A viruses may be inherently more pathogenic than others due to an enhanced ability to infect and spread in the infected host, a feature determined largely by the affinity of the Env for the viral receptor. Indeed, such a virus (FeLV-945) has been described and shown to have a higher binding affinity for its receptor . Mapping the determinants of the enhanced binding of the FeLV-945 SU to feline cells suggested that the major determinant of the enhanced binding of the 945 SU resided in variable region B (VRB) of gp70. FeLV-945 is a D83:D91 virus, similar to the DD mutant we examined in this study; however, inserting the VRA of 945 in the background of FeLV-61E (an SU that binds with a lower affinity to its receptor) did not confer an enhanced binding upon the 61E SU, suggesting that multiple determinants in gp70 may contribute to the receptor binding affinity. However, it should also be noted that the 61E SU varies from the Glasgow-1 SU at a number of other residues across gp70; the context in which D83:D91 is expressed may be critical to its effect on binding affinity. Moreover, in this study, we expressed SU fusion proteins with a C-terminal IgG Fc-tag (dimeric) and binding was assessed on both mouse and guinea-pig cells expressing individual receptors, whereas the 945-SU proteins  were expressed as C-terminal HA tagged proteins (monomers) and binding assessed on the feline lymphosarcoma cell line 3201, a cell line that produces a soluble 35 kDa endogenous FeLV Env protein capable of viral interference . Such experimental differences may modulate both the affinity and the specificity of the Env-receptor interaction in the two systems. For example, it has been shown that the context in which the receptor THTR1 was expressed altered the efficiency of receptor usage by FeLV  while soluble endogenous FeLV Env produced from 3201 cells conferred infectivity on the otherwise defective FeLV-T Env . Irrespective of the differences in the experimental systems, the enhanced binding of the Glasgow-1 DD mutant SU-Fc to THTR1 was consistent with the enhanced entry and replication of the virus, while the high affinity binding of the 945-SU was consistent with enhanced binding of intact virus particles from FeLV-945 to the same cells .
It is possible that both the affinity of the FeLV SU for its cognate receptor and its ability to induce fusion once bound combine to determine the eventual route of viral entry. Our data may predict that during FeLV-C evolution, additional mutations accumulate during long-term viral replication and that these mutations decrease SU affinity for THTR1 whilst increasing the relative affinity for the FLVCR1 homologues. Such viral evolution would eventually result in an Env capable of mediating fusion and entry via FLVCR1, producing the FeLV-C phenotype and associated PRCA symptoms. This theory is supported by the observation that FeLV-C (Sarma) displayed a severely limited ability to bind to THTR1 despite possessing the D83:D91 motif. However this altered binding affinity may be mediated by a range of mutations across the SU, not comprising a single binding motif, explaining why individual FeLV-C Env proteins are functionally but not genetically conserved. It is possible that the overall final Env conformation, rather than specific individual residues, permits FLVCR1-mediated membrane fusion and cellular entry. The deletions which we observed in multiple FeLV-C env clones from our primary isolates may be essential for decreasing the affinity of the THTR1-SU interaction and allowing FLVCR1-mediated entry. This deletion may represent the final mutation in the progression from FeLV-A to -C. The mechanism(s) underlying the deletion of residues within the RBD-encoding region of env remain to be ascertained; however, a scenario could be envisaged wherein the RNA structure within some regions may be less stable and inherently more likely to break and reform.
Having established that subtle variations in the VRA of FeLV-A could have a significant impact upon the biological properties of subgroup A FeLV Glasgow-1, we asked whether we could mimic the in vivo selective pressure exerted upon FeLV by the humoral immune system. By culturing virus in the presence of sub-optimal concentrations of either monoclonal anti-gp70 antibody or pooled serum from FeLV-infected cats, we demonstrated the acquisition of non-synonymous mutations with time in culture, although we were not able to demonstrate a shift from subgroup A (THTR1-using) to subgroup C (FLVCR1-using). Subgroup C viruses emerge in an estimated 1% of anaemic cats, suggesting that a relatively rare set of circumstances combines to drive their evolution. It is possible that the epitope specificity of the antibody response elicited following infection may prove critical in determining the composition of variants that evolve, and so serum from cats from which FeLV-C had been isolated rather than a diverse pool of FeLV-infected cat sera would be required to mimic this response in vitro. While our working hypothesis is that the humoral immune response influences the likelihood of FeLV-C emerging in infected cats, other factors may have a significant impact upon viral evolution; for example VRA may constitute a T cell epitope in some cats and pressure to escape a cellular immune response may play a role in driving variation in VRA. Alternatively, variation amongst cats in both the levels of receptor expression and the cell types upon which the receptors are expressed may determine sites of viral replication. A scenario could be envisaged whereby co-expression of both THTR1 and FLVCR in the same cell may permit the emergence of variants with dual tropisms and ultimately a specific tropism for FLVCR1-expressing cells.
A detailed analysis of the mutations that were identified allowed several inferences to be made. Firstly, the rate of mutation appeared consistent across all cultures, indicating viral genetic drift occurred at similar rates regardless of the presence or absence of VNAs. Additionally, few mutations were identified more than once, indicating a single FeLV genome had not emerged as the dominant viral species in any culture. However, the limited capacity of our env selection method (cloning as opposed to deep-sequencing) must be taken into account as it is possible that the amplification of more env sequences, or the use of alternative techniques, may have produced different results. Amplification and cloning of env genes may not provide a sufficiently broad picture of the genomes present in the culture. It remains possible that dual-tropic virus strains, or those containing mutations indicative of a FeLV-C phenotype, were present in some cultures, but were not detected during our analysis. The results presented herein therefore represent a “snapshot” of the viral genomes present at 50 days post-infection. We were unable to determine whether specific viruses formed prominent subpopulations during the course of infection, indicative of a species with a replicative advantage expanding.
Apart from viral genetic drift, there are numerous alternative mechanisms for the observed mutations; for example, some mutations may have arisen at sites as a result of structural biochemistry. There is evidence that adenine-thymine tracts (consisting of four consecutive A or T nucleotides) are associated with “bends” in the nascent DNA strand, and are more likely to be mutated through misincorporation . As 13 of the 59 mutations (~22%) observed in the long-term replication study occurred in AT-rich tracts (nucleotide data not shown), it is possible these are due to this phenomenon, rather than either neutralisation escape or receptor tropism expansion as originally predicted. In addition, some mutations may have arisen as the result of apolipoprotein B mRNA-editing enzyme catalytic polypeptide (APOBEC) activity. APOBECs mutate retroviral genomes by deaminating the cytosines within DNA, eventually causing an accumulation of G to A mutations in the RNA genome. However, there was minimal evidence of APOBEC activity in the env sequences analysed in this study; of the 59 mutations identified only 7 were G-to-A transitions. Additionally, only 3 of these 7 were found within likely targets for feline APOBECs (AGG or GGG motifs) . This is in accordance with other reports that APOBECs exhibit weak restriction of FeLV in natural infections [42, 43].
The inclusion of VNAs in our long-term replication study was based upon the assumption that FeLV-C evolution may be a result of the replicating virus escaping antibody-mediated neutralisation. This hypothesis is not without precedent as there are numerous instances of retroviral antigenic variation and receptor usage alterations being driven by pressure from the host immune response. VNA play a role in the selection and expansion of viral variants in both simple and complex retroviruses, and this has been mimicked successfully in vitro in numerous cases [44–46]. In the example of equine infectious anaemia virus (EIAV), an in vitro model of viral evolution found that 13 viral passages were required to obtain an antibody-escape mutant. This phenotype was conferred by only two altered epitopes in the SU domain . This is a similar timespan to that used in our study, indicating it was sufficient to observe antibody-escape mutants. Despite the lack of FeLV-C associated mutations observed here, development of FeLV-C as a consequence of antibody escape remains a plausible theory. There are numerous variables in this process which could not be replicated in our model, including the broad range of antibodies produced in a competent immune response. Before definitive conclusions can be drawn about the role of VNAs in FeLV-C development, this experiment should be repeated using a more extensive range of both polyclonal and monoclonal antibodies.