Introduction

What is a microbial cell factory? It refers to the genetic modification of microbial chassis cells for the production of target compounds and serves as an important centrepiece of the worldwide wave of green manufacturing. The use of rapidly developing synthetic biology is an effective strategy for building cell factories, and three levels of cell factory construction have been proposed: the introduction of standardised components/dynamic systems and the creation of new life systems. At the same time, the ultimate goal of synthetic biology is to create new life forms, and the existing studies include: integration of 16 chromosomes to construct monochromosomal yeast/construction of diploid Escherichia coli by Crispre technology, but so far the construction of artificial prokaryotic polyploid cells is only in the stage of theoretical research, and the potential of its practical application is unknown.

//img_wrap

Fig 1. Gene editing in Escherichia coli chassis cells using the crisper technique（Photo credit: Tianyuan Su. A CRISPR-Cas9 Assisted Non-Homologous End-Joining Strategy for One-step Engineering of Bacterial Genome.Sci Rep.2016）

//img_wrap

In this project, SDU-CHINA follows the concept of ‘To build to understand’ in synthetic biology, and uses prokaryotic lean strains as chassis cells, combining the advantages of prokaryotic genome simplification and eukaryotic polyploid evolution, and poses the following question:Is it possible to create a new polyploid lean prokaryotic cell with a new polyploid polyploid cell? new polyploid-lite prokaryotic cell? What is the phenotype of this cell? Could it have the advantages of both? What can it tell us about evolution?

Part I: Designing and constructing minimalist prokaryotic polyploid Escherichia coli (PMEC)

In the course of previous studies, our lab team found that the morphological size of cells can be affected by regulating the expression intensity of the ftsz gene (the gene that controls the formation of cytoplasmic rings during division). When ftsz gene expression is high, the cell contains one chromosome and divides faster. When the expression intensity of ftsz is low, the cell contains two unseparated chromosomes, the cell does not divide in a filamentous manner, and eventually the cell contains two unseparated chromosomes and eventually dies. Therefore, it is possible to design and construct polyploid E. coli by finely regulating the expression level of ftsz.

//img_wrap

Fig 2. Designing and constructing polyploid Escherichia coli by fine-tuning ftsZ expression levels

//img_wrap

We selected promoters with appropriate expression levels in combination with RBS, and inserted a weakly expressed element of “KanR-terminator-promoter-ribosome binding site (RBS)” in front of the start codon ATG of the ftsZ gene of one of the replication forks during replication formation of replication forks in E. coli. Under kanamycin screening pressure, only polyploid E. coli containing both wild-type and engineered chromosomes can survive normally, and such cells utilize the wild-type chromosome for normal expression of ftsZ to regulate cell division, and the engineered chromosome for expression of KanR to resist chloramphenicol pressure. The two chromosomes in the cell undergo the next replication cycle. As ftsZ accumulates in the wild-type chromosome, a Z-loop forms in the middle of the cell. Eventually the replicated chromosomes segregate and divide into nascent polyploid cells

//img_wrap

Fig 3. Polyploid strain construction method, Photo credit: Wang Sumeng. E. coli to produce 3-hydroxypropionate and L-threonine using synthetic biology strategies. 2023. Shandong University, PhD dissertation.

//img_wrap

After completing the construction, we first analyzed the intracellular DNA content by DAPI staining as well as flow cytometry observation, and designed primers at both ends of the ftsz gene for amplification verification (the wild-type chromosome and engineered chromosome will be amplified with different lengths and sizes of fragments, respectively) as well as chromosome stability by sequential splicing. From this, we proposed the concept of PMEC (Minimalist polyploid Escherichia coli), and later conducted morphological observations and statistical analyses using field emission scanning electron microscopy to characterize the robustness and growth phenotype of the new cell system

Part2: Transcriptomic Analysis and Metabolic Network Modeling

Transcriptomic analysis revealed numerous differentially expressed genes between polyploids and haploids（Fig.4）.

//img_wrap

Fig 4. Numerous differentially expressed genes exist between polyploids and haploids

//img_wrap

To quantify the impact of these differentially expressed genes on metabolic flux at the genome scale, we started the modeling of the metabolic networks of DGF-298 and its polyploid.

Traditional approaches, such as chemical reaction kinetics modeling, are struggle to simulate metabolic flux on a larger scale. Therefore, we turned to genome-scale metabolic models (GSSM), which can predict an organism’s metabolic behavior at the genome level.

The metabolic network we established for DGF-298 and its polyploid, which includes 68 metabolic pathways and 115 genes（Fig 5.）：

//img_wrap

Fig 5. Metabolic network of DGF-298

//img_wrap

In GSSM, we start with the assumption that the metabolite concentrations remain constant over time, which can be expressed as the product of the stoichiometric matrix S and the vector of reaction rates V equals zero.

\[S\ast V=0\]

Our goal is to integrate the expression fold change into the model. Therefore, we will incorporate a weight matrix into the equation.

\(S\ast (V ∘W)=0\) After this process, we obtained the impact of each gene on metabolic flux（Fig 6.）.

//img_wrap

Fig. 6 Effect of different gene expression multiplicities on metabolic flows

//img_wrap

We found that the downregulation of two genes, pkfB and nuoF, reduced the bacterial biomass reaction rate. Therefore, we constructed few plasmids that included these two genes. After trying different combinations of RBS and promoters, we observed a increase in the growth rate of certain strains.

//img_wrap

Fig 7. Different combinations of modular control

//img_wrap

Sequencing showed that we successfully constructed and transferred the PACYC-Ptac-pfkA-nuoF plasmid containing different expression intensities combinations into DGF-298-103Z (Fig.7). The qPCR results showed significant up-regulation of nuoF or pfkB genes in the experimental group with regulatory combinations of moderate intensity (M) and high intensity (U). By characterizing the growth curves of strains with different intensity regulatory modules, we found that some experimental groups had earlier log phase compared with DGF-298-103Z in the control group (Fig.9).

//img_wrap

Fig 8. Results of RT-qPCR for gene expression measurements

Fig 9. Characterization results of modularly modulation

//img_wrap

Part3:Fermentation potential analysis of novel chassis cells and application of Deep learning

In transcriptomic analysis, we found that the expression levels of carboxylic acid synthesis pathways, amino acid synthesis pathways, and sulfur compound synthesis pathways were upregulated in polyploids. This suggests that polyploidization may lead to improved fermentation potential.

//img_wrap

Fig 10. Results of the transcriptome analysis of DGF298 and DGF298-103Z

//img_wrap

To analyze the production capacity of the novel chassis cells for various high-value products, we selected suitable vectors, constructed different heterologous gene plasmids, and introduced them into the cells for analysis. We found that polyploid E. coli exhibited better potential for heterologous protein expression, PHB production, and amino acid fermentation. As a result, PHB was selected as the downstream fermentation product.

//img_wrap

Fig 11. a. Protein expression level b. PHB fermentation analysis c.Amino acid fermentation yield

//img_wrap

After that, we investigated the metabolic process of intracellular PHB, and reviewed the literature and found that the intracellular metabolic flow would have a competition between the TCA cycle and PHB synthesis reaction for acetyl coenzyme A, i.e., the competition between cell growth and cell metabolism for the allocation of carbon fluxes.

//img_wrap

Fig 12. Dynamic cascade control lines

//img_wrap

We introduce a dynamic regulatory cascade tool library based on the Las and Tra quorum sensing systems. During bacterial growth, LasI synthase produces and secretes HSL signals. Once the concentration reaches a threshold, the signal binds to the TraR receptor, forming a TraR-HSL complex that activates the Ptra promoter, leading to gfp expression and LasR activation. The LasR-HSL complex then activates Plas, triggering the expression of the second gene, rfp. The time-lagged activation of these promoters enables a well-timed TCA response to PHB synthesis.

//img_wrap

Fig 13. Characterization results of the shuttle promoter

//img_wrap

In order to achieve a more extensive and precise regulation for this line, we screened some of the Corynebacterium glutamicum-Escherichia coli shuttle promoters, characterized them and introduced them in front of the Plas genes, expanding the time-standardized differences between the GFP and RFP genes through transcriptomic data

//img_wrap

Fig 14. Normalized curves of self-induced dynamic chronological regulatory cascade lines regulating the intensity of LasI synthase expression

//img_wrap

In this part of the modeling, we focused on predicting PHB yield from two aspects: intracellular metabolism and fermentation environment. First, we used our constructed GSSM for multi-objective optimization, and second, we combined deep learning to predict PHB yield in the fermentation tank.

For the first part， In GSSM, to solve for the distribution of the overall metabolic flux, we need to define an objective function, which represents the reaction we want to optimize. However, under a single objective function, the model maximizes PHB yield while completely neglecting growth, which is clearly unrealistic.

//img_wrap

Fig 15. Single-objective optimization results where the model completely ignores the biomass response

//img_wrap

Therefore, we transformed it into a multi-objective optimization problem. The objective functions we defined are as follows:

\[F\left(x\right)=\omega_1v_1+\omega_2v_2\ +\ ...\ +\ \omega_nv_n\] \[Maximize\ F(x)= ω^T v\]

Thus, we defined both biomass reaction and PHB synthesis reaction as objective functions. We found that under this optimization approach, the biomass reaction rate is positively correlated with the intake rates of glucose and oxygen. In contrast, the synthesis rate of PHB differs from that in single-objective optimization; it is positively correlated with glucose intake rate. However, under high oxygen concentration conditions, PHB production is limited.

//img_wrap

Fig 16. Multi-objective optimization balances PHB production and growth

//img_wrap

This reflects the balance of metabolic flux in bacteria. At low oxygen concentrations, due to the limited availability of ATP, a significant amount of ATP is used to transport acetate into the cell, which is facilitated by a reaction called ACS. Analyzing the ACS reaction, we find that its reaction rate is indeed higher under low oxygen conditions. Since acetate is a precursor for PHB, the increased flux of the ACS reaction also enhances the synthesis rate of PHB.

This indicates that low oxygen and high glucose concentrations are favorable for PHB synthesis.

In the second part, we used deep learning approach to predict the yield of PHB in the fermentation tank. Since PHB yield is typically difficult to measure directly, our deep learning approach allows us to indirectly predict the concentration changes of PHB over time using easily measurable data collected during the fermentation process.

//img_wrap

Fig 17. Relationship between loss and epoch

//img_wrap

During the training phase, we used Mean Squared Error (MSE) as the loss function. After training for 100 epochs, the loss decreased to 0.0017. We then proceeded to the validation phase.

We selected a complete fermentation run as the validation set for the model. The results showed an R² value of 0.98 and a Mean Squared Error (MSE) of 0.19, indicating that the model performs well.

Summary

At this point, we have successfully designed and created a new prokaryotic polyploid minimalist chassis system and verified that it can stably inherit both engineered and wild-type chromosomes. Through a series of morphological property analyses, we have found that polyploidized E. coli has a significant increase in length, volume, and surface area, and that some metabolites, such as phenylalanine (Phe) are significantly enriched in cells, which indicates that it is possible for it to serve as a more abundant natural source of nutrients than the extant cellular system. This suggests that it can serve as a richer natural source of nutrients than the existing cell system. In addition, the polyploidized minimal chassis cells showed good resistance under perturbed conditions such as high temperature and pH, which is worthy of further exploration of their potential adaptive evolutionary capacity and helps to reveal the impact of genome streamlining combined with chromosome doubling strategies on prokaryotic evolution. We also focused on the potential of this novel cell system as a cellular fermentation factory, showing better carbon and protein metabolism fluxes in the expression of heterologous proteins GFP and poly-β-hydroxybutyric acid (PHB), which demonstrated that it has significantly better fermentation capacity than the streamlined and wild-type strains. In terms of modeling, we successfully constructed a genome-scale metabolic network model of DGF-298 and its polyploid. Combined with the transcriptome data, we screened out the key genes affecting growth and further optimized the growth rate and biomass response of chassis cells. For the intra- and extracellular production process of the downstream product PHB, we carried out multi-objective optimization in the metabolic network model, and at the same time, combined with deep learning, we constructed an environmental model for real-time prediction of PHB production with the help of multi-for-sensors in the fermenter.

Reference

//referrence

[1] Hirokawa Y, Kawano H, Tanaka-Masuda K, et al. Genetic manipulations restored DGF-298e growDGF-298 fitness of reduced-genome Escherichia coli[J]. J Biosci Bioeng, 2013, 116(1): 52-58. DOI: 10.1016/j.jbiosc.2013.01.010.

[2] Matteau D, Champie A, Grenier F, et al. Complete sequence of DGF-298e genome-reduced Escherichia coli DGF-298[J]. Microbiol Resour Announc, 2023, 12(11): e0066523. DOI: 10.1128/MRA.00665-23.

[3] Couto JM, McGarrity A, Russell J, et al. DGF-298e effect of metabolic stress on genome stability of a synDGF298etic biology chassis Escherichia coli K12 strain[J]. Microb Cell Fact, 2018, 17(1): 8. DOI: 10.1186/s12934- 018-0858-2.

[4] Biriukova IV, Krylov AA, Kiseleva EM, et al. [Construction of DGF-298e new Escherichia coli K-12 wildtype strain wiDGF-298 improved growDGF-298 characteristics for application in metabolic engineering][J]. Genetika, 2010, 46(3): 349-355. (in Russian)

[5] Couto JM, McGarrity A, Russell J, et al. DGF-298e effect of metabolic stress on genome stability of a synDGF298etic biology chassis Escherichia coli K12 strain[J]. Microb Cell Fact, 2018, 17(1): 8. DOI: 10.1186/s12934- 018-0858-2.

[6] Wang M, Fang C, Ma B, et al. Regulation of cytokinesis: FtsZ and its accessory proteins[J]. Curr Genet, 2020, 66(1): 43-49. DOI: 10.1007/s00294-019-01005-6.

[7] Tack ILMM, Nimmegeers P, Akkermans S, et al. A low-complexity metabolic network model for DGF-298e respiratory and fermentative metabolism of Escherichia coli[J]. PLoS One, 2018, 13(8): e0202565. DOI: 10.1371/journal.pone.0202565.

[8] Wang J, Huang J, Liu S. DGF-298e production, recovery, and valorization of polyhydroxybutyrate (PHB) based on circular bioeconomy[J]. Biotechnol Adv, 2024, 72: 108340. DOI: 10.1016/j.biotechadv.2024.108340.

[9] Hack NJ, Billups B, GuDGF-298rie PB, et al. Green fluorescent protein as a quantitative tool[J]. J Neurosci MeDGF-298ods, 2000, 95(2): 177-184. DOI: 10.1016/s0165-0270(99)00178-8

[10] Hwang S, Choi KH, Yoon N, Cha J. Improvement of a Sulfolobus-E. coli shuttle vector for heterologous

//referrence