Glucose limitation in Lactococcus shapes a single-peaked fitness landscape exposing membrane occupancy as a constraint

A central theme in biology is to understand the molecular basis of fitness: which strategies succeed under which conditions; how are they mechanistically implemented; and which constraints shape trade-offs between alternative strategies. We approached these questions with parallel bacterial evolution experiments in chemostats. Chemostats provide a constant environment with a defined resource limitation (glucose), in which the growth rate can be controlled. Using Lactococcus lactis, we found a single mutation in a global regulator of carbon metabolism, CcpA, to confer predictable fitness improvements across multiple growth rates. In silico protein structural analysis complemented with biochemical and phenotypic assays, show that the mutation reprograms the CcpA regulon, specifically targeting transporters. This supports that membrane occupancy, rather than biosynthetic capacity, is the dominant constraint for the observed fitness enhancement. It also demonstrates that cells can modulate a pleiotropic regulator to work around limiting constraints.

Additionally, a requirement of laboratory evolution is that sources of technical variability 128 between replicates are minimized. Our initial focus was set on carefully defining the 129 cultivation conditions (chemostat and medium) and on standardizing experimental procedures 130 (inoculation, sampling, and cell storage). We developed a chemically defined medium 131 particularly suited for prolonged cultivation (CDMPC) based on the nutrient requirements 132    Parallel laboratory evolution in chemostat cultures 145 We exposed L. lactis MG1363 to parallel chemostat cultivations: Genr0 was cultivated 146 continuously in quadruplicate, each using 60 mL glucose-limited chemostats at a growth rate 147 of 0.5h -1 for over 300 volume changes. Culture samples were harvested periodically every 148 10-15 generations from the effluent to determine cell density and the fluxes of fermentation 149 substrates and products (an example is given in Figure 1C). Cell samples were stored as 150 glycerol stocks at -80°C to be used as snapshots of the evolutionary process. Throughout the prolonged cultivation, all four replicates (referred hereon as 309C1 through 152 309C4) displayed very similar behavior. Biomass concentration did not vary significantly 153 indicating that biomass yield on substrate remained fairly constant (Supplementary Figure 1). 154 L. lactis can catabolize glucose through homolactic or mixed acid fermentation, excreting, 155 respectively, lactate, or acetate, formate and ethanol . 156 At a dilution rate of 0.5 h -1 , Genr0 produces mainly lactate (lactate:acetate ratio ~16). 157 Throughout the course of the experiment, all parallel reactors shifted gradually towards 158 mixed acid fermentation leading to a ratio of approximately 7. Despite the fact that this 159 fermentative mode leads to a theoretical 50% increase in ATP yield, the biomass 160 concentration did not change. The latter can be explained by the observation that the evolved 161 cells also gradually excreted more pyruvate, the final shared precursor of the homolactic and 162 the mixed acid fermentative pathways (Supplementary Figure 1). 163 When evolved cells were grown on glucose in batch, we observed differences with Genr0 in 164 the maximal growth rate (μ max ), length of the lag phase and sedimentation. We compared to 165 Genr0 both the population samples collected after 309 generations and single colony isolates 166 from the evolved population samples ( Supplementary Figures 1 and 6). Irrespective of 167 whether the evolved populations or the isolates were compared, they were clearly 168 outperformed by Genr0 in batch. Most notably, μ max dropped to a value close to the dilution 169 rate at which cells were evolved, which falls nearly 25% below the μ max of the parent strain. 170 The evolved strains also exhibited extended lag phases and sedimentation during batch 171 cultivations. The sedimentation was reminiscent of the AcmA-deficient phenotype previously 172 described . 173 Identification of the mutations arising during the evolutionary experiment 174 We sequenced single-colony isolates from the original strain stock of Genr0 and from the end 175 of the four replicate evolution experiments performed at a dilution rate of 0.5h -1 (i.e. 309C1,  Table 2). In comparison 179 with Genr0, we found SNPs unique to the evolved strains ( Figure 2). The number of SNPs 180 accumulated per genome was between one (309C1) and six (309C4). This corresponds to a 181 mutation rate of 1.3 to 7.8 x 10 -9 (per base pair per generation) and is in line with currently  mutation must be sufficient to confer a growth advantage to the evolved strains when 204 compared to Genr0, and to a great extent be the genetic basis of the common phenotypic 205 differences observed. These include the growth kinetics and sedimentation phenotype 206 ( Supplementary Figures 1 and 6) and, as addressed below, altered glucose utilization kinetics.

207
The sequences were also analyzed for insertions and/or deletions and when compared to the 208 Genr0 sequence; the evolved strains showed no major frameshifts (see Supplementary Table   209 2).

210
Effect of CcpA M19I on global gene expression patterns 211 The evolved strains were revived in CDMPC and grown until mid-exponential phase at 212 which point microarray analyses were performed (Kuipers et al., 2002). Gene expression in 213 the evolved strains was compared to that of Genr0. Evolved strain 309C1, which contains  Table 3, Supplementary Table 4). CcpA controls the preferential use of 217 glucose over other sugars (Deutscher, Francke and Postma, 2006) and regulon analysis of the 218 10 differentially expressed genes revealed that CcpA-regulated genes were indeed over-219 represented in the data set.   The ability to utilize other carbon sources was diminished for the evolved strains 242 (Supplementary Figure 4A). This coincided with the down regulation of, amongst others, 243 genes encoding for maltose transporters, ABC sugar importers and sugar utilization enzymes 244 (Supplementary Table 3), many of which are CcpA-regulated.

245
In general, CcpA is a versatile global regulator that can act as a transcriptional repressor or 246 activator. An amino acid substitution in the DNA binding region may therefore have 247 qualitatively different effects at different target promoters. This is indeed what we observe: 248 15 CcpA-regulated genes are down regulated, 6 are up-regulated, while 11 remain 249 unchanged. We next examined the molecular basis of the differential effects on expression.  We then tested whether the phenotypic changes imposed by the CcpA mutation would be 306 sufficient to explain the wash-in kinetics observed using the glycerol stocks. For this purpose, 307 a simple model of chemostat growth with two competing strains was used to simulate the 308 14 prolonged experiment at a D of 0.5h -1 . The model was able to accurately fit the observed 309 population dynamics, suggesting that the underlying mechanism can be fully understood in once again the only mutation that was consistent across all isolates was located in residue 19 331 of CcpA. This time all possible codons that encode Ile were found, suggesting that the amino 332 acid substitution is the only factor that confers a strong fitness benefit. 333 We further tested whether the same mutation is also advantageous for dilution rates that are 334 far lower than μ max by performing competition experiments in chemostats between Genr0 and 335 309C1 (harboring only the CcpA-Met19 mutation). We used our growth model parameterized 336 with the data from the evolution experiments at a D of 0.5h -1 , to simulate the population   Replicate chemostats show high evolvability and low plasticity 357 We studied the molecular mechanism of the adaptive strategy employed by L. lactis during 358 growth in a glucose-limited chemostat. In this way, we kept the identity of the selective for a specific quantitative change in fitness then evolvability is low, and vice-versa.

373
Evolvability can be assessed by determining the number of generations that it takes fitter 374 mutants to fix (Colegrave and Collins, 2008;Pigliucci, 2008). High plasticity and high 375 evolvability occur when alternative mutants fix at similar rates. High evolvability can also 376 occur at low plasticity: then one mutation fixes across replicates and dilution rates. The latter 377 was clearly the case in our study, and it appears that there is a single-peaked fitness 378 landscape, suggesting a strong preference for this specific mechanism, independent of the 379 strength of the selection pressure. As we will discuss shortly, our data suggest that alleviating 380 constraints on membrane occupancy may be that mechanism. 381 Mutation in a global regulator rather than specifically in a glucose transporter 382 In the chemostat, fitter mutants outgrow their competitors because they can grow at a rate 383 equal to the dilution rate, but at a lower concentration of the limiting nutrient (Gresham and  407 This allows the overall fine-tuning rather than the optimization of a particular trait (e.g. 408 increasing Ks of a single transporter), in response to imposed selection pressure. 409 CcpA was not the only global regulator to accumulate mutations. We also found a Phe  Fitness enhancement independent from growth rate requires phenotypic plasticity 415 The fact that we find a single mutation and that we can capture the population dynamics of growth rates also studies here (0.2 -0.6 h -1 ). The presence of such metabolic phenotypic 430 plasticity seems therefore a requirement for our finding of low genetic plasticity and high 431 evolvability. It shows that the intracellular proteome does not impose major constraints on the 432 organism under these conditions, but rather the proteome of the membrane, as we will discuss 433 next. 434   provide an experimental case-study that suggests that indeed uptake constraints are the main 459 constraint for fitness in the chemostat, and that global regulators may be able to overcome 460 them, through pleiotropic effects.

461
In line with this, we found that the growth rate of the CcpA-Met19 mutant was lower than in 462 wildtype when grown in batch (glucose excess), despite higher glucose uptake capacity -463 something that is rather common for microorganisms evolved to nutrient-limited conditions 464 (Gresham and Hong, 2014). Also serial batch growth of L. lactis in emulsion-a method to 465 select for high cell number -selected for a mutation in the glucose transporter itself, rather 466 than in a regulator thereof (Bachmann et al., 2013). In that case, the mutation reduced 467 glucose transport activity, thereby mimicking a low-glucose environment associated with 468 acetogenic metabolism, a high ATP-yield strategy.

469
In conclusion, our work shows that evolution to constant growth conditions can be mediated   Table 1). The major differences compared to previously published        36 The sequences of Genr0 and the evolved strains were compared to those published for L.   Table 3). Many of these genes are CcpA-regulated.  Table 6).

129
The binding of CcpA from Genr0 and 309F1 to 4 DNA operators was examined: synthetic    using Ni-NTA resin (Qiagen, Germantown, MD, USA) as previously described (Kowalczyk 311 and Bardowski 2003). The protein concentration was determined with the DC Protein assay 312 kit (Bio-Rad, Hercules, CA, USA) using BSA as a standard.