Temporal dynamics of viral fitness and the adaptive immune response in HCV infection

Reviewer #1 (Public review):

Summary:

The authors examine CD8 T cell selective pressure in early HCV infection using. They propose that after initial CD8-T mediated loss of virus fitness, in some participants around 3 months after infection, HCV acquires compensatory mutations and improved fitness leading to virus progression.

Strengths:

Throughout the paper, the authors apply well-established approaches in studies of acute to chronic HIV infection for studies of HCV infection. This lends rigor the to the authors' work.

Weaknesses:

(1) The Discussion could be strengthened by a direct discussion of the parallels/differences in results between HIV and HCV infections in terms of T cell selection, entropy, and fitness.

(2) In the Results, please describe the Barton model functionality and why the fitness landscape model was most applicable for studies of HCV viral diversity.

(3) Recognize the caveats of the HCV mapping data presented.

(4) The authors should provide more data or cite publications to support the authors' statement that HCV-specific CD8 T cell responses decline following infection.

(5) Similarly, as the authors' measurements of HCV T and humoral responses were not exhaustive, the text describing the decline of T cells with the onset of humoral immunity needs caveats or more rigorous discussion with citations (Discussion lines 319-321).

(6) What role does antigen drive play in these data -for both T can and antibody induction?

(7) Figure 3 - are the X and Y axes wrongly labelled? The Divergent ranges of population fitness do not make sense.

(8) Figure S3 - is the green line, average virus fitness?

(9) Use the term antibody epitopes, not B cell epitopes.

Reviewer #2 (Public review):

Summary:

In this work, Walker and collaborators study the evolution of hepatitis C virus (HCV) in a cohort of 14 subjects with recent HCV infections. They focus in particular on the interplay between HCV and the immune system, including the accumulation of mutations in CD8+ T cell epitopes to evade immunity. Using a computational method to estimate the fitness effects of HCV mutations, they find that viral fitness declines as the virus mutates to escape T-cell responses. In long-term infections, they found that viral fitness can rebound later in infection as HCV accumulates additional mutations.

Strengths:

This work is especially interesting for several reasons. Individuals who developed chronic infections were followed over fairly long times and, in most cases, samples of the viral population were obtained frequently. At the same time, the authors also measured CD8+ T cell and antibody responses to infection. The analysis of HCV evolution focused not only on variation within particular CD8+ T cell epitopes but also on the surrounding proteins. Overall, this work is notable for integrating information about HCV sequence evolution, host immune responses, and computational metrics of fitness and sequence variation. The evidence presented by the authors supports the main conclusions of the paper described above.

Weaknesses:

One notable weakness of the present version of the manuscript is a lack of clarity in the description of the method of fitness estimation. In the previous studies of HIV and HCV cited by the authors, fitness models were derived by fitting the model (equation between lines 435 and 436) to viral sequence data collected from many different individuals. In the section "Estimating survival fitness of viral variants," it is not entirely clear if Walker and collaborators have used the same approach (i.e., fitting the model to viral sequences from many individuals), or whether they have used the sequence data from each individual to produce models that are specific to each subject. If it is the former, then the authors should describe where these sequences were obtained and the statistics of the data.

If the fitness models were inferred based on the data from each subject, then more explanation is needed. In prior work, the use of these models to estimate fitness was justified by arguing that sequence variants common to many individuals are likely to be well-tolerated by the virus, while ones that are rare are likely to have high fitness costs. This justification is less clear for sequence variation within a single individual, where the viral population has had much less time to "explore" the sequence landscape. Nonetheless, there is precedent for this kind of analysis (see, e.g., Asti et al., PLoS Comput Biol 2016). If the authors took this approach, then this point should be discussed clearly and contrasted with the prior HIV and HCV studies.

Another important point for clarification is the definition of fitness. In the abstract, the authors note that multiple studies have shown that viral escape variants can have reduced fitness, "diminishing the survival of the viral strain within the host, and the capacity of the variant to survive future transmission events." It would be helpful to distinguish between this notion of fitness, which has sometimes been referred to as "intrinsic fitness," and a definition of fitness that describes the success of different viral strains within a particular individual, including the potential benefits of immune escape. In many cases, escape variants displace variants without escape mutations, showing that their ability to survive and replicate within a specific host is actually improved relative to variants without escape mutations. However, escape mutations may harm the virus's ability to replicate in other contexts. Given the major role that fitness plays in this paper, it would be helpful for readers to clearly discuss how fitness is defined and to distinguish between fitness within and between hosts (potentially also mentioning relevant concepts such as "transmission fitness," i.e., the relative ability of a particular variant to establish new infections).

One concern about the analysis is in the test of Shannon entropy as a way to quantify the rate of escape. The authors describe computing the entropy at multiple time points preceding the time when escape mutations were observed to fix in a particular epitope. Which entropy values were used to compare with the escape rate? If just the time point directly preceding the fixation of escape mutations, could escape mutations have already been present in the population at that time, increasing the entropy and thus drawing an association with the rate of escape? It would also be helpful for readers to include a definition of entropy in the methods, in addition to a reference to prior work. For example, it is not clear what is being averaged when "average SE" is described.