Reducing the sensitivity of Halomonas sp. to oxygen availability through adaptive laboratory evolution

TL;DR

Scientists evolved a salt-loving bacterium to grow better in low-oxygen conditions, improving its usefulness for industrial processes.

Contribution

A new framework for reducing microbial sensitivity to oxygen through adaptive evolution is proposed.

Findings

Adaptive evolution improved Halomonas sp. growth robustness by 62% in micro-aerobic environments.

PHB production yield increased by 21%, while ectoine production decreased by 79%.

Genetic mutations in transport and enzymatic pathways were linked to improved oxygen tolerance.

Abstract

Halophiles attract increasing attention to serve as sustainable industrial hosts for prolonged continuous processes or in open-vat fermentations owing to the reduced risk of contamination enabled by high salt concentration in the medium. Despite growing interest, their application in large-scale manufacturing remains limited partly due to bioprocessing challenges. A robust host that performs consistently well across scales should withstand variations in oxygen availability since local hypoxic regions often manifest in large-scale tanks even under strict operational control. In this work, we modified Halomonas sp. to achieve robust growth profiles in micro-aerobic environments by gradually exposing a continuous culture to stress induced by reduced oxygen availability using adaptive laboratory evolution. Nominal contamination was observed at 8.5% salt concentration even during non-aseptic operation, offering a suitable environment for prolonged continuous cultivation. Following adaptive laboratory evolution, Halomonas sp. achieved comparable growth at low-to-moderate (0.5%-3.0%) and moderate-to-high (3.0–5.0%) dissolved oxygen availability (OD600 = 4.46 ± 0.29 & µ = 0.40 h− 1 and 4.96 ± 0.81 & µ = 0.35 h− 1, respectively). Adaptive evolution improved the growth robustness of Halomonas sp. by 62% and its PHB production yield by 21%, while reducing ectoine production by 79% indicting improved metabolic perception of oxygen availability and reduced stress response. Mutations in cellular transport mechanisms and enzymatic pathways enabled this response as indicated by whole genome sequencing. The proposed framework is an effective strategy to reduce the oxygen sensitivity of microbial strains in general to improve their physiological properties, and to genetically design populations suitable for large scale bioprocess operations.

The online version contains supplementary material available at 10.1186/s12934-026-02951-w.

Figures

Determination of optimal salt concentration for the ALE experiment. Effect of NaCl concentration on A growth profiles, B the maximum specific growth rates, and C the maximum OD_600_ of wild-type *Halomonas* sp. DSM7328 is displayed for salt concentrations ranging from 3.5% to 9.5% compared to 1.0% NaCl in LB medium used as control. All error bars indicate the standard deviation (SD) of six replicates (*n* = 6). Statistical analysis was performed using student’s t-test (p* < 0.05 and *p* < 0.01). These analyses were carried out by a semi-permeable film that induces micro-aerated conditions in t

Growth characterisation of *Halomonas *sp. DSM7328 grown in LB85 at different dissolved oxygen concentrations. A Growth profiles, B the maximum specific growth rates (µ_max_), and C the maximum OD_600_ when the bioreactor DO level was controlled at 30%, 10%, 5.0%, 1.0%, or no oxygen influx was allowed are shown in the subsequent plots. Each condition was tested in 4 replicates (*n* = 4) and all error bars indicate the standard deviation (SD) of these four replicates (*n* = 4) with no outliers detected. Statistical analysis was performed using student’s t-test (p* < 0.05 and *p* < 0.01)

Adaptive laboratory evolution (ALE) experiment performance of *Halomonas* sp. DSM7328. A Overview of the chemostat operation by stepwise reduction of DO availability over time (black) and the corresponding dilution rate adjustments (blue) that were introduced. The dilution rates were adjusted to maintain consistent optical density in the outlet stream and were always set at or below the maximum specific growth rate of the original strain. B Culture OD_600_ (black dots) in response to the changes in dilution rate presented for the complete course of ALE. The small image in the right lower corne

Growth evaluation of the evolved population (EV) and the original strain (ST). A Growth profile in 96-microwell plates, B the specific growth rates, and C the maximum OD_600_ of ST (light shades) and EV (dark shades) in micro-aerated (blue) (*n* = 78 for ST and *n* = 81 for EV) and aerated (grey) conditions (*n* = 90 for both ST and EV) were tested in 96-microwell plates, and D in 250 mL shake flasks (*n* = 3) with baffled flasks with vented caps for mimicking aerobic environments (grey) and with non-baffled flat caps to mimic a reduction in oxygen availability (blue)

PHB, ectoine, and hydroxyectoine production of the evolved population. A Intracellular PHB accumulation in mg per unit OD_600_, B intracellular ectoine concentration, and C intracellular hydroxyectoine production of the original strain (grey) and the evolved population (blue) grown in aerated or micro-aerated environments. All error bars indicate the standard deviation (SD) of three replicates (*n* = 3). Statistical analysis was performed using student’s t-test (p* < 0.05 and *p* < 0.01)

Whole genome sequencing summary for identifying the mutational load of *Halomonas* post-ALE. The breakdown of the different types of mutations acquired in the whole genome are provided in (A), and the relative distribution of these mutations in the coding and the non-coding genome is given in (B). The breakdown of different types of mutations including insertion, deletion, single nucleotide variant, multi-nucleotide variant, and replacement acquired by the population were provided for the coding (C) and the non-coding (D) genome separately. E displays the relative distribution of the total num

Full text

Introduction

Continuous biomanufacturing has been receiving increasing interest owing to its several advantages over batch and fed-batch processing. The high productivity and the high maximum specific growth rates that can be achieved [1] as a consequence of the sustained addition of fresh culture medium and the removal of the spent medium [2], as well as the reduced capital costs [3] and energy consumption [4] renders it a promising platform to produce variety of value-added products more effectively, aligning closely with the United Nations sustainable bioeconomy goals.

Industrial implementation of continuous operations relies heavily on developing appropriate strategies for process optimisation to maximise productivity and to overcome process-specific challenges, which include increased risk of contamination over prolonged operation periods and the acquisition of host strain mutations during long process runs [5]. Developing an open or non-aseptic continuous process that utilises extremophilic microorganisms as hosts offers one possible option to achieve cost efficiency while simultaneously adhering to sustainability goals [6].

The Halomonas genus has recently attracted attention for developing such processes suitable for industrial applications. The genus comprises halophilic bacteria that can thrive in high salinity environments. The species of the genus are currently investigated for their potential to act as alternative chassis owing to their competitive advantages such as their suitability for open-vat processes [7, 8]. In addition to their process-specific advantages, several species of the genus such as Halomonas boliviensis [9] and Halomonas salina [10] were reported to produce and intracellularly accumulate valuable native metabolites such as poly(3-hydroxybutyrate) (PHB) and 4,5,6-tetrahydro-2-methyl-4-pyrimidinecarboxylic acid (ectoine). Ectoine is a biocompatible solute, which serves as an osmoprotectant in the presence of high extracellular salt concentration [11, 12], and is employed as a protectant in cosmetics and biomedical industries [13]. Some species of the genus were specifically implicated in the production of 5-hydroxyectoine, a hydroxylated derivative of ectoine with similar but superior functional properties [14]. Polyhydroxyalkanoates (PHAs), a class of metabolites including PHB, are highly sought after as they are precursors of bioplastic compounds [15]. PHA accumulation in Halomonas was shown to be strictly impacted by dissolved oxygen level of the environment [16, 17].

More recently, Halomonas-based research diversified to incorporate previously unexplored, or recently discovered species to tap onto the potential exhibited by the diverse characteristics that the members of the genus display. An increasing number of Halomonas species with a varying range of capabilities are being discovered, isolated and characterised; such as the crude oil-utilising Halomonas shengliensis sp. nov [18]. , or naphthalene degrading Halomonas dongshanensis sp. nov [19]. and Halomonas pacifica strain Cnaph3 [20], . Halomonas species are assessed for their ability to utilise non-conventional feedstocks for chemical production and growth. Examples include Halomonas alkalicola growth on fruit peels [21], or Halomonas sp. YLGW01 growth on crude glycerol [22].

Novel genetic tools were developed for editing Halomonas genome such as the implementation of a cas9 gene disruption system in Halomonas sp. KM-1 [23] and the engineering of a salt-inducible ectoine promoter region in Halomonas elongata [24]. The physiological properties of the genus were altered by genomic modifications to endow the cells with the capability to perform self-flocculation by deletion of exopolysaccharides and O-antigen in Halomonas bluephagenesis [25], or to produce bacterial Vitreoscilla haemoglobin (VHb) for increased production of poly(3-hydroxybutyrate), P3HB [26]. The introduction of a bacterial haemoglobin was a strategy for improving cellular oxygen availability for P3HB production. The investigation of the impact of oxygen on Halomonas species has been very limited to date; the challenge is further compounded by the complexity of the oxygen response at the cellular level rendering the success of targeted genetic engineering strategies that involve specific genes or genomic sites limited.

One strategy for conditioning bacteria to surviving in harsh environments is Adaptive Laboratory Evolution (ALE), which is based on the concept of random acquisition and long term selection of desirable traits over the course of gradually increased exposure to harsh or stressing environmental conditions [27]. It is typically performed through sequential serial passages [28] or continuous cultivation [29] and is a suitable method for improving the metabolic capabilities of a broad range of species; ethanol production of Saccharomyces cerevisiae was increased in the presence of toxic compounds, Bacillus coagulans survival was improved at low pH, Escherichia coli survival and recombinant protein production was improved at high pH [28], or the growth rate of Cupriavidus necator was increased in the presence of 0.5% glycerol [30]. ALE was adopted as a strategy to improve Schizochytrium sp. survival in response to environmental stress induced by oxygen. The growth of this microalgae was improved by adapting the cells to high oxygen availability, which had a direct impact on its docosahexaenoic acid productivity and cognate enzyme activities, although the oxygen levels were neither monitored or controlled in this batch serial dilution setup [31].Similar non-targeted genetic engineering strategies have not yet been tested to enhance the micro-aeration resistance of Halomonas for the purposes of improving host resilience for bioprocess operations.

In this work, we deployed continuous ALE for the first time for any Halomonas species in a strictly controlled bioreactor to create a healthy Halomonas sp. population that can grow robustly despite oxygen limitation. We will show in the following sections that the population was able to maintain its robustness at dissolved oxygen levels as low as 0.5%. For this purpose, the optimal ALE conditions were initially identified. As a first step, the optimal salt concentration of the medium was determined to ensure minimal risk of contamination in chemostats operating for extended periods of time. This was followed by the determination of the starting dissolved oxygen (DO) level for the ALE experiment. The optimised conditions established in the preliminary studies were used as the initial setup in the chemostat, where the cell population was gradually subjected to increasingly limited dissolved oxygen availability. Consequently, the population acquired and retained mutations that enhanced its robustness under the near-hypoxic conditions. The evolved population was assessed for growth robustness by means of reduced phenotypic variability, as well as its ectoine and PHB production. Finally, whole-genome sequencing was performed to identify the genetic changes acquired during this adaptation.

Materials and methods

Bacterial strain

Halomonas sp. DSM7328 was purchased from Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures GmbH (DSMZ). The culture temperature for this strain was reported at 30 °C [32]. Master stock was stored in 20% glycerol at -80 °C. This glycerol stock was inoculated overnight in LB broth supplemented with 60 g·L^− 1^ NaCl at 30 °C for pre-culturing.

Growth profiling

Halomonas sp. DSM7328 was grown in LB Broth (10 g·L^− 1^ casein enzymic hydrolysate, 5 g·L^− 1^ yeast extract, and 10 g·L^− 1^ NaCl) from Millipore, Merck. For determining the optimal NaCl concentration, cells were tested with NaCl supplemented at varying concentrations in LB broth totalling 3.5%, 4.0%, 5.0%, 6.0%, 7.0%, 8.0%, 8.5%, 9.0% and 9.5%. The control LB medium contained 1.0% NaCl as per the standard recipe. The experiment was conducted using both sterile and non-sterile medium in parallel studies exposed to the same environment. The overnight cell cultures were inoculated in fresh NaCl-supplemented LB media to a starting OD_600_ of 0.05. Cultures were aliquoted to 200 µL in each well of 96-well plates (Catalogue no. 82.1581, SARSTEDT), and the microwell plates were covered with Greiner EASYseal™ clear film (Catalogue no. A5596, Sigma-Aldrich). Each condition was assessed in 6 replicates. Culture plates were incubated at 30 °C, agitated at 200 rpm in CLARIOstar^®^ Plus Plate Reader (BMG Labtech, Germany). OD_600_ measurement was evaluated in the plate reader in 5-minute intervals until 10 h of cultivation. Gram staining was performed followed by the manufacturer’s protocol (Catalogue no. PL.8055/25, Prolab Diagnostics, UK). The specimen was observed under the microscope (Zeiss ICS KF2, Germany).

Growth rates were calculated by plotting the natural logarithm of OD_600_ (ln(OD_600_)) against cultivation time. The slope of linear range at the exponential phase was reported as the specific growth rate.

Bacterial cultivation in bioreactor

Cells were cultured in 900 mL LB medium supplemented with 85 g·L^− 1^ NaCl in total (LB85 medium) in a 1.2-litre Applikon bioreactor (Getinge, Germany). Batch fermentation was performed with starting OD_600_ adjusted to 0.05. Culture conditions were set at 30 °C, pH 7.0, and 400 rpm agitation speed. 5 M NaOH base and 20% phosphoric acid were used to control pH. The antifoam, 10% polypropylene glycol, was added at the beginning of cultivation. The different target DO levels were tested in individual cultures in 4 replicates at 0.0, 1.0, 5.0, 10, 20 and 30% DO. Absorbance was measured using OD (Dencytee^®^) sensor (Catalogue no. 243756, Hamilton, Germany) at 12-minute intervals and monitored by Hamilton Arc View. Absorbance values from the probe were calibrated with OD_600_ measurement. The absorbance was converted into OD_600_ measurement using the following equation:

\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\mathrm{OD}}_{{{\mathrm{600}}~}} {\text{ = }}\left( {{\mathrm{0}}{\mathrm{.42}} \times {\mathrm{Absorbance}}} \right) \pm {\mathrm{offset}} $$\end{document}

where 0.42 was the ‘cell factor’ and the offset assumed the value in the range of 0.8–2.7 in different measurements, as denoted by the software. The specific growth rates were calculated from the slope of linear regression plot between natural log of cell densities against cultivation time.

Adaptive laboratory evolution via continuous fermentation

The Adaptive Laboratory Evolution (ALE) experiment was conducted in 1.2-litre Applikon bioreactors operated in continuous mode with a working volume of 750 mL. The initial settings were the same as those described in the previous section. Frozen stocks were grown in LB85 medium at 30 °C, 200 rpm overnight before they were used to inoculate the preculture. Cells inoculated to an OD_600_ of 0.05 in LB85 medium were grown at 30 °C, 200 rpm for 10 h. The bioreactor was then inoculated to a starting OD_600_ of 0.05 from this preculture. The cells were cultured at 5.0% DO in batch mode until they reached mid-exponential growth phase. A response was observed in DO levels in the form of a sharp peak in the software traces indicating the instigation of oxygen limitation for the culture, which was used as a signal to switch from batch mode to continuous mode of operation. Fresh sterile feed of LB85 medium was continuously fed into the bioreactor in continuous mode. DO level was set at 5.0% as the starting culture setting. The dilution rate was increased stepwise in 0.05 h^− 1^ intervals from 0.1 h^− 1^ to 0.4 h^− 1^, which was identified as the critical dilution rate, determined from the maximum specific growth rate of the culture. Cell density was maintained in the range of 4.0–5.0 OD_600_ as the dilution rate was increased. Cell culture was monitored by OD_600_ readings using OD (Dencytee^®^) sensor and spectrophotometer (Jenway). DO level was reduced after the culture remained at a steady state for 100–200 generations within ± 10% OD_600_ for the set DO level. DO setpoint was stepwise reduced from 5.0% to 4.0%, 3.0%, 2.0%, 1.0%, and 0.5%, subsequently. The generation number at each DO level was calculated according to Eqs. (2) and (3). Population sample was collected after the culture spent 185 generations at 0.5% DO. Cells were harvested in 20% glycerol and stored at -80 °C for further analysis.

\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{Generation number = }}\frac{{{\text{Cultivation time}}}}{{{\text{Doubling time}}}} $$\end{document} \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\text{Doubling time}\text{}\mathrm{=}\frac{\mathrm{ln(2)}}{\text{Dilution rate}}$$\end{document}

where dilution rate was selected to be the specific growth rate (µ).

Culture conditions for testing evolved population and original strain

For the preculture, the evolved population and the original strain were grown overnight at 30 °C, 200 rpm. Cells were inoculated into 200 µL sterile LB85 medium containing 96-microwell plates to OD_600_ of 0.05. The air permeability was controlled using Greiner EASYseal™ clear film (Catalogue no. A5596, Sigma-Aldrich) and sterile Aeraseal™ film (Catalogue no. Z721573, EXCEL Scientific, USA) for micro-aerated and aerated conditions, respectively. Each condition was assessed in at least 60 replicates. Plates were incubated at 30 °C, 200 rpm and the OD_600_ was measured in 5-minute intervals in CLARIOstar^®^ Plus Plate Reader (BMG Labtech). Growth rates were calculated as previously described for the aerated condition. The linear range was selected as the timeframe of 8–10 h post-inoculation for the aerated conditions and 4–6 h post-inoculation for the micro-aerated condition. The cultures achieved their maximum OD_600_ values at the stationary phase of growth and consequently the maximum OD_600_ values were calculated by averaging the OD_600_ readings collected in the timeframe of 20 to 25 h.

For scaling purposes, the original strain and the evolved population were cultured in two different types of 250 mL shake flasks in triplicates; baffled flasks with vented cap (Catalogue no. BBV250, Fisher Scientific) and non-baffled flasks with flat cap (Catalogue no. PBNV250, Fisher Scientific) mimicking the aerated and the micro-aerated conditions, respectively. Flasks were inoculated to an OD_600_ of 0.05 in 50 mL LB85 medium. Growth rates and the maximum OD_600_ were determined as described above. Samples were collected at the late exponential phase (after 10 h) for intracellular PHB and ectoine measurements.

Intracellular PHB and ectoine measurements

PHB measurement protocol was adapted from [33]. Briefly, the cell pellets were collected by centrifugation at 12,000 rpm for 5 min using Eppendorf centrifuge 5424 R. Pellets were resuspended in 1 mL sterile water. Then, 40 µL of Nile Red dye (Catalogue no. 19123, Sigma-Aldrich) dissolved in DMSO was introduced into the suspension. The mixture was incubated at room temperature for 30 min and centrifuged at 12,000 rpm for 10 min; the supernatant was discarded. The pellet was resuspended in 1 mL sterile water by vortexing vigorously. A total of 300 µL of this solution was aliquoted into 0.5 mL reaction tubes. The fluorescence intensity was measured at 525 nm excitation and 565–650 nm emission using fluorometer (DeNovix, USA). Then, PHB concentration was normalised by OD_600_ from cell cultures to achieve a unit of mg PHB per OD_600_. The PHB Standard poly[(R)-3-hydroxybutyric acid] (Catalogue no. 363502) was purchased from Sigma-Aldrich, Merck.

Ectoine and 5-hydroxyectoine measurements were performed by reverse-phase high performance liquid chromatography (RP-HPLC). Ectoine and 5-hydroxyectoine were extracted from cell pellets by ethanol [34]. Cultures were centrifuged at 4,000 rpm, 4 °C for 30 min to collect the pellets. Pellets were resuspended in 90% v/v ethanol (Catalogue no. 459836, Sigma-Aldrich) by vigorous vortexing and shaking at 1,000 rpm for 30 min. Supernatant was collected and filtered through PTFE syringe filter 0.22 μm (Catalogue no. ANP1322, Gilson Scientific). Filtered samples were analysed via RP-HPLC (Agilent) using acetonitrile as a mobile phase in ZORBAX Eclipse XDB-C18 column 4.6 × 12.5 mm (Catalogue no. 990967-902). The mobile phase was 40% acetonitrile/MilliQ water, and the flow rate was 1 mL·min^− 1^. Both ectoine and 5-hydroxyectoine were monitored at 210 nm using UV/Vis detector. The standards for ectoine (Catalogue no. 81619) and 5-hydroxyectoine (Catalogue no. 70709) were purchased from Sigma-Aldrich, Merck.

Whole genome sequencing, data preprocessing and processing

The evolved population and the starting population of Halomonas sp. DSM7328 were cultured in 10 mL LB85 medium overnight at 30 °C, 200 rpm. Genomic DNA of both strains were isolated using DNeasy Blood & Tissue Kit as per the manufacturer’s instructions (Catalog no. 69504, Qiagen). DNA concentrations and purities (A_260/230_) were measured using NanoDrop Microvolume Spectrophotometer (Thermo Fisher Scientific). Approximately 300 ng DNA was provided for processing.

Illumina DNA Prep kit with Unique Dual Indexes was used; libraries of equal volumes were pooled prior to the final clean-up. The concentration of the library pool was estimated using the Qubit dsDNA HS assay (Life Technologies) and library pool size was confirmed on the Agilent TapeStation 4200 (Agilent HS D1000 assay). The library pool was denatured and sequenced on the MiSeq instrument (Illumina, San Diego, US) at 12pM, using a V3 600-cycle kit (paired end reads; 2 × 300 bp).

Samples were quality and adapter trimmed using fastp. The trimmed reads were run through Kraken [35] using the standard-8 database. Kraken output was further processed using Bracken [36] to estimate abundances. Samples were confirmed to belong to the Halomonas genus prior to analysis to avoid any contamination-related misinterpretations. Raw reads for the original strain sample were then processed through Bactopia pipeline (https://bactopia.github.io/). Reads were quality trimmed and assembled using Unicycler [37] and annotated with Prokka within the Bactopia framework. The annotated consensus was imported to CLC Genomics Workbench v11.01. Both the original strain and the evolved population sequence quality trimmed reads were mapped back to the original strain consensus, duplicates were removed, and variants were called at minimum 10 reads coverage and 2% frequency threshold using CLC genomics workbench. SNPs that were identified were annotated based on the Prokka annotated original strain. The mutated genes were identified and categorised based on the frequency of the mutation acquired in the sample. The raw and preprocessed data are available in EBI’s ArrayExpress Database under the accession number: E-MTAB-15,059.

Statistical analysis

Significance analysis was conducted using student’s unpaired t-test. Statistical analysis was performed at 95% (p < 0.05) and 99% (p < 0.01) confidence intervals. All statistical analyses were carried out with the summary plots generated by GraphPad Prism software 10 for Mac, GraphPad Software, Boston, Massachusetts USA (www.graphpad.com). Two-way ANOVA was used to assess the significance of the growth characteristics after ALE. Outliers were identified and removed based on the Chauvenet’s criterion as necessary. Relative Standard Deviation (RSD) was calculated by dividing the standard deviation of the sample by its mean converted into percentage.

Results and discussion

This study investigated the sensitivity of Halomonas sp. to oxygen availability, with the goal of enhancing its robustness under oxygen-limited conditions commonly observed in large-scale bioprocesses. Due to the complexity of the hierarchical and metabolic response of cells to oxygen availability, a holistic approach was selected to address this challenge by exposing the cell culture to gradually increasing strictly controlled levels of hypoxic stress to allow genomic rewiring through adaptive laboratory evolution. Prior to performing the experiment, two parameters, which were essential to the ALE configuration, namely the NaCl concentration of the culture and the starting dissolved oxygen (DO) level were determined. Different concentrations of NaCl were tested to identify the best growth-supporting salt concentration, which would also minimise the risk of contamination. This was followed by a screening of DO setpoints to investigate the minimum setpoint level that would successfully support the growth of the starting culture in the bioreactors. Following the parameter fine-tuning, ALE was commenced at the selected salt concentration starting from the DO setpoint that was determined in the preceding experiments. The investigation was concluded by the phenotypical and genomic characterisation of the evolved Halomonas population through analytical methods and whole genome sequencing.

Optimisation of the medium salt concentration for Halomonas sp. DSM7328 cultivation

LB medium with a range of NaCl concentrations were tested to identify an optimal concentration that maximises biomass growth with low contamination risk. This analysis revealed that the concentration of NaCl in the medium had a direct impact on the duration of the lag phase and the maximum cell density (measured as OD_600_) achieved in batch cultures of Halomonas sp. DSM7328 (Fig. 1, Supplementary Data S1). The NaCl concentration, final biomass density and the length of the lag phase were all directly proportional to the amount of salt in the growth environment. In particular, the responses were classified into three main groups based on the tested NaCl concentrations; Low salt: 3.5-5.0%, Moderate salt: 6.0–8.0%, and High salt: 8.5–9.5% groups (Fig. 1(A)). The lag phase was prolonged when cells were exposed to high salt concentrations, and the cultures achieved higher cell density. Remarkably, the variability between the replicates was also the lowest for this group. The observed differences in the growth characteristics of these cultures in the presence of high or moderate-to-low salt concentrations were statistically significant (Fig. 1(B) and 1(C)). The extension of the lag phase during growth in high NaCl concentration was reported for different bacteria such as E. coli [38], the halotolerant bacterium Marinobacter hydrocarbonoclasticus [39] and Halomonas sp. AAD12 [40] corroborating with the current findings.

Fig. 1. Determination of optimal salt concentration for the ALE experiment. Effect of NaCl concentration on A growth profiles, B the maximum specific growth rates, and C the maximum OD_600_ of wild-type Halomonas sp. DSM7328 is displayed for salt concentrations ranging from 3.5% to 9.5% compared to 1.0% NaCl in LB medium used as control. All error bars indicate the standard deviation (SD) of six replicates (n = 6). Statistical analysis was performed using student’s t-test (*p < 0.05 and **p < 0.01). These analyses were carried out by a semi-permeable film that induces micro-aerated conditions in the wells. D Gram staining of Halomonas sp. DSM7328 cultures was performed for cultures grown in sterile and non-sterile LB medium with 1.0%, 3.5%, 7.0% and 9.0% NaCl. Scale bar indicates 20 μm

The contamination-proneness of the cultures was assessed by the detection of Gram-positive bacterial contamination using Gram staining (Fig. 1(D)). Low salt concentration was not sufficient to prevent the growth of contaminants present in or that got introduced to the cultivation environment. This observation was not specific to experimental setup or due to unplanned accidental exposure to contaminants. It was exclusively caused by the presence of salt in insufficient concentrations to prevent predominantly non-halotolerant growth since all control cultures that were initially sterilised remained contaminant-free throughout the experiment. These results favoured the selection of NaCl concentration from among the high salt group. The lowest concentration within this group was selected for resource management and for the minimisation of possible equipment corrosivity risks in the prolonged ALE experiment.

The optimal condition was selected as the LB medium supplemented with 8.5% NaCl in total (LB85 medium). For this condition, the specific growth rate of the culture and the maximum cell density measured as optical density were 0.55 ± 0.10 h^− 1^ and 1.00 ± 0.05 OD_600_, respectively. Even though the highest mean specific growth rate (1.32 ± 0.20 h^− 1^) was achieved in low salt (4.0% NaCl) concentration, LB85 was still considered more advantageous for the ALE experimental setup owing to lower risks of contamination, high cell density and reduced variability over a long-term continuous experiment. This process decision to utilise a high salt concentration as a preventative strategy to reduce risk of contamination was also adopted as a strategy previously [41].

Effect of dissolved oxygen on wild-type Halomonas sp. DSM7328 growth

Once the salt concentration of the cultivation medium was determined as 8.5% at micro-scale, the next step was to scale the process up to benchtop bioreactors and to determine the starting level of DO for ALE. Identification of this critical DO level beyond which the cells are exposed to stress induced by oxygen limitation was important from a process efficiency point and from mutation load perspective: Starting the process at the critical DO would lead to the shortest possible process duration to be achieved, reducing both risks of contamination as well as the possible acquisition of additional mutations that are not relevant for adapting to low oxygen levels in the growth environment.

In the absence of any oxygen limitation, we confirmed that LB medium was limiting for the primary carbon source among the four macro-nutrients (Supplementary Figure S1). This indicated that during the ALE experiment to follow, the same limitation would hold at the growth rates set by the dilution rate. Following the identification of the nutrient limitation, we proceeded with the assessment of the limiting threshold for oxygen availability.

In order to identify the critical DO level, Halomonas sp. DSM7328 was cultured in LB85 medium at gradually reduced DO levels starting from the typical setpoint of 30% DO for microbial fermentations followed by reduced DO levels of 10%, 5%, 1% and 0% (Supplementary Data S2). Growth profiles were similar at 30%, 10%, and 5% DO whereas differences in growth were emerging at 1% DO with prolonged lag phase of ca. 2 h and reduced growth rate from 0.43 ± 0.06 h^− 1^ at 5.0% DO to 0.26 ± 0.05 h^− 1^ at 1.0% DO (Fig. 2(A)). Growth was severely impeded (0.09 ± 0.03 h^− 1^) at 0% DO (Fig. 2(B)). The maximum OD_600_ reached when DO level was controlled at 10%, 5% or 1% was not significantly different from that measured when DO was controlled at 30%, but it decreased dramatically to 0.62 ± 0.21 OD_600_ when cells were cultivated when no influx of oxygen was allowed (Fig. 2(C)).

Fig. 2. Growth characterisation of Halomonas sp. DSM7328 grown in LB85 at different dissolved oxygen concentrations. A** Growth profiles, B the maximum specific growth rates (µ_max_), and C the maximum OD_600_ when the bioreactor DO level was controlled at 30%, 10%, 5.0%, 1.0%, or no oxygen influx was allowed are shown in the subsequent plots. Each condition was tested in 4 replicates (n = 4) and all error bars indicate the standard deviation (SD) of these four replicates (n = 4) with no outliers detected. Statistical analysis was performed using student’s t-test (*p < 0.05 and **p < 0.01)

Oxygen transfer in the bioreactor is the rate-limiting step during aerobic fermentation because oxygen has low solubility in the culture broth [42]. Therefore, it needs to be continuously supplied into bioreactors for maintaining a steady and controllable process, rendering DO a key parameter that is determinant of culture performance with respect to cell growth and production yield in a multitude of microorganisms. One such example is the anti-tumour product TL1-1 production by the fungus, Daldinia eschscholzii, where controlling DO levels by cascading the speed of agitation and the rate of aeration improved TL1-1 production by 15.4 fold for 500-litre bioreactors compared to the initial production [43]. Halophilic bacteria are no exception to this; Halobacterium salinarum growth was improved by increased oxygen availability [44].

Assessment of the sensitivity of Halomonas sp. DSM7328 growth to oxygen availability elucidated 5% DO as a coarse-grained critical threshold for oxygen sensitivity; consequently, this was selected as the initial condition for the ALE experiment.

Adaptive laboratory evolution for robust growth under oxygen limitation

The extent of oxygen sensitivity of a strain can be defined by the variability in its growth profile exhibited when exposed to limiting levels of oxygen. Removal of this sensitivity will render dissolved oxygen availability less influential to the cultivation process. The ALE experiment was performed for this purpose. Halomonas sp. DSM7328 culture growth was monitored throughout the ALE experiment as DO levels were decreased stepwise starting from 5.0% DO for the population cultivated in LB85 medium. Growth (or phenotypic) robustness was employed as the principal criterion to assess the performance of the evolved population. We evaluated growth robustness by measuring variability of growth rates of repeated cultivations calculated by the relative standard deviation (RSD) expressed as a percentage. A low percentage of RSD is regarded as an indicator of high robustness of growth.

ALE was carried out for 955 generations in total. The maximum and the critical dilution rates were defined and optimised based on the maximum specific growth rate (µ_max_) of the starting strain Halomonas sp. DSM7328. The µ_max_ at 5.0% DO was ca. 0.43 h^− 1^, which was equivalent to the maximum dilution rate that can be employed safely in a continuous culture without causing washout. The µ_max_ of the tested cultures varied between 0.37 and 0.49 h^− 1^ in batch cultivation at of 5.0% DO, therefore the highest dilution rate at which the cultures were operated at 5.0% DO was selected as 0.35 h^− 1^ to account for any culture-to-culture variability as well as to prevent any possibility of washout in the event that unaccountable operational fluctuations might occur. The dilution rate was increased to 0.40 h^− 1^ from 2.0% DO onwards as the level dropped down to 0.5% DO because the evolved cultures were observed to have higher cell densities than that of the initial culture and this adjustment was introduced to compensate for the effect of higher cell densities attained at steady state. The growth rate, hence, the dilution rate, was maintained within the 0.35–0.40 h^− 1^ range at all times.

The cell population adaptively modified its growth characteristics in response to changes in DO availability ranging from 5.0% to 0.5% (Fig. 3(A, B)). As the DO level of the culture was decreased from 5.0% to 4.0% maintaining the same dilution rate, its optical density initially decreased by 61%. This response, as evidenced by the reduction in optical density, showed that the 5.0% DO availability was indeed near-critical on cell growth of this population. When the DO setpoint was lowered from 5.0% to 4.0%, this coincided with the increase of the dilution rate from 0.3 h^− 1^ to 0.35 h^-1^ to lower the OD_600_. While it may appear as unclear whether the observed decrease in OD_600_ was due to oxygen limitation or to a reduction in biomass concentration caused by the increased dilution rate, the following steps of the ALE further reducing the DO down to 3% or lower did not result in any decrease in OD_600_ leading us to conclude that the reduction was not indeed a response to lowering of oxygen availability but an immediate response to increased dilution rate.

Fig. 3. Adaptive laboratory evolution (ALE) experiment performance of Halomonas sp. DSM7328.** A** Overview of the chemostat operation by stepwise reduction of DO availability over time (black) and the corresponding dilution rate adjustments (blue) that were introduced. The dilution rates were adjusted to maintain consistent optical density in the outlet stream and were always set at or below the maximum specific growth rate of the original strain. B Culture OD_600_ (black dots) in response to the changes in dilution rate presented for the complete course of ALE. The small image in the right lower corner displays and isolated section where the change in dilution rate from 0.35 to 0.40 h^− 1^ (black line) at 3.0%DO was indicated by the increase of OD_600_ (black dots). C Number of generations (grey columns) and the corresponding total residence times the population spent (blue squares) at each DO level. D Mean OD_600_ measurements at different DO levels during ALE. At each condition, OD_600_ readings were recorded at least 3 times (n ≥ 3). All error bars indicate the standard deviation (SD) of the repeated measurements. Statistical analysis was performed using student’s t-test (*p < 0.05 and **p < 0.01)

Cells were maintained at this level of oxygen availability for 96 generations until robust cell densities could be achieved close to that observed at 5.0% DO. Typically, the population took longer to adapt to the new stress level as DO was reduced stepwise until 1.0%, although a further reduction in DO availability down to 0.5% was not as disruptive, possibly suggesting that the key mutations for survival under oxygen limitation were acquired by the population in earlier stages, rendering the population robust against these adverse conditions at 0.5% DO. The culture approximately spent 61, 96, 149, 227, 237, and 185 generations at 5.0, 4.0, 3.0, 2.0, 1.0, and 0.5% DO availability, respectively (Fig. 3(C)). Low number of generations spent at 5.0% DO availability was further indication that the cells were not struggling, and overall displayed robust population dynamics with limited selective pressure for growth. At 3.0% DO availability, the population adapted to the condition effectively, possibly having already acquired some of the key mutations, and the dilution rate needed to be increased in order to ensure steady non-increasing biomass concentration (Fig. 3(B)). It would be important to note that the increase in OD_600_ at any point during ALE should be interpreted as leading to an increase in growth rate as this would merely represent an increase in biomass yield. However, since this increase in yield was indicative of superior adjustment in that setting, the growth rate was increased as a countermeasure to ensure that the OD_600_ remained consistent across the duration of the ALE.

At 0.5% DO, the decision was made to terminate the ALE experiment as the population demonstrated robust growth with no significant difference in optical density measurements at steady state with the growth performance of the population kept at relatively high levels of DO (Fig. 3(D)). It would be realistic to assume that while in a large-scale operation ‘pockets’ or regions of temporary oxygen limitation could be observed, the bioreactor itself would rarely be exposed to extended periods of complete anaerobicity. Therefore, further reduction in DO levels was not considered to be essential for the specific bioprocess related context of this study thus leading to termination of the continuous bioreactor run after ca. 3 months. The evolved population was then genetically characterised for its mutational load, its growth robustness as well as its PHB and ectoine production capabilities.

Physiological characterisation of the evolved population in response to low oxygen availability

In order to assess the performance of the evolved population under low oxygen availability, the growth profiles and the growth rates of this population were investigated at scale. At microscale (Fig. 4(A)-(C)) and in shake flasks (Fig. 4(D)), both the evolved population and the original strain achieved higher cell densities in aerated environments than in microaerobic conditions. The peak cell density was typically achieved faster, at about 20 h of cultivation of the evolved population under aerobic conditions compared to 24 h for the original strain indicating the improved fitness characteristics of the evolved population (Fig. 4A).

Fig. 4. Growth evaluation of the evolved population (EV) and the original strain (ST).** A** Growth profile in 96-microwell plates, B the specific growth rates, and C the maximum OD_600_ of ST (light shades) and EV (dark shades) in micro-aerated (blue) (n = 78 for ST and n = 81 for EV) and aerated (grey) conditions (n = 90 for both ST and EV) were tested in 96-microwell plates, and D in 250 mL shake flasks (n = 3) with baffled flasks with vented caps for mimicking aerobic environments (grey) and with non-baffled flat caps to mimic a reduction in oxygen availability (blue)

While the average specific growth rate of the original strain was apparently higher than the one achieved by the evolved population in aerated environments (0.24 ± 0.14 and 0.16 ± 0.05 h^− 1^, respectively), this difference was statistically insignificant due to the high variation across replicate growth profiles generated for the original strain as denoted by the large relative standard deviation (RSD) of ca. 58% as opposed to the reduced variability for the evolved population (RSD of ca. 31%). While the specific growth rates of both the original strain (0.32 ± 0.19 h^− 1^) and the evolved population (0.33 ± 0.07 h^− 1^) were similar under oxygen limitation, the variation was substantially reduced by evolution, from the RSD of 59% for the starting strain down to an RSD of 21% for the evolved population (Fig. 4(B)). This substantial reduction of the relative standard deviation, indicative of the reduction of fermentation variability, indicated an improved growth robustness of more than 30% following adaptive evolution providing better predictability of operation and improved process performance. ANOVA analysis highlighted the availability of oxygen as the main factor in explaining these differences shown (p-value < 0.0001) followed by the interaction effect between oxygen availability and the impact of induced adaptive evolution (p-value < 0.005), all aligned with the intended design of the experiment.

The highest cell densities indicated by the maximum OD_600_ were measured during the stationary phase of growth. The mean maximum optical cell density of the original strain was higher than that of the evolved population under oxygen limitation (OD_600_ at 0.77 ± 0.25 and 0.55 ± 0.11, respectively) as well as under aerobic conditions (OD_600_ at 2.06 ± 0.24 and 1.49 ± 0.19, respectively). While the variability in the maximum optical cell densities achieved under aerobic conditions was reasonably low and similar for both the original strain and the evolved population (RSD of 11% and 13%, respectively), a substantial reduction was noted for the evolved population under oxygen limitation (relative standard deviation of 32% vs. 20%), indicative of superior robustness in performance for microaerobic batch growth as a result of the ALE (Fig. 4(C)). Availability of oxygen and the impact of induced adaptive evolution as well as the interaction effect between these two factors were identified as highly significant (p-values < 0.0001) by ANOVA.

At a relatively larger scale than microwell plates, the growth performance of the starting strain was similar to that of the evolved population in micro-aerated shake flask environments, reaching comparable cell densities at stationary phase (6.37 ± 0.07 and 6.18 ± 0.03 OD_600_, respectively), also similar to the performance of the evolved population with no imposed oxygen limitation (6.56 ± 0.14 OD_600_), which were all substantially lower than the growth performance of the original strain cultured with no imposed oxygen limitation (8.72 ± 0.16 OD_600_). Nevertheless, our experiments confirmed that oxygen no longer serves as a limitation factor for this population after ALE (Fig. 4(D)).

These experiments were conducted with a limited number of replicates and limited data sampling; therefore, the true variability of the growth profiles was not assessed in this particular context. It would also be important to note that the extent of oxygen limitation, which could be imposed by the microtiter plate setting and the shake-flask setting, were not comparable. The volumetric mass transfer coefficient for oxygen, k_L_a, for the shake flask experimental settings employed in this study (50 mL working volume in 250 mL Erlenmeyer flasks, operating at 30 °C, and 200 rpm) was previously reported to be in the range of 100–150 h^− 1^ [45, 46], whereas the k_L_a for a 96-well plate under the experimental conditions for this work (200 µl culture volume in 390 µl wells, operating at 30 °C, and 200 rpm) was reported to be in the range of 20–25 h^− 1^ [47]. Even without taking the impact of baffles and vented caps or air-permeable filter lids into consideration, the volumetric mass transfer coefficients indicate a favourable environment for oxygen transfer in shake flasks, likely less representative of the micro-aerobic conditions that are of interest for the purposes of this analysis, preventing a direct assessment of true micro-aeration to be made in the shake flask environment.

Our results confirmed that Halomonas sp. growth was indeed sensitive to the availability of oxygen in the fermentation environment and that the mutations acquired through evolution indeed created a phenotypically discernible response. The difference in these responses was exacerbated by limitations in the environmental oxygen availability; the growth behaviour of the evolved population was less prone to variability in response to changes in the oxygen availability than the original strain. This has an important implication from a bioprocess development perspective. DO control in a bioreactor may present a great challenge as the influx of air (or oxygen) into a bioreactor is achieved by sensor-driven automatic opening and/or closing of the gas valve. When the DO level drops below the selected set point, its recovery is typically achieved by a combination of increasing gas input and the stirring speed, the two factors affecting the DO level in a fermentation process [48]. Especially in large scale, this action may introduce substantial variability into the process manifesting as variations in local DO concentrations within the bioreactor. A population that is less susceptible to instantaneous changes in oxygen availability in a vessel will likely yield low variability in production as well. Biomass and product variability were reported to incur profound impact on production costs and, ultimately, the feasibility of a bioprocess [49, 50], hence the advantage of working with a robust population against inherent process variabilities is indisputable.

The original strain and the evolved population of Halomonas sp. DSM7328 were investigated for any changes in the PHB and ectoine production capability. Intracellular PHB, ectoine, and 5-hydroxyectoine concentrations were determined after 10 h of batch cultivation in shake flasks, which corresponded to the exponential phase of cell growth, with no apparent nutritional limitations yet imposed in the cultures. The PHB concentration of the evolved population was 20.77% higher than that of the original strain under micro-aerated conditions (5.56 ± 0.28 mg· OD_600_^−1^ vs. 4.95 ± 0.10 mg· OD_600_^−1^ PHB, Fig. 5(A)). On the other hand, there was no significant difference in production under the aerated condition. Availability of oxygen was reported to affect the production of PHB. Polyhydroxyalkanoate (PHA) productivity was increased in low-oxygen cultivation of Bacillus endophyticus [17]. DO was controlled in the range of 1–5% in Halomonas campisalis MCM B-1027 producing PHA [51] and Pseudomonas putida LS46 was grown in microaerophilic environments to facilitate PHA production [52]. While the increase in PHB production under oxygen limitation was minimal for the original strain, ALE significantly improved the culture production yield under oxygen limitation. The improved PHB production observed post-ALE was a successful production improvement strategy, which could, in the future, be used in combination with metabolic evolution to develop strains of industrial significance. The exponential phase of batch growth, which was assessed in this analysis is a realistic representative of quasi-steady state behaviour and could therefore be considered as illustrative of the culture performance during continuous operation, for which Halomonas genus is of particular interest.

Fig. 5PHB, ectoine, and hydroxyectoine production of the evolved population.** A** Intracellular PHB accumulation in mg per unit OD_600_, B intracellular ectoine concentration, and C intracellular hydroxyectoine production of the original strain (grey) and the evolved population (blue) grown in aerated or micro-aerated environments. All error bars indicate the standard deviation (SD) of three replicates (n = 3). Statistical analysis was performed using student’s t-test (*p < 0.05 and **p < 0.01)

Ectoine production was reduced in the evolved population compared to the original strain both under aerated and micro-aerated conditions (73.7 ± 20.7 mg·L^− 1^ vs. 48.9 ± 2.2 mg·L^− 1^ and 55.3 ± 6.0 mg·mL^− 1^ vs. 11.6 ± 0.8 mg·mL^− 1^), respectively. Relatively speaking, ectoine accumulation in the cell was higher for both the original strain and the evolved population in aerated environments, implicating the presence of a potential metabolic or environmental stressor when exposed to oxygen-rich environments, although this difference was not significant for the original strain. Since ectoine is a stress-relieving cytoprotectant [12], reduced requirement for its intracellular accumulation, particularly in the case of the evolved population growing under oxygen limitation, would contribute reasonable evidence towards the removal of this perceived-stress otherwise (Fig. 5(B)). Despite the differences observed in ectoine production, the intracellular concentration of 5-hydroxyectoine was not significantly affected (Fig. 5(C)). This was thought to be due to the fact that EctD, a non-haem iron(II)- and 2-oxoglutarate-dependent dioxygenase, uses oxygen to hydroxylate ectoine into 5-hydroxyectoine [12]. The adaptation to low oxygen concentration was not thought to modify this enzymatic step as it did not have a direct contribution towards improved fitness of the Halomonas population.

These results indicated that the evolved population accumulated high levels of PHB intracellularly as a metabolic response to adaptation to low oxygen availability. On the other hand, ectoine accumulation was reduced for the evolved population. Ectoine allows bacteria to deal with stress conditions such as osmotic pressure [53]. The reduction in its production by the evolved population indicated that the population no longer perceived their environment as a stress inducing environment, thus turning the stress response pathways down for metabolic efficiency and instead, directing the resources towards PHB production. This trade-off, decrease in ectoine production as a result of reduced stress, presents a challenge for biotechnological applications where ectoine is the principal product of interest. Recovery of high production capability, as and when needed, may require the implementation of further targeted or untargeted genetic engineering strategies. While the reduction in the production of the stress response metabolite, ectoine, could be explained by the cells’ perception of a reduced-stress environment, it should be noted that this hypothesis does not completely rule out the possibility of novel metabolic rewiring nor new resource allocations taking place independent of stress.

Analysis of the mutational load of the evolved population by whole genome sequencing

In order to identify the mutations acquired during the ALE experiment, whole genome sequencing was performed (Supplementary Figure S2). Aligning the whole genome sequencing data with the genome database confirmed that the extended duration of the experiment did not introduce any contamination to the culture (Supplementary material Fig. S1). This supported initial hypothesis that the utilisation of high NaCl concentration would act as a preventative measure to reduce the risk of contamination for processes that reside over extended durations, which would typically be relevant to continuous fermentation. Mutations were identified in 1,706 sites in the genome (Fig. 6(A)). Among these mutations, 591 (ca. 35%) were identified to be allocated to the protein encoding regions of the genome (Fig. 6(B)). Of these sites, more than 30% (178/591) were located in regions that were annotated as hypothetical proteins with yet unelucidated functionality. Considering that ca. 65% of the total number of mutations were acquired in the non-coding genome, the results collectively indicated that the evolved population acquired a high number of modifications in genomic regions of either non-specific sites or regions of unknown functionality for enhancing its robustness in response to low oxygen availability. The most frequently encountered type of mutation in both the coding and the non-coding genome was identified as single nucleotide variations (SNV), approximately accounting for 47% and 69% of the total number of mutations identified, respectively, followed by insertions (Fig. 6(C) and 6(D)).

Fig. 6. Whole genome sequencing summary for identifying the mutational load of Halomonas post-ALE. The breakdown of the different types of mutations acquired in the whole genome are provided in (A), and the relative distribution of these mutations in the coding and the non-coding genome is given in (B). The breakdown of different types of mutations including insertion, deletion, single nucleotide variant, multi-nucleotide variant, and replacement acquired by the population were provided for the coding (C) and the non-coding (D) genome separately. E displays the relative distribution of the total number of mutations as acquired by a relative fraction of the population represented in 10%-bins

The mutations identified in the genomic regions of known process and functionality were located in the ATP binding cassette (ABC) transporters that import sugar, multidrug, sulphate, D-allose, phosphate, zinc, and iron [54] and other miscellaneous transporters that facilitate the import and export of substrates and metabolites across the cellular membrane. The adaptive evolution of the population to limitations in oxygen availability was achieved by triggering a collective genetic response heavily impacting metabolic activity through metabolic transport across the cellular boundary. Therefore, cellular transport emerged as the prominent functionality that played a role in relieving the stress associated with oxygen limitation, thus allowing superior adaptive capabilities, as discussed above. We should also note that this indirect relief from the stress imposed by hypoxia does not necessarily mean that mutations in genes related to metabolic transport contributed to overcoming oxygen transfer limitations.

A noteworthy observation is that while some of these mutations were identified in protein-encoding regions associated with functionalities impacted by the availability of oxygen, many others were not, indicating that the improved robustness of the evolved population resulted from the collective and concerted action of a series of genetic modifications on predominantly adjunct functionalities resulting in the observed phenotype. This phenotype may be specific to oxygen availability or indicative of a high-level modification that enabled improved resilience in accordance with generalised stress response. However, this was not possible to assess definitively since a considerable fraction of the mutations was acquired in genomic regions of unverified functionality.

Whole genome sequencing elucidated a non-uniform pattern in the acquisition of mutations, leading to the emergence of a heterogeneous population at the end of the process where some mutations were acquired by all or most of the cells (> 90%) whereas other were identified in a lower fraction of the population (Fig. 6(E), Table 1, Supplementary material Table S1). In fact, most mutations (61%) were acquired by less than 10% of the population. Based on this, we propose a possible explanation for how this genetic heterogeneity aligns with the observed phenotypical outcome. We hypothesise that the genetically distinct subpopulations, each carrying different levels of mutational load, might be collectively required to establish a stable, high-performing population. Although genetically heterogeneous, this evolved population consistently behaved phenotypically as a single, robust strain, a unit entity, in other words, as repeatedly confirmed in subsequent analyses.

From an evolutionary perspective, epistasis, the interaction where one gene modifies the effect of another, profoundly shapes evolution by altering fitness landscapes, determining accessible evolutionary pathways, and influencing adaptation, often by making beneficial mutations conditional on the presence of other specific mutations [55–57]. In this particular instance, we observe a similar phenomenon whereby the evolved population composed of genetically heterogeneous subpopulations exhibited low phenotypic variability when cultured, emphasising the important role of a mechanism that acted potentially as a simile to complex epistatic interactions leading to unforeseeable phenotypic outcomes.

Notably, the aggregate phenotypic performance of this composite population is proposed to exceed that of any individual genotype, including those carrying one or several of the identified mutations aligning with previous work by Fortuna et al., who showed that the mapping between genotypic and phenotypic variation determines which variants become available to natural selection, thereby constraining possible evolutionary trajectories. The well-established robustness of biological systems to genotypic change, and the existence of genotype networks yielding identical phenotypes, support the view that pervasive epistasis strongly shapes evolutionary outcomes. Under such conditions, the phenotypic effect of any individual mutation depends heavily on the genetic background in which it occurs [58–63].

The evolved population characterised in the present work exhibits an analogous form of epistatic interaction, but operating across subpopulations generated through adaptive laboratory evolution (ALE) rather than within a single genome. This cross-population epistasis likely underpins the superior collective phenotype observed.

Table 1. Top 10 most frequently acquired mutations in the evolved populationRegionLocusTagTypeFrequency**Amino acid changeGeneProtein name82,2683765SNV100.0Gly335AspmgtE_2magnesium transporter90824544SNV92.9Ala226ValHlyD family efflux transporter periplasmic adaptor subunit14,7782924SNV91.2Leu1550ProtdhL-threonine 3-dehydrogenase240,2883425SNV88.6Tyr58Cyshypothetical protein113,5594285SNV88.1Gln520mexImultidrug efflux RND transporter permease MexI144,4241944SNV88.1Val32Alaresponse regulator transcription factor33,7641836SNV87.9Phe354Leuhypothetical protein81,0603968SNV87.8Val276Alamethyl-accepting chemotaxis protein77,6283761SNV87.7Tyr309CysatpAF0F1 ATP synthase subunit alpha477,130446SNV87.6Val904AlapvdLpyoverdine non-ribosomal peptide synthetase /polyketide synthase PvdL1,004,686928SNV87.6Val320AlaY-family DNA polymerase361,6012145SNV87.5Asp142Glyhypothetical protein*SNV: single nucleotide variation, ** refers to the fraction of population in which this mutation was detected

A mutation (Gly335Asp) was identified in the magnesium transporter (MgtE) gene across all cells in the population (Table 1). Magnesium transport is a key process [64] since magnesium ions were reported to play a role in stabilising membranes and ribosomes in bacteria [65], and inhibiting or stimulating anaerobic microbial activity [66]. The mutation acquired in the gene encoding MgtE could potentially be associated with this functionality in response to maintaining microbial activity under limited oxygen availability although this association would necessitate further confirmation. A mutation (Ala226Val) in the HlyD family efflux transporter periplasmic adaptor subunit, which is a membrane fusion protein, was identified in 93% of the population. This protein is a component of the type I secretion system that is found in Gram-negative bacteria [67] and has function in exporting different compounds [68], although the exact reason or connection between low-oxygen adaptation and a mutation in this gene is yet unclear.

The mutation (Leu1550Pro) of the gene encoding L-threonine 3-dehydrogenase, which catalyses L-threonine to 2-amino-3-ketobutyrate [69], was detected in 91% of the evolved population. The metabolic reaction catalysed by the enzyme encoded by this gene is important in an anaerobic process to generate energy for the cells [70]. Consequently, threonine was previously used as the sole substrate for the fermentation of the anaerobic bacterium, Clostidium sticklandii [71]. Threonine metabolism was reported to influence the production of the two key native metabolites of Halomonas spp.; ectoine and PHB. An upregulation of threonine biosynthetic pathway was consequently reported to decrease ectoine production [72]. Although, high intracellular L-threonine concentration was detected when phaCAB was overexpressed for PHB production in E. coli [73], no direct link could be established between the mutational load of the PHB biosynthetic genes and the threonine metabolism. However, the emergence of this mutation in conjunction to the differences observed in the PHB and ectoine production capabilities of the evolved population provided further supporting evidence for this notion, though the direct role of this mutation for this population yet remains unclear.

The mutations observed in the DNA repair machinery mutations (mutL) were retained in nearly 70% of the population, possibly indicating its early adoption during ALE. Hypoxia can downregulate various DNA repair pathways, impacting the cell’s capacity for high-fidelity repair and potentially increasing mutation rates [74, 75]. Partial impairment of DNA repair pathways could possibly be an early adaptation to induce the acquirement of diverse mutational load for an effective evolutionary strategy.

Albeit being low abundance a particularly interesting point mutation was detected in EctA, the diaminobutyrate acetyltransferase, which is the enzyme that catalyses the acetylation of L-2,4-diaminobutyrate using acetyl-CoA as a substrate. The threonine at position 9 was mutated to alanine in approximately 20% of the population (Supplementary material Table S1). This reaction is a crucial step in the biosynthesis of the osmoprotectant, ectoine. Considering that a marked decrease was observed in the intracellular accumulation of ectoine in the evolved population without causing any loss in fitness, this could potentially indicate a functional adaptation towards low oxygen availability manifesting itself complementary to this genetic modification.

It should be noted that these mutations and the cognate functional implications of the modifications they might have imposed would necessitate reverse engineering or functional genomic data to establish direct causal relationships, leaving the possibility of alternative reasons as to the role of these mutations to achieve the observed phenotype open to further interpretation.

As a final note, there are certain practical considerations that raise further open questions to establish the potential of the industrial applicability of this strategy and for the potential of Halomonas as novel industrial biotechnology hosts. The resistance of the host to contamination by non-halophilic microorganisms, as also confirmed in this study, renders Halomonas a particularly interesting candidate host for open vat or continuous fermentations, both of which are notoriously known to be prone to contamination. That said, the presence of high salt concentrations would have to be carefully assessed for economies of scale and for the purposes of the possible impact of salt corrosion on the lifecycle of a process plant. These challenges are further exacerbated by the impact of high salt concentration on oxygen transfer and conventional downstream separation processes necessitating novel alternatives to be explored.

Conclusion

Oxygen is essential for aerobic growth and metabolism but its low solubility in aqueous systems often limits biomass formation and product synthesis in oxygen-dependent species. In this study, continuous ALE was employed for the first time to generate Halomonas sp. strain DSM7328 capable of robust growth under severe oxygen limitation. ALE induced whole-cell adaptations through the accumulation of multiple genetic changes, collectively reducing cellular stress responses, population heterogeneity, and batch-to-batch variability, with only a minor reduction in biomass yield. Notably, the identified mutations were not predominantly associated with oxygen-specific pathways and were largely located in non-coding or poorly characterised genomic regions. This highlights ALE’s capacity to expand metabolic capabilities beyond what can be readily achieved through targeted engineering approaches. Interestingly, the evolved population was genetically heterogeneous yet phenotypically stable proposing a higher-level epistatic effect between genetically heterogeneous subpopulations. By generating complex and non-intuitive genotype–phenotype relationships, ALE represented a powerful strategy for addressing multifactorial challenges in strain design, particularly for emerging bioprocessing hosts with limited prior cellular knowledge such as Halomonas. The prevalence of mutations in genes of unknown function further underscores the need for comprehensive functional characterisation to enable effective exploitation of Halomonas genus for bioprocessing applications.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1