Comprehensive mutational scanning of a kinase in vivo reveals context dependent fitness landscapes
Alexandre Melnikov, Peter Rogov, Li Wang, Andreas Gnirke, Tarjei S Mikkelsen
This paper was recently published in bioRxiv by Tarjei Mikkelsen’s group. They use high-throughput screening and sequencing to map the distribution of mutation in a kinase evolved in the presence of increasing concentrations of different aminoglycoside antibiotics. The mutational analysis reveals that mutations accumulate differently according to the substrate (and concentration) used, suggesting a fitness landscapes can be context dependent.
I had a couple of comments, and Tarjei has been so kind to reply to me publicly in bioRxiv. Our exchange is also reported below.
Deep mutational scanning has emerged as a promising tool for mapping sequence-activity relationships in proteins, RNA and DNA. In this approach, diverse variants of a sequence of interest are first ranked according to their activities in a relevant pooled assay, and this ranking is then used to infer the shape of the fitness landscape around the wild-type sequence. Little is currently know, however, about the degree to which such fitness landscapes are dependent on the specific assay conditions from which they are inferred. To explore this issue, we performed deep mutational scanning of APH(3′)II, a Tn5 transposon-derived kinase that confers resistance to aminoglycoside antibiotics, in E. coli under selection with each of six structurally diverse antibiotics at a range of inhibitory concentrations. We found that the resulting fitness landscapes showed significant dependence on both antibiotic structure and concentration. This shows that the notion of essential amino acid residues is context-dependent, but also that this dependence can be exploited to guide protein engineering. Specifically, we found that differential analysis of fitness landscapes allowed us to generate synthetic APH(3′)II variants with orthogonal substrate specificities.
I was particularly interested by the change in tolerance you observe at different selection stringencies. However, I was wondering if it might be possible for you to test enzyme activity more accurately on some variants to prove this. Indeed, although mutation on catalytic residues might be tolerated in vivo (possibly by a binding process favoured by the very high enzyme relative to antibiotic concentration), differences in vitro could be significantly higher.
Did shotgun sequencing allowed you to rebuild individual full length sequences? If so, did you look at epistatic interaction between different positions? (you hint at it at some point, but it would be a long discussion to do here, so I’ll leave it for now pending interest on your side). If not, how did you go about rebuilding individual mutants?
Finally, your analysis of substrate-dependent mutational tolerance is very interesting from an evolutionary and functional point of view. Besides comparing mutants with all substrate-specific mutation-tolerant positions mutated, did you try to limit comparison to those that are in closer proximity to substrate/active site (there are some useful pymol script that can help for this)? Major functional changes could be brought about by fewer mutations, and the rest be noise responsible for the general decrease in overall activity you report. Here again, testing enzymes in vitro would tell you whether activity or solubility was primarily compromised.
As a very vague warning, I found the last part of conclusions slightly unfair. Partially because I work with (surprise surprise) directed evolution and hence believe that the technique itself is suited for similar experiments, although it hasn’t been used much still (alas, our draft isn’t quite ready yet!). But also because the functional and evolutionary results you report here are (to me) very interesting and I believe you could conclude on those directly instead than finishing on the potential of the method.
1) We have not yet performed comparisons of in vivo and in vitro activities, but I should clarify that the concentration-dependent tolerances we observe do not appear to extend to the catalytic residues previously identified by Nurizzo et al. and others. In particular Asp190 does not appear to tolerate mutations in any context – see for example Supplementary Figure 3. The concentration-dependent mutation tolerance is indeed likely to primarily reflect stability/solubility requirements as a function of antibiotic concentration. (Note that the substrate-dependence analysis attempts to control for this by matching selective pressures).
2) Shotgun sequencing does not allow reconstruction of full mutants, but we work around this limitation by constructing the libraries in such a way that the vast majority of mutants only have one substitution (see Figure 1 and Methods). The upside (or downside, depending on what you’re interested in) of this is that we greatly reduce confounding effects from epistatic interactions.
3) We did do some initial exploration of the effect of single vs multiple substitutions – see Supplementary Table 2 and Supplementary Figure 12. In general, the changes in specificity observed from combining mutations are not fully recapitulated by any single substitution. (But we’re certainly not arguing that the greedy optimization is the optimal strategy – in most of the cases there probably are better performing subsets).
4) No slight against directed evolution was intended! In fact, think selection and screening are very complementary approaches. The point was simply that it is often easier to set up two separate assays for activities A and B, rather than a single assay for “A but not B”. Thus, in many cases it may be easier to score all members of a low to moderate complexity library for their A and B activities separately in order to find the best “A but not B” candidates afterwards, rather than to try to directly select for “A but not B” from a large library. Whether or not this is true will of course be very dependent on the specific assays and molecules.