On-line analysis and in situ pH monitoring of mixed acid fermentation by Escherichia coli using combined FTIR and Raman techniques

We introduce an experimental setup allowing continuous monitoring of bacterial fermentation processes by simultaneous optical density (OD) measurements, long-path FTIR headspace monitoring of CO2, acetaldehyde and ethanol, and liquid Raman spectroscopy of acetate, formate, and phosphate anions, without sampling. We discuss which spectral features are best suited for detection, and how to obtain partial pressures and concentrations by integrations and least squares fitting of spectral features. Noise equivalent detection limits are about 2.6 mM for acetate and 3.6 mM for formate at 5 min integration time, improving to 0.75 mM for acetate and 1.0 mM for formate at 1 h integration. The analytical range extends to at least 1 M with a standard deviation of percentage error of about 8%. The measurement of the anions of the phosphate buffer allows the spectroscopic, in situ determination of the pH of the bacterial suspension via a modified Henderson-Hasselbalch equation in the 6–8 pH range with an accuracy better than 0.1. The 4 m White cell FTIR measurements provide noise equivalent detection limits of 0.21 μbar for acetaldehyde and 0.26 μbar for ethanol in the gas phase, corresponding to 3.2 μM acetaldehyde and 22 μM ethanol in solution, using Henry’s law. The analytical dynamic range exceeds 1 mbar ethanol corresponding to 85 mM in solution. As an application example, the mixed acid fermentation of Escherichia coli is studied. The production of CO2, ethanol, acetaldehyde, acids such as formate and acetate, and the changes in pH are discussed in the context of the mixed acid fermentation pathways. Formate decomposition into CO2 and H2 is found to be governed by a zeroth-order kinetic rate law, showing that adding exogenous formate to a bioreactor with E. coli is expected to have no beneficial effect on the rate of formate decomposition and biohydrogen production. Electronic supplementary material The online version of this article (10.1007/s00216-020-02865-5) contains supplementary material, which is available to authorized users.

described in refs [S3,S4]. Briefly, a frequency doubled Nd:YAG laser, 532.2 nm, 20 mW (Lasos, GL3dT) emits green excitation light that is turned by 90° by a small mirror and coupled into a microscope objective. The small mirror was a 2 mm × 3 mm oval film deposited in the centre of a glass slide so as not to take away too much of the Raman backscattered light. The microscope objective is a 20x, 0.50 NA achromatic objective (OptoSigma, 028-0220) with a large clear aperture (8.2 mm). The objective focused the laser light very tightly at 2 mm distance from the objective front into the glass tube, as well as collimating the resulting Raman backscattered light.
The sample volume is essentially the focus volume with an estimated spatial resolution below 100 μm. The backscattered light passed through the glass slide and was coupled into a lens and transmitted to the monochromator (Shamrock SR-750-A) equipped with 1200 l/mm grating, 750 nm blaze, and CCD camera (Andor i-Dus DU420A-OE at -80 °C). The grating provided a 880 cm -1 spectral range at about 0.8 cm -1 resolution. After wavenumber calibration, Raman peak position accuracy is estimated to be ± 3 cm -1 . Raman reference spectra were obtained in borosilicate NMR test tubes. A scheme of the Raman setup is part of Fig. 2 (main text). In addition, see below for two photos of the Raman spectrometer. S-5

S.2.2. Gas Chromatography
For reference, concentrations of ethanol and acetic acid have also been measured by gas chromatography. 0.2 mL of sample was dissolved in 0.8 mL acetone, and 0.3 µL of this solution injected into a standard GC instrument (temperature programmed Agilent DB-WAX UI, with 1.4 mL/min H 2 carrier gas and FID detector). Retention times for ethanol and acetic acid were 4.5 and 11 minutes, respectively. From the GC peak integral, the concentration of the sample was determined after a calibration. The calibration plots are shown below, demonstrating good linearity and dynamic range. Error bars, as represented by the standard deviation of repeat measurements, are approximately the size of the symbols used or smaller and are therefore not included in the calibration plots.

S.2.3. Gas Phase FTIR Spectroscopy with White cell
To determine gas phase concentrations (partial pressures), we applied the Beer-Lambert law with absorption cross sections from literature databases (HITRAN [S5] or PNNL [S6]), and compared integrated absorbances (band integrals) in the ranges specified in the main text. The spectral features (overtones, combination bands) were chosen to have absorbances well below ln (I 0 /I) < 2, to avoid saturation effects. We validated this approach using N 2 O as a test gas with the White cell set at 8 m absorption path length. The first calibration plot shows calculated band integrals using the known partial pressures and 8 m path length vs experimental integrated absorbance of the 2ν 1 overtone of N 2 O (centred at 2563.5 cm -1 ). For each filling pressure, a simulated absorption spectrum was generated assuming an 8 m path length. The figure below shows the measured 2ν 1 band integral for each filling pressure against the integrals for the same band in simulated spectra generated using cross-sections from HITRAN and assuming an 8 m absorption path length. The gradient of the plot is 0.995 ± 0.009, indicating that the predicted and measured spectra show very good agreement and that the White cell path length must be close to 8 m, as predicted by counting the total number of reflections of the HeNe alignment laser.
Excellent linearity and 1:1 correspondence is seen. Error bars are approximately the size of the symbols used.

Fig. S4
Measured 2ν 1 overtone band integral for different filling pressure of N 2 O against the integrals for the same band in simulated spectra generated using cross-sections from HITRAN and assuming an 8 m absorption path length In a further calibration plot to corroborate our procedure, the figure below shows a nine point calibration based on the integral of the N 2 O ν 1 +2ν 2 combination band. As can be seen, within this pressure range, the band shows excellent linearity with increasing N 2 O filling pressure.

Fig. S5
Calibration based on the experimental integral of the N 2 O ν 1 +2ν 2 combination band versus N 2 O filling pressure