2.1.

Strain and Culture Medium

C. necator H16 strain was used in this work.

Agar plate medium contained the following: $10\mathrm{g}/\mathrm{L}$ yeast extract, $10\mathrm{g}/\mathrm{L}$ trypton, $5\mathrm{g}/\mathrm{L}$ beef extract, $5\mathrm{g}/\mathrm{L}$ NaCl, and $20\mathrm{g}/\mathrm{L}$ agar, pH 7.0. Seed medium consisted of the same composition as the plate medium except agar.

Liquid fermentation medium was composed of the following: 20 g/L fructose, $9\mathrm{g}/\mathrm{L}{\mathrm{Na}}_2{\mathrm{HPO}}_4·7{\mathrm{H}}_2\mathrm{O}$, $1.5\mathrm{g}/\mathrm{L}{\mathrm{KH}}_2{\mathrm{PO}}_4$, $0.2\mathrm{g}/\mathrm{L}{\mathrm{MgSO}}_4·7{\mathrm{H}}_2\mathrm{O}$, $0.002\mathrm{g}/\mathrm{L}{\mathrm{CaCl}}_2·2{\mathrm{H}}_2\mathrm{O}$, $0.5\mathrm{g}/\mathrm{L}{\mathrm{NaHCO}}_3$, $0.005\mathrm{g}/\mathrm{L}$ ammonium ferric citrate, and 1 mL of trace element solution. To investigate the effect of C/N ratio on PHB production, ammonium sulfate was added as a nitrogen source at final concentrations of 0.5, 1.0, 2.0, 4.0, and $6.0\mathrm{g}/\mathrm{L}$. The trace element solution contained the following: $0.2\mathrm{g}/\mathrm{L}{\mathrm{CoCl}}_2·6{\mathrm{H}}_2\mathrm{O}$, $0.1\mathrm{g}/\mathrm{L}{\mathrm{ZnSO}}_4·7{\mathrm{H}}_2\mathrm{O}$, $0.03\mathrm{g}/\mathrm{L}{\mathrm{MnCl}}_2·4{\mathrm{H}}_2\mathrm{O}$, $0.03\mathrm{g}/\mathrm{L}{\mathrm{NaMoO}}_4·2{\mathrm{H}}_2\mathrm{O}$, $0.02\mathrm{g}/\mathrm{L}{\mathrm{NiCl}}_2·6{\mathrm{H}}_2\mathrm{O}$, $0.01\mathrm{g}/\mathrm{L}{\mathrm{CuSO}}_4·5{\mathrm{H}}_2\mathrm{O}$, $0.02\mathrm{g}/\mathrm{L}{\mathrm{FeSO}}_4·7{\mathrm{H}}_2\mathrm{O}$, and $0.3\mathrm{g}/\mathrm{L}{\mathrm{H}}_3{\mathrm{BO}}_3$.

2.2.

Bacterial Culture and Sampling

Single colonies of C. necator from the agar plate were inoculated into 150-mL shake flasks containing 50 mL of seed medium, and then cultivated at 28°C with shaking at 200 rpm for 36 h. The preculture was transferred into 250-mL shake flasks containing 80 mL of fermentation culture medium, and then incubated at 28°C and 200 rpm for 72 h. A 2-mL sample was withdrawn from the culture broth in 3 or 6 h intervals. Each sample was divided into two portions: 1.5 mL for measuring cellular density and 0.5 mL for collecting Raman spectra.

2.3.

Analytical Methods

Cell growth was monitored by measuring the absorbance of the culture broth at 600 nm on a Beckman DU800S after suitable dilution with 0.9% NaCl. The fructose concentration in the culture broth was estimated through the dinitrosalicylic acid reagent method.¹⁴ Berthlot reaction was used to determine residual ammonia nitrogen concentration.¹⁵

2.4.

Experimental Setup and Raman Measurements

Details of the LTRS method have been published elsewhere.¹⁶ In brief, a laser beam from a wavelength-stabilized diode laser (780 nm, 500 mW, TEC300, Sacher Lasertechnik Group, Marburg, Germany) was circularized with a pair of anamorphic prisms, spatially filtered, and then introduced into an inverted differential interference contrast microscope (TE2000U, Nikon, Kanagawa, Japan). The equipped high numerical aperture objective ($100\times$, $\mathrm{N.A.}=1.30$) was used to focus the laser beam forming an optical trap. The same laser beam was used as the excitation source. The Raman scattering light from a single cell was collected with the same objective, focused onto the entrance slit of an imaging spectrograph, and then recorded with a wind-cooled ($-70°\mathrm{C}$) charge-coupled detector (PIXIS 400BR, Princeton Instruments, Trenton). The spectral resolution of the Raman system was $\sim 6{\mathrm{cm}}^{-1}$. A polystyrene bead of $2.0\mu \mathrm{m}$ in diameter suspended in water was used to calibrate the LTRS.

The culture broth was diluted 3000 fold with sterile water and loaded into the hole of the sample holder, which was then sealed with a cover slip. A single bacterial cell was captured by the laser beam and levitated $5\mu \mathrm{m}$ above the quartz cover slip to reduce the florescence background from the substrate. The Raman spectrum of the trapped cell was acquired with a 45-mW laser power and 60-s exposure time. The background spectrum without a bacterial cell in the trap was recorded under the same acquisition conditions. The Raman spectra of 50 cells and five background spectra were collected for each sample. All Raman measurements were performed at room temperature (24°C).

2.5.

Raman Data Analysis

Raw spectral data were saved in ASCII format and imported into Visual Basic 6.0 for data analysis. Background subtraction and response function calibration were conducted as follows: $S_{\mathrm{act}}(v)=[S_{\mathrm{acq}}(v)-S_{\mathrm{bg}}(v)]/R(v)$, where $S_{\mathrm{act}}(v)$ is the actual spectrum of a single cell, $S_{\mathrm{bg}}(v)$ is the background spectrum, $S_{\mathrm{acq}}(v)$ is the raw spectrum acquired by LTRS, and $R(v)$ is the response function of the instrument. A Savitzky–Golay smoothing filter with a window width of nine points was used to remove the noise, and a baseline calibration was performed by using the polynomial fitting method.¹⁷

3.1.

Raman Spectra of Cupriavidus necator Cells and Pure Poly(3-Hydroxybutyrate)

In this study, the PHB production inside C. necator cells was monitored through LTRS. All Raman-active molecules, including PHB and other biological macromolecules, contribute to the spectra of bacterial cells, resulting in an overlap of various signals. Therefore, a detailed comparison of the bacterial spectra and the reference spectrum of pure PHB is necessary. Figure 1 shows the Raman spectra of pure PHB (curve D) and bacterial samples cultured at 0 h (curve B) and 20 h (curve A) during fermentation. The differentiated spectrum (curve C) between the spectra of the two bacterial samples above-mentioned (curve A minus curve B) was calculated. As indicated by the reference spectrum of pure PHB, the major peaks of PHB were located at 835, 901, 1058, 1104, 1354, 1456, and $1732{\mathrm{cm}}^{-1}$⁵. Among these Raman bands, the $1732\text{-}{\mathrm{cm}}^{-1}$ characteristic band, which was assigned to a $\gamma (\mathrm{C}=\mathrm{O})$ stretching vibration, indicated a strong signal with no peaks from other intracellular components as observed around this region. Hence, this peak was used to evaluate the relative PHB content in C. necator cells. Apart from the above bands originating from PHB, bands associated with nucleic acids (located at 725, 782, and $1575{\mathrm{cm}}^{-1}$)^18–20 and proteins (located at 1004 and $1657{\mathrm{cm}}^{-1}$)^18,20 were observed in the Raman spectra of C. necator cells. These bands could provide information on the metabolic activity of individual bacterial cells.

Fig. 1

Raman spectra of C. necator cultured for 24 h (curve A) and 0 h (curve B); curve C represents the differentiated spectrum between A and B; and the reference Raman spectrum of PHB is presented by curve D.

3.2.

Kinetics of Poly(3-Hydroxybutyrate) Fermentation in Cupriavidus necator Cells Cultivated at Various Carbon-to-Nitrogen Ratios

Unlike carbon source, nitrogen source is not used as a substrate for PHB production in C. necator cells. However, nitrogen as an essential nutrient for bacterial growth exerts an indirect but significant influence on PHB biosynthesis. Suitable C/N ratio is also important for effective PHB production.²¹ In this study, the time-dependent cellular growth, substrate consumption, and PHB formation in C. necator cultivated at various initial C/N ratios were monitored using LTRS and other conventional analytic approaches (Fig. 2). C/N molar ratios of 29.3 to 2.4 were obtained by varying the ammonium sulfate concentration from 0.5 to $6.0\mathrm{g}/\mathrm{L}$ while fixing fructose at $20\mathrm{g}/\mathrm{L}$. Figure 2(a) illustrates the influence of different C/N ratios on C. necator growth, which was measured as optical density at 600 nm. The final bacterial cellular density increased as the initial ammonium sulfate concentrations increased from 0.5 to $2.0\mathrm{g}/\mathrm{L}$ (corresponding to C/N ratios of 29.3 to 7.3). By contrast, the final cellular density decreased when the initial ammonium sulfate concentrations were further increased to 4.0 and $6.0\mathrm{g}/\mathrm{L}$ (corresponding to C/N ratios of 3.7 and 2.4, respectively). This result suggests that a moderate concentration of nitrogen source favors C. necator growth, whereas too low or too high concentrations can negatively affect microorganism growth.

Fig. 2

Time course of PHB fermentation in C. necator cells at various initial C/N ratios. (a) Growth of C. necator, (b) residual ammonia sulfate, (c) residual fructose, and (d) intensities of Raman band at $1732{\mathrm{cm}}^{-1}$.

The nitrogen source in the culture broth was completely consumed after 9 to 30 h of cultivation at initial C/N ratios of 29.3 to 7.3. However, 46% and 72% of ammonium sulfate were not utilized and remained in the final culture medium at initial C/N ratios of 3.7 and 2.4, respectively [Fig. 2(b)]. The utilization of carbon source was also influenced by the initial C/N ratio [Fig. 2(c)]. The fructose was metabolized from 42 to 54 h within the C/N ratio of 14.7 to 2.4 but was incompletely consumed until the end of fermentation (72 h) at a C/N ratio of 29.3.

As described above, the intensity of the Raman band at $1732{\mathrm{cm}}^{-1}$ can be used to measure PHB content in bacterial cells. Figure 2(d) depicts the time-dependent changes in the band intensity at $1732{\mathrm{cm}}^{-1}$; thus, the relative content of PHB in C. necator cells during cultivation at various C/N ratios. For each sample, the Raman spectra of 50 randomly selected individual cells were acquired and their averages were calculated. The band intensity at $1732{\mathrm{cm}}^{-1}$ was plotted against its corresponding sampling time. Figure 2(d) shows that the $1732\text{-}{\mathrm{cm}}^{-1}$ band intensity and the PHB accumulation trajectories can be divided into two types: (1) nitrogen-limiting cultivation when the initial C/N ratio ranged from 29.3 to 7.3 and the nitrogen source was depleted during fermentation and (2) balanced growth when the initial C/N ratio was at or below 3.7 and the residual nitrogen source existed in the final culture medium. Under nitrogen-limiting conditions, PHB biosynthesis was triggered when the nitrogen source in the broth was consumed and decreased to a certain suboptimal level. Afterward, the PHB was accumulated dramatically and continuously until the exhaustion of the carbon source. However, under balanced growth conditions, carbon was utilized mainly for cellular growth rather than PHB production because the nitrogen supply was sufficient throughout the cultivation period. Small amounts of PHB polymerization were detected during the late exponential phase and stationary phase probably because of the slight limitation of oxygen or other nutrients.

3.3.

Dynamic Changes in Nucleic Acid and Protein Content Inside Cupriavidus necator Cells During Cultivation

Nitrogen source plays a vital role in nucleic acid and protein metabolism. Thus, it affects the distribution of carbon flux on cellular growth and PHB accumulation. Tracing the changes in nucleic acid and protein content within C. necator cells will enrich our understanding of the regulation of PHB biosynthesis. Figure 3 shows the dynamic changes in intensities of the Raman bands specific to proteins and nucleic acids. The Raman peak at $1657{\mathrm{cm}}^{-1}$ was attributed to protein amide I band, and its intensity indicated protein abundance in cells. At relatively high C/N ratios (29.3 to 7.3), the $1657\text{-}{\mathrm{cm}}^{-1}$ band intensity rapidly increased to a high level in a short period after inoculation (6 to 24 h) and then continuously decreased to the lowest value ($\sim 50\%$ of the initial intensity). However, at low C/N ratios (3.7 and 2.4), the $1657\text{-}{\mathrm{cm}}^{-1}$ peak intensity elevated to the highest value during the first 6 h of cultivation and was maintained in such a high level throughout the remaining period [Fig. 3(a)]. The Raman band at $782{\mathrm{cm}}^{-1}$ is associated with nucleic acids consisting of DNA and RNA. RNA is considerably more abundant than DNA in a prokaryotic cell;²² thus, the $782\text{-}{\mathrm{cm}}^{-1}$ peak intensity is mainly related to the RNA content of bacterial cells. Figure 3(b) illustrates that the intensities of the $782\text{-}{\mathrm{cm}}^{-1}$ band exhibited a similar trend to those of the $1657\text{-}{\mathrm{cm}}^{-1}$ band, i.e., the band intensity at relatively high C/N ratios rapidly increased and then declined, whereas the $782\text{-}{\mathrm{cm}}^{-1}$ peak intensity at low C/N ratios maintained a high level throughout the fermentation period except at the early growth phase.

Fig. 3

Time course of intensities of Raman bands at (a) $1657{\mathrm{cm}}^{-1}$ and (b) $782{\mathrm{cm}}^{-1}$ from C. necator.

To elucidate the relationship among the metabolic activities of PHB, proteins, and nucleic acids, we calculated the correlations between the Raman bands associated with these substances in bacterial cells cultivated at initial C/N ratios of 29.3 to 2.4 and sampled at 0 to 72 h (Fig. 4). A linear relationship was observed between the intensities of the 782- and $1657\text{-}{\mathrm{cm}}^{-1}$ bands with a correlation coefficient of 0.85 [Fig. 4(a)]. This finding implies that the protein content positively correlated with the RNA level in C. necator cells. The reason behind this correlation is that ribosomal RNA, which constitutes 80% to 85% of total RNA, forms structural and functional components of ribosomes that serve as the subcellular units responsible for protein synthesis. Therefore, the intensity of the 782- or $1657\text{-}{\mathrm{cm}}^{-1}$ band may indicate the metabolic activity involved in cellular growth. Furthermore, the $1732\text{-}{\mathrm{cm}}^{-1}$ peak intensities plotted against the $1657\text{-}{\mathrm{cm}}^{-1}$ peak intensities showed a linear relationship with a correlation coefficient of $-0.86$ [Fig. 4(b)]. This result suggests that the PHB production and protein content negatively correlated in individual cells. After completing a comprehensive analysis of the data from Figs. 2, 3, and 4, the changes in PHB, nucleic acid, and protein contents inside C. necator cells can be explained as follows. On the one hand, when the bacterial cells were cultured under balanced growth conditions (low C/N ratios), the carbon source was mostly used for cellular growth rather than PHB production. On the other hand, under nitrogen-limiting growth conditions, the syntheses of proteins and/or nucleic acids were blocked, and cellular growth was inhibited. This phenomenon triggered the shift of carbon flux from the tricarboxylic acid cycle to the PHB biosynthetic pathway.

Fig. 4

Correlation between the intensities of Raman bands at (a) 1657 and $782{\mathrm{cm}}^{-1}$ and at (b) 1732 and $1657{\mathrm{cm}}^{-1}$.

3.4.

Single-Cell Analysis of Poly(3-Hydroxybutyrate) Production in Cupriavidus necator

Current knowledge on the growth characteristics of C. necator and conditions of PHB synthesis was mostly obtained by monitoring fermentation parameters with conventional chemical analytic methods. However, these methods only provide average values over the population of a culture consisting of millions of cells. In many cases, individual cells in a population, even if derived from the same clone, show different responses to environmental changes because of population heterogeneity. Hence, single-cell-based analytic techniques are required to reveal this heterogeneity in the responses of individual cells. In this work, we employed LTRS to assess the PHB levels in individual cells by measuring the Raman scattering intensities of PHB-associated bands and evaluated the variation in PHB levels per cell over 50 single cells of C. necator at varying time points of the batch growth. Figure 5 presents the time-dependent changes in the coefficients of variation (CVs, standard deviation/mean) of $1732{\mathrm{cm}}^{-1}$ Raman band intensities of bacterial cells cultivated within the C/N ratio of 29.3 to 7.3. Under all the culture conditions, the highest CV values (above 1.30) were observed at the beginning of the cultivation (0 h), suggesting an extremely inhomogeous nature of the PHB content among the individual cells of the inoculums. During the lag phase, the PHB that accumulated during the preculture degraded in most of the cells within the microbial populations exposed to relatively rich nitrogen supply in the fresh fermentation medium. This phenomenon resulted in homogeneous, low PHB levels inside the cells and, thus, decreased the CV to low values (0.31 to 0.76). The second highest CV values (0.83 to 0.91) were observed at the early stage of PHB synthesis (12 to 24 h), when the residual nitrogen source in the culture broth was consumed to a suboptimal concentration. Bacterial cells at varying stages in their life cycle respond differently to nitrogen limitation stress. Thus, the asynchronous behavior of the individual cells in the populations caused a great heterogeneity in PHB accumulation, i.e., some of the cells initiated PHB accumulation instantaneously, whereas the others needed a relatively long preparation time for PHB biosynthesis. After the middle exponential phase of growth, most cells initiated PHB synthesis and PHB content variation; thus, the CV value drastically decreased to low levels (0.20 to 0.38) and was maintained at such a level for the remaining cultivation period. The above results show that LTRS method can reveal the heterogeneity within a C. necator cell population under dynamic process conditions and give us a more comprehensive view of the physiological state of a growing microbial population. Because the phenotypic heterogeneity has important practical consequences for the productivity and stability of industrial fermentations,²³ integrating such information with microbial population characteristics measured via conventional analysis methods will ultimately allow one to develop, control, and enhance microbial performances in PHB fermentation.

Fig. 5

Time-dependent changes in CVs for the $1732\text{-}{\mathrm{cm}}^{-1}$ Raman band intensity of C. necator cells cultivated at C/N ratios of 29.3, 14.7, and 7.3.

In PHB fermentation, each cell contributes to the final product yield, although dead, inactive, or weakly active cells limit this productivity. Therefore, detailed information on single-cell behavior within a cell population will be beneficial to the evaluation and control of such processes. Figure 6 shows the violin plots representing the PHB content distribution of single cells cultured at an initial C/N ratio of 29.3. Violin plots are similar to box plots, except that they also show the probability density of the data at different values.²⁴ The white dots in Fig. 6 indicate the median, allowing quick comparisons of average PHB levels among multiple bacterial samples, whereas the black bars indicate the first and third quartiles. The surrounding violin shell consists of a mirrored local kernel density estimation that graphically shows the distributional characteristics of PHB content at the single-cell level within each sample. The bacterial cell populations sampled at certain cultivation stages (e.g., at 12, 36, 66, or 72 h) were evidently composed of two or three subpopulations on the basis of PHB abundance in individual cells. Numerous factors, including cell cycle phases, individual cell age, microenvironmental conditions, and mutations,²⁵ can cause this multimodal distributions of metabolite content in individual cells within a population. Further investigations can be conducted in the future by combining LTRS with other sophisticated and high-throughput techniques, such as transcriptomic or proteomic methods, to disintegrate microbial populations and to understand their innate molecular heterogeneity. For instance, LTRS may be used to sort the subpopulations of interest on the basis of PHB content in individual cells, and then the gene expression profiles of the sorted subpopulations may be obtained through transcriptomic or proteomic analyses. Such a strategy may help identify the nature of these subpopulations.

Fig. 6

Violin plots for the distributions of $1732\text{-}{\mathrm{cm}}^{-1}$ Raman band intensities of C. necator cultured for different times at an initial C/N ratio of 29.3. Raman spectra of $\sim 50$ individual cells were recorded for each sample. In the $X$-axis, the categorical variables of $t1$ to $t72$ correspond to bacterial samples taken at time points of 0 to 72 h, respectively.