See discussions, stats, and author profiles for this publication at: [Link]

net/publication/268878227

Quantifying the effects of light intensity on bioproduction and maintenance

energy during photosynthetic growth of Rhodobacter sphaeroides

Article in Photosynthesis Research · November 2014

DOI: 10.1007/s11120-014-0061-1 · Source: PubMed

CITATIONS READS

14 255

5 authors, including:

Saheed Imam Colin M Fitzgerald

Synthetic Genomics University of Wisconsin–Madison
38 PUBLICATIONS 730 CITATIONS 7 PUBLICATIONS 192 CITATIONS

SEE PROFILE SEE PROFILE

Timothy J Donohue Daniel R Noguera

University of Wisconsin–Madison University of Wisconsin–Madison
280 PUBLICATIONS 7,563 CITATIONS 312 PUBLICATIONS 8,261 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Low dissolved oxygen nutrient removal View project

All content following this page was uploaded by Timothy J Donohue on 26 January 2015.

The user has requested enhancement of the downloaded file.

Photosynth Res (2015) 123:167–182
DOI 10.1007/s11120-014-0061-1

REGULAR PAPER

Quantifying the effects of light intensity on bioproduction

and maintenance energy during photosynthetic growth
of Rhodobacter sphaeroides
Saheed Imam • Colin M. Fitzgerald • Emily M. Cook •

Timothy J. Donohue • Daniel R. Noguera

Received: 1 April 2014 / Accepted: 19 November 2014 / Published online: 27 November 2014
Ó Springer Science+Business Media Dordrecht 2014

Abstract Obtaining a better understanding of the physi- light intensities. Photosynthetic maintenance energy
ology and bioenergetics of photosynthetic microbes is an requirements were also significantly affected by light
important step toward optimizing these systems for light intensity, with links to differences in biomass composition
energy capture or production of valuable commodities. In and the need to maintain redox homeostasis. Inclusion of
this work, we analyzed the effect of light intensity on the measured condition-dependent biomass and mainte-
bioproduction, biomass formation, and maintenance energy nance energy parameters and the measured photon uptake
during photoheterotrophic growth of Rhodobacter sph- rate into a genome-scale metabolic model for R. sphaero-
aeroides. Using data obtained from steady-state bioreactors ides (iRsp1140) significantly improved its predictive per-
operated at varying dilution rates and light intensities, we formance. We discuss how our analyses provide new
found that irradiance had a significant impact on biomass insights into the light-dependent changes in bioenergetic
yield and composition, with significant changes in photo- requirements and physiology during photosynthetic growth
pigment, phospholipid, and biopolymer storage contents. of R. sphaeroides and potentially other photosynthetic
We also observed a linear relationship between incident organisms.
light intensity and H2 production rate between 3 and
10 W m-2, with saturation observed at 100 W m-2. The Keywords Photosynthesis  Maintenance energy 
light conversion efficiency to H2 was also higher at lower Rhodobacter sphaeroides  Metabolic modeling 
Bioenergetics  Hydrogen

Electronic supplementary material The online version of this

article (doi:10.1007/s11120-014-0061-1) contains supplementary Background
material, which is available to authorized users.

S. Imam Microbes hold great potential as industrial systems for

Program in Cellular and Molecular Biology, University of economical production of high-value commodities (Gro-
Wisconsin, Madison, WI, USA nenberg et al. 2013; Peralta-Yahya et al. 2012). To opti-
mize these biological systems for efficient bioproduction, it
S. Imam  T. J. Donohue
Department of Bacteriology, Wisconsin Energy Institute, is imperative we gain better knowledge of their physio-
University of Wisconsin – Madison, Suite 5166, 1552 University logical and bioenergetic processes, so improvements can be
Avenue, Madison, WI 53726-4084, USA made in culture conditions or strain design. Photosynthetic
microbes, with their ability to harness light energy, are the
S. Imam  C. M. Fitzgerald  T. J. Donohue  D. R. Noguera
DOE Great Lakes Bioenergy Research Center, University of major contributor to global carbon cycling and can be
Wisconsin, Madison, WI, USA instrumental in development of bioprocesses for sustain-
able energy and bioremediation (Atsumi et al. 2009; Gro-
C. M. Fitzgerald  E. M. Cook  D. R. Noguera (&)
nenberg et al. 2013; Peralta-Yahya et al. 2012).
Department of Civil and Environmental Engineering, University
of Wisconsin, Madison, WI, USA Rhodobacter sphaeroides is perhaps the most well-stud-
e-mail: noguera@[Link] ied photosynthetic bacterium and has been used as a model

123
168 Photosynth Res (2015) 123:167–182

system for investigating various aspects of photosynthesis production using constraint-based models, and therefore,
(Imam et al. 2011; Mackenzie et al. 2007). In addition to its can be crucial for downstream strain design efforts.
ability to grow under a variety of conditions (aerobic In this study, we aimed to (i) quantify the impact of light
respiratory, anaerobic respiratory, and anoxygenic photo- intensity on biomass formation, biomass composition,
synthetic) and utilize a diverse array of growth substrates biomolecule production, and cellular maintenance energy
(Imam et al. 2013; Mackenzie et al. 2007), R. sphaeroides is requirements in R. sphaeroides; and (ii) use parameters
also capable of producing significant quantities of valuable estimated from these analyses to improve the calibration of
commodities including H2 (Kim et al. 2008; Kontur et al. iRsp1140 by updating biomass reactions and maintenance-
2011; Yilmaz et al. 2010), polyhydroxybutyrate (PHB) associated parameters. Using a series of chemostats run at
(Khatipov et al. 1999), and ubiquinone (Kien et al. 2010), as varying dilution rates, we observed that biomass compo-
well as being able to fix CO2 and N2 (Atsumi et al. 2009; sition, as well as production of several target biomolecules
Connor and Atsumi 2010; Masepohl and Hallenbeck 2010; (lipid, PHB, and H2) changed with variation in light
Wahlund et al. 1996). While several aspects of R. sph- intensity. We also found that the photosynthetic mainte-
aeroides metabolic processes have been studied in detail, no nance energy requirements were also significantly affected
detailed systems-level analysis of light absorption and its by changes in light intensity. The data were useful to
effects on bioenergetics and physiology has been conducted. understand the impact of light intensity on growth and
In recent years, constraint-based modeling has proved to metabolic flux distributions. Furthermore, the R. sphaero-
be a very useful tool for the systems-level analysis of ides metabolic model, iRsp1140, updated with the experi-
cellular metabolism, yielding new insights into condition- mentally determined biomass and maintenance energy
dependent metabolic flux distributions and cellular physi- parameters, was used to evaluate the impact of the model
ology. Furthermore, constraint-based analysis can guide improvements in the prediction of biomolecule production.
metabolic engineering efforts to optimize the production of The incorporation of condition-dependent biomass and
cell biomass and metabolites of interest (Oberhardt et al. maintenance energy parameters significantly improved the
2009). Previous work led to the construction and refine- predictive capability of iRsp1140. Overall, these analyses
ment of high-quality genome-scale metabolic models for R. provide new insights into the light-dependent changes in
sphaeroides, iRsp1095 (Imam et al. 2011) and iRsp1140 bioenergetic requirements and cellular behavior during
(Imam et al. 2013), which have been used for analysis of its photosynthetic growth of R. sphaeroides.
metabolic processes (Carapezza et al. 2013; Imam et al.
2011, 2013). However, previous application of these and
many other models of photosynthetic organisms have, with Results and discussion
a few exceptions (Vu et al. 2012; Kliphuis et al. 2011), not
considered measured light uptake as an important modeling The ability to efficiently convert available nutrient and
constraint, generally relying on arbitrary thresholds and energy sources into cell biomass or economically valuable
estimates (Montagud et al. 2010; Rex et al. 2013; end products is a desirable characteristic for microbes to be
Dal’Molin et al. 2011; Nogales et al. 2012). used in bioproduction. Thus, an evaluation of biomass and
An important factor in determining the efficiency of a product yields is key to determining the suitability of can-
cell for bioprocessing is the amount of metabolic energy didate systems. Previous efforts at modeling R. sphaeroides
utilized for biomass production and maintenance of cellular metabolism resulted in the formulation of iRsp1140 (Imam
homeostasis (Stouthamer and van Verseveld 1987; Russell et al. 2013). As with any model, its predictive ability depends
and Cook 1995). In general, an ideal microbial cell factory on good calibration to experimental observations. Since light
would have low maintenance energy demands, allowing for intensity can influence metabolic processes in photosynthetic
the direction of more of its metabolic energy toward pro- organisms, we studied its effects on biomass yield and
duction of commodities of interest (Stouthamer and van composition, biomolecule production, and ATP requirements
Verseveld 1987; Tannler et al. 2008). There are several for maintenance to better understand the impact of light on R.
studies of how changes in light intensity impact growth of sphaeroides physiology. We also aimed to obtain light-
R. sphaeroides and other phototrophic microbes in batch or dependent parameters from these analyses to better constrain
continuous culture (Aiking and Sojka 1979; Schumacher iRsp1140 and thereby improve its predictions.
and Drews 1979; Golecki et al. 1980; Oelze 1988; Biel
1986; Campbell and Lueking 1983). However, the steady- Quantifying the effects of light and nutrient availability
state maintenance energy requirements for R. sphaeroides on biomass yield
at different light intensities have not been determined.
Knowledge of energy maintenance parameters is needed To assess the impact of light intensity on R. sphaeroides
for the accurate prediction of biomass and metabolite physiology and bioenergetics, we set up bioreactors

123
Photosynth Res (2015) 123:167–182 169

irradiated at low (3 and 5 W m-2), moderate (10 W m-2), majority of the supplied light was not being used by the
and high (100 W m-2) light intensities (corresponding cells.
to *20, 33.5, 66.9, and 669 lmol photons m-2 s-1, The biomass yields on light energy (Yx,E) (i.e., the amount
respectively, at 800 nm). At each light intensity, multiple of biomass formed per mol of photons supplied) were cal-
chemostats were run at retention times ranging between 8 culated from the inverse of the slopes in Fig. 1a (Zijffers
and 39 h (i.e., dilution rates between 0.026 and 0.125 h-1). et al. 2010) to be 0.0016, 0.0015, and 0.0012 gDW mmol
Cells were harvested to determine biomass and PHB photons-1 at 3, 5, and 10 W m-2, respectively. These rep-
compositions. The uptake rates for the components in the resent high biomass yields that are close to theoretical
media with highest concentration, succinate, glutamate, maximum biomass yields predicted for green algae (Zijffers
aspartate, and sulfate were empirically determined, while et al. 2010), and are similar to those reported for other
H2 production was measured prior to harvesting cells. The phototrophs (Kliphuis et al. 2011; Zijffers et al. 2010),
specific light supply rate (SLSR), in mmol photons though the growth mode used here is photoheterotrophic as
gDW-1 h-1, was also calculated from the measured bio- opposed to photoautotrophic in the other studies. The lower
mass and light intensities (Table S1). biomass yield observed at 10 W m-2 could reflect the
nitrogen source becoming limiting at this light intensity,
Effect of light energy on biomass yield which was not the case at the lower intensities (Table S1).
Alternatively, it could indicate a drop in efficiency of light
During photoheterotrophic growth, the cell’s ability to utilization. In addition, the average biomass yield on light at
produce biomass is dependent on both available light and 100 W m-2 was 0.00013 gDW mmol photons-1, which is
organic substrate(s). Under light-saturating conditions, only *10 % of that observed at lower to moderate light
biomass yield would be expected to be dependent on intensity. This again is consistent with light saturation at this
organic substrate availability. Conversely, under light intensity.
limiting (and nutrient-replete conditions), biomass yield While most of the supplied light, under sub-saturating
would be a function of light availability. To determine how conditions, is apparently diverted toward biomass forma-
varying light intensities affect growth yields, we assessed tion, cells also require additional energy for non-growth-
the relationship between SLSR and the dilution rate associated processes including maintenance of cellular
(D) (i.e., the specific growth rate). We observed a linear homeostasis. This additional energy requirement is
relationship for reactors illuminated at 3, 5, and 10 W m-2, obtained from the y-intercept in regression plots of biomass
with R2 of 0.9, 0.95, and 0.8 (p values of 0.0016, 0.024, and yield on light (Fig. 1a) (Zijffers et al. 2010). At sub-satu-
0.0018), respectively (Fig. 1a), indicating that the supplied rating light intensities, the light-derived maintenance
light was a limiting factor on steady-state growth. How- energy requirements were 5.4, 0.83, and 16.7 mmol pho-
ever, there was no correlation between the SLSR and D at tons gDW-1 h-1 at 3, 5, and 10 W m-2, respectively.
100 W m-2 (R2 = 0.007) (Fig. S1a), indicating that light These values suggest that a greater amount of energy is
is saturating at this intensity and suggesting that the required for cellular maintenance at higher light intensities.

a 120 b 2500
y = 838.65x + 16.708
Cell dry weight (mgCOD/L)
(mmol photons gDW-1 h-1)
Light supply rate (SLSR)

100 R² = 0.8
2000
R² = 0.92
80
1500
3 W m-2
y = 632.01x + 5.4468 3 W m-2
60 R² = 0.92
R² = 0.9 10 W m-2 1000 10 W m-2
40 y = 651.76x + 0.8287 5 W m-2
5 W m-2
R² = 0.95
500
20 R² = 0.8

0 0
0 0.05 0.1 0.15 0 0.02 0.04 0.06 0.08 0.1 0.12
Dilution rate (h-1) Dilution rate (h-1)

Fig. 1 Relationship of dilution rate to SLSR and biomass production. (green-filled triangle), and 10 (red-filled rectangle) W m-2. There
Regression of a specific light supply rate (SLSR) against dilution rate was no statistically significant correlation between these measure-
D and b cell dry weight against D, at 3 (blue-filled diamond), 5 ments at 100 W m-2 (Fig. S1)

123
170 Photosynth Res (2015) 123:167–182

This could be a reflection of increased turnover of cellular strong correlation between nutrient uptake and biomass
components induced by light-dependent stress or an observed under these conditions.
increased amount of energy diverted toward energy-
dependent maintenance of redox homeostasis via hydrogen Effect of light intensity on steady-state biomass
production. composition
We also observed a high correlation between the mea-
sured cell dry weight (DW) and D for reactors run at sub- Previous analyses of batch cultures have shown that light
saturating light intensities, with R2 values ranging from 0.8 intensity significantly impacts R. sphaeroides physiology.
to 0.92 (Fig. 1b). The observed negative correlation In particular, the amount of light-capturing pigments pro-
between cell mass and D is consistent with cells growing duced increases significantly with decreases in incident
under limiting conditions, as higher dilution rates result in light intensity (Aagaard and Sistrom 1972; Chory and
less dense cultures. In these cultures, organic substrates in Kaplan 1983). However, as these analyses were not con-
the media were abundant (Table S1), indicating that light ducted under steady-state conditions, we assessed the
intensity was the main limiting factor under these condi- impact of changing light regimes on R. sphaeroides bio-
tions. At the saturating light intensity of 100 W m-2, we mass composition in continuous culture.
did not observe a significant correlation between biomass Incident light intensity had only a small effect on cel-
and D (Fig. S1b), an observation that suggests light is not lular protein content, which was slightly higher at low light
limiting under these conditions. These data indicate that intensities (3 and 5 W m-2) compared to moderate and
light is the limiting factor on biomass yield at 3, 5, and high light intensities (10 and 100 W m-2) (Fig. 2a). On the
10 W m-2 and that R. sphaeroides is capable of achieving other hand, light intensity had a significant, and previously
relatively high biomass yields under these conditions. unreported, impact on PHB production in R. sphaeroides.
These data also indicate that as the incident light increases, At low light intensities (3 and 5 W m-2), PHB made up
an increasing amount of light energy is devoted toward only *3 % of the DW, which was more than fourfold less
non-growth-associated processes, which could include than that observed at moderate and high light intensities,
maintenance of redox poise and coping with light-depen- where PHB accounted for 13–14.8 % of the total cell DW,
dent stress. Alternatively, some supplied light might be respectively (Fig. 2b). PHB serves as a major route for
unutilized or dissipated through mechanisms for non-pho- dissipating excess electrons in R. sphaeroides (Yilmaz
tochemical quenching. et al. 2010), as its production requires significant input of
reducing power in the form of NADPH, which is mainly
generated via the energy-dependent transhydrogenase
Effect of nutrient availability on biomass yield reaction (Imam et al. 2013). Thus, the low levels of PHB
formed at low light intensities may reflect the energy
In contrast to the relationships between SLSR and D, the limitations under these conditions.
correlation between nutrient uptake and D was poor at sub- Consistent with previous observations in batch cultures,
saturating light intensities (Fig. S2). There was very little the cellular bacteriochlorophyll (Bchl) content was
correlation between uptake of succinate (the major carbon dependent on incident light intensity. The cellular Bchl
source) or sulfate and D at these light intensities, indicating content was 0.71 and 0.91 % of DW at 3 and 5 W m-2,
that neither of these was a limiting nutrient. However, respectively. In contrast, a significant drop in Bchl content
glutamate (the main nitrogen source) uptake showed a was observed at moderate (0.53 %) and high (0.17 %) light
better correlation to D (R2 [ 0.6) at these light intensities intensities, representing an *5 fold difference between
(Fig. S2). On the other hand, there were high correlations high and low light conditions (Fig. 2c). These observations
between uptake of all nutrients measured and D at an are consistent with the need for more photopigments at
incident light intensity of 100 W m-2 (R2 [ 0.93) (Fig. lower light intensities to absorb the limited amount of
S2). The observed biomass yields from carbon uptake incident light.
(Yx,C) at 5, 10, and 100 W m-2 were 0.64, 0.61, and During photosynthetic growth, R. sphaeroides signifi-
0.57 gCODx gCOD-1 c , respectively (Fig. S3). However, cantly increases its membrane surface area by forming
the fit was much poorer at 3 W m-2, indicating organic intracytoplasmic membranes to house its photosynthetic
substrates were not limiting. Overall, these observations apparatus (Kiley and Kaplan 1988). While varying light
are consistent with a model in which incident light intensity intensity resulted in significant changes in Bchl content, the
is the main constraint on growth at sub-saturating light relative magnitude of the change in cellular phospholipid
intensities of 10 W m-2 and below. Under saturating light levels was smaller. The highest amount of phospholipids
conditions (100 W m-2), the major limitation to growth was obtained for cells grown at 5 W m-2 (*6 % of DW),
becomes the rate of organic substrate utilization; hence, the while the lowest was observed for cell grown at

123
Photosynth Res (2015) 123:167–182 171

a 80 b
16
% Cell dry weight

% Cell dry weight

60
12

40
8

20 4

0 0
3W m-2 5W m-2 10 W m-2 100 W m-2 3 W m-2 5 W m-2 10 W m-2 100 W m-2
Light intensity Light intensity

c 1.0 d 7

% Cell dry weight

0.8
% Cell dry weight

0.6 4

0.4 3

2
0.2
1

0.0 0
3 W m-2 5 W m-2 10 W m-2 100 W m-2 3 W m-2 5 W m-2 10 W m-2 100 W m-2
Light intensity Light intensity

Fig. 2 Biomass composition is significantly affected by light inten- represent the average amounts of the various biomass components
sity. Assessment of the contribution of a protein, b PHB, c bacterio- across all chemostats run at each light intensity. The error bars
chlorophyll, and d phospholipids to cell dry weight at 3, 5, 10, and represent the standard error of mean of all chemostats at each light
100 W m-2 light intensities. The values plotted in the bar charts intensity

100 W m-2 (*4 % of DW), only a 1.5-fold difference production helps maintain redox homeostasis by serving as
(Fig. 2d). The light-dependent changes in biomass com- a sink for excess reducing equivalents (Imam et al. 2011;
position identified here are useful for generating condition- Kontur et al. 2012; Yilmaz et al. 2010). Given the high
dependent biomass reactions (Table S2) for simulation of energy demands of this process, the amount of H2 produced
growth and flux distributions at different light intensities. might be altered by changing light intensity. We observed a
positive linear relationship between light intensity and the
Effect of light intensity on metabolite secretion amount of H2 produced. At low (3–5 W m-2) to moderate
(10 W m-2) light intensities, the increase in H2 production
In addition to determining the effect of light intensity on R. appeared to be directly proportional to incident light,
sphaeroides biomass composition, we also analyzed its increasing at a rate of *0.25 mmol gDW-1 h-1 of H2
effects on selected metabolic end products at steady state. production per unit of irradiance (*0.1 ml h-1 with the
experimental conditions used; R2 [ 0.99; Fig. 3a). This
H2 production is dependent on light intensity observation indicates that light is a limiting factor for
steady-state H2 production, consistent with previous
Rhodobacter sphaeroides is capable of producing large observations in batch cultures (Miyake and Kawamur
amounts of H2 under photoheterotrophic conditions via 1987), and it predicts that cells are using a proportional
ATP-dependent nitrogenase activity (Kontur et al. 2011; amount of the incident light for H2 production over this
Masepohl and Hallenbeck 2010; Yilmaz et al. 2010). H2 range. While there was an *1.7-fold increase in the

123
172 Photosynth Res (2015) 123:167–182

a 4 b 6

H2 production (mmol gDW -1 h-1)

H2 production (mmol gDW -1 h-1) 5
3
4

2 3

y = 0.2454x + 0.4034
R² = 0.9994 2
1
1

0 0
0 5 10 15 0 50 100
Incident light (W m-2) Incident light (W m-2)

Fig. 3 Relationship of H2 production rate to light intensity. a Regres- extended to 100 W m-2, showing loss of linear relationship due to
sion of average H2 production rate against light intensity between 3 light saturation
and 10 W m-2. b Plot of average H2 production against light intensity

amount of H2 produced when comparing moderate Light intensity alters end product profiles
(10 W m-2) to high (100 W m-2) light, the increase was
not proportional to light intensity, providing another In addition to H2, R. sphaeroides excretes soluble
indication of light saturation at 100 W m-2 (Fig. 3b), microbial products (SMP), such as TCA cycle interme-
and suggesting that H2 production had reached its diates, lactate, and pyruvate, when growing in batch
maximum rate [an average of *4.9 mmol gDW-1 h-1 cultures (Yilmaz et al. 2010). To determine if the amount
(or *1.28 ml h-1)] under these conditions. of SMPs produced during steady-state growth was affec-
The efficiency of light energy conversion is another ted by light intensity, we compared the measured chem-
useful parameter for assessing the effectiveness of photo- ical oxygen demand (COD) of the spent media to its
biological H2 production (Koku et al. 2002; Miyake and calculated COD (based on measured extracellular con-
Kawamur 1987). We found that the light energy conversion centrations of succinate, glutamate, and aspartate; see
efficiency of steady-state cultures was relatively high at ‘‘Materials and methods’’ section). For chemostats oper-
low and moderate light intensities with average efficiencies ated at low light intensities (3 and 5 W m-2), the mea-
of 4.75, 6.35, and 5.77 % at 3, 5, and 10 W m-2, respec- sured and estimated COD were within 10 % of each
tively. The energy conversion efficiencies obtained at these other. Conversely, at moderate to high light intensities (10
light intensities are in general agreement with reported and 100 W m-2), the difference in measured and esti-
values for cells grown in batch cultures (Koku et al. 2002; mated COD was *20 %. When we analyzed the spent
Miyake and Kawamur 1987). On the other hand, at reactor media for organics, only small amounts of pyru-
100 W m-2, the light conversion efficiency dropped down vate, lactate, glycerol, xylitol, and xylose were observed.
to just 0.77 %, consistent with light being in excess. Thus, the chemical identity of most of the SMPs in
In addition, to the light conversion efficiency, another steady-state cultures remains unresolved. In addition, the
important parameter is the substrate conversion efficiency levels of these metabolites did not change significantly
(i.e., ratio of observed H2 produced to the theoretical across the reactors. The increased extracellular COD
maximum obtainable with the main growth substrate, observed at higher light intensities could reflect altered
succinate in our chemostats) (Koku et al. 2002; Sasikala metabolic flexibility in these cultures, where conserving
et al. 1993). At 3 W m-2, the substrate conversion effi- metabolic energy by nutrient oxidation is less of a con-
ciency was *21 %, while the efficiencies significantly straint on the cells due to the increased availability of
increased at higher light intensities—40, 34 and 34 % at 5, light energy. Alternatively, it could be an indication of
10, and 100 W m-2, respectively. This suggests a signifi- increased cellular stress or photo-damage at higher light
cant energy limitation at 3 W m-2 that negatively impacts intensities, with the increased extracellular COD coming
the cell’s ability to convert the growth substrate into H2. from secretion of cellular constituents into the media.

123
Photosynth Res (2015) 123:167–182 173

Effect of light intensity on maintenance energy our analysis, these differences did not appear to be
requirements significant.

The energy a cell expends for growth and for maintaining Maintenance energy requirements vary with light intensity
cellular homeostasis, is an important factor in determining
its performance in the laboratory or in industrial settings At light intensities of 3 and 5 W m-2, we estimated the
(Stouthamer and van Verseveld 1987). Thus, we assessed GAM to be 282 and 297 mmol ATP gDW-1, respectively,
the maintenance energy requirements of R. sphaeroides at while NGAM was estimated to be 2.6 and 0.5 mmol ATP
varying light intensities to gain a better understanding of its gDW-1 h-1, respectively (Fig. 4b). Overall, there did not
bioenergetic requirements and limitations. appear to be significant differences between these param-
In metabolic flux models, maintenance energy is divi- eters at these two light intensities, which might be expected
ded into two major components: the growth-associated given the small difference in light intensities and the
maintenance (GAM) and non-growth-associated mainte- similarities in steady-state biomass composition under
nance (NGAM) energy requirements, used for biomass these conditions (Fig. 2).
formation and maintenance of cellular homeostasis, At a light intensity of 10 W m-2, the calculated GAM
respectively (Thiele and Palsson 2010). For photosyn- was 386 mmol ATP gDW-1, which is significantly higher
thetic organisms, the total amount of ATP predicted to be than that obtained at lower light intensities (Fig. 4b). This
produced will depend on the efficiency with which the likely reflects the significant differences in biomass com-
incident light is utilized by the cells. This efficiency of position between cells grown under these conditions. Fur-
light utilization (U), which is dependent on the light thermore, the estimated NGAM was 8.4 mmol
absorbing capacity of the photosynthetic apparatus, can ATP gDW-1 h-1, significantly higher than that observed
also be significantly impacted by cell shading, back at the two lower light intensities. This larger NGAM value
scatter of light by cells, and reactor configuration (depth, could be the result of increased stress (heat, photochemical,
etc.). Since U is an unknown parameter that influences the or others) at higher light intensities. Alternatively, this
calculation of maintenance energy requirements (Table difference could reflect the increased ATP required to
S3), we used iRsp1140 to computationally assess its maintain homeostasis, for instance, via ATP-dependent H2
impact on modeling predictions and to select an appro- production (Kontur et al. 2011), which is significantly
priate value. higher at 10 W m-2. Incorporating ATP-dependent H2
We constrained iRsp1140 using (i) the determined production constraints into iRsp1140 with the calculated
condition-dependent biomass compositions at varying light GAM and NGAM parameters showed that requiring addi-
intensities; (ii) the measured uptake rates for succinate, tional ATP for this process significantly deteriorates the
glutamate, aspartate, and sulfate; and (iii) the observed model’s predictions as light (and thus ATP) is limiting
steady-state growth rates to predict maximum ATP (Fig. S5). This suggests that ATP utilized for H2 production
hydrolysis rates at candidate Us of 10, 25, 50, 75, and is already accounted for in the estimated maintenance
100 %, and at incident light intensities of 3, 5, and energy parameters.
10 W m-2. Since the relationship between maximum ATP As expected, the iRsp1140-derived maintenance energy
hydrolysis and observed growth rate is expected to result in values are very similar to the related, previously estimated,
a good linear fit, we selected U = 75 % as this produced light-derived maintenance energy values obtained from
the best fit for the sub-saturating light intensities (Fig. S4). regressing SLSR on D (Fig. 1a). Given that two photons
Light utilization efficiencies of 10 and 25 % resulted in lead to generation of *1 ATP during cyclic photophos-
significantly poorer fits between the predicted ATP phorylation (Allen 2003), the GAM (NGAM) calculated
hydrolysis rates and l, whereas 75 % efficiency provided from that analysis is equivalent to 316 (2.7), 326 (0.4), and
marginally better fits than 50 and 100 % (Fig. 4a, Fig. S4). 419 (8.4) mmol ATP gDW-1 (mmol ATP gDW-1 h-1) at
Interestingly, the selected value of U is similar to the 3, 5, and 10 W m-2, respectively, very similar to that
maximum light utilization efficiency of 80 % estimated for predicted by iRsp1140. The small differences between
algal cells (Kliphuis et al. 2011). Nevertheless, given that these values likely result from aspects of ATP utilization
the goodness of fit was relatively insensitive to U between accounted for elsewhere in iRsp1140 (such as charging of
50 and 100 %, the selected U should not be interpreted as a tRNAs with amino acids), as well as the inclusion of
unique descriptor of the physiological state of the cells. It nutrient-derived ATP in the model’s predictions and the
should also be noted that U can vary with light intensity rotational fold symmetry of R. sphaeroides ATP synthase.
used, as efficiency might be expected to drop at light Overall, this indicates that the vast majority of energy used
intensities that exceed the photosynthetic capacity of the for growth and maintenance of cellular homeostasis during
cell. However, at the sub-saturating light intensities used in photoheterotrophic growth is derived from photon

123
174 Photosynth Res (2015) 123:167–182

a 1 b 60

0.9 y = 386.1x + 8.4

50

Max. ATP hydrolysis

Goodness of fit (R2) 0.8 R² = 0.74

(mmol gDW-1 h-1)

0.7 40
0.6
0.5 3 W m-2 30 3 W m-2
5 W m-2 y = 282.5x + 2.6
0.4 10 W m-2
R² = 0.86
10 W m-2 20 5 W m-2
0.3
y = 297.6x + 0.5
0.2
R² = 0.93
10
0.1
0
0 25 50 75 100 0
0 0.05 0.1 0.15
% Efficiency of light utilization
Specific growth rate μ (h-1)

Fig. 4 Assessing photosynthetic maintenance energy in R. sphaero- at 3 (blue-filled diamond), 5 (green-filled triangle), and 10 (red-filled
ides. a Comparison of the goodness of fit (i.e., R2 values) of the rectangle) W m-2 light intensity. The best fit obtained across all data
regression of maximum ATP hydrolysis on specific growth rate l was at 75 % efficiency. b Regression of maximum ATP hydrolysis
considering a range of efficiencies of light utilization (i.e., 10, 25, 50, against specific growth rate l at 3 (blue-filled diamond), 5 (green-
75, and 100 % of measured light intensity) in each simulation. The filled triangle), and 10 (red-filled rectangle) W m-2 using a U of
analysis was conducted with data obtained from chemostats operated 75 %

absorption, leaving energy in organic substrates to be uti- photons being supplied in excess under these conditions
lized for bioproduct formation. (Fig. 5a).
We could not use similar methods to accurately calcu- The above analysis also suggests that LUR can have a
late maintenance energy requirements at 100 W m-2 as a significant impact on the observed and predicted behavior
plot of maximum ATP hydrolysis versus growth rate under of photosynthetic cells. Thus, if LUR is ignored or inac-
these conditions did not show any significant correlation curately measured, it could have a negative effect on the
(Fig. S6). This observation might reflect the fact that light validity of computational predictions. Indeed, plotting the
is saturating at this intensity, so only a small fraction of predicted optimal growth rates against observed specific
incident light is being used by cells under these conditions. growth rates (i.e., dilution rate), when LUR and all mea-
However, utilizing more conservative estimates of light sured uptake and production rates are taken into account
absorption efficiency did not improve these results (Fig. (Fig. 5b), and comparing this to a plot of predictions
S6). An alternative explanation is that there is increased generated by assuming that light is not limiting (i.e.,
cellular damage from light, heat, photochemical stress, and omitting LUR), showed that inclusion of the LUR signifi-
other factors at high light intensities, which would com- cantly improves the predictions of iRsp1140 (Fig. 5c). The
plicate attempts to estimate maintenance energy under fit between predicted and observed growth rates improves
these conditions. from an R2 of 0.63 (p = 2.9 9 10-5) to 0.85
(p = 7.2 9 10-9) with the inclusion of the measured LUR.
Improvements in model predictions Only cultures grown at low to moderate light intensities
(3–10 W m-2) were considered for this analysis since at
Our comprehensive analysis of light effects on biomass saturating light (100 W m-2), the SLSR is not an adequate
yield, composition, and maintenance energy requirements estimation of optimal LUR (Fig. 5a). It should also be
indicates that, at low and moderate light intensities, pho- noted that inclusion of the experimentally determined
tons are a significant growth limiting factor, while this is parameters for GAM and NGAM also significantly
not the case at high light intensities. Consistent with those improved the predictions of iRsp1140 compared to using
observations, iRsp1140 predicts that at low to moderate the previously estimated maintenance energy requirements
light intensities (3–10 W m-2), the measured SLSR and (Imam et al. 2011) (Fig. 5d).
the specific light uptake rate (LUR) required to achieve Since the determined GAM and NGAM parameters
optimal growth (optimal LUR), given the substrate uptake were estimated using all of the data generated in this study
rates, are comparable (Fig. 5a). In most cases, the mea- (at 3, 5, and 10 W m-2), a cross-validation of these
sured SLSR is less than the predicted optimal LUR, con- parameters was carried out (Geisser 1975; Stone 1974). For
sistent with photons being a growth limiting factor. each light intensity, we re-computed the GAM and NGAM
Conversely, the measured SLSR was about 10 times the values using all but one of the chemostat measurements for
predicted optimal LUR at 100 W m-2, consistent with that light intensity. We then used the newly computed

123
Photosynth Res (2015) 123:167–182 175

5 W m-2
a 10000 b 0.12
R² = 0.851
Light (mmol photons gDW-1 h -1) 3 W m-2 10 W m-2 100 W m-2

Predicted growth rate (h-1)

0.1
1000
0.08

100 0.06

0.04

10
0.02
Optimal LUR
SLSR 0
1
0 0.05 0.1 0.15
Increasing growth rate
Observed specific growth rate μ (h-1)

c 0.12 d 0.12
R² = 0.6303 R² = 0.6256

Predicted growth rate (h-1)

0.1
Predicted growth rate (h-1)

0.1

0.08 0.08

0.06 0.06

0.04 0.04

0.02 0.02

0 0
0 0.05 0.1 0.15 0 0.05 0.1 0.15
Observed specific growth rate μ (h-1) Observed specific growth rate μ (h-1)

Fig. 5 Assessing the impact of SLSR, LUR, and maintenance energy growth rates using all measured parameters including LUR (b) or
on modeling predictions. a Plot comparing the measured SLSR and excluding LUR (c) as a constraint for FBA predictions. d Comparison
the specific LUR predicted by iRsp1140 to be required for optimal of iRsp1140 growth rate predictions to the observed growth rates
growth via FBA (optimal LUR), given the measured substrate uptake using previously estimated maintenance energy parameters (Imam
rates. Comparison of the observed and iRsp1140 predicted specific et al. 2011)

GAM and NGAM values as input for the model, which was 0.00015 respectively). Thus, within the limits of repro-
then constrained with uptake data from the chemostat that ducibility of experimental data generated in this study, the
was left-out and used to predict the growth rate for that model updated with the newly estimated GAM and
experiment. This rotation estimation analysis was con- NGAM, and constrained with LUR, shows improved pre-
ducted for all data points at each light intensity (Fig. S7). dictive capabilities.
While we observed an R2 reduction from 0.85 to 0.80 We also assessed iRsp1140’s ability to predict the pro-
(Fig. 5b and Fig. S7), the correlation was still significant duction rate of metabolites such as H2 and PHB, which
(p = 8.7 9 10-8), indicating the ability of the trained were measured as part of our analyses, but not used as input
model to reproduce the observation for experiments not for calculating GAM and NGAM. We observed a good
included in the training dataset. In addition, the mean correlation between the predicted and observed values
squared errors, calculated as the mean of the squared dif- (R2 = 0.53, p = 0.005) for simulations conducted with
ference between predicted and observed growth rates LUR included as a constraint (Fig. 6a). Conversely, when
across all light intensities, were smaller for both the cross- LUR was excluded, no significant correlation was observed
validation and final model predictions (0.0001 and between the measured and predicted H2 production rates
0.00008, respectively) compared to predictions without the (R2 = 0.001) (Fig. 6b). Furthermore, simulations con-
LUR constraint (i.e., Fig. 5c) or with the previously esti- ducted using the previously estimated biomass and main-
mated maintenance parameters (i.e., Fig. 5d) (0.0002 and tenance energy requirements (Imam et al. 2011) also

123
176 Photosynth Res (2015) 123:167–182

a b

Predicted H2 (mmol gDW -1 h-1)

Predicted H2 (mmol gDW -1 h-1)

8 25
R² = 0.53 R² = 0.001
6 20

15
4
10
2
5

0 0
0 2 4 6 0 2 4 6
Observed H2 (mmol gDW -1 h-1) Observed H2 (mmol gDW -1 h-1)
c d

Predicted H2 (mmol gDW -1 h-1)

Predicted H2 (mmol gDW -1 h-1)

25 25
R² = 0.0043 R² = 0.004
20 20

15 15

10 10

5 5

0 0
0 2 4 6 0 2 4 6
Observed H2 (mmol gDW -1 h-1) Observed H2 (mmol gDW -1 h-1)

Fig. 6 Prediction of H2 production. Plots comparing the observed parameters (Imam et al. 2011), with and without LUR as a constraint,
and iRsp1140 predicted maximum H2 production rate using all respectively. For each simulation, the observed growth rate was used
measured parameters including LUR (a) or excluding LUR (b) as a as an additional constraint, while H2 production was used as the
constraint for FBA predictions. c and d show the equivalent objective function
simulations using the previously estimated maintenance energy

showed poor correlation to observed H2 production rates, (Mahadevan and Schilling 2003). However, with the
regardless of whether LUR was used as a constraint or not. application of additional relevant constraints, one can sig-
These results point to an improved ability of the updated nificantly reduce the size of the solution space (Reed
instance of iRsp1140 to predict H2 production in R. sph- 2012). We observed that LUR, in particular, has a signifi-
aeroides and highlight the importance of using experi- cant impact on the solution space of iRsp1140 predictions.
mentally determined LUR as a constraint. Using flux variability analysis (FVA) (Mahadevan and
While inclusion of these newly estimated parameters Schilling 2003), we assessed the range of predicted flux
and LUR also significantly improved iRsp1140’s ability to values for each reaction at optimal growth rate under light-
predict PHB production rate (Fig. S8), the agreement limiting and -saturating conditions (i.e., with and without
between maximum predicted and observed production use of the specific LUR as an additional constraint). Using
rates was less than that for H2 production (R2 = 0.19). This all available nutrient uptake data from a representative
suggests that iRsp1140 is not yet sufficiently well con- chemostat operated at a light intensity of 10 W m-2, a total
strained to accurately predict the production rate of this of 795 reactions were able to carry flux under light-satu-
particular metabolite. rating conditions, compared to 448 when the LUR con-
straint was imposed. Of these 448 reactions, which carried
Impact of light limitation on the metabolic solution space flux under both conditions, a total of 342 reactions
(76.3 %) showed no variability in flux under light-limiting
While constraint-based analysis can be useful for studying conditions, compared to only 123 reactions (27.5 %) with
metabolism on a genome-scale, one of its limitations lies in invariable fluxes under light-saturating conditions. These
the large solution space of feasible flux distributions that results indicate that there is a significant reduction in the
can result from optimizing of a given objective function solution space when the LUR constraint is included.

123
Photosynth Res (2015) 123:167–182 177

3 A B C D E F G H I J K L M N O P Q

Predicted flux range (mmol gDW-1 h-1)

2

-1

-2

TCA cylce Glycolysis

Fig. 7 Variability of predicted flux distributions under light-limiting In both sets of simulations, the model was constrained with the uptake
and -saturating conditions. Box plots of flux distribution for reactions rates for succinate, glutamate, aspartate, sulfate, as well as production
of the TCA cycle (A–J) and glycolytic pathway (K–Q) in R. rates of PHB and H2. For the light-constrained simulation, LUR was set
sphaeroides under photoheterotrophic conditions. The first box plot to the measured SLSR. A citrate synthase; B aconitase; C isocitrate
in each group represents the flux distribution under light-constrained dehydrogenase; D and E a-ketoglutarate dehydrogenase; F succinyl-
conditions, while the second box represents flux distributions with CoA synthetase; G succinate dehydrogenase; H fumarate reductase;
unconstrained light. Distributions were obtained by sampling 1,000 I malate dehydrogenase; J phosphoenolpyruvate carboxykinase; K tri-
alternative optimal solutions. It should be noted that outliers were osephosphate isomerase; L glyceraldehydes-3-phosphate dehydroge-
omitted from the box plots for the unconstrained light analysis to allow nase; M phosphoglycerate kinase; N phosphoglyceromutase;
better visualization of the differences in predicted flux ranges. Data O phosphopyruvate hydratase; P fructose-bisphosphate aldolase;
used in the simulations corresponded to a chemostat run at 10 W m-2. Q D-fructose 1,6-bisphosphatase

Examination of the fluxes through central metabolism potential as contributors to biotechnological innovation, as
helps illustrate this finding (Fig. 7), as the fluxes through their ability to efficiently harness solar energy and produce
most TCA cycle reactions show no change over the optimal value-added commodities can help develop a more sus-
solution space under light-constrained conditions, whereas tainable society. Well-studied photosynthetic microbes,
significant variability in these fluxes is predicted with such as R. sphaeroides, can provide high-value products
unconstrained light. This reduced solution space likely and guide future research given its well-studied photo-
reflects the energy limitations under light-limiting condi- chemical cycle and the available genetic, genomic, and
tions and predicts that flux through energy demanding computational tools available to study this organism. To
reactions will be avoided or minimized when light is take advantage of photosynthetic systems, it is imperative
limiting. we gain an improved understanding of their physiology and
energetics, and the contribution of light intensity to these
processes.
Conclusions In this study, we gained a better quantitative under-
standing of the impact of light on a photosynthetic organ-
Photosynthetic organisms play a key role in global carbon ism by examining its effects on the physiology and
cycling and the planetary food chain. They also have bioenergetics of R. sphaeroides. We found significant light-

123
178 Photosynth Res (2015) 123:167–182

Table 1 Properties of R. sphaeroides during photoheterotrophic growth

3 W m-2 5 W m-2 10 W m-2 100 W m-2

Substrates
Carbon Succinate Succinate Succinate Succinate
Nitrogen Glutamate Glutamate Glutamate Glutamate
Yields
Biomass yield on light, Yx,E (gDW mmol photons-1) 0.0016 ± 0.0002 0.0015 ± 0.0002 0.0012 ± 0.0002 ND
Biomass yield on substrate, Yx,C (gCODx gCOD-1
c ) ND 0.64 ± 0.17 0.61 ± 0.07 0.57 ± 0.04
Maintenance energy
Light maintenance (mmol photons gDW-1 h-1) 5.4 ± 6.4 0.8 ± 5.4 16.7 ± 11.7 ND
GAM (mmol ATP gDW-1) 282 ± 51 298 ± 58 386 ± 86 ND
NGAM (mmol ATP gDW-1 h-1) 2.6 ± 3.2 0.5 ± 3 8.4 ± 5.9 ND
Hydrogen
Average H2 production (mmol gDW-1 h-1) 1.1 ± 0.2 1.7 ± 0.1 2.9 ± 0.4 4.9 ± 0.3
Substrate conversion efficiency (%) 21 ± 5 40 ± 7 34 ± 3 34 ± 4
Light conversion efficiency (%) 4.8 ± 0.7 6.4 ± 0.4 5.8 ± 0.2 0.8 ± 0.05
Biomass (% DW)
Protein 71.1 ± 2.0 73.3 ± 1.3 60.1 ± 4.6 63.8 ± 3.6
PHB 3.0 ± 1.0 3.3 ± 0.7 13.0 ± 2.0 14.8 ± 0.9
Pigment 0.7 ± 0.2 0.9 ± 0.06 0.5 ± 0.05 0.2 ± 0.05
Lipid 4.5 ± 0.6 6.1 ± 0.2 5.5 ± 0.3 4.0 ± 0.3

dependent changes in biomass composition as well as PHB used to improve the modeling and performance of this and
and H2 production in steady-state cells. Furthermore, our related photosynthetic organisms for industrial uses.
analysis showed that the metabolic energy required for
cellular maintenance under photosynthetic conditions is
significantly affected by light intensity, but it is also linked Materials and methods
to resultant differences in biomass composition and the
need to maintain redox homeostasis. At low light intensi- Chemostat operation
ties, R. sphaeroides maintenance energy requirements were
relatively low, suggesting that it could serve as an excellent Rhodobacter sphaeroides 2.4.1 cells were grown on Si-
microbial factory for bioproduction under these conditions. strom’s Minimal Medium (SMM) (Sistrom 1960) with
Maintenance energy increased significantly at higher light 8.1 mM glutamate used to replace ammonia as the main
intensities, but with a tradeoff of producing significantly nitrogen source. Under these conditions, H2 production via
larger amounts of potentially useful end products like PHB nitrogenase activity is promoted (Kontur et al. 2011; Yilmaz
and H2. et al. 2010). Cultures (16.5 ml) were maintained in contin-
We also showed that use of the condition-dependent uously stirred, sealed 20 ml chemostats at 30 °C, which
parameters for biomass and maintenance energy can sig- were illuminated by an incandescent light box at 3, 5, or
nificantly improve predictions of iRsp1140. By incorpo- 10 W m-2, or with an incandescent flood light source at
rating measured photon uptake for the first time in 100 W m-2. Reported light intensities were measured with
modeling photosynthetic growth of R. sphaeroides, we also a Yellow–Springs–Kettering model 65-A radiometer
significantly improved the predictive ability of iRsp1140. through a Corning 7-69 filter [transmission 750–900 nm,
Furthermore, we showed that the solution space of mod- i.e., the photosynthetically active radiation (PAR) for R.
eling predictions is significantly reduced when light-limit- sphaeroides]. Culture turbidity was monitored using a Klett–
ing conditions are considered, with only a handful of Summerson photoelectric colorimeter (Klett MFG Co., NY).
reactions in the network having variable fluxes under these The chemostats were operated in batch-fed mode, in
conditions. Overall, these analyses have improved our which a portion of spent media was wasted for 1 min
understanding of the effects of light on R. sphaeroides immediately followed by a 3-min feeding cycle of an
metabolism, bioenergetics, and physiology. In addition, it equivalent volume of new media. This cycle was repeated
has provided several new parameters (Table 1) that can be every 20 min. The feeding and wasting rates were varied to

123
Photosynth Res (2015) 123:167–182 179

attain a range of retention times varying from 8 to 39 h. where qH2 is the density of H2 (0.08988 g/l); VH2 is the
Reactors were started in batch mode until cells reached an volume of H2 produced per hour in l hr-1; I is the irradi-
optical density of [100 Klett units (1 Klett unit ance in W m-2; and A is the total surface area of the
equals *107 cells ml-1) (Imam et al. 2011). Cells were bioreactor in m2. Substrate conversion to H2 efficiency was
continuously fed with medium using Watson Marlow model calculated using succinate as the substrate (Koku et al.
120U peristaltic pumps (Watson Marlow Inc., Wilmington, 2002; Sasikala et al. 1993).
MA), and media were wasted using Masterflex model C/L
peristaltic pumps (Cole-Palmer Instrument Co., Vernon Quantification of biomass and cell dry weight
Hills, IL). Reactors were checked approximately every 12 h,
and when necessary, pumping was manually adjusted to Samples from individual chemostats were centrifuged at
correct for any changes in reactor volume due to imbalances 5,500 rcf at 4 °C to obtain cell pellets for biomass analysis.
of liquid inflow and outflow. Cultures were grown for at Total cellular protein was quantified via the Lowry assay
least five retention times and analyzed when steady state was (Hartree 1972; Lowry et al. 1951), while total cellular Bchl
established (verified from constant turbidity measures and/or was determined spectrophotometrically (Cohen-Bazire
constant gas production rates). et al. 1957). The phospholipid component was determined
by total phosphorus assay on lipids extracted via standard
Light supply rates chloroform/methanol extraction (Rouser et al. 1970). The
PHB content of cells was determined by GC–MS (GC-
The measured incident light intensity at the PAR measured 2010 gas chromatograph coupled to a QP-2010S mass
in W m-2 was converted into standard units of photon flux spectrometer detector; Shimadzu Scientific) (Yilmaz et al.
(PF) of lmol photon m-2 s-1 at a wavelength of 800 nm 2010). Other biomass components were assumed to remain
using the equation: constant based on previous measurements (Imam et al.
2011). Cell DW was calculated using the COD mass bal-
I  k  109
PF ¼ ; ance approach (Imam et al. 2011; Yilmaz et al. 2010).
h  c  NA
where I is the irradiance in W m-2; k is the wavelength of Quantification of nutrient uptake and biomolecule
light in nm; h is Planck’s constant (6.626068 9 10-34 secretion
m2 kg s-1); c is the speed of light (2.997925 9 108 m s-1);
and NA is Avogadro’s number (6.02 x 1017 lmol-1). To analyze the presence of media nutrients and products,
The PF was then converted into a SLSR in mmol culture supernatants were filtered through a 0.22-lm Nylon
gDW-1 h-1 for model simulations with the equation: Filter (Fisher Scientific, Pittsburg, PA) and frozen at
PF  3600  Areactor  U -80 °C until analysis. The uptake rates for succinate,
SLSR ¼ ; glutamate, and aspartate were measured as previously
DW  Vreactor
described (Yilmaz et al. 2010) by comparing their con-
where PF is the photon flux in lmol photon m-2 s-1; centrations in the filtered supernatant from the chemostat
Areactor is the surface area of the bioreactor in m2; DW is cultures to that in the initial media. Sulfate was quantified
the cell DW in g l-1; Vreactor is the volume of the bioreactor using Dionex 2100 Series Ion Chromatogram (Thermo
in ml; and U is the efficiency of light utilization taken to be Scientific) using EPA method 300.1. Ammonia concen-
0.75 based on best fits from regression of max. ATP pro- trations were analyzed using the salicylate method (Method
duction on l (see ‘‘Effect of light intensity on maintenance 10031, Hach Company, Loveland, CO). COD was ana-
energy requirements’’ section). lyzed using the reactor digestion method (Method 8000,
The biomass yield on light at 100 W m-2, which could Hach Company, Loveland, CO).
not be determined from the regression of SLSR and D, was Lactate, pyruvate, and malate were measured using GC–
calculated as the average biomass yield as previously MS (Yilmaz et al. 2010). Glucose, xylose, pyruvate, xyli-
described (Zijffers et al. 2010). tol, glycerol, formate, acetate, and ethanol were quantify-
 
Biomass yield gDW mmol photons1 ing using high-performance liquid chromatography
DW  l  Vreactor (HPLC) with a C18 column and a refractive index detector
¼ (RID-10A) on an Agilent system (Agilent Technologies,
PF  Areactor  3600
Santa Clara, CA), in collaboration with the Metabolomics
Light conversion to H2 efficiency was calculated (Koku Facility of the Great Lakes Bioenergy Research Center.
et al. 2002) with the equation: Total gas production and the gas composition of the
33:61  qH2  VH2 headspace were determined as previously described (Yil-
% Efficiency ¼ ;
IA maz et al. 2010).

123
180 Photosynth Res (2015) 123:167–182

Assessment of media for secretion of untargeted SMPs summary statistics obtained from these were used generate
the box plots in Fig. 7. All simulations were conducted
To test if other SMPs not targeted in our analysis might be under the GAMS programming environment (GAMS
present in the media, the measured COD of the culture Development Corporation, Cologne, Germany) using the
supernatants was compared to the estimated COD of the CPLEX solver.
supernatant. The estimated COD of the supernatant was
calculated based upon the measured concentrations of Determination of maintenance energy requirements
succinate, glutamate, and aspartate in the culture superna-
tant. The estimated COD contribution for each compound To determine the cellular maintenance energy requirements
was calculated using the following formula: at a given light intensity, the maximum ATP hydrolysis
X X rate for each reactor was calculated by constraining the
ThCODi ¼ Ci þ 0:25  Hi þ 0:5 model with uptake rates for light, succinate, glutamate,
X X  aspartate, and sulfate and the experimentally determined
 Ni  0:75  Oi
growth rate. At each light intensity, the photosynthetic
 ð32 mg O2 =mmolÞ  ½Conci ; biomass reaction of iRsp1140 (RXN1306) was replaced
with a new biomass reaction corrected for condition-spe-
where RCi, RHi, RNi and ROi are the number of carbon,
cific differences in biomass composition and removing the
hydrogen, nitrogen, and oxygen molecules, respectively,
previously estimated GAM value of 53.65 mmol ATP
present in compound i; [Conci] is the concentration of
gDW-1 from the reaction. Simulations were conducted by
compound i in units of mmol l-1. ThCODi is the estimated
maximizing the ATP hydrolysis reaction (RXN0765) under
CODi in units of mgCOD l-1 (Rittmann and McCarty
these constraints. For each light intensity, the maximum
2001).
ATP hydrolysis rate (qATP-total) for each reactor was plotted
against the specific growth rate (l). The ATP balance for
Constraint-based analysis anaerobic photoheterotrophic growth can be considered as
qATPcarbon þ qATPlight ¼ qATPtotal
The genome-scale metabolic model for R. sphaeroides, ¼ GAMATP  l þ NGAMATP
iRsp1140 (Imam et al. 2013), was used for all simulations.
Flux balance analysis (FBA) (Varma and Palsson 1994) where qATP-carbon is the ATP production rate from carbon
was used to simulate in silico growth by solving the linear sources taken up from the media (succinate, glutamate, and
programming problem: aspartate), qATP-light is the ATP production rate from light,
and qATP-total is the ATP production/hydrolysis rate all in
max vbiomass mmol gDW-1 h-1. GAMATP is the GAM energy require-
s:t: ment (in mmol gDW-1), and NGAMATP is the NGAM
S  v¼0 energy requirement (in mmol gDW-1 h-1). GAMATP and
NGAMATP were determined from the slope and y-intercept,
vmin  v  vmax ;
respectively, of the regression of qATP-total on l.
where mbiomass is the flux through biomass objective func-
tion; m is the vector of steady-state reaction fluxes; and mmin Acknowledgments This work was funded in part by the Depart-
ment of Energy, Office of Science, Great Lakes Bioenergy Research
and mmax are the minimum and maximum allowable fluxes Center (DE-FC02-07ER64494), and the Genomics:GTL and SciDAC
set to -1,000 and 1,000 mmol gDW-1 h-1 for reversible Programs (DE-FG02-04ER25627). SI was supported during part of
reactions and 0 and 1,000 mmol gDW-1 h-1 for forward this work by a William H. Peterson Predoctoral Fellowship from the
only reactions. Measured uptake rates for light, succinate, University of Wisconsin-Madison Bacteriology Department.
glutamate, aspartate, and sulfate, as well as the measured
growth rates and the production rates of PHB and H2 were
used as initial input constraints for modeling when
References
required. Other media components were allowed to be
freely exchanged with the extracellular space. FVA (Ma- Aagaard J, Sistrom WR (1972) Control of synthesis of reaction centre
hadevan and Schilling 2003) and alternative optima ana- bacteriochlorophyll in photosynthetic bacteria. Photochem Pho-
lysis (Lee et al. 2000; Reed and Palsson 2004) were tobiol 15(2):209–225
conducted as previously described (Imam et al. 2011). For Aiking H, Sojka G (1979) Response of Rhodopseudomonas capsulata
to illumination and growth rate in a light-limited continuous
assessing differences in flux distribution between light- culture. J Bacteriol 139(2):530–536
limiting and -saturating conditions, 1,000 alternative solu- Allen JF (2003) Cyclic, pseudocyclic and noncyclic photophosphor-
tions were sampled from the optimal solution space, and ylation: new links in the chain. Trends Plant Sci 8(1):15–19

123
Photosynth Res (2015) 123:167–182 181

Atsumi S, Higashide W, Liao JC (2009) Direct photosynthetic during photobiological H2 production by Rhodobacter sphaero-
recycling of carbon dioxide to isobutyraldehyde. Nat Biotechnol ides. Appl Environ Microbiol 77(20):7425–7429
27(12):1177–1180 Kontur WS, Schackwitz WS, Ivanova N, Martin J, Labutti K, Deshpande S,
Biel AJ (1986) Control of bacteriochlorophyll accumulation by light Tice HN, Pennacchio C, Sodergren E, Weinstock GM, Noguera DR,
in Rhodobacter capsulatus. J Bacteriol 168(2):655–659 Donohue TJ (2012) Revised sequence and annotation of the Rhodob-
Campbell TB, Lueking DR (1983) Light-mediated regulation of acter sphaeroides 2.4.1 genome. J Bacteriol 194(24):7016–7017
phospholipid synthesis in Rhodopseudomonas sphaeroides. Lee S, Phalakornkule C, Domach MM, Grossmann IE (2000)
J Bacteriol 155(2):806–816 Recursive MILP model for finding all the alternate optima in
Carapezza G, Umeton R, Costanza J, Angione C, Stracquadanio G, LP models for metabolic networks. Comput Chem Eng
Papini A, Lio P, Nicosia G (2013) Efficient behavior of 24(2–7):711–716
photosynthetic organelles via Pareto optimality, identifiability, Lowry OH, Rosebrough NJ, Farr AL, Randall RJ (1951) Protein
and sensitivity analysis. ACS Synth Biol 2(5):274–288 measurement with the Folin phenol reagent. J Biol Chem
Chory J, Kaplan S (1983) Light-dependent regulation of the synthesis 193(1):265–275
of soluble and intracytoplasmic membrane proteins of Rhodo- Mackenzie C, Eraso JM, Choudhary M, Roh JH, Zeng X, Bruscella P,
pseudomonas sphaeroides. J Bacteriol 153(1):465–474 Puskas A, Kaplan S (2007) Postgenomic adventures with
Cohen-Bazire G, Sistrom WR, Stanier RY (1957) Kinetic studies of Rhodobacter sphaeroides. Annu Rev Microbiol 61:283–307
pigment synthesis by non-sulfur purple bacteria. J Cell Physiol Mahadevan R, Schilling CH (2003) The effects of alternate optimal
49(1):25–68 solutions in constraint-based genome-scale metabolic models.
Connor MR, Atsumi S (2010) Synthetic biology guides biofuel Metab Eng 5(4):264–276
production. J Biomed Biotechnol. doi:10.1155/2010/541698 Masepohl B, Hallenbeck PC (2010) Nitrogen and molybdenum
Dal’Molin CG, Quek LE, Palfreyman RW, Nielsen LK (2011) control of nitrogen fixation in the phototrophic bacterium
AlgaGEM—a genome-scale metabolic reconstruction of algae Rhodobacter capsulatus. Adv Exp Med Biol 675:49–70
based on the Chlamydomonas reinhardtii genome. BMC Genom Miyake J, Kawamur S (1987) Efficiency of light energy conversion to
12(suppl 4):S5 hydrogen by the photosynthetic bacterium Rhodobacter sph-
Geisser S (1975) The predictive sample reuse method with applica- aeroides. Int J Hydrogen Energy 12(3):147–149
tions. J Am Stat Assoc 70(350):320–328 Montagud A, Navarro E, de Cordoba PF, Urchueguia JF, Patil KR
Golecki JR, Schumacher A, Drews G (1980) The differentiation of the (2010) Reconstruction and analysis of genome-scale metabolic
photosynthetic apparatus and the intracytoplasmic membrane in model of a photosynthetic bacterium. BMC Syst Biol 4:156
cells of Rhodopseudomonas capsulata upon variation of light Nogales J, Gudmundsson S, Knight EM, Palsson BO, Thiele I (2012)
intensity. Eur J Cell Biol 23(1):1–5 Detailing the optimality of photosynthesis in cyanobacteria
Gronenberg LS, Marcheschi RJ, Liao JC (2013) Next generation through systems biology analysis. Proc Nat Acad Sci USA
biofuel engineering in prokaryotes. Curr Opin Chem Biol 109(7):2678–2683
17(3):462–471 Oberhardt MA, Palsson BO, Papin JA (2009) Applications of
Hartree EF (1972) Determination of protein: a modification of the genome-scale metabolic reconstructions. Mol Syst Biol 5:320
Lowry method that gives a linear photometric response. Anal Oelze J (1988) Regulation of tetrapyrrole synthesis by light in
Biochem 48(2):422–427 chemostat cultures of Rhodobacter sphaeroides. J Bacteriol
Imam S, Yilmaz S, Sohmen U, Gorzalski AS, Reed JL, Noguera DR, 170(10):4652–4657
Donohue TJ (2011) iRsp1095: a genome-scale reconstruction of the Peralta-Yahya PP, Zhang F, del Cardayre SB, Keasling JD (2012)
Rhodobacter sphaeroides metabolic network. BMC Syst Biol 5:116 Microbial engineering for the production of advanced biofuels.
Imam S, Noguera DR, Donohue TJ (2013) Global insights into Nature 488(7411):320–328
energetic and metabolic networks in Rhodobacter sphaeroides. Reed JL (2012) Shrinking the metabolic solution space using
BMC Syst Biol 7(1):89 experimental datasets. PLoS Comput Biol 8(8):e1002662
Khatipov E, Miyake M, Miyake J, Asada Y (1998) Polyhydroxybu- Reed JL, Palsson BO (2004) Genome-scale in silico models of E. coli
tyrate accumulation and hydrogen evolution by Rhodobacter have multiple equivalent phenotypic states: assessment of
sphaeroides as a function of nitrogen availability. In: Zaborsky correlated reaction subsets that comprise network states.
O, Benemann J, Matsunaga T, Miyake J, San Pietro A (eds) Genome Res 14(9):1797–1805
BioHydrogen. Springer, US, pp 157–161 Rex R, Bill N, Schmidt-Hohagen K, Schomburg D (2013) Swimming
Kien NB, Kong IS, Lee MG, Kim JK (2010) Coenzyme Q10 in light: a large-scale computational analysis of the metabolism
production in a 150-l reactor by a mutant strain of Rhodobacter of Dinoroseobacter shibae. PLoS Comput Biol 9(10):e1003224
sphaeroides. J Ind Microbiol Biotechnol 37(5):521–529 Rittmann B, McCarty PL (2001) Environmental biotechnology:
Kiley PJ, Kaplan S (1988) Molecular genetics of photosynthetic principles and applications. McGraw-Hill Science Engineering,
membrane biosynthesis in Rhodobacter sphaeroides. Microbiol New York
Rev 52(1):50–69 Rouser G, Fkeischer S, Yamamoto A (1970) Two dimensional then
Kim E, Lee M, Kim M, Lee JK (2008) Molecular hydrogen layer chromatographic separation of polar lipids and determina-
production by nitrogenase of Rhodobacter sphaeroides and by tion of phospholipids by phosphorus analysis of spots. Lipids
Fe-only hydrogenase of Rhodospirillum rubrum. Int J Hydrog 5(5):494–496
Energy 33(5):1516–1521 Russell JB, Cook GM (1995) Energetics of bacterial growth: balance
Kliphuis AM, Klok AJ, Martens DE, Lamers PP, Janssen M, Wijffels of anabolic and catabolic reactions. Microbiol Rev 59(1):48–62
RH (2011) Metabolic modeling of Chlamydomonas reinhardtii: Sasikala K, Ramana VC, Rao RP, Kovacs KL (1993) Anoxygenic
energy requirements for photoautotrophic growth and mainte- phototrophic bacteria: physiology and advances in hydrogen
nance. J Appl Phycol 24(2):253–266 production technology. Adv Appl Microbiol 38:211–295
Koku H, Eroglu I, Gunduz U, Yucel M, Turker L (2002) Aspects of Schumacher A, Drews G (1979) Effects of light intensity on
the metabolism of hydrogen production by Rhodobacter sph- membrane differentiation in Rhodopseudomonas capsulata.
aeroides. Int J Hydrogen Energy 27(11–12):1315–1329 Biochim Biophys Acta 547(3):417–428
Kontur WS, Ziegelhoffer EC, Spero MA, Imam S, Noguera DR, Sistrom WR (1960) A requirement for sodium in the growth of
Donohue TJ (2011) Pathways involved in reductant distribution Rhodopseudomonas spheroides. J Gen Microbiol 22:778–785

123
182 Photosynth Res (2015) 123:167–182

Stone M (1974) Cross-validatory choice and assessment of statistical AS, Reed JL (2012) Genome-scale modeling of light-driven
predictions. J Royal Stat Soc 36(2):111–147 reductant partitioning and carbon fluxes in diazotrophic unicel-
Stouthamer AH, van Verseveld HW (1987) Microbial energetics lular cyanobacterium Cyanothece sp. ATCC 51142. PLoS
should be considered in manipulating metabolism for biotech- Comput Biol 8(4):e1002460
nological purposes. Trends Biotechnol 5(5):149–155 Wahlund TM, Conway T, Tabita FR (1996) Bioconversion of CO2 to
Tannler S, Decasper S, Sauer U (2008) Maintenance metabolism and ethanol and other compounds. American Chemical Society
carbon fluxes in Bacillus species. Microb Cell Fact 7:19 Division of Fuel Chemistry 3:1403–1405
Thiele I, Palsson BO (2010) A protocol for generating a high-quality Yilmaz LS, Kontur WS, Sanders AP, Sohmen U, Donohue TJ,
genome-scale metabolic reconstruction. Nat Protoc 5(1):93–121 Noguera DR (2010) Electron partitioning during light- and
Varma A, Palsson BO (1994) Metabolic flux balancing: basic nutrient-powered hydrogen production by Rhodobacter sphaero-
concepts, scientific and practical use. Nat Biotechnol ides. Bioenerg Res 1:55–66
12:994–998 Zijffers JW, Schippers KJ, Zheng K, Janssen M, Tramper J, Wijffels
Vu TT, Stolyar SM, Pinchuk GE, Hill EA, Kucek LA, Brown RN, RH (2010) Maximum photosynthetic yield of green microalgae
Lipton MS, Osterman A, Fredrickson JK, Konopka AE, Beliaev in photobioreactors. Mar Biotechnol (NY) 12(6):708–718

123

View publication stats