Abstract
The bacterium E. coli is widely used to produce recombinant proteins such as growth hormone and insulin. One inconvenience with E. coli cultures is the secretion of acetate through overflow metabolism. Acetate inhibits cell growth and represents a carbon diversion, which results in several negative effects on protein production. One way to overcome this problem is the use of a synthetic consortium of two different E. coli strains, one producing recombinant proteins and one reducing the acetate concentration. In this paper, we study a mathematical model of such a synthetic community in a chemostat where both strains are allowed to produce recombinant proteins. We give necessary and sufficient conditions for the existence of a coexistence equilibrium and show that it is unique. Based on this equilibrium, we define a multi-objective optimization problem for the maximization of two important bioprocess performance metrics, process yield and productivity. Solving numerically this problem, we find the best available trade-offs between the metrics. Under optimal operation of the mixed community, both strains must produce the protein of interest, and not only one (distribution instead of division of labor). Moreover, in this regime acetate secretion by one strain is necessary for the survival of the other (syntrophy). The results thus illustrate how complex multi-level dynamics shape the optimal production of recombinant proteins by synthetic microbial consortia.
Similar content being viewed by others
Data availability statement
Data sharing not applicable to this article as no datasets were generated or analysed during the current study.
Notes
This can be seen by noting that d is decreasing and that \(d(l)=0\).
Let \(\xi \) be a state variable of the model. A steady-state is detected when \(|\xi (t+\Delta t)-\xi (t)|<\delta \). In our simulations we choose \(\delta =10^{-6}\) g L\(^{-1}\) and \(\Delta t=10\,\) days.
References
Baeshen MN, Al-Hejin AM, Bora RS, Ahmed M, Ramadan H, Saini KS, Baeshen NA, Redwan EM (2015) Production of biopharmaceuticals in E. coli: current scenario and future perspectives. J Microbiol Biotechnol 25(7):953–962
Basan M, Hui S, Okano H, Zhang Z, Shen Y, Williamson JR, Hwa T (2015) Overflow metabolism in E. coli results from efficient proteome allocation. Nature 528(7580):99
Bernstein HC, Paulson SD, Carlson RP (2012) Synthetic Escherichia coli consortia engineered for syntrophy demonstrate enhanced biomass productivity. J Biotechnol 157(1):159–166
Bremer H, Dennis P (2013) Modulation of chemical composition and other parameters of the cell at different exponential growth rates. In: Slauch J (ed) Ecosal plus: cellular and molecular biology of E. coli, Salmonella, and the Enterobacteriaceae, ASM Press, Washington, DC
Byrd RH, Hribar ME, Nocedal J (1999) An interior point algorithm for large-scale nonlinear programming. SIAM J Optim 9(4):877–900
Chesson P (2000) Mechanisms of maintenance of species diversity. Annu Rev Ecol Syst 31(1):343–366
De Mey M, De Maeseneire S, Soetaert W, Vandamme E (2007) Minimizing acetate formation in E. coli fermentations. J Ind Microbiol Biotechnol 34(11):689–700
Deutscher J, Francke C, Postma PW (2006) How phosphotransferase system-related protein phosphorylation regulates carbohydrate metabolism in bacteria. Microbiol Mol Biol Rev 70(4):939–1031
Doran PM (1995) Bioprocess engineering principles. Elsevier, Amsterdam
Eiteman MA, Altman E (2006) Overcoming acetate in Escherichia coli recombinant protein fermentations. Trends Biotechnol 24(11):530–536
Enjalbert B, Millard P, Dinclaux M, Portais JC, Létisse F (2017) Acetate fluxes in Escherichia coli are determined by the thermodynamic control of the Pta-AckA pathway. Sci Rep 7(1):1–11
Fuchs C, Köster D, Wiebusch S, Mahr K, Eisbrenner G, Märkl H (2002) Scale-up of dialysis fermentation for high cell density cultivation of Escherichia coli. J Biotechnol 93(3):243–251
Görke B, Stülke J (2008) Carbon catabolite repression in bacteria: many ways to make the most out of nutrients. Nat Rev Microbiol 6(8):613–24
Gunantara N (2018) A review of multi-objective optimization: methods and its applications. Cogent Eng 5(1):1502242
Hansen SR, Hubbell SP (1980) Single-nutrient microbial competition: qualitative agreement between experimental and theoretically forecast outcomes. Science 207(4438):1491–1493
Harvey E, Heys J, Gedeon T (2014) Quantifying the effects of the division of labor in metabolic pathways. J Theor Biol 360:222–242
Hays SG, Patrick WG, Ziesack M, Oxman N, Silver PA (2015) Better together: engineering and application of microbial symbioses. Curr Opin Biotechnol 36:40–49
Heßeler J, Schmidt JK, Reichl U, Flockerzi D (2006) Coexistence in the chemostat as a result of metabolic by-products. J Math Biol 53(4):556–584
Huang CJ, Lin H, Yang X (2012) Industrial production of recombinant therapeutics in Escherichia coli and its recent advancements. J Ind Microbiol Biotechnol 39(3):383–399
Izard J, Gomez Balderas CD, Ropers D, Lacour S, Song X, Yang Y, Lindner AB, Geiselmann J, de Jong H (2015) A synthetic growth switch based on controlled expression of rna polymerase. Mol Syst Biol 11(11):840
Jagmann N, Philipp B (2014) Design of synthetic microbial communities for biotechnological production processes. J Biotechnol 184:209–218
Kremling A, Geiselmann J, Ropers D, de Jong H (2015) Understanding carbon catabolite repression in Escherichia coli using quantitative models. Trends Microbiol 23(2):99–109
Kurland C, Dong H (1996) Bacterial growth inhibition by overproduction of protein. Mol Microbiol 21(1):1–4
Lin H, Castro NM, Bennett GN, San KY (2006) Acetyl-CoA synthetase overexpression in Escherichia coli demonstrates more efficient acetate assimilation and lower acetate accumulation: a potential tool in metabolic engineering. Appl Microbiol Biotechnol 71(6):870–874
Liu X, Li XB, Jiang J, Liu ZN, Qiao B, Li FF, Cheng JS, Sun X, Yuan YJ, Qiao J et al (2018) Convergent engineering of syntrophic Escherichia coli coculture for efficient production of glycosides. Metab Eng 47:243–253
Luli GW, Strohl WR (1990) Comparison of growth, acetate production, and acetate inhibition of Escherichia coli strains in batch and fed-batch fermentations. Appl Environ Microbiol 56(4):1004–1011
Martínez C, Gouzé JL (2021) Global dynamics of the chemostat with overflow metabolism. J Math Biol 82(3):1–21
Mauri M, Gouzé JL, De Jong H, Cinquemani E (2020) Enhanced production of heterologous proteins by a synthetic microbial community: Conditions and trade-offs. PLoS Comput Biol 16(4):e1007795
Miettinen K (2012) Nonlinear Multiobjective Optimization, vol 12. Springer Science & Business Media
Milias-Argeitis A, Rullan M, Aoki S, Buchmann P, Khammash M (2016) Automated optogenetic feedback control for precise and robust regulation of gene expression and cell growth. Nat Commun 7:12546
Pandhal J, Noirel J (2014) Synthetic microbial ecosystems for biotechnology. Biotechnol Lett 36(6):1141–1151
Pirt S (1965) The maintenance energy of bacteria in growing cultures. Proc R Soc Lond B Biol Sci 163(991):224–31
Roell GW, Zha J, Carr RR, Koffas MA, Fong SS, Tang YJ (2019) Engineering microbial consortia by division of labor. Microb Cell Factories 18(1):1–11
Rosenzweig RF, Sharp R, Treves DS, Adams J (1994) Microbial evolution in a simple unstructured environment: genetic differentiation in Escherichia coli. Genetics 137(4):903–917
Sari T, El Hajji M, Harmand J (2012) The mathematical analysis of a syntrophic relationship between two microbial species in a chemostat. Math Biosci Eng 9(3):627–645
Shong J, Diaz MRJ, Collins CH (2012) Towards synthetic microbial consortia for bioprocessing. Curr Opin Biotechnol 23(5):798–802
Smith HL, Waltman P (1995) The theory of the chemostat: dynamics of microbial competition, vol 13. Cambridge University Press, Cambridge
Stump SM, Klausmeier CA (2016) Competition and coexistence between a syntrophic consortium and a metabolic generalist, and its effect on productivity. J Theor Biol 404:348–360
Tsoi R, Wu F, Zhang C, Bewick S, Karig D, You L (2018) Metabolic division of labor in microbial systems. Proc Natl Acad Sci USA 115(10):2526–2531
Van Dien S (2013) From the first drop to the first truckload: commercialization of microbial processes for renewable chemicals. Curr Opin Biotechnol 24(6):1061–1068
Wolfe AJ (2005) The acetate switch. Microbiol Mol Biol Rev 69(1):12–50
Wu G, Yan Q, Jones JA, Tang YJ, Fong SS, Koffas MA (2016) Metabolic burden: cornerstones in synthetic biology and metabolic engineering applications. Trends Biotechnol 34(8):652–664
Acknowledgements
This work was supported by the INRIA (IPL CoSy) and by the ANR projects Maximic (ANR-11-LABX-0028-01) and Ctrl-AB (ANR-20-CE45-0014). Additional support for Carlos Martínez was provided by the European Union within ESIF in the framework of the Operational Programme “Research, Development and Education” (CZ.02.2.69/0.0/ 0.0/18_053/0016982).
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
A Proofs
1.1 A.1 Change of notation
In this appendix, we present the proofs of all the statements from Sect. 3, and we present a series of technical results needed for their proofs. We begin by simplifying the notations presented in Sect. 2 and recalling the hypotheses that are necessary for the proofs.
To simplify the notation of uptake rates, we denote \(r_{up,p}^g\) by \(r_g\) and we define
Using property (8) and the definition of \(r_{over,p}^a\), we have that
Thus, \(r_a\) represents an acetate flux between producers and the medium, which can be negative or positive depending on the sign of \(r_g-l\). We also define
It is important to note that \(\rho _a(0, r_g)=0\) and \(\rho _a(A, r_g)\ge 0\) for all \(A, r_g \ge 0\), and that \(A \longmapsto \rho _a(A,r_g)\) is strictly increasing. The following parameters simplify the notations of product yields:
Finally, to simplify the notation of (9), we define
The system (9) can be rewritten as follows:
This new notation is a compromise between a simple notation to address the proofs and a notation that is closed to that presented in Sect. 2. Note that we have defined \(\mu _p\), \(\mu _c\), \(r_a\), and \(\rho _a\) in terms of A and \(r_g\). The dilution rates defined in Propositions 1 and 2 can be rewritten as:
We also define the following constant:
We recall the hypotheses of the model in terms of the new notation. Hypothesis (11) is equivalent to
and Hypothesis (12) is equivalent to
From now on, we assume that (22) and (23) are true.
Our main task is to prove Theorem 1, that is, to find necessary and sufficient conditions for the existence of a coexistence equilibrium of (19). Therefore, we must study the existence and uniqueness of solutions to
with \(B_p,B_c>0\) and \(G,A \ge 0\).
1.2 A.2 Results
Our first lemma establishes some properties of any non-negative solution to (24) with \(B_p>0\). These properties will be repeatedly used along the appendix.
Lemma 2
Let \(D>0\), and let \(B_p\), \(B_c\), A, and \(r_g\) be such that
with \(B_p>0\) and \(B_c, A, G, r_g \ge 0\). Let \(D^a\) and \(\alpha \) be given by (20) and (21), respectively, and let us define
The following statements are equivalent:
-
(a)
\(A>0\),
-
(b)
\(r_a(A, r_g)<0\),
-
(c)
\(r_g>l\),
-
(d)
\(\mu _p(A, r_g)=\alpha (r_g-l)+D^a\) and \(r_g> l\),
-
(e)
\(r_g= \gamma \) and \(\gamma >l\),
-
(f)
\(D> D^a\).
Proof
If (a) holds, then \(DA>0\) and \(\rho _a(A,r_g)B_c \ge 0\). From the second equation in (25), we obtain that \(r_a(A,r_g) B_p<0\). Since \(B_p>0\), we conclude that \(r_a(A,r_g)<0\), and therefore (b) holds. If (b) holds, then \(r_g\) cannot be equal or lower than l, otherwise otherwise \(r_a(A,r_g)\) would be non-negative. Thus, (c) holds. Now, if (c) holds, by definition, \(r_a(A, r_g)=-k_{over}(r_g-l)\). Using the definition of \(\mu _p\), we obtain
which is equivalent to
and (d) holds. If (d) holds, from the first equation in (25), we obtain that
from where \(r_g=\gamma \), and (e) holds. Now, if (e) holds, we have that \(\gamma > l\), or equivalently
from where \(D>D^a\), and (f) holds. We prove the last part indirectly, that is, we prove that if (a) is false, then (f) is also false. Indeed, if \(A=0\), then \(\rho _a(A, r_g)=0\), and from the second equation in (25), we conclude that \(r_a(A, r_g)=0\). Thus, necessarily \(r_g\le l\), otherwise \(r_a(A, r_g)\) is negative. Consequently, by definition of \(\mu _p\),
Combining the previous equation with the first equation in (25), we conclude that \(D \le D^a\). This completes the proof. \(\square \)
The following lemma states some properties related to \(D^p\).
Lemma 3
Let \(\gamma \) be given by (26) and let \(D^a\) and \(D^p\) be given by (20). We have:
-
(a)
\(D^p=\alpha [r_g(G_{in},0)-l]+D^a\).
-
(b)
\(D^p>D^a\).
-
(c)
If \(D<D^p\), then \(\gamma <r_g(G_{in},0)\).
Proof
From (23), we have that \(r_g(G_{in},0)>l\). Then, we have that
From the previous equation and the definitions of \(D^p\) (see (20)) and \(\mu _p\), we obtain that
which proves (a) and implies
and (b) is proved. For part (c), if \(D<D^p\), using (27) and the definition of \(\gamma \), we have that
This completes the proof. \(\square \)
The proof of Proposition 1 in Sect. 3 follows directly from the following proposition.
Proposition 4
Let \(D^a\) and \(D^p\) be given by (20). We have:
-
(a)
If \(D^p>D\), then (19) admits a unique equilibrium of the form \(E^p=(B_p^p,0,G^p,A^p)\), with \(B_p^p>0\). Moreover:
-
(I)
If \(D > D^a\), then \(A^p>0\) and \(r_g(G^p,A^p)>l\).
-
(II)
If \(D \le D^a\), then \(A^p=0\) and \(r_g(G^p,0)\le l\).
-
(I)
-
(b)
If \(D^p \le D\), then (19) has no equilibrium with \(B_p>0\) and \(B_c=0\).
Proof
The equilibria of (19) with \(B_c=0\) and \(B_p>0\) are given by the solutions of the following system
From the equivalence among (a), (c) and (f) in Lemma 2, we have that
-
(1)
If \(D > D^a\), then any solution of (28) satisfies \(A>0\) and \(r_g(G,A)>l\).
-
(2)
If \(D \le D^a\), then any solution of (28) satisfies \(A=0\) and \(r_g(G,A)\le l\).
If statement (1) holds, using Lemma 2 part (d), we obtain that (28) is equivalent to
From the first equation in (29) (or Lemma 2 part (e)), we have that \(r_g(G,A)=\gamma \), with \(\gamma \) defined by (26). Thus, from the second and third equations in (29), we have
Replacing (30) and (31) in the first equation of (29), we obtain the following equation for \(B_p\):
From the monotonicity of \(r_g\), we have that g is strictly decreasing. Since \(g(0)=D^p-D\) (see Lemma 3 part (a)), and \(g(G_{in}/\beta _g)=D^a-D<0\), we conclude that (32) admits a unique solution \(B_p^p>0\) if \(D<D^p\), and has no positive solution if \(D\ge D^p\). In particular, this proves (b).
If statement (2) holds, after replacing A by 0, we obtain that (28) is equivalent to
From the third equation in (33), we obtain that \(r_a(G,A)=0\). Thus, (33) is reduced to:
From the second equation in (34), we have that \(G \in [0, G_{in})\), otherwise \(B_p\le 0\). Now, since \(G \longmapsto h(G)\) is strictly increasing and \(h(0)=-k_{def}-D<0\), (34) admits a unique positive equilibrium if and only if \(h(G_{in})>0\). We have
Therefore, \(h(G_{in})>0\) and the proof is completed. \(\square \)
The following proposition is equivalent to Proposition 2 in Sect. 3.
Proposition 5
(Cleaner steady state) Let \(D^c\) be given by (20). We have:
-
(a)
If \(D^c>D\), then (19) admits a unique equilibrium with \(B_c>0\) and \(B_p=0\), which is of the form \(E^c=(0,B_c^c,G^c,0)\).
-
(b)
If \(D^c\le D\), then (19) has no equilibrium with \(B_c>0\) and \(B_p=0\).
Proof
Any equilibrium with \(B_c>0\) and \(B_p=0\) is a solution of (replace \(B_p\) by 0 in (24)):
From the third equation in (35), we have that \(A=0\), otherwise we have an equality between a positive and a negative number. Using the definition of \(\mu _c\) and the fact that \(\rho _a(0,r_g)=0\), (35) can be rewritten as
From the second equation in (36), we have that \(G \in [0, G_{in})\), otherwise \(B_c\le 0\). Now, since h is strictly increasing and \(h(0)=-k_{deg}-D<0\), we have that (36) admits a solution, which is unique, if and only if \(h(G_{in})>0\). The rest of the proof follows from noting that \(h(G_{in})=D^c-D\). \(\square \)
The following lemma shows that if \(D \le D^a\), then there is a coexistence equilibrium if and only if some parameters are perfectly balanced. Note that this lemma does not ensure the uniqueness of a coexistence equilibrium.
Lemma 4
Let \(D^a\) and \(D^p\) be given by (20). If \(D \le D^a\), then (19) admits a coexistence equilibrium if, and only if
Proof
Since \(D\le D^a\), from the equivalence among the statements (a), (c), and (f) in Lemma 2, we have that for any solution to (24) with \(B_p>0\), \(A=0\) and \(r_g(G,0) \le l\). Therefore, \(r_a(0, r_g(G,0))=\rho _a(0, r_g(G,0))=0\) and (24) is equivalent to
If \(Y_{g/p}\ne \beta Y_{g/c}\), then the first and second equation cannot be satisfied at the same time. Thus, there cannot be a coexistence equilbrium. On the other hand, if \(Y_{g/p}= \beta Y_{g/c}\), then G is a solution of \(h(G)=0\) with \(h(G):=Y_{g/p} r_g(G,0)-k_{deg}-D\). Note that h is strictly increasing, \(h(0)=-k_{deg}-D<0\), and
Thus \(h(G)=0\) admits a unique solution \(G^* \in (0, G_{in})\). Replacing G by \(G^*\) in the third equation in (37), we obtain the existence of infinity coexistence equilibria. \(\square \)
The following lemma shows that if the dilution rate is too high then coexistence is impossible.
Lemma 5
Let \(D^p\) be given by (20). If \(D\ge D^p\), then (19) has no coexistence equilibrium.
Proof
We prove this by contradiction. Let us assume that \(D\ge D^p\) and that (19) admits a coexistence equilibrium \((B_p^*, B_c^*, G^*,A^*)\). Since \(D^p>D^a\) (see Lemma 3 part (b)), we can use the equivalence between (d) and (f) from Lemma 2 to conclude that
From Lemma 2 part (a), we obtain that
From the third equation in (24), we have that \(G^*<G_{in}\). Using the monotonicity of \(r_{g}\), we can combine (38) and (39) to obtain \(\mu _p(A^*,r_g(G^*,A^*))<D^p\). Finally, from the first equation in (24), we conclude that \(D<D^p\), which contradicts our initial hypothesis (\(D \ge D^p\)). This completes the proof. \(\square \)
The following establishes some necessary conditions for coexistence.
Lemma 6
Let \(D^a\), \(D^p\), and \(D^c\) be given by (20) and let \(G^c\) be given by Proposition 5. Let us assume that \(D^a<D<D^p\) and \(D<D^c\). If (19) admits a coexistence equilibrium, then
Proof
Let us assume that (19) admits a coexistence equilibrium \((B_p^*, B_c^*,G^*,A^*)\). Since \(D>D^a\), from Lemma 2 part (e), we know that \(r_g(G^*,A^*)=\gamma \) with \(\gamma \) defined by (26). Now, from the second equation in (24) and the definition of \(G^c\) we have
which is equivalent to
Since \(D>D^a\), from Lemma 2, we have that \(A^*>0\), hence \(\rho _a(A^*, \gamma )>0\). Consequently, from (41), we conclude that
which implies
Since \(r_g(G^c,0)>l\) (see (42)), from the definitions of \(r_a\) and \(\mu _p\), we conclude that
Combining the previous equation with (43), we conclude that (40) holds and the proof is completed. \(\square \)
The following lemma gives upper bounds for any coexistence equilibrium.
Lemma 7
Let \(D^a\) and \(D^p\) be given by (20). Let us assume that \(D^a< D <D^p\). Then, any solution of (24) with \(B_p,B_c>0\) satisfies
with \(A^p\) and \(G^p\) given by Proposition 4.
Proof
Let us assume that (19) admits a solution \((B_p^*,B_c^*,G^*,A^*)\). Since \(D>D^a\), we have that \(\gamma = r_g(G^*, A^*)= r_g(G^p, A^p)\) (see Lemma 2). Thus, we have the following equations:
and
We prove that \(A^*<A^p\) by contradiction. Let us assume that \(A^*\ge A^p\). Using the monotonicity of \(r_g\) and the fact that \(r_g(G^p,A^p)=r_g(G^*,A^*)\), we obtain that \(G^p\le G^*\). From the first equation in (46) and the first equation in (45), we obtain that:
from where \(0\le D(G^*-G^p)<\gamma (B_p^p-B_p^*)\), which implies
From the second equation in (46) and the second equation in (45), we obtain that:
Now, since \(\gamma >l\), we have that \(r_a(A^*, \gamma )=r_a(A^p, \gamma )=-k_{over}(\gamma -l)<0\). Thus, from (48), we obtain that
which implies \(B_p^*>B_p^p\). This contradicts (47). Then, \(A^p> A^*\) and consequently \(G^p>G^*\). \(\square \)
The following proposition is equivalent to Proposition 3 in Sect. 3.
Proposition 6
Let \(G^c\) and \(A^c\) be given by Proposition 4 and let \(D^a\) and \(D^p\) be given by (20). If (19) admits a coexistence equilibrium \((B_p^*, B_c^*, G^*, A^*)\), then:
-
(a)
\(D \le D^a\) implies \(A^*=0\) and \(r_g(G^*,0)\le l\).
-
(b)
\(D^a<D<D^p\) implies \(0<A^*<A^p\) and \(r_g(G^*,A^*)>l\).
Proof
From the equivalence among (a), (c), and (f) in Lemma 2, we obtain that immediately
-
(I)
\(D \le D^a\) implies \(A^*=0\) and \(r_g(G^*,A^*)\le l\).
-
(II)
\(D^a<D<D^p\) implies \(0<A^*\) and \(r_g(G^*,A^*)>l\).
And from Lemma 7, we obtain that \(D^a<D<D^p\) implies \(A^*<A^p\). This completes the proof. \(\square \)
The following result states some necessary and sufficient conditions for the coexistence of a unique coexistence equilibrium.
Lemma 8
Let \(D^p\), \(D^a\), and \(A^p\) be given by Proposition 4 and let \(\gamma \) be defined by (26). If \(D^a<D<D^p\), then (9) admits a coexistence equilibrium (unique) if, and only if,
Proof
Since \(D>D^a\), from Lemma 2, we know that any solution to (24) satisfies
Thus, from the second equation in (24), we obtain that \(\phi (A):=\mu _c(A, \gamma )-D =0\). We note that (49) is equivalent to \(\phi (A^p)>0\) and \(\phi (0)<0\). If (19) has a coexistence equilibrium, say \((B_p^*, B_c^*, G^*, A^*)\), then \(\phi (A^*)=0\). Since \(\phi \) is strictly increasing and \(0<A^*<A^p\) (see Lemma 7), we conclude that \(\phi (A^p)>0\) and \(\phi (0)<0\). Thus, (49) is a necessary condition for the existence of a coexistence equilibrium.
We prove now that (49) is also a sufficient condition, that is, if (49) holds, then (24) admits a unique solution. Since (49) holds and \(\phi \) is strictly increasing, there exists a unique \(A^* \in (0, A^p)\) such that \(\phi (A^*)=0\).
Now, replacing A by \(A^*\) in (50), we obtain the following equation for G:
It is clear that \(\varphi \) is strictly increasing and that \(\varphi (0)=-\gamma <0\). Let \(G^p\) be given by Proposition 4. Since \(A^p>A^*\) ( see Lemma 7), we have that \(\varphi (G^p)>r_g(G^p,A^p)-\gamma \). From the definition of \(G^p\) and \(A^p\) (e.g. see (45)), we have that \(r_g(G^p,A^p)-\gamma =0\), and hence \(\varphi (G^p)>0\). Consequently, there is a unique \(G^* \in (0,G^p)\) such that \(\varphi (G^*)=0\). It remains to prove the existence and uniqueness of a positive solution of the following linear system for \((B_p,B_c)\) obtained from the third and fourth equations in (24):
From the first equation in (51) we have that
Combining the previous equation with the second equation in (51), we obtain:
Since \(A^p>A^*\) and \(G^p>G^*\), we have that
Finally, if we isolate \(B_p^p\) in the second equation in (45), and we replace it in the first equation, we obtain that
This shows that \(\kappa >0\). Then (51) has a positive solution. \(\square \)
The following result states the conditions presented in statements (b) and (c) in Theorem 1.
Proposition 7
Let \(D^p\), \(D^a\), and \(D^c\) be given by (20) and let \(\gamma \) be defined by (26). We have that:
-
(a)
If \(D^a<D<D^p\) and \(D<D^c\) then (9) admits a (unique) coexistence equilibrium, if and only if
$$\begin{aligned} \mu _p(0, r_g(G^c,0))>D\,\,\hbox {and}\,\, \mu _c(A^p, \gamma )>D, \end{aligned}$$with \(G^p\) and \(A^p\) defined in Proposition 4, and \(G^c\) defined in Proposition 5.
-
(b)
If \(D^a<D<D^p\) and \(D\ge D^c\) then (9) admits a (unique) coexistence equilibrium, if and only if
$$\begin{aligned} \mu _c(A^p, \gamma )>D, \end{aligned}$$with \(G^p\) and \(A^p\) defined in Proposition 4.
Proof
Let us assume that \(D^a<D<D^p\). From Lemmas 6 and 8, we know that (a) and (b) provide necessary conditions for the existence of a coexistence equilibrium. It remains to prove that they also provide sufficient conditions. Using Lemma 8, we know that \(\mu _c(A^p, \gamma )>D\) and \(\mu _c(0, \gamma )<D\) are sufficient conditions. In statements (a) and (b), it is direct to see that \(\mu _c(A^p, \gamma )>D\) holds, therefore we must prove that each statement, (a) and (b), also implies \(\mu _c(0, \gamma )<D\). Thus, the proof of parts (a) and (b) follows from proving that
-
(I)
if \(D<D^c\), \(D^a<D<D^p\), and \(\mu _p(0,r_g(G^c,0))>D\), then \(\mu _c(0, \gamma )<D\), and
-
(II)
if \(D\ge D^c\) and \(D^a<D<D^p\), then \(\mu _c(0, \gamma )<D\),
respectively.
For (I), from the hypotheses, we have that \(\mu _p(0,r_g(G^c,0))>D^a\). Using the definition of \(\mu _p\) and \(D^a\), we have
which can be rearranged as
From the definition of \(r_a\) (see (16)), we have that \(r_a(0,r_g(G^c,0))\) cannot be positive (evaluate \(r_a(A,r_g)\) at \(A=0\)). Therefore, (52) implies that \(r_g(G^c,0)>l\). Using again the definition of \(r_a\), we have that \(r_a(0,r_g(G^c,0))=-k_{over}[r_g(G^c,0)-l]\). Hence, from the definition of \(\mu _p\) (see (18)), we have
with \(\alpha \) defined by (21). Now, since \(D>D^a\), according to Proposition 4, \(r_g(G^p, A^p)>l\). Thus, using Lemma 2 and the definition of \(\gamma \), we have that
By definition of \(G^p\) and \(A^p\), we have that \(\mu _p(G^p, A^p)=D\). Thus, from the hypotheses we have that
Combining (53), (54), and (55), we conclude that
Now, using the definition of \(G^c\) (dilution rate equal to growth rate) and (56), we have that
and (I) is proved.
For (II), using the definition of \(D^c\) and \(\mu _c\), and Lemma 3 part (c), we have
Now, using the fact that \(D \ge D^c\), from the previous equation, we conclude that \(D>\mu _c(0,\gamma )\). This completes the proof. \(\square \)
B Algorithm to find the coexistence equilibrium
Let \(D^a\) and \(D^p\) be given by Proposition 1. From now on, we assume that \(D \in (D^a, D^p)\), otherwise there is no coexistence equilibrium (see Theorem 1). The first step is to determine the equilibrium \(E^p=(B_p^p,0,G^p,A^p)\) given by Proposition 1. The instructions on how to do so are dictated by the proof of Proposition 1. Indeed, \(B^p_p\) is obtained as the solution of
with g defined by
with \(\alpha =(1-Y_{h/p})(Y_g-k_{over}Y_a)\), \(\beta _a=\gamma /D\), \(\beta _a=k_{over}(\gamma -l)/D\), and
Equation (57) has a unique solution on the interval \([0,G_{in}/\beta _g]\). Moreover, \(g(0)>0\) and \(g(G_{in}/\beta _g)<0\), which provides an interval to look for the solution. Thus, equation (57) can be easily solved, for example, with the solver fzero in MATLAB. The values of \(A^p\) and \(G^p\) are obtained from
We also need the value of \(G^c\), the glucose concentration associated with the equilibrium \(E^c\) given by Proposition 2. Let \(D^c\) be given by Proposition 2. If \(D<D^c\), then \(G^c\) is the unique solution of \((1-Y_{h/c})f_c(G,0)-k_{deg}-D=0\). This equation is easily solved explicitly. If \(D\ge D^c\), we will take \(G^c\) as \(G_{in}\). This is useful to distinguish the cases (b) and (c) in Theorem 1 (see Remark 7).
Now, we have to determine the coexistence equilibrium. The instructions on how to do so are dictated by the proof of Lemma 8:
-
1)
Determine \(c_1=(1-Y_{h/p})f_p(G^c,0)-k_{deg}-D\) and \(c_2=(1-Y_{h/c})f_c(G^p,A^p)-k_{deg}-D\).
-
2)
If \(c_1\le 0\) or \(c_2 \le 0\), then there is no coexistence equilibrium. The algorithm ends. However, if \(c_1\) and \(c_2\) are positive, go to the next step.
-
3)
Find \(A^* \in [0,A^p]\) as the unique solution of \(\phi (A)=0\), with \(\phi \) defined by
$$\begin{aligned} \phi (A)=(1-Y_{h/c})\left[ Y_{g} \beta \gamma +Y_a\left( k_a \frac{A}{A+K_a}d(\beta \gamma )+k_{Acs} \frac{A}{A+K_{Acs}} \right) \right] -k_{deg}-D, \end{aligned}$$with \(\gamma \) given by (58). We have that \(\phi (0)<0\) and \(\phi (A^p)>0\).
-
4)
Find \(G^*\in [0,G_{in}]\) as the unique solution of \(\varphi (G)=0\) with \(\varphi (G):=(1-Y_{h/p})f_p(G,A^*)-k_{deg}-D\). We have that \(\varphi (0)<0\) and \(\varphi (G_{in})>0\).
-
5)
Find \(B_p^*\) and \(B_c^*\) as the unique solution of the following linear system:
$$\begin{aligned} \left[ \begin{array}{cc} r_{up,p}^g(G^*,A^*) &{} r_{up,c}^g(G^*,A^*)\\ r_{over,p}^a(G^*,A^*) &{} r_{up,c}^a(G^*,A^*) \end{array}\right] \left[ \begin{array}{c} B_p \\ B_c \end{array}\right] =\left[ \begin{array}{c} D(G_{in}-G^*) \\ DA^* \end{array}\right] . \end{aligned}$$
C Algorithm to solve the MOP
Problem (15) is solved numerically with the interior point algorithm implemented in the toolbox fmincon of MATLAB (Byrd et al. 1999). To use fmincon, the objective function must be continuous on the feasible region, which must be defined through equalities and non-strict inequalities. In the following, we adapt (15) to use fmincon.
Following Theorem 1, the set \(\Omega \) defined in Sect. 4 can be described such that each element \(v=(Y_{h/p},Y_{h/c},D)\) on \(\Omega \) satisfies
where \(D^a\), \(D^p\), \(G^p\), and \(A^p\) are given by Proposition 1, \(D^c\) and \(A^c\) are given by Proposition 2, and \(G^c\) is given by (13). We define the set \(\bar{\Omega }\) such that each element \(v=(Y_{h/p},Y_{h/c},D)\) on \(\bar{\Omega }\) satisfies
The function \(\Phi \) defined in Sect. 4 is defined on \(\Omega \) and we must extend its definition on \(\bar{\Omega }\). If \(v \in \Omega \), we can determine the coexistence equilibrium using the algorithm presented in Appendix B, and therefore evaluate the function \(\Phi \) defined in Sect. 4. For \(v \in \bar{\Omega }-\Omega \), we solve the system (9)–(10) with an initial condition satisfying \(B_p(0), B_c(0)>0\). We run then the model in the long-term until an equilibrium is reached. We will denote the total protein concentration associated to this equilibrium by \({\hat{H}}^*(v)\). Thus, the extension of \(\Phi \) on \(\bar{\Omega }\), is given by:
The continuity of \(\hat{\Phi }\) can be observed in Figs. 3C and D. In the region of coexistence (shaded region), we have that \(v\in \Omega \), while on the boundary of this interval, we have that \(v\in \bar{\Omega }-\Omega \). In this situation, we observe a continuous transition between a coexistence equilibrium and a non-coexistence equilibrium as D approaches the boundary of this region.
We solve numerically the following problem instead of (15):
where
D Robust choice of the down-regulation function
Let us assume that in (9) the down-regulation function d is replaced by \({\hat{d}}\), defined in (6). We then run long-term simulations of (9)–(10) until an equilibrium is reached (which is always the case).Footnote 2 Then, we evaluate the the process yield and the productivity of the system at the time at which steady state has been reached. Figure 7 shows the result of this experiment for 1000 different values of \((Y_{h/p},Y_{h/c},D)\). As we can see, the POF obtained when d is given by (7) represents a good approximation of the POF when d is given by (6). This shows that the choice of d in this paper is adequate to study the model proposed by Mauri et al. (2020).
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Martínez, C., Cinquemani, E., Jong, H.d. et al. Optimal protein production by a synthetic microbial consortium: coexistence, distribution of labor, and syntrophy. J. Math. Biol. 87, 23 (2023). https://doi.org/10.1007/s00285-023-01935-3
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00285-023-01935-3