Materials
The strain was purchased as Acetobacter aceti CIP 58.66 from the Biological Resource Center of Pasteur Institute (Paris, France). However, 16S rRNA gene analysis performed for this study revealed that the strain does not belong to the aceti species but to the cerevisiae species. The strain is thus referred to as Acetobacter sp. CIP 58.66 in this paper. Its 16S rRNA nucleotide sequence is available on the GenBank database with the accession number MZ242087.
Yeast extract and Bacto™ Peptone were obtained respectively from Organotechnie (La Courneurve, France) and BD-France (Le Pont de Claix, France). Monohydrate citric acid was purchased from Acros Organics (Geel, Belgium). Glycerol and trichloroacetic acid were acquired from VWR Chemicals (Leuven, Belgium). 1,3-PDO, K2HPO4, ammonium hydroxide and sulfuric acid were obtained from Sigma-Aldrich (Saint Louis, USA). 3-HP (30% w/v) was purchased from TCI Europe (Zwijndrecht, Belgium). Lastly, 3-HPA was chemically synthesized at URD Agro-Biotechnologies Industrielles (Pomacle, France) according to (Burgé et al. 2015).
Preparation of the inocula
The strain was stored at – 80 °C in 20% (w/v) glycerol. For inocula preparation, 1 mL of stock culture was added to 25 mL of sterile, basal medium (5 g L−1 yeast extract, 8.71 g L−1 K2HPO4, 3 g L−1 peptone; pH was adjusted to 6.5 with a 5.5 mol L−1 H2SO4 solution) in 250 mL baffled shake flasks, and incubated for 63 h in an incubator shaker (Infors Multitron, Bottmingen, Switzerlan) at 30 °C and 200 rotations per minute (rpm). This inoculum was used directly for batch experiments in shake flasks. For bioreactor experiments, a second culture was carried out for volume expansion: 1 mL of the first culture was inoculated to 50 mL of sterile basal medium supplemented with glycerol (10 g L−1), in 500 mL baffled shake flasks. This culture was incubated 24 h at 30 °C and 200 rpm. Sterility was assured by autoclaving all medium-containing flasks at 120 °C for 20 min.
Batch growth experiments in shake flasks
Batch cultures on 1,3-PDO were carried out in 500 mL baffled shake flasks containing 50 mL of sterile basal medium. Prior to inoculation, the medium was supplied with filter-sterilized 1,3-PDO, in order to reach concentrations of 5, 10 or 20 g L−1. Cultures were inoculated with an initial cell dry weight (DW) of 0.04 gDW L−1. Cultures were then incubated at 30 °C and 200 rpm. Each condition was tested in duplicate. A control condition was also tested without any addition of 1,3-PDO to the basal medium.
Batch cultures on glycerol were performed similarly. Sterile basal medium was supplemented with 10 g L−1 of glycerol and a citrate–phosphate buffer was used (9 g L−1 K2HPO4 and 4.4 g L−1 citric acid). Initial pH was adjusted to 6.5, 5.0, 4.5 or 4.0, using H2SO4 (5.5 mol L−1). Each condition was tested in three replicates.
Semi-continuous bioconversion experiments in a bioreactor
Semi-continuous bioconversions were carried out at 30 °C in a 3.6 L Labfors 4 bioreactor (Infors, Bottmingen, Switzerland), with an initial working volume of 1 L. The bioreactor was autoclaved with the medium at 120 °C during 20 min. Online measurements (temperature, pH, partial pressure of O2, stirring rate, air flow rate) were saved through Iris v.5 software (Infors). Partial pressure of O2 (pO2) and pH were monitored with 405-DPAS-SC and InPro 6800 probes respectively (Mettler Toledo). The pO2 probe was calibrated in the medium at 30 °C, just before inoculation. 100% and 0% values were calibrated by successively saturating the medium with dry air at 4 normal litres per minute (NL min−1) and with N2 (> 99%) at 1.5 bar, with a constant agitation speed of 400 rpm, during 20 min each. During bacterial cultures, agitation speed (100–800 rpm, Rushton turbine) and airflow rate (1–4 NL min−1) were automatically controlled in order to maintain pO2 above 40%.
First, a growth phase on glycerol was carried out in batch mode. The initial medium (1 L) was composed of 10 g L−1 glycerol, 5 g L−1 yeast extract, 3 g L−1 peptone. pH was initially adjusted to 5.0 using H2SO4 (5.5 mol L−1) and then left uncontrolled during growth on glycerol. The bioreactor was inoculated so that the initial DW was around 0.02 gDW L−1. Due to the inoculation, the initial pH rose to 5.2. Once the late exponential growth phase was reached, pH was adjusted to 4.0, 4.5 or 5.0 using H2SO4 (5.5 mol L−1). Filter-sterilized 1,3-PDO (5 mL) was then added to the medium in order to trigger bioconversion, and an equimolar solution of NH4OH and 1,3-PDO was plugged to the bioreactor, for both pH control and 1,3-PDO feeding purposes. Each condition was tested in duplicates. A schematic representation of the process is shown on Fig. 2.
Analytical methods
Bacterial biomass measurements
Cell growth was monitored by off-line optical density (OD) measurements at 600 nm using an Evolution 201 spectrophotometer (ThermoScientific, Madison, USA). OD was correlated to cell dry weight (DW) using five dedicated shake flask cultures on glycerol. At 8, 24, 30 and 48 h of culture, 10 mL broth samples were taken and filtered on pre-dried PES filters (pore size: 0.2 µm). These filters were then dried in an oven at 90 °C for at least 24 h. Filters were weighted just before filtration and after drying, on a precision scale (Sartorius ED224S, Goettingen, Germany). DW was derived from the difference between the two weightings. Measurements were also performed with sterile, basal medium. The following relationship was established:
$${\mathrm{DW}}= 0.59 \cdot {\mathrm{ OD }}({\mathrm{R}}^2 = 0.96)$$
(1)
Chemical analysis
Glycerol, 1,3-PDO, 3-HP and 3-HPA were quantified by HPLC with a refractive index detector (Waters 2414, Guyancourt, France). 750 µL of each sample was mixed with 750 µL of a trichloroacetic acid solution (6% v/w) and put at 0 °C for at least 45 min, for protein precipitation. Samples were then centrifuged at 10,000g during 4 min at 4 °C. Supernatants were then filtered through nylon filters (pore size: 0.2 µm). Analysis was performed on an Aminex HPX-87H column (300 mm × 7.8 mm; Bio-Rad, Richmond, USA). 3-HPA was analyzed at a 0.6 mL min−1 flow rate at 35 °C, with 5 mmol L−1 H2SO4 as mobile phase. 1,3-PDO and 3-HP from samples not containing glycerol were analyzed in the same conditions, while samples containing glycerol were analyzed at a 0.4 mL min−1 flow rate at 65 °C, with H2SO4 at 0.5 mmol L−1 as mobile phase, allowing better resolution between 3-HP and glycerol peaks. Chromatograms were analyzed using Empower 3 software (Waters). Regarding 3-HP and 3-HPA, no high purity, analytical-grade product can be found commercially. Their quantifications are therefore likely to be subjected to higher measurements errors, which could explain why the overall carbon recovery of some experiments was found slightly above 100%.
Calculations
Working volume correction
In bioreactor experiments, despite the use of a condenser on the gas outlet pipe, a volume loss still occurred, due to dry air sparging and to long culture times. It was thus necessary to take this phenomenon into account in the analysis. For each experiment, the final working volume was measured and the overall volume loss was calculated proportionally to the cumulative air flow.
Kinetics analysis
Kinetics were analysed using the Gompertz model, as modified by Zwietering et al. (1990) (Eq. 2):
$$G\left(t\right)=a\cdot {\text{exp}}\left\{-{\text{exp}}\left[\frac{b\cdot e}{a} \left(c-t\right)+1\right]\right\},$$
(2)
where e represents exp(1), t is the time (h) and a, b, c are the parameters to be fitted. The model was first fitted to the natural logarithm of the relative population size ln(DWt/DW0). In this case, parameter b is the maximal specific growth rate µmax (h−1). Results are given as parameter estimation ± standard error of estimation. Model fitting was validated by computing the Root Mean Square Error (RMSE). RMSE were normalised by the mean value of the experimental data, in order to compare them for different models.
The same model was also fitted to 3-HP concentrations: the time derivative of this is the volumetric productivity r3-HP (g3-HP L−1 h−1). Specific 3-HP productivities, q3-HP (g3-HP gDW−1 h−1), were estimated by first fitting dedicated Gompertz models to 3-HP and DW amounts in grams, in order to take into account the variation of volume. Then, using both fitted models, q3-HP was computed using Eq. 3, with a 0.01 h time step:
$${q}_{\text{3-HP}}\left(t\right)= \frac{1}{DW(t)}\cdot \frac{d\text{3-HP}}{dt}.$$
(3)
Parameter optimisation was performed with the Levenberg–Marquardt least-squares method, using Python 3.6 and the “curve_fit” function from package scipy.optimize.
Because of a biphasic exponential growth on glycerol, specific growth rates were estimated for both phases separately by performing linear regression of ln(DWt/DW0) values against time. These regressions were done in Python 3.6, with function “OLS” from package statsmodel.api.
Moreover, number of generations Ng of a growth phase were computed as log2(DWf/DWi), where DWi and DWf are respectively the initial and the final DW values of the considered growth phase.
Statistical analysis
Results are all given as mean ± sample standard deviation, and the sample size is reminded as “n = ”. For n = 2, results are given as mean ± half the amplitude between the duplicates. Means were compared using the Fisher-Pitman permutation test, for which no distribution hypothesis is required. These tests were carried out with the “oneway_test” function (with “distribution” option set to “asymptomatic”) from “coin” package on R 3.3.1 software. Differences were considered significant when p-value was lower than 0.05.