Microscopy is one of the first equipment used to study single-cell properties and population heterogeneities. Light, atomic force and fluorescence microscopies are all adapted to analyze single-cell properties during fermentation at different levels. Cell suspensions are commonly diluted for microscope observation. Even if this method allows observing each cell individually (e.g. morphology, viability, fluorescence), microscopic analysis is limited concerning the amount of data collected (Fernandes et al. 2011, Zhang et al. 2015).
Light microscopy is mostly used to observe population heterogeneity based on cell morphology (i.e.
cell size and shape) (Fernandes et al. 2011).
Some specific fluorescence related observations can only be made by fluorescence microscopy. Several fluorescent dyes developed for fluorescent microscopy are nowadays used in flow cytometry that offers the advantages of a high-throughput method. With fluorescent microscopy, it is possible to use multiple stains simultaneously which is a good strategy to obtain more data about single-cell physiological state, as long as the emission spectrum of the fluorescent dyes do not overlay. This method also allows observing fluorescent protein producing strains. On one hand, the main advantages of using fluorescent microscopy in single-cell analysis are its specificity, sensitivity, temporal and spatial resolution. On the other hand, the main drawbacks are the fading and photobleaching of the dye and fluorescence quenching (Fernandes et al. 2011).
Cytometry refers to the technologies used to characterize biological single-cells by counting, measuring or comparing them (Brehm-Stecher et al. 2004). Flow cytometry measurement (abbr. FCM) is used to measure the detailed properties of individuals in a whole population by sorting, counting and examining.
The first step of the analysis after sampling is based on hydrodynamics properties. Single cells are indeed focused in a fluid stream (i.e. sheath fluid). Then, this stream is intercepted by a beam of light which is most of the time a laser beam. The laser beam illuminates each cell to measure light scattering properties and to excite fluorescent molecules. Data concerning optical properties are collected in real-time and saved (Brehm-Stecher et al. 2004, Tracy et al. 2010, Fernandes et al. 2011).
This analysis relies on light scattering, excitation and emission properties. If the light deviation angle through the particles is small (0°-5°), it will be defined as forward scatter (abbr. FSC) which is related to cell size or volume. Otherwise, if the light deviation angle is big (90°) it will be defined as side scatter (abbr. SSC) and can be related to intracellular content called also granularity (Fernandes et al. 2011). Cell number, size and content can be measured simultaneously by this method as well as the fluorescent probe response of the cells (Brehm-Stecher et al. 2004). Analyzing at the single-cell level allows analyzing each individual in a cell population which enables to access distributions within a short time of analysis and large number of cells (Figure 2). Therefore, it gives a more precise information about cell populations behavior than mean-populations analysis (Fernandes et al. 2011).
Figure 2: Results of a flow cytometry analysis of a heterogenic population of E. coli strain expressing GFP (SSC vs. FSC) Fernandes et al. 2011)
Cell properties measurable by flow cytometry can be divided into two groups: intrinsic and extrinsic properties. On one hand, intrinsic properties (e.g. membrane composition, size) can simply be measured without cell labelling with the FSC and SSC channels. On the other hand, extrinsic properties (e.g. membrane integrity or potential) require fluorescent strains or probes to be characterized (Fernandes et al. 2011). So, flow cytometry provides information-rich data sets on cell populations at the single-cell level (Brehm-Stecher et al. 2004). As this analysis method allow visualizing single-cells from different subpopulations, flow cytometry is more and more used to study population heterogeneity in bioprocesses, especially with reporter strains and fluorescent dyes (Heins et al. 2018).
Even if flow cytometry has been mainly used to study mammalian cells, it is also possible to apply this method to bacteria. However, smaller cell size constitutes a real challenge (Brehm-Stecher et al. 2004). Accurate FCM analysis requires that the number of events counted in bacterial sample really corresponds to the number of cells of interest within the sample. In the work of Bahl et al. (2004) the results obtained by FCM were compared to those obtained by plate counting from the same bacterial sample (Bahl et al. 2004).
Fluorescence-activated cell sorting
Fluorescence-activated cell sorting (abbr. FACS) is an interesting extension of flow cytometry principles as it enables to physically sort cells according to their light scattering and fluorescence. Cells are ordered in a sheath fluid and pass a laser beam where the fluorescent proteins are excited. Fluorescence emission signal of single-cells allows highly specific downstream cell sorting. The method set to capture cells consists in a vibrating mechanism to split the continuous flow into cell-containing droplets. These droplets are selectively charged and electrostatically diverted into a collecting container (Ma et al. 2017). Therefore, FACS permits to physically separate subpopulations of interest from a heterogenic population (Fernandes et al. 2011). FACS systems are efficient and accurate, but they also are expensive and bulky. Plus, the viability of cells might be disrupted because of the strong electric fields encountered during sorting (Ma et al. 2017).
As FACS is a rather common method among eucaryotic cell (e.g. isolation of cell type-specific apoptotic bodies from mice (Atkin-Smith et al. 2017), selective breeding of live oil-rich microalgua Euglena gracilis (Yamada et al. 2016). It is still less used in bacterial cells studies (e.g. selection of pyruvate variants of Corynebacterium glutamicum enabling improved lysine production from glucose (Kortmann et al. 2019) because of certain challenges (i.e. small cell size, lower protein content per cell) (Tracy et al. 2010).
Fluorescence in situ hybridization: FISH
Fluorescence in situ hybridization is a high throughput method using nucleotide probes tagged with a fluorescent stain to hybridize specific RNA or DNA sequences. So, this method allows the visualization of cell subpopulation. Historically, stained cells are then analyzed by microscopy. Few examples using flow cytometry or even sorted with fluorescence activated cell sorting (FACS) (Haroon et al. 2013) have been developed. Cells are fixated and permeabilized for the need of this technique (Brehm-Stecher et al. 2004).
For mammalian cells, this method has already been associated with flow cytometry and the newly created method is often referred as Flow-FISH (Tracy et al. 2010). Flow-FISH requires cell fixation, permeabilization and hybridization with a set of fluorescently labeled oligonucleotide probes. Samples are then analyzed by flow cytometry (Arrigucci et al. 2017).
Ribosomes are a privileged target because they are largely amplified in growing cells. So, fluorescent tagged oligonucleotides (about 15-20 bp long) which target rRNAs are hybridized to whole cells (Brehm-Stecher et al. 2004, Haroon et al. 2013). Other RNA targets can be used in FISH methods like mRNA or tmRNA, but in these cases an amplification step is often necessary. The abundance of these RNA is indeed lower than with rRNA so, amplifying the sequences of interest is a way to make the fluorescent signal significant and precisely detectable (Brehm-Stecher et al. 2004).
FISH-based analyzing methods can also be used to detect target DNA sequences on low-copy-number plasmids (101-103 copies per cell) or on chromosomes (< 10 copies per cell). In this case, polynucleotide probes are used and they are introduced in higher concentration (x 1000) and present higher hybridization times because of their increased length (50 to 1200 nucleotides). The name of the method Ring-FISH comes from the resulting fluorescent signal that forms a halo around the fluorescence-emitter cells (Brehm-Stecher et al. 2004).
It has also been possible to used FISH-methods to detect antibodies and other characteristic diagnostic binding events like capsular, flagellar or cell-wall antigens. These are whole cell methods (Brehm-Stecher et al. 2004).
Plasmid instability is a major concern of industrial recombinant strains. Plasmid instability has three main causes. First, asymmetric segregation during cell division causes plasmid loss for a certain portion of cells. Then, instability of genome structure can change in certain cells when all of them have the plasmid. Last, compared to plasmid-free cells, plasmid-bearing cells present a growth disadvantage (Alhumaizi et al. 2006).
Causes and consequences of plasmid loss
Metabolic load of recombinant plasmids
The introduction of a recombinant plasmid in a host cell requires resources to be maintained, replicated and to produce recombinant proteins (De Gelder et al. 2007). So, two main biological mechanisms are competing in plasmid-bearing cells: plasmid replication and cell growth (Silva et al. 2012). The metabolic load is defined by the amount of host cell resources (e.g. raw material, energy) derived from host metabolism to synthesize recombinant RNA or recombinant proteins (Glick et al. 1995). Plasmid stability and metabolic burden problematics are difficult to dissociate from each other.
In fact, the introduction of a plasmid into the host might induce a metabolic burden.
Higher plasmid copy number or plasmid size increase metabolic load. Indeed, the amount of energy required to replicate the plasmid in host cells is higher in this case, causing lower growth rates and lower gene expression levels (Bentley et al. 1990, De Gelder et al. 2007, Million-Weaver et al. 2014). For instance, it was demonstrated for recombinant plasmids with RSF1050 and pFH118-backbone in E. coli HB101 (i.e. 1 copy, 0.92 h-1; 34 copies, 0.77 h-1) (Bailey et al. 1993), as well as for plasmid pGFPuv in E. coli JM101-derivatives (i.e. 129 copies, 0.76 h-1; 537 copies, 0.59 h-1) (Cunningham et al. 2009). Nevertheless, plasmid maintenance and replication have a lower impact on metabolic burden than recombinant protein expression itself (Silva et al. 2012). Indeed, protein biosynthesis requires energy. More precisely, several GTP (abbr. Guanosine Triphosphate) are consumed for every amino acid added to the protein chain and one ATP (abbr. Adenosine triphosphate) molecule is consumed per aminoacyl-tRNA (Glick et al. 1995). Plus, recombinant protein overproduction might also cause depletion of some aminoacyl-tRNA or amino acids (i.e. codon bias)(Glick et al. 1995).
Metabolic load may also lead to size and shape heterogeneity among isogenic host cells (Glick et al. 1995), as well as decrease in host viability (Silva et al. 2012). Problems in plasmid carriage in the host cell can occur when recombinant plasmid proteins are interfering with the actions of host natural proteins. Moreover, the presence of a recombinant plasmid in a host cell has been associated with drastic variations in concentration of cellular enzymes involved in carbon, amino acid, nucleotide metabolism, translation and also with ribosome content decrease (De Gelder et al. 2007).
So, the relative importance of metabolic load due to recombinant protein production depends on the amount of protein produced, the plasmid copy number and its size, the host cell metabolic state and the growth medium composition (Bentley et al. 1990, Glick et al. 1995). As a result, plasmid-free cells grow faster than plasmid-bearing cells because they are not submitted to metabolic load anymore (Glick et al. 1995). The resulting growth rate difference intensifies segregational instability (Bentley et al. 1990, De Gelder et al. 2007). As a result, plasmid loss causes decrease in encoded protein expression (Glick et al. 1995). Therefore, in the absence of selection pressure, recombinant plasmids are more likely to be lost during culture. It has been demonstrated that plasmid-induced metabolic load is the cause not only of physiological and metabolic alteration in the host cell, but also of several stress responses (Silva et al. 2012).
Most microorganisms possess a complex biochemical response mechanism to face diverse environmental stress (Glick et al. 1995). The presence of a heterogeneous plasmid might induce stress responses in host cells. Such mechanisms might lead to disrupted cell growth, plasmid expression and to lower plasmid DNA yields, which might favor plasmid instability. First, heat-shock is a stress response mechanism characterized by the expression of so-called heat-shock proteins. They are encoded on the σ32 operon on Escherichia coli and a lot of them are either chaperones ensuring proper protein folding or proteases degrading proteins. Second, stringent response is induced by amino acid starvation and limitation (Silva et al. 2012) or aminoacyl-tRNA pool depletion because of recombinant protein overexpression. In case of aminoacyl-tRNA limitation because of recombinant protein overexpression, the probability to insert incorrect amino acids instead of limiting amino acids increases. Consequently, translational errors increase during foreign protein overexpression because of codon bias (Glick et al. 1995). Stringent response might either inhibit energy consuming synthesis (e.g. proteins, rRNA, tRNA, plasmid DNA synthesis), or activate synthesis of proteases or other proteins to manage the stress situation (Silva et al. 2012). Third, the SOS response consists in a network of reactions insuring DNA repair in Escherichia coli. One of the reactions related to the SOS response is the nucleotide excision repair (abbr. NER) ensuring the detection and removal of unusual DNA structures (Silva et al. 2012).
Segregational plasmid stability
Plasmid segregational instability is the main cause of plasmid loss in a growing culture (Mathur et al. 2009).
Active plasmid partitioning of low copy number plasmids
Partitioning systems insure the efficiency of plasmid copy distribution to daughter cells throughout cell division (Schwartz et al. 2003). They also ensure that no foreign plasmid is transmitted to daughter cells and that only pre-existing copies are properly segregated. Two types of active partitioning systems are found among low copy plasmids: type I and type II. Active partitioning systems like the ones described below have not been observed within high copy plasmid (Million-Weaver et al. 2014). Type I and type II par systems are both based on three main components (Figure 3): a DNA-binding protein serving as adaptor (i.e. SopB for F plasmid and ParR for R1 plasmid), a motor protein (e.g. SopA-ATP for F plasmid, ParM-ATP for R1 plasmid) and a cis-acting centromeric sequence element (e.g. sopC for F plasmid, parC for R1 plasmid). First, the DNA-binding protein serves as adaptor protein and binds to the centromere. Then, adaptor protein binding to the centromere nucleates the motor polarization. The formation of a filament occurs and pulls plasmids apart (par type I) or push plasmids to opposite poles of the cell (par type II). To polymerize the motor protein energy derives from hydrolysis of nucleotide triphosphate. Par systems are autoregulated by their own DNA-binding protein (Million-Weaver et al. 2014).
Partitioning mechanisms for high copy plasmid segregation
Like explained above, no active partitioning system has been found among high copy plasmids yet. Plus, high copy plasmids segregational stability is negatively impacted by their higher metabolic burden (Mathur et al. 2009). Thus, plasmid-free cells stochastically arising from segregational instability will have a fitness advantage compared to plasmid-bearing cells and overwhelm them in whole cell population (Million-Weaver et al. 2014).
There are several hypotheses to explain plasmid segregation among high copy number plasmids.
● Random distribution hypothesis:
During cell division, high copy number plasmids are partitioned among daughter cells by random plasmid diffusion and nucleoid exclusion. Therefore, daughter cells might not receive the same plasmid copies from the mother cell (Pogliano et al. 2002, Million-Weaver et al. 2014).
Theoretically, because high copy number plasmids have a large plasmid copy number, they have low probability of plasmid-free cells to emerge and low plasmid loss frequency. However, several clues suggest that other phenomena might affect segregational stability of high copy number plasmids (Million-Weaver et al. 2014). In fact, there are proofs that not all plasmids diffuse freely through the membrane. For instance, ColE1 plasmids in E. coli have been shown to be excluded from the nucleoid region and are being localized at poles and mid-cell. In this case, the mid-cell region of the mother cell becomes a pole of one of the daughter cells (Pogliano et al. 2001).
Plus, high copy number plasmids are more exposed to plasmid dimerization which is an increasing factor of plasmid instability. Based on these observations, the hypothesis of a normal distribution of the plasmid copy number is less than likely (Million-Weaver et al. 2014).
● Alternative hypothesis:
Studies demonstrated that elements like clustering and plasmid loss dynamics as well as recombinant gene expression are not consistent with a randomly plasmid segregation, but rather with a regulated distribution mechanism like active partitioning. For high copy number plasmids, this mechanism might depend on chromosome-encoded proteins (Million-Weaver et al. 2014). But all of this is still at the hypothetical state and the existence of a regulating mechanism for high copy number plasmid still needs to be proven. But it is also possible that the host has a negative effect on the active partitioning system (De Gelder et al. 2007, Ponciano et al. 2007).
Structural plasmid stability
Structural plasmid instability is defined as the modification in the nucleotide sequence of the plasmid by any kind of mutations. In case of frame shift mutation, proteins with the wrong amino acid sequences might be synthesized (Friehs et al. 2004). Structural stability has a lower impact on plasmid stability than segregational stability and is commonly maintained at a low frequency (Silva et al. 2012).
There are various causes for structural instability like (Friehs et al. 2004, Silva et al. 2012):
● Plasmid size,
● PolyA sequences, inverted and direct repeats,
● Mutagenes or free oxygen radicals causing modifications in the nucleotide sequence,
● DNA polymerase mistakes or insufficient repair mechanism,
● Recombination events occurring between the plasmid and the chromosome,
● High expression of transposons and insertion sequences.
To increase structural plasmid stability, parameters that one can act on are growth conditions and the choice of the host strain (Friehs et al. 2004).
Conjugation (or mating) consists in the transfer of DNA by direct cell to cell contact by a conjugative plasmid (Harrison et al. 2012). It has been proven that plasmid re-uptake by segregants through conjugation with neighboring plasmid-bearing cells occurs and influences plasmid stability (De Gelder et al. 2007). Indeed, the insertion of a heterologous plasmid in a host cells through conjugation mechanisms might lead to metabolic load due to plasmid maintenance and the production of plasmid-encoded proteins, as detailed above (Harrison et al. 2012).
Importance of the plasmid-host interaction
The interaction between the bacterial host and the plasmid vector is of crucial importance, because it can influence the ability of the plasmid vector to colonize new hosts or simply to be maintained and transmitted. In fact, most plasmid vectors depend on the host replication machinery for their own expression. Plasmid stability of a particular plasmid vector can be highly variable within the same genus and even the same species. Patterns of plasmid loss can be different from one strain to another, suggesting that plasmid loss causes might be different (De Gelder et al. 2007).
All plasmids have a different host range. For instance, broad-host range plasmids can be replicated and transferred among phylogenetically distant organisms, whereas narrow-host range plasmids are only compatible with organisms from the same species. First, host range can be affected by replication and maintenance mechanisms. In a same host, the chromosome and the plasmid(s) share the replication system of the host. So, plasmids with similar replication initiation systems are supposed to have similar host ranges. However, if two plasmids share comparable replication and / or partitioning systems, they have higher chances not to be stably propagated in host cells. This phenomenon is called plasmid incompatibility, and is a classical method for plasmid classification (Shintani et al. 2019).
The main cause of plasmid loss, for the same plasmid vector from one host to the other can be different (De Gelder et al. 2007, Ponciano et al. 2007). For instance, the plasmid pCAR1 encodes for degradation of carbazole to catechol. It has been shown that different growth rates were obtained with different hosts cells (e.g. P. fluorescens Pf0-1, P. aeruginosa PAO1, P. putida KT2440) with carbazole as the sole carbon source on minimal media. Plus, DNA re-arrangements (i.e. gene deletion) were found in the slower growing strains. Kinetic differences were due to cumulated catechol toxicity and to differences in catechol metabolism between host cells (Takahashi et al. 2009). Another example with the plasmid pB10 in 19 different hosts is given in De Gelder et al. (2007). It was shown that stability of pB10 was impacted by disrupted plasmid replication leading to lower plasmid copy numbers, higher segregation frequency causing less efficient active partitioning, important differential growth between plasmid bearing and free cells because of increased metabolic load and conjugative transfer allowing conjugative plasmid-reuptake by cells.
So, the main factors influencing plasmid persistence into a bacterial population are : segregational plasmid loss, conjugative plasmid transfer, plasmid cost (i.e. metabolic burden and differential growth rate between plasmid-free and plasmid-bearing cells) and selection pressure (De Gelder et al. 2007).
Table of contents :
Introduction and context of the study
Part 1: Literature review
1 Population heterogeneity
1.1 Form of heterogeneity
1.2 Cause of heterogeneity
1.2.1 Gene expression stochasticity
1.2.2 Cell cycle
1.2.3 Age distribution
1.2.5 Extracellular micro-environment
1.3 Heterogeneity analysis at single-cell level
1.3.1 Single-cell isolation / individualization
1.3.2 Single-cell analysis
2 Plasmid stability
2.1 Causes and consequences of plasmid loss
2.1.1 Metabolic load of recombinant plasmids
2.1.2 Segregational plasmid stability
2.1.3 Structural plasmid stability
2.1.5 Importance of the plasmid-host interaction
2.2 Segregational plasmid stability monitoring
2.2.1 Plate count
2.2.2 Reporter protein expression
2.2.3 Plasmid copy number determination
2.3 Plasmid stabilization strategies
2.3.1 Recombinant plasmid genetic construction
2.3.2 Mechanisms of plasmid stable maintenance
3 Study model: isopropanol production by Cupriavidus necator
3.1 Genome description
3.1.1 The two chromosomes
3.1.2 The megaplasmid
3.2 Metabolism description
3.2.1 Heterotrophic metabolism
3.2.2 Autotrophic metabolism
3.2.3 Central carbon metabolism
3.2.4 Lithotrophic metabolism
3.2.5 Respiratory mechanism
3.2.6 PHB biosynthesis
3.3 C. necator as a recombinant bioproduction platform
3.4 Population heterogeneity in Cupriavidus necator
3.5 Plasmid stability strategies in Cupriavidus necator
3.6 Isopropanol production by recombinant Cupriavidus necator strains
3.6.1 Natural biological production
3.6.2 Recombinant biological production
4 Conclusion and objective of the study
Part 2: Material and methods
1 Strains, plasmids and media
1.2.1 Rich media
1.2.2 Mineral media
1.3.1 Description of plasmids
1.3.2 Plasmid construction
2 Culture conditions
2.1 Glycerol stock preparation
2.2 Preculture scheme and flask cultivation on fructose
2.2.1 Preculture scheme
2.2.2 Flask cultivations on fructose
2.2.3 Plasmid curing subcultures in flasks
2.2.4 Bioreactor inoculation
2.3 Bioreactor fermentations
2.3.1 Experimental set-up
2.3.2 Process control and regulation
2.3.3 Gas analysis
2.3.4 Batch cultivations
2.3.5 Fed-Batch cultivations
2.3.6 Continuous cultures
2.3.7 Sampling procedure description
3 Analytical procedure
3.1 Biomass characterization
3.1.1 Optical density
3.1.2 Cell Dry Weight
3.1.3 Optical microscopy
3.2 Plate count
3.3 Flow Cytometry
3.3.1 Working principle of the flow cytometer
3.3.2 Experimental protocol for population heterogeneity assessment
3.4 Fluorescence Activated Cell Sorting (FACS)
3.5 Fluorescence measurement in the medium
3.6 Determination of fructose, organic acids and ammonium concentrations
3.6.1 High-performance liquid chromatography (HPLC)
3.6.2 Gas chromatography (GC)
3.6.3 High Pressure Ionic Chromatography (HPIC)
4 Methodologies for data treatment
4.1 Rate expression for gas-phase reactions
4.1.1 Nitrogen balance
4.1.2 Dioxygen balance
4.1.3 Carbon dioxide balance
4.2 Volume determination during fed-batch cultures
4.3 Rate expressions for liquid-phase reactions
4.4 Determination of instantaneous and overall yields
4.5 Carbon and elemental balances
4.5.1 Carbon balance
4.5.2 Nitrogen balance
4.5.3 Elemental balance
4.6 Smoothing of experimental data
4.7 Statistical analysis: Normality of distribution functions by BoxPlot representation
Part 3: Results and discussion
Chapter 1: Plasmid expression stability during heterologous isopropanol production in fed-batch bioreactor1.1 Abstract
1.3 Material and method
1.6 Results synthesis
Chapter 2: Identification of heterologous subpopulations from a pure culture in a bioreactor
Subchapter 1: Plasmid expression level heterogeneity monitoring via heterologeous eGFP production at
the single-cell level in Cupriavidus necator
Subchapter 2: Study of plasmid expression level heterogeneity under plasmid-curing like conditions
2.2.3 Material and Methods
2.2.6 Results synthesis
Subchapter 3: Investigation of the robustness of Cupriavidus necator engineered strains during fed-batch
2.3.3 Material and methods
2.3.6 Results synthesis
Chapter 3: Plasmid expression level heterogeneity during heterogeneous isopropanol production studied by
an eGFP monitoring system in fed-batch bioreactor
3.3 Material and method
3.6 Results synthesis
Part 4: General discussion, conclusions and perspectives