Parallel evolution during the SREE is associated with negative regulators

In strong contrast to the REE, the most frequent targets of mutation outside of wss or PFLU0085 in the SREE were not genes encoding putative DGCs. The removal of wsp, aws and mws appears to have increased the relative frequency of non-WS types arising from this selective regime. These types were either unique to the SREE, as in the case of mutations to PFLU4939, or occurred at a higher frequency than in the REE, as in the case of the mutation to nlpD.

Similar to the common pathways in the REE – wsp, aws and mws – these convergent pathways in the SREE appear to cause an adaptive trait following a loss-of-function

mutation. This provides a possible example of loci, additional to wsp, aws and mws, at which encoded regulation increases the translation of mutation into adaptive phenotypic change. An explanation of the functionality of each pathway, and how this functionality interacts with deleterious mutations to provide adaptive outcomes in the SREE, is presented below.

4.4.5.1 Parallelism at locus PFLU4939 is associated with loss-of-function mutations

Mutations to PFLU4939 arose in three separate lines of the SREE and in all cases arose in shaken environments following mutations to c-di-GMP regulators. As described in Section 4.3.5.1 above, this locus encodes a transcription factor homologous to MvaT - a regulator for multiple exoproducts across Pseudomonas spp. The identification of three independent mutations at this locus, one of which is a nonsense mutation, suggests these three mutants all involve the loss of function of MvaT, which results in the increased expression of these exoproducts. The regulatory nature of the regulator encoded at this locus appears to make this gene a target of repeated deleterious mutations. This again appears to provide an example of the role of negative regulation in biasing the rate at which loci cause phenotypic change.

Mutations to PFLU4939 were not identified in the original REE. This is surprising given that these mutants arose in environments similar to that which wss mutants evolved in both the original REE and the SREE. As with mutations to wss, mutations to PFLU4939 arose in shaken environments with WS types as direct ancestors. The adaptive advantage of mutations in mvaT in the SREE – as compared to the REE – can be explained via three, non-exclusive, mechanisms.

1) Molecular epistasis. The mutated product of PFLU4939 may produce specific epistatic interactions with preceding mutations, which may increase the fitness of the PFLU4939 mutants. These interactions result either from the three deletions in the ancestral genotype, or from mutations that have arisen during the course of the SREE. This is supported by the identification of PFLU0956 as the direct ancestor of two PFLU4939 mutants. MvaT may act as a transcriptional regulator of DGCs such as PFLU0956, in which case loss of function of MvaT may be adaptive via reduction of expression of these DGCs. This hypothesis may be interrogated by introducing the PFLU4939 mutation in genotypes with and without DGC mutations (including mutations

of PFLU0956). The fitness of genotypes featuring PFLU4939 mutations in combination with PFLU0956 mutations should be significantly higher than the fitness caused by PFLU4939 mutations in combination with other DGC mutations.

2) Target size of the ancestral WS-causing genes. WS types may be reverted to SM by mutations at non-GGDEF domain loci that regulate the WS phenotype (such as wss) or by the mutation of genes that reduce the phenotypic changes caused by c-di-GMP production (here called ‘masking mutants’). In the genotype in which PFLU4939 mutation arose, the preceding GGDEF domain-encoding genes have relatively small target sizes. This small size may preferentially increase the chance evolution of ‘masking mutants’ such as mvaT mutants, compared to mutants that target the preceding GGDEF domain-encoding genes. In lines 4 and 6, mutations to PFLU0956 were identified in the direct ancestor to the PFLU4939 mutation. PFLU0956 is a relatively small gene (1494 bp) compared to the operons such as wsp (~8.4 kb), aws (~2.3 kb) and mwsR (3.9 kb). In line 5, the preceding mutations are loss-of-function PFLU458 and PFLU4744 mutants – these genes are thus unlikely to be the target of further reverting mutations. The small target size of these preceding WS-causing genes may reduce the rate at which loss-of-function mutations to these c-di-GMP regulators arise in shaken environments. This reduced target size may allow the evolution of alternative mutants, via mutations to targets such as PFLU4939 that may ‘mask’ the effects of the WS phenotype.

3) The relatively high fitness of ancestral WS types in shaken environments. The ancestral WS prior to PFLU4939 mutants may exhibit a high degree of fitness in the shaken environment (compared to wsp, aws or mws mutants). A reduced cost of the WS phenotype in the shaken environment will decrease the frequency in which loss-of- function mutations are selected and may also result in the evolution of alternative phenotypes that increase fitness in the shaken environment, without affecting cellulose production. In such a historical context, PFLU4939 mutants may arise that would not be adaptive with a preceding mutation to wsp, aws or mws. This hypothesis could be investigated by comparative fitness assays of different WS types, including PFLU0956, when grown in shaken microcosms.

Without further detailed work exploring the molecular interactions and the phenotype caused by the PFLU4939 mutations, such hypotheses explaining the repeated evolution of PFLU4939 mutants remain untested.

4.4.5.2 Accounting for parallelism in the nlpD gene

The propensity for the nlpD gene to affect a phenotype via a loss-of-function mutation provides an explanation as to why this gene is the target of mutation in multiple instances in the SREE. It is interesting to note that nlpD mutants arose in two of the three cases following deleterious mutation to the wss operon, suggesting that these cell-division mutants have an increased chance of arising provided an absence of the WS type. This is consistent with the REE, in which nlpD mutation arose in five independent lines, with deleterious mutations to the wss operon occurring in the ancestors of four of these nlpD mutants. However, despite nlpD appearing as a target for loss-of-function mutations, there is extreme molecular parallelism at this locus, with mutations occurring in three independent instances in the SREE at an identical nucleotide (C565T). The only alternative loss-of-function mutation to nlpD was found in line 8, and only after a mutation to C565T had previously occurred. This strongly suggests a secondary feature biases either the mutation rate to this region or constrains the possible spectrum of adaptive mutations to this particular locus. Given the highly unusual nature of this molecular parallelism, this mutation was the subject of further study (see Chapter 5).