Approximating wild-type metabolism with FBrAtio
Obvious futile cycles allowing artificial NAD(P)+/NAD(P)H balancing were located and corrected, but the problems described above remained. Given the ability of the metabolic network to artificially balance NAD(P)+/NAD(P)H without using the acid/solvent production pathways, additional constraints and flux ratio constraints were implemented. First, the reactions involving the ferredoxins were further constrained for irreversibility. These reactions and their updated constraints are given in Table 1. Simulation of metabolism, given these and glucose uptake and SPF constraints are shown in Figure 2. In this simulation, the following results do not coincide with experimental observations: (i) succinate was produced in high levels during acidogenesis, (ii) the production of acetate and lactate far exceed butyrate production, (iii) only butyrate was re-consumed (negative flux values), (iv) the amount of butyrate re-consumed exceeded the amount of butyrate produced, (v) butanol was produced well before acetone, and (vi) hydrogen (H2) production fell to zero and rose during solventogenesis. These results show that during early production of butanol (before production of acetone), butyrate was consumed through the production pathway. This is consistent with previous findings
7
58
, and the thermodynamic reversibility calculations
54
insist the reactions catalyzed by the butyrate kinase (BK) and the phosphotransbutyrylase (PTB) (see Figure 1) remain reversible.
<p>Table 1</p>
Reactions and updated constraints involving the ferredoxins
Enzyme Name
Reaction
Lower Bound
Upper Bound
*The reduced ferredoxin (FdRed) and the oxidized ferredoxin (FdOx).
PFO: pyruvate ferredoxin oxidoreductase
C
o
A
+
p
y
r
u
v
a
t
e
+
F
d
Ox
→
C
O
2
+
a
c
e
t
y
l
−
C
o
A
+
H
+
+
F
d
Red
0
1000
HYDA: hydrogenase
2
H
+
+
F
d
Red
→
H
2
+
F
d
Ox
0
1000
FNO: ferredoxin NAD+ oxidoreducatase
N
A
D
+
+
H
+
+
F
d
Red
→
N
A
D
H
+
F
d
Ox
0
1000
FNPO: ferredoxin NADP+ oxidoreductase
N
A
D
P
+
+
H
+
+
F
d
Red
→
N
A
D
P
H
+
F
d
Ox
0
1000
<p>Figure 2</p>FBA results of wild-type metabolism using iCAC490
FBA results of wild-type metabolism using iCAC490. The model was simulated given (i) a glucose uptake rate of 10
m
m
o
l
h
r
·
g
D
C
W
, (ii) varied SPF, and (iii) constraints listed in Table 1. These results do not coincide with experimental observation.
![]()
A total of five flux ratio constraints were found necessary to generate simulations of metabolism that were qualitatively consistent with experimental observations. First, the butyrate to acetate uptake ratio was constrained according to previously published findings
8
. The assumption was made that the extracellular butyrate concentration was twice that of the extracellular acetate concentration. From the original relationship
8
, this meant that the flux of butyrate uptake through the CoAT, f(rCoAT, butyrate), to acetate uptake, f(rCoAT, acetate), was equal to 0.63, as shown in Equation 5. This flux ratio is referred to as f(rCoAT, butyrate):f(rCoAT, acetate).
f
r
C
o
A
T
,
b
u
t
y
r
a
t
e
:
f
r
C
o
A
T
,
a
c
e
t
a
t
e
=
f
C
o
A
T
,
b
u
t
y
r
a
t
e
f
C
o
A
T
,
a
c
e
t
a
t
e
=
0.63
Implementing this flux ratio alone led to high initial production of succinate, and hydrogen production increased during solventogenesis. Over-production of ethanol and lactate was also observed. So, a flux ratio was installed to direct the conversion of pyruvate to either lactate, through the lactate dehydrogenase (LDH), or to acetyl-CoA through the pyruvate ferredoxin oxidoreductase (PFO) (see Table 1). This flux ratio is shown as Equation 6. This ratio was set equal to 10 to coincide with published experimental observations that lactate production is much less than acetate and butyrate production
2
.
f
r
P
F
O
f
r
L
D
H
=
10
Simulations with these two flux ratios were characterized by (i) high acetate and ethanol production, (ii) low butyrate and butanol production, (iii) high initial secretion of succinate, and (iv) low initial production of hydrogen. The secretion of succinate was investigated next. This is indicative of high fluxes through the TCA cycle. No significant succinate export has been reported for C. acetobutylicum, so an optimized metabolic model must produce only minimal succinate (if any). In the iCAC490 model, production of oxaloacetate from pyruvate requires HCO-
3. When CO2 is produced, it is either (i) transported out of the cell or (ii) converted to HCO-
3. The fate of CO2 is determined by physicochemical properties of the intracellular environment and has a significant impact on intracellular metabolism. Thus, a ratio constraint for CO2 export against conversion was derived. This ratio is shown in Equation 7 and was set equal to 5 to approximate intracellular conditions. This value was chosen because it led to effective simulations. The physicochemical nature of this flux ratio constraint is currently under investigation.
f
C
O
2
e
x
p
o
r
t
f
C
O
2
c
o
n
v
e
r
s
i
o
n
=
5
This flux ratio constraint corrected hydrogen production (i.e., high hydrogen production during acidogenesis and reduced during solventogenesis) and minimized succinate secretion, which is consistent with experimental observations. This constraint also resulted in increased CO2 production during solventogenesis, relative to production during acidogenesis. However, metabolic activity was still characterized by (i) high acetate and ethanol production and (ii) low butyrate and butanol production (results not shown). To address this, flux ratios were constructed around the use of the acetyl-CoA metabolite pool in clostridial metabolism (see Figure 1). Acetyl-CoA can be utilized to (i) produce ATP through acetate production, (ii) regenerate NAD+ through ethanol production, or (iii) balance both ATP and NAD+ by producing butyrate. However, the shorter metabolic pathways result in acetate and ethanol production and accommodate minimizing the total flux of the metabolic network. To approximate wild-type clostridial metabolism, flux must proceed from acetyl-CoA through the thiolase (THL) with greater flux than through the phosphotransacetylase (PTA) towards acetate or through the bifunctional aldehyde/alcohol dehydrogenase (AAD). To ensure this, additional flux ratio constraints were derived. The ratio of metabolic flux through the THL relative to the PTA, f(rTHL):f(rPTA), was set equal to 2, and the flux through the THL relative to the AAD, f(rTHL):f(rAAD_1), was set equal to 10. This ensured the majority of acetyl-CoA was sent to the butyrate producing pathway while a greater amount of acetyl-CoA was converted to acetate than was converted to ethanol. These flux ratios are shown in Equations 8 and 9. Simulation results are shown in Figure 3. Due to the constant glucose uptake rate, solutions to the flux balance equation were only possible for SPF values greater than −17
m
m
o
l
H
+
h
r
·
g
D
C
W
. In a previous study
12
, much higher glucose uptake rates enabled flux solutions at SPF values approaching −55
m
m
o
l
H
+
h
r
·
g
D
C
W
. In the simulations shown in Figure 3, a roughly 1:1 production ratio of H2/CO2 was observed during exponential growth, and the production of H2 decreased in solventogenesis, while the production of CO2 increased. This is consistent with experimental observations
59
. Butyrate was produced in greater amounts than acetate, and both acids were taken up during solventogenesis. The production of butanol was greater than that of ethanol and was similar to that of acetone. The production of lactate was minimal, as was the secretion of succinate. The maximum specific growth rate of the culture occurred towards the end of acidogenesis but prior to solventogenesis onset. Thus, when the flux ratio constraints of Equations 59 were applied through FBrAtio, the iCAC490 genome-scale model was able to capture the major properties of wild-type metabolism. This was done for the first time without constraining acids and solvents production rates directly using constraints on transporters or exchange fluxes.
f
r
T
H
L
:
f
r
P
T
A
=
f
r
T
H
L
f
r
P
T
A
=
2
f
r
T
H
L
:
f
r
A
A
D
_
1
=
f
r
T
H
L
f
r
A
A
D
_
1
=
10
<p>Figure 3</p>FBrAtio results of wild-type metabolism using iCAC490
FBrAtio results of wild-type metabolism using iCAC490. The model was simulated model given (i) a glucose uptake rate of 10
m
m
o
l
h
r
·
g
D
C
W
, (ii) varied SPF, (iii) constraints listed in Table 1, (iv) unidirectional acetate and butyrate secretion by diffusion, and (v) f (rTHL) f (rPTA) = 2, (vi) f (rTHL): f (rAAD_1) = 10, (vii) f (CO2export): f (CO2conversion) = 5, (viii) f (rPFO): f (rLDH) = 10, (ix) f (rCoAT, butyrate):f (rCoAT, acetate) = 0.63. Results qualitatively fit wild-type metabolism.
![]()
asRNA knockdown of CoAT only
Previous research found that knockdown of the CoA transferase using asRNA technology resulted in a more favorable butanol to acetone selectivity. Although, overall reduced butanol yields were observed
50
, possibly due to simultaneous AAD knockdown. To determine if FBrAtio can predict of these findings, a flux ratio constraint was derived to simulate knockdown of the CoAT. It was assumed this knockdown led to a decreased flux of acetate and butyrate re-uptake. As shown in Figure 1, the CoAT converts acetoacetyl-CoA to acetoacetate while transporting acetate or butyrate into the cell and converting them to acetyl-CoA or butyryl-CoA, respectively. Acetoacetyl-CoA, on the other hand, can also be converted to β-hydroxybutyryl-CoA by the β-hydroxybutyryl-CoA dehydrogenase (BHBD). So, a flux ratio constraint was derived to specify the distribution of β-hydroxybutyryl-CoA utilized by the CoAT and the BHBD. This flux ratio is called f(rCoAT): f(rBHBD), where the flux through the CoAT includes fluxes for both acetate and butyrate re-uptake. In the wild-type simulations shown in Figure 3, given an SPF of 5
m
m
o
l
H
+
h
r
·
g
D
C
W
, the flux through the CoAT was 2.03
m
m
o
l
h
r
·
g
D
C
W
for acetate re-uptake and 1.28
m
m
o
l
h
r
·
g
D
C
W
for butyrate re-uptake (total CoAT flux of 3.31
m
m
o
l
h
r
·
g
D
C
W
). The wild-type flux through BHBD was 3.04
m
m
o
l
h
r
·
g
D
C
W
to yield a wild-type f(rCoAT): f(rBHBD) flux ratio of 1.09. The flux ratio is given in Equations 10. However, this flux ratio in the wild-type was a function of the SPF, and this relationship is shown in Figure 4. To simulate knockdown of the CoAT, two separate approaches were attempted. The first approach continued to treat this flux ratio as a function of the SPF. The flux ratio was constrained to values of (i) 75%, (ii) 50%, and (iii) 25% of the wild-type value, as shown in Figure 4. The second approach was to fix the flux ratio to a specified value for all values of the SPF. The flux ratios chosen were: (i) 1, (ii) 0.5, (iii) 0.1, and (iv) 0.01.
f
r
C
o
A
T
f
r
B
H
B
D
=
1.09
<p>Figure 4</p>The wild-type f(rCoAT):f(rBHBD) flux ratio
The wild-type f(rCoAT):f(rBHBD) flux ratio. Knockdowns of the CoAT of (i) 25%, (ii) 50%, and (iii) 75% are shown.
![]()
FBrAtio results with the f(rCoAT):f(rBHBD) flux ratios of Figure 4 are shown in Figure 5. In particular, results of acetone, butanol, and ethanol predictions are shown to focus on the impact of CoAT asRNA knockdown on solvent production. The SPF range shown is from −10 to 5
m
m
o
l
H
+
h
r
·
g
D
C
W
. Consistent with published experimental results
43
50
, simulated knockdown of the CoAT resulted in an increased butanol to acetone ratio. Significantly lower acetone production (~50% reduction) was predicted with increased butanol production (~25% increase) when CoAT activity was down-regulated by 75%. Significant about these simulation results is that the impact of “fine-tuning” an asRNA construct can be observed through simulations using flux ratio constraints.
<p>Figure 5</p>FBrAtio predictions of solvent production using the iCAC490 model and ratios of Figure 4
FBrAtio predictions of solvent production using the iCAC490 model and ratios of Figure4. The model was simulated given (i) a glucose uptake rate of
10
m
m
o
l
h
r
·
g
D
C
W
, (ii) varied SPF, (iii) constraints listed in Table 1, (iv) unidirectional acetate and butyrate secretion by diffusion, (v) f (rTHL): f (rPTA) = 2, (vi) f (rTHL): f (rAAD_1) = 10, (vii) f (CO2export): f (CO2conversion) = 5, (viii) f (rPFO): f (rLDH) = 10, (ix) f (rCoAT, butrate): f (rCoAT, acetate) = 0.63, and (x) f (rBHBD) flux ratios defined in Figure 4 as a function of the SPF. The following curves are shown: acetone (cyan), butanol (red), and ethanol (green).
![]()
It was important to determine whether the f(rCoAT):f(rBHBD) flux ratio must be represented as a function of the SPF (Figure 4). In further simulations, this flux ratio constraint was artificially set and held constant for all values of the SPF. FBrAtio results are shown in Figure 6 for flux ratios of: (i) 1, (ii) 0.5, (iii) 0.1, and (iv) 0.01. Similar trends were obtained. As the f(rCoAT): f(rBHBD) flux ratio decreased, metabolic flux was forced through the BHBD enzyme (rather than through CoAT), resulting in (i) decreased acetate/butyrate re-uptake, (ii) decreased acetone production, and (iii) increased flux through the butanol production pathway. The exaggerated flux ratio simulations of Figure 6 (ratios of 0.1 and 0.01) show the potential of effective asRNA or gene knockout of the CoAT. However, even though CoAT knockdown simulations resulted in predicted phenotypes with improved butanol to acetone selectivity, these phenotypes did not show overall reduced ethanol and butanol production. This suggests AAD knockdown may be required to impact alcohol yields, which is discussed in the next section.
<p>Figure 6</p>FBrAtio predictions of solvent production using the iCAC490 model and fixed ratios
FBrAtio predictions of solvent production using the iCAC490 model and fixed ratios. The model was simulated given (i) a glucose uptake rate of
10
m
m
o
l
h
r
·
g
D
C
W
, (ii) varied SPF, (iii) constraints listed in Table 1, (iv) unidirectional acetate and butyrate secretion by diffusion, (v) f (rTHL): f (rPTA) = 2, (vi) f (rTHL): f (rAAD_1) = 10, (vii) f (CO2export): f (CO2 conversion) = 5, (viii) f (rPFO): f (rLDH) = 10, (ix) f (rCoAT, butrate): f (rCoAT, acetate) = 0.63, and (x) constant f (rCoAT): f (rBHBD) flux ratios of 1, 0.5, 0.1, and 0.01. The following curves are shown: acetone (cyan), butanol (red), and ethanol (green).
![]()
Knockdown of both CoAT and AAD
In published research
43
50
, it was suspected that the aad gene was knocked-down when the asRNA for cftB (CoAT) was applied. The resulting strain produced very little ethanol or butanol. Since these two genes reside in the same operon, this hypothesis is valid. The previous results (Figures 5 and 6) showed that the butanol to acetone ratio was increased with CoAT knockdown. Simulations of the knockdown of both CoAT and AAD to defined levels using FBrAtio are shown in Figure 7. The AAD was knocked-down by the use of two flux ratio constraints. In these simulations, knockdown of AAD must be accommodated at two metabolic branch points (i) acetyl-CoA and (ii) butyryl-CoA (see Figure 1). To simulate the AAD knockdown, the ratio of flux diverted through AAD, relative to its other choices, was decreased by 80%. This factor was chosen based on the published effectiveness of the originally designed asRNA
50
. To incorporate the knockdown into flux ratio constraints, the f(rTHL):f(rAAD_1) flux ratio constraint was increased from a value of 10 to 50. To implement the flux ratio constraint at butyryl-CoA, a new f(rPTB):f(rAAD_2) flux ratio constraint was created. In wild-type simulations (see Figure 3), this flux ratio was unconstrained and had an average value of 0.52 at an SPF value of 0
m
m
o
l
H
+
h
r
·
g
D
C
W
and 0 at an SPF value of 5
m
m
o
l
H
+
h
r
·
g
D
C
W
(assumed value of 0.25 during solventogenesis). At highly negative values of the SPF, this ratio becomes infinite (since no butanol is produced). The f(rPTB):f(rAAD_2) was constrained to 1.25 (80% knockdown over 0.25) over all values of the SPF. The f(rTHL):f(rAAD_1) flux ratio constraint was held constant at 50 (also 80% knockdown). The CoAT was knocked-down 80% by adjusting the f(rCoAT):f(rBHBD) flux ratio to 0.2 over all values of the SPF. Results for solventogenesis (SPF = 5) are shown in Figure 7. Initial simulations revealed a large ethanol production that was not consistent with experimental findings. A closer inspection of genome-wide metabolic fluxes revealed several reactions were involved in futile cycles to produce excess acetaldehyde. To correct this, a new flux ratio constraint was created between AAD_1 and AAD_3 and is shown in Equation 11. This new flux ratio constraint was set to 1.2 since it is known that acetaldehyde is not produced exclusively through AAD_1 (although the exact ratio has not been measured). FBrAtio results given the f(rAAD_3):f(rAAD_1) flux ratio constraint yielded significantly reduced ethanol production. Results with 80% knockdown of CoAT and AAD still showed significant butanol production; however, this was significantly reduced (towards zero) as the percent knockdown was increased.
f
(
r
A
A
D
_
3
)
f
(
r
A
A
D
_
1
)
=
1.2
<p>Figure 7</p>FBrAtio predictions using the iCAC490 model given various levels of CoAT and AAD knockdown
FBrAtio predictions using the iCAC490 model given various levels of CoAT and AAD knockdown. Shown are predictions of growth (x10) (blue), acetone (cyan), butanol (red), and ethanol (green) during solventogenesis (SPF = 5). The following flux ratio constraints were applied: (i) f(rTHL):f(rAAD_1), (ii) f(rCoAT):f(rBHBD), (iii) f(rPTB):f(rAAD_2), and (iv) f(rAAD_3):f(rAAD_1). The following were held constant: (i) glucose uptake rate of
10
m
m
l
h
r
·
g
D
C
W
, (ii) constraints listed in Table 1, (iii) f(rTHL):f(rPTA) = 2, (iv) f(CO2export):f(CO2conversion) = 5, (v) f(rPFO):f(rLDH) = 10, and (vi) f(rCoAT, butyrate):f(rCoAT, acetate) = 0.63.
![]()