Download as pdf or txt
Download as pdf or txt
You are on page 1of 45

Article type : Original Article

Accepted Article
Concurrent isotope-assisted metabolic flux analysis and transcriptome
profiling reveal responses of poplar cells to altered nitrogen and carbon
supply

Xiaofeng Zhang1, ■ , &, Ashish Misra1,#,&, Shilpa Nargund1,&, Gary D. Coleman2 and Ganesh Sriram1,
1
Department of Chemical and Biomolecular Engineering and 2Department of Plant Science and
Landscape Architecture, University of Maryland, College Park, MD 20742, USA

Current affiliations: ■ Biology Department, Brookhaven National Laboratory, Upton, NY 11973, USA,
#
Department of Biochemical Engineering and Biotechnology, Indian Institute of Technology, Delhi,
India
&
These three authors contributed equally to this work.


Corresponding author: Ganesh Sriram. Address: 1208D, Chemical and Nuclear Engineering Building
090, College Park, MD 20742. Phone: +1 301 405-1261; Fax: +1 301 405-0523; Email:
gsriram@umd.edu.

Running title: Metabolic response to altered nitrogen and carbon supply in poplar cells

Keywords: Poplar, nitrogen use efficiency, metabolic flux analysis, microarray

Summary

Reduced nitrogen is indispensable to plants. However, its limited availability in soil combined with

the energetic and environmental impacts of nitrogen fertilizers motivates research into molecular

mechanisms toward improving plant nitrogen use efficiency (NUE). We performed a systems-level

investigation of this problem by employing multiple ’omics methodologies on cell suspensions of

This article has been accepted for publication and undergone full peer review but has not
been through the copyediting, typesetting, pagination and proofreading process, which may
lead to differences between this version and the Version of Record. Please cite this article as
doi: 10.1111/tpj.13792
This article is protected by copyright. All rights reserved.
hybrid poplar (Populus tremula x Populus alba). Acclimation and growth of the cell suspensions in

four nutrient regimes ranging from abundant to deficient supplies of carbon and nitrogen revealed
Accepted Article
that cell growth under low-nitrogen levels was associated with substantially higher NUE. To

investigate the underlying metabolic and molecular mechanisms, we concurrently performed

steady-state 13C metabolic flux analysis with multiple isotope labels and transcriptomic profiling with

cDNA microarrays. The 13C flux analysis revealed that the absolute flux through the oxidative

pentose phosphate pathway (oxPPP) was substantially lower (~3-fold) under low-nitrogen

conditions. Additionally, the flux partitioning ratio between the tricarboxylic acid (TCA) cycle and

anaplerotic pathways varied from 84%:16% under abundant carbon and nitrogen to 55%:45% under

deficient carbon and nitrogen. Gene expression data, together with the flux results, suggested a

plastidic localization of the oxPPP as well as transcriptional regulation of certain metabolic

branchpoints including those between glycolysis and the oxPPP. The transcriptome data also

indicated that NUE-improving mechanisms may involve a redirection of excess carbon to aromatic

metabolic pathways and extensive downregulation of potentially redundant genes (in these

heterotrophic cells) that encode photosynthetic and light-harvesting proteins, suggesting the

recruitment of these proteins as nitrogen sinks in nitrogen-abundant conditions.

1. Introduction

Nitrogen plays a critical role in life as it is contained in the building blocks of proteins and nucleic

acids, and is a component of many specialized metabolites involved in stress responses (Heldt and

Piechulla, 2010). Nitrogen deficiency in plants can result in reduced yields, decreased nutritional

quality, alterations in postharvest physiology and increased pathogen susceptibility (Amtmann and

Armengaud, 2009). Despite the importance of nitrogen to plant nutrition, predicting the relationship

between nitrogen deficiency and plant performance has often been difficult (Amtmann and

This article is protected by copyright. All rights reserved.


Armengaud, 2009). This is likely because of the complexity of nitrogen metabolism as well as the

interactions between carbon and nitrogen metabolism. On the one hand, nitrogen uptake and
Accepted Article
metabolism requires carbon skeletons, energy carriers, reductants and cotransporters, which are

contributed by carbon metabolism (Kruger and von Schaewen, 2003; Sahrawy et al., 2004; Gao et al.,

2008; Nunes-Nesi et al., 2010). On the other hand, crucial carbon metabolic pathways including

photosynthesis, photorespiration and mitochondrial respiration are sensitive to nitrogen availability

(Nunes-Nesi et al., 2010). Interactions between carbon and nitrogen metabolism are influenced by

environmental factors including light, water availability, nutrient levels and pH. These interactions

also involve various signaling molecules such as nitrate, ammonium, sugars, amino acids, and organic

acids (Nunes-Nesi et al., 2010). Because nitrogen and carbon metabolism can compete for

metabolites such as ATP (Sahrawy et al., 2004) and reductants (Escobar et al., 2006), their control is

likely to be tightly linked and coordinated at the level of gene expression (e.g. Escobar et al., 2006).

Quantitative measurements of fluxes and transcripts are valuable in elucidating nitrogen metabolism

(Amtmann and Armengaud, 2009). Previous research directed at understanding the molecular

impact of different nitrogen sources or availability on plant metabolism has mainly employed

metabolomics, transcriptomics or proteomics (including enzyme activity measurements) (Junker et

al., 2007; Allen et al., 2011; Hockin et al., 2011; Noorhana, 2011; Truong et al., 2013; Chen et al.,

2016). Recently, 13C metabolic flux analysis (13C MFA) has been used to study the influence of

nitrogen sources on carbon flow in metabolism (Allen and Young, 2013; Masakapalli et al., 2013). For

example, Masakapalli et al. (2013) found that in Arabidopsis cells, replacing ammonium with nitrate

in the culture media increased the oxidative pentose phosphate pathway flux by 50%. Allen and

Young (2013) found that reducing the carbon to nitrogen ratio from 37:1 to 13:1 resulted in a 90%

increase in the TCA cycle flux, which increased the contribution of malic acid to pyruvate production.

Studying the already existing metabolic machinery used by plants for internal nitrogen recycling and

genetically eliminating possible constraints could facilitate the development of strategies that utilize

This article is protected by copyright. All rights reserved.


nitrogen more efficiently. Poplar, a lignocellulosic feedstock, can recycle up to 80% of the nitrogen in

leaves to perennial tissues during leaf senescence (Pregitzer et al., 1990). This internal nitrogen
Accepted Article
recycling involves the accumulation of bark storage proteins (BSPs) in bark phloem parenchyma and

xylem ray cells and predominantly employs glutamine as the nitrogen transporter (Cantón et al.,

2005; Islam et al., 2015). Because seasonal nitrogen recycling is a major factor contributing to NUE,

there is interest in understanding this process and the carbon-nitrogen interactions in this tree

species. Previously, we employed 13C MFA to poplar cell suspensions and observed that unlabeled

biomass initially present prior to 13C addition significantly refluxes into the metabolic network and

modeling of this phenomenon required the development of a comprehensive metabolite reflux

model (Nargund et al., 2014).

Here, we report a concurrent investigation of central carbon metabolic fluxes and genomewide

transcript levels in poplar cells under various conditions of carbon and nitrogen availability. We

systematically designed four carbon and nitrogen source regimes (hereafter, “environmental

conditions” or “conditions”): (i) high carbon-high nitrogen (CN), (ii) high carbon-low nitrogen (Cn),

(iii) low carbon-high nitrogen (cN) and (iv) low carbon-low nitrogen (cn), which led to substantial

differences in cellular growth and nitrogen uptake rates. Differences in growth provided an initial

assessment of how the cells responded to carbon and nitrogen limitations. We then performed

three parallel 13C isotopic labeling experiments (LEs) and evaluated fluxes by using our previously

developed metabolic network model (Nargund et al., 2014). In addition, the flux analysis was

complemented by assaying transcript levels via cDNA microarrays. The altered metabolic flux

distributions and gene expression results revealed the capacity of poplar cells to adapt to low

nitrogen availability by improving NUE.

This article is protected by copyright. All rights reserved.


2. Results
Accepted Article
2.1 Enhanced nitrogen use efficiency (NUE) is associated with low nitrogen availability

We measured several macroscopically observable metabolic properties of the poplar cells under the

four environmental conditions listed in Section 1. These measurements included biomass

composition (Table 1), cell growth (Figure 1a), carbon and nitrogen uptake rates (Figure 1b-d), and

abundances of proteinogenic amino acids (Figure 1e). Cell growth measurements (Figure 1a)

showed a good fit to exponential curves (R2 > 0.95 for all conditions), which is indicative of metabolic

steady state (Zamboni et al., 2009). Cell growth was also influenced by both low-carbon and low-

nitrogen conditions, consistent with previous studies (Truong et al., 2013; Noorhana, 2011). Notably,

at the initial nutrient concentrations used in this study, reduced nitrogen availability resulted in a

more pronounced effect on cell growth than carbon deficiency. Furthermore, cells under the high-

nitrogen conditions exhibited more rapid initial growth than cells under the low-nitrogen conditions.

Analysis of carbon and nitrogen uptake rates showed that the yield of biomass relative to carbon

(Figure 1c) remained relatively unchanged across the four conditions, ranging from 1.3 mg biomass

(mg carbon)-1 to 1.5 mg biomass (mg carbon)-1. By contrast, the yield of biomass relative to nitrogen

(Figure 1d) was significantly higher under the low-nitrogen conditions (21 mg biomass [mg nitrogen]-
1
for the Cn condition, 95% confidence interval = [17, 25]; 23 mg biomass [mg nitrogen]-1 for the cn

condition, 95% confidence interval = [19, 25]) compared to the high-nitrogen conditions (11 mg

biomass [mg nitrogen]-1 for the cN condition, 95% confidence interval = [8, 14]; 15 mg biomass [mg

nitrogen]-1 for the cn condition, 95% confidence interval = [13, 17]). These differences in biomass

yield relative to nitrogen uptake indicate higher NUE under nitrogen-limited conditions, irrespective

of carbon level. To further substantiate this, we carried out another biomass yield analysis focusing

on nitrogen amount incorporated into synthesized proteins, which led to the same conclusion

(Supporting note 1). The linearity of the data points in Figures 1c and 1d indicate that the yield of

biomass relative to carbon or nitrogen was constant from 0 d to 7 d, which further supports that

This article is protected by copyright. All rights reserved.


metabolic steady state was maintained during the experimental period. A comparison of carbon and

nitrogen consumption (Figure 1b) further shows that under the low-nitrogen conditions, the cells
Accepted Article
were able to utilize less nitrogen while consuming the same amount of carbon, suggesting an altered

carbon-nitrogen balance. The abundances of proteinogenic amino acids were also responsive to

nitrogen availability (Figure 1e, Table S1). For example, Leu constituted more than 15% of

proteinogenic amino acids under high-nitrogen conditions, but constituted less than 4% of

proteinogenic amino acids under low-nitrogen conditions. Similarly, the combined proportion of Glu

and Gln (Glx) was 20% under the high-nitrogen conditions but fell to less than 10% under the low-

nitrogen conditions. Since the biomass was continuously formed as the cells grew, the higher

abundance of certain proteinogenic amino acids in biomass indicates higher flux towards their

synthesis.

2.2 Differences in mass isotopomer abundances across the environmental conditions suggest

nontrivial flux alterations in oxidative pentose phosphate pathway (oxPPP), TCA cycle and acetyl-

CoA reflux

To resolve intracellular fluxes, we performed three-week 13C LEs as described in Section 5.2. Briefly,

cells growing under each environmental condition were fed either (i) 98% U-13C glucose, (ii) 98% 1-
13
C glucose or (iii) a mixture of 28% U-13C and 72% naturally abundant glucose in separate, parallel

LEs. Our experimental design ensured that the isotopic steady-state was approximated. To verify this,

we measured the mass isotopomer distribution (MIDs) of intracellular soluble amino acids obtained

from cells harvested after each week of growth throughout the three-week growth period. An

example of this analysis for cells growing under the CN condition with 28% U-13C LE is shown in

Figure 2. These results illustrate that after two weeks of growth, total 13C enrichment of most amino

acids was nearly 28% (Figure 2a). In addition, unsupervised clustering analysis confirmed that MIDs

of amino acids from the two-week old and three-week old cells were similar (Figure 2b). This

comparison supports our application of a steady state model to estimate fluxes.

This article is protected by copyright. All rights reserved.


The entire isotopomer dataset consisted of about 400 amino acid mass isotopomers for each

condition (Table S2-S4). A comparison of the isotopomer data across the four environmental
Accepted Article
conditions showed that the largest number of significantly different isotopomer abundances lay

between the CN and the cn condition [85 isotopomer abundances with difference > 0.03, p < 0.05

(via t-test) and 29 isotopomer abundances with difference > 0.06, p < 0.05] (Figure S1a). In mass

isotopomer space, a difference of 0.06 (on a scale from 0 to 1) is substantial, considering that many

recent 13C MFA studies reported mass isotopomer differences less than 0.06 even between different

genotypes that showed significant flux differences (e.g. Lonien and Jörg Schwender, 2009; Kind et

al., 2013; He et al., 2014; Wasylenko and Stephanopoulos, 2014). For example, Lonien and

Schwender et al. (2009) observed four-fold flux alterations in Arabidopsis genotypes resulting from a

transcription factor mutation, with a maximum difference of mass isotopomer abundances of about

0.06. Thus, the differences in mass isotopomer abundances across conditions observed in this study

suggest nontrivial flux alterations.

Comparison of MIDs of specific amino acid fragments across the four environmental conditions

qualitatively indicated flux differences in the metabolic network. For example, in the 98% 1-13C LE,

53% ± 1% of the His[2345] fragment (i.e., the His fragment consisting of carbons C2-C5; same

notation hereafter) and the His[23456] fragment were unlabeled under the Cn condition whereas

61% ± 1% of the His[2345] and His[23456] fragments were unlabeled under the CN condition (Figure

3a). As His originates from pentose phosphates, the observed difference of MIDs suggests different

flux patterns towards pentose phosphate synthesis in the different conditions. Similarly, in the 98%

U-13C glucose LE, the different MIDs of Glx (Glu + Gln, synthesized from α-ketoglutarate) and Thr

(synthesized from aspartic acid) between CN and Cn conditions pointed to different flux patterns

through the TCA cycle and anaplerotic reactions (Figure 3b, 3c). These results are indicative of

differences in metabolic fluxes associated with carbon and nitrogen levels, yet accurate

quantification of these flux differences between the four conditions requires full isotopomer

modeling and statistical analysis as reported in the following section (Section 2.3).

This article is protected by copyright. All rights reserved.


2.3 Nitrogen deficiency reduced oxPPP flux, increased anaplerotic flux and increased acetyl-CoA

reflux
Accepted Article
Our metabolic network model, similar to the one used in our previous study (Nargund et al., 2014),

consisted of glycolysis, PPP, TCA cycle, anaplerotic reactions and reflux of acetyl-CoA. The inclusion

of acetyl-CoA reflux was supported by measured Glx isotopomer abundances (Section 3.1). Based on

our previous study (Nargund et al., 2014), glucose and anaplerotically fixed CO2 were the two carbon

sources included in the model. Three intracellular compartments were included in the model:

cytosol, mitochondrion and plastid, with glycolysis and pentose phosphate pathways duplicated in

both the cytosol and plastid. The complete metabolic network model used for MFA is shown in

Figure 4 and Table S5.

For each of the four environmental conditions, we evaluated metabolic fluxes from the entire

dataset of approximately 400 isotopomer abundances from the three LEs. The sum-of-squared

residuals (SSR) between measured and calculated mass isotopomer abundances fell within the

acceptable ranges for the four conditions (achieved SSR: CN – 228, cN – 294, Cn – 343, cn – 370;

acceptable SSR: CN – 384, cN – 375, Cn – 373, cn – 379, Supporting note 2). The flux estimations

(Table S6) showed differences in the fluxes through the glycolytic and PPP between the cytosol and

plastid. For all the four conditions the glycolytic flux (pgif) was greater than the oxPPP flux (g6pdh) in

the cytosol (e.g. CN – pgif: 1.96 ± 0.33; g6pdh: 0.23 ± 0.15). However, in the plastid there was a high

flux through the oxPPP (g6pdhp) that was realized by the reversal of plastidic upper glycolysis (pgifp)

(e.g. CN – g6pdhp: 6.37 ± 0.89; pgifp: -6.04 ± 0.96). More than 90% of the total oxPPP fluxes were

found to occur in the plastid (e.g. CN – g6pdhp: 6.37 ± 0.89; g6pdh: 0.23 ± 0.15). By comparing the

flux maps across the four environmental conditions (Figure 4a, 4b), we found reduced plastidic

oxPPP flux resulting from lower carbon or nitrogen levels, and this reduction was greater for

nitrogen than carbon (g6pdhp flux – CN: 6.37 ± 0.89 μmol day-1 [mg biomass]-1, cN: 3.79 ± 1.00 μmol

day-1 [mg biomass]-1, Cn: 1.88 ± 0.65 μmol day-1 [mg biomass]-1, cn: 2.54 ± 0.62 μmol day-1 [mg

This article is protected by copyright. All rights reserved.


biomass]-1). A plot of the metabolic flux space for the four environmental conditions (Figure 5a) also

showed that the high- and low-nitrogen levels were mainly distinguished by different oxPPP fluxes.
Accepted Article
This is consistent with previous studies with soybean embryos showing that higher ratios of carbon

to nitrogen supply resulted in lower PPP flux (Allen and Young, 2013).

We also observed a greater forward flux of anaplerotic reactions under low-carbon and low-nitrogen

conditions (anapf – CN: 0.36 ± 0.23 μmol day-1 [mg biomass]-1, cN: 1.55 ± 0.27 μmol day-1 [mg

biomass]-1, Cn: 1.35 ± 0.22 μmol day-1 [mg biomass]-1, cn: 0.92 ± 0.11 μmol day-1 [mg biomass]-1).

Higher influx of unlabeled acetyl-CoA [evaluated as described in Nargund et al. (2014)] was also

observed under low-nitrogen conditions (acain – CN: 0.14 ± 0.03 μmol day-1 [mg biomass]-1, cN: 0.27

± 0.04 μmol day-1 [mg biomass]-1, Cn: 0.35 ± 0.04 μmol day-1 [mg biomass]-1, cn: 0.21 ± 0.04 μmol

day-1 [mg biomass]-1). From our previous work (Nargund et al., 2014), we concluded that poplar cells

were able to obtain additional carbon sources apart from glucose through the assimilation of CO2

into malate and oxaloacetate and recycled acetyl-CoA produced from degradation of lipids and

various amino acids. Here, a comparison across the four environmental conditions further indicated

that carbon or nitrogen deficiency led to higher acetyl-CoA reflux.

In addition, we calculated the overall carbon partitioning (Figure 5b) and relative flux ratios at the

G6P and PEP-Pyr (i.e., a consolidated metabolite that combines PEP and Pyr) nodes (Figure 5c, 5d).

Firstly, the ratio of CO2 outflux to glucose influx was similar across the four conditions [100-115%, μg

day-1 (mg biomass)-1]. At the G6P node (Figure 5c), the reversal of pgi (phosphoglucose isomerase)

reaction accounted for about 50% of the oxPPP flux for the two low-carbon conditions. However, a

comparison between the two high-carbon conditions (CN vs. Cn) revealed that when nitrogen was

abundant, the pgi reaction was reversed to a greater extent which maintained a high oxPPP flux

This article is protected by copyright. All rights reserved.


(ratio of pgib flux to glucose uptake – CN: 156 ± 26%, Cn: 36 ± 14%). At the PEP-Pyr node (Figure 5d),

comparison of flux ratio revealed a substantial proportion of carbon flow being redirected from the
Accepted Article
TCA cycle to anaplerotic reactions when either carbon or nitrogen levels was reduced (ratio of

anaplerosis flux to TCA flux – CN: 16:84, cN: 41:59, Cn: 42:58 and cn: 45:55).

2.4 Gene expression was influenced more by nitrogen supply than by carbon supply

To gain greater insight into how poplar cells adapt to carbon or nitrogen limitations in relation to

NUE, we measured transcripts via a whole-genome cDNA microarray. From this we performed an

extensive comparison of transcript levels for a subset of 892 genes encoding enzymes involved in

metabolism between the four environmental conditions. Similar to isotopomer abundances, reduced

nitrogen level tended to result in a greater number of differentially expressed genes than reduced

carbon level (Figure S1b). Furthermore, unsupervised hierarchical clustering of these genes by

principal component analysis showed a similar trend (Figure S1c). The dendrogram clearly clustered

the three biological replicates of each condition at the first level (variance between 5 to 15), and the

two low-nitrogen conditions clustered together while the two high-nitrogen conditions clustered

together at the second level (variance between 25 to 30).

This subset of 892 genes predicted to encode enzymes involved in metabolism, based on biological

function annotations in PoplarCyc and UniProt (Apweiler et al., 2004), was used to examine their

expression under the four environmental conditions. In Figure 6, we show a comparison between

the expression of genes encoding enzymes in central carbon metabolism and the fluxes of

corresponding pathways. All expression or flux changes (log2 scale) were calculated relative to the

CN condition. This comparison shows a general consistency between enzyme transcript levels and

flux calculations. For example, a decline in transcript abundance for genes encoding plastidic G6PDH

(Potri.014G166800) (cN, -2.4 ± 0.2; Cn, -3.9 ± 0.3; cn, -4.1 ± 0.1) and PYRK (Potri.010G254900) (cN, -

This article is protected by copyright. All rights reserved.


1.2 ± 0.2; Cn, -1.9 ± 0.2; cn, -2.1 ± 0.2) was consistent with reduced flux through the oxPPP as

described in Section 2.3. We also observed an increase in transcript abundance for a gene encoding
Accepted Article
PEPC (Potri.007G011200) under low-nitrogen conditions (cN, 0.1 ± 0.2; Cn, 3.2 ± 0.2; cn, 2.8 ± 0.2),

which is consistent with the higher anaplerotic fluxes estimated by MFA (Section 2.3). In MFA, we

sometimes consolidated sequential chemical reactions that each have a single influx and outflux (e.g.

“oxPPP” represents G6PDH, gluconolactonase and 6-phosphogluconate). In such cases, the

comparisons above are between fluxes through the consolidated reactions and transcript levels of

genes encoding enzymes involved in one of the reactions (e.g. oxPPP flux with transcript level of

Potri.014G166800 encoding G6PDH).

Although we observed an overall agreement between transcript levels and metabolic fluxes, we also

found that the transcript levels of some genes associated with metabolism were inconsistent with

the calculated fluxes. For example, transcript levels of genes encoding glyceraldehyde-3-phosphate

dehydrogenase (GAPDH, Potri.014G140500) and pyruvate decarboxylase (PDC, Potri.017G151900)

were reduced under the low-carbon or low-nitrogen conditions (Table S7: Potri.014G140500 – cN, -

3.9 ± 0.1; Cn, -5.5 ± 0.1; cn, -5.6 ± 0.4; Potri.004G054100 [EC: 4.1.1.1] – CN, -3.4 ± 0.3; Cn, -4.0 ± 0.3;

cn, -3.9 ± 0.4), while the corresponding fluxes showed little difference across the four environmental

conditions (Table S6: pfk – CN, 0.97 ± 0.26; cN, 0.95 ± 0.21; Cn, 0.86 ± 0.16; cn, 0.59 ± 0.14; pdh –

CN, 1.05 ± 0.25; cN, 1.48 ± 0.31; Cn, 1.34 ± 0.19; cn, 0.62 ± 0.18). We believe it was coincidental that

most reactions producing reducing equivalents (GAPDH and PDC) exhibited disagreement between

flux and transcript level trends. A notable exception to this trend is the g6pdh reaction, whose flux

showed very good agreement with the transcript level of one of the genes encoding plastidic G6PDH

(Potri.014G166800).

We also examined the expression of genes that encode enzymes involved in pathways beyond

central carbon metabolism. Figure 7 shows the expression of genes grouped into six categories: (a)

primary metabolism, (b) amino acid synthesis and degradation, (c) sugar synthesis, (d)

This article is protected by copyright. All rights reserved.


photosynthesis/light capture, (e) lignin and lignin related and (f) redox and cytochrome (the

complete dataset of expression values is contained in Table S7). Because the focus of this study was
Accepted Article
to examine the influence of nitrogen supply on transcriptome changes, we concentrated on

comparisons between the Cn and CN conditions. Apart from the aforelisted examples of differential

gene expressions relevant to central carbon metabolism, we observed significantly altered transcript

abundances for one gene that encodes acyl-CoA thioesterase (ACT) and two genes that encode acyl-

CoA ligase (ACLs), whose transcript abundances increased under the Cn condition (ACT: Cn, 4.4 ± 0.1;

ACL: Cn, 1.9 ± 0.1 and 1.8 ± 0.1) (Figure 7a). The transcriptomic analysis also revealed little influence

of nitrogen availability on the expression of genes encoding enzymes involved in Gln, Glu, Cys and

Gly synthesis (Figure 7b). Additionally, this analysis indicated that the transcript levels of genes

encoding enzymes involved in specialized metabolite production were also influenced by nitrogen

deficiency. For example, transcript abundance of genes encoding three inositol-3-phosphate

synthase enzymes was substantially reduced (Figure 7c, I3PS: Cn, -7.5 ± 0.1, -3.4 ± 0.5 and -4.4 ± 0.2).

The transcript abundances of genes encoding components of the photosystem and light harvesting

processes were also reduced in cells supplied with low nitrogen (Figure 7d, LHCB, PSA, PSOE: Cn, fold

changes ranging from -8.3 ± 0.5 to -0.5 ± 0.1). In contrast to photosystem related genes, the

transcript abundances of genes encoding enzymes involved in lignin and phenylalanine metabolism

were significantly increased under low-nitrogen conditions (Figure 7e, CCAR: Cn, 4.8 ± 0.2; CA3MT:

Cn, 1.7 ± 0.1 and 1.8 ± 0.1; LnCA: Cn, 5.2 ± 0.1, 5.4 ± 0.1 and 4.4 ± 0.1; PheAL: Cn, 4.6 ± 0.2, 2.0 ± 0.7,

1.6 ± 0.2; APDH: Cn, 2.2 ± 0.2, 2.4 ± 0.1, 2.1 ± 0.4, 1.5 ± 0.6). Cytochrome P450 defense-related

response genes also showed substantial upregulation (Figure 7f, CYP450D: Cn, 4.4 ± 0.7, 3.4 ± 0.1,

4.9 ± 0.1 and 4.7 ± 0.2), while two cytochrome P450 oxidation-reduction genes showed substantial

downregulation (CYP450OR: Cn, -5.8 ± 0.2 and -5.7 ± 0.3). Together, these results are consistent with

those of Gutierrez et al. (2007), where the major conclusion was that nitrate deprivation resulted in

(i) reduced mRNA levels of genes encoding enzymes involved in photosynthesis and (ii) increased

mRNA levels of genes encoding enzymes involved in the metabolism of many secondary metabolites.

This article is protected by copyright. All rights reserved.


3. Discussion
Accepted Article
Nitrogen is an essential nutrient for plants. Its flow through central metabolism is catalyzed by the

products of a complex network of genes and is tightly linked to both carbon assimilation and energy

metabolism (Krouk et al., 2010). Because of the importance of nitrogen to plant growth, deciphering

how plant cells respond to altered nitrogen availability is necessary to understand cellular nitrogen

homeostasis. At the transcriptome level, this response has been studied extensively in Arabidopsis

(e.g. Gutierrez, Gifford, et al., 2007; Krouk et al., 2009; Canales et al., 2014). Although less research

has been performed in plants at the metabolic flux level, three recent studies have shown that (i)

Arabidopsis cell suspensions undergo redistribution of metabolic fluxes when supplied with different

sources of inorganic nitrogen (Murashige-Skoog versus ammonium-free media) (Masakapalli et al.,

2013), (ii) the ratio of carbon to nitrogen affected intracellular metabolism in soybean embryos

(Allen and Young, 2013) and (iii) fluxes in central carbon metabolism vary significantly between

Brassica napus embryos growing with organic nitrogen sources and inorganic nitrogen sources,

independent from enzyme activity (Junker et al., 2007). In the current study, we obtained insights

into carbon-nitrogen interactions by simultaneously employing transcriptomic analysis and 13C MFA

to probe intracellular metabolic responses to changes in nutrient (especially nitrogen) supply.

3.1 Metabolic steady-state and minimization of metabolite reflux

In this study, we carried out LEs in shake flasks to study the influence of different (initial) carbon and

nitrogen source concentrations on intracellular metabolism in poplar cells. Alternatively, a

chemostat could have been employed to maintain the carbon and nitrogen source concentrations

constant at the initial levels. However, plant cell cultures are not always amenable to continuous

culture in chemostat. Many recent studies on plant systems also employed batch culture (Allen and

Young, 2013; Masakapalli et al., 2013; Masakapalli, Bryant, et al., 2014). Additionally, compared to

This article is protected by copyright. All rights reserved.


continuous cultures, batch operation has a lower risk of infection and mutation as well as greater

flexibility to switch between different processes (Villadsen et al., 2011). During the LEs, although the
Accepted Article
concentration of carbon and nitrogen sources gradually decreased during cell growth, our strategy

of transferring cells to fresh medium every 7 d (Section 5.1 and 5.2) ensured that the carbon and

nitrogen sources were not depleted. Moreover, these cells were acclimated under the four

conditions for more than six months before LEs were carried out, which helped to stabilize their

response to altered nutrient levels. The achievement of metabolic steady-state is corroborated by

two observations: (i) Near exponential growth throughout each growth cycle (7 days) (Figure 1a),

and (ii) constant biomass yield on carbon and nitrogen (Figure 1c, 1d).

In previous research (Nargund et al., 2014), we studied poplar cell suspensions that grew for one

week under one of the conditions (CN) of the current study. We found that metabolism in these cells

was characterized by mixing and reflux between initially present naturally abundant (“unlabeled”)

biomass and biomass that was newly synthesized during the LE. Amongst various metabolic models,

only a model involving both reflux and mixing of metabolites (chiefly amino acids) could satisfactorily

account for the labeling data. In contrast to this previous study that focused on this “reflux”

phenomenon, the current study was designed with LEs to minimize the influence of the reflux on

isotopomer measurements. This was accomplished by using three-week LEs that reduced most of

the initially present unlabeled-biomass and by measuring amino acids in soluble metabolite pools

that are less affected by reflux compared to proteinogenic amino acids (Nargund et al., 2014). In fact,

the 13C enrichment of most amino acids measured in the current study was substantially higher than

those we reported previously, indicating that a much smaller proportion of initially present,

unlabeled-biomass remained after three weeks of growth. Nevertheless, we still observed a subset

of isotopomer signatures (e.g. MIDs of Glu[12345]: M+2, 8% ± 1%; M+3, 11% ± 1%) that were

consistent with acetyl-CoA reflux (Nargund et al., 2014).

This article is protected by copyright. All rights reserved.


3.2 Isotopomer abundances and transcriptomics data supported the duplication of glycolysis and

PPP in poplar cell suspensions


Accepted Article
The duplication of central carbon metabolism in multiple cellular compartments has been studied

extensively for many plant cultures (e.g. Sriram et al., 2004; Sriram et al., 2007). For example, in

Arabidopsis cell suspensions Masakapalli, et al. (2010) concluded that labeling data alone was not

sufficient to distinguish the cytosolic and plastidic PPP. A similar conclusion was also reached for

soybean embryos (Allen and Young, 2013). In contrast to the aforementioned studies,

compartmented central carbon metabolism was indicated using 13C labeling experiments in

Arabidopsis leaves based on the different labeling patterns of ADP-glucose and UDP-glucose (Ma et

al., 2014). Furthermore, enzyme localization would also suggest that glycolysis and PPP are very

likely to be present in both cytosol and plastid (Kruger and von Schaewen, 2003). In our study, oxPPP

compartmentation in poplar cells is supported by multiple observations. Firstly, a comparison of

MIDs between Ser[123] and glycerol[123] fragments (Figure S2a) under the CN condition showed

stark differences. Although both Ser and glycerol originate from triose phosphates, they exhibited

more than 10% difference in MIDs in 98% 1-13C glucose LE and 28% U-13C glucose LE. This

observation is explained by duplicated triose phosphate pools in the cytosol and plastid and the

compartmentalization of pathways, such as the oxPPP, that influence the MIDs of the triose

phosphates. We also performed MFA using an uncompartmented model in which cytosolic and

plastidic pathways were not distinguished and this model showed poor agreement with measured

isotopomer data (Supporting Note 2). The different performance in fitting measured isotopomer

data by compartmented and uncompartmented models is also illustrated by the measured and

predicted isotopomers obtained from 28% U-13C glucose LE for two amino acid fragments, His[2345]

and Met[2345] (Figure S2b).

This article is protected by copyright. All rights reserved.


To further substantiate our compartmented model, we examined the predicted compartmental

location of the G6PDH enzymes encoded in the poplar genome using PLEXdb (Wise et al., 2007).
Accepted Article
Transcript levels corresponding to the plastid-localized G6PDH were considerably reduced under

low-nitrogen conditions while those corresponding to the cytosol-localized G6PDH showed little

difference between all the environmental conditions (Table S7, Supporting note 2). This difference

in gene expression is consistent with the flux results where plastidic G6PDH flux accounted for the

major difference in total G6PDH flux.

3.3 Upregulation of genes encoding enzymes involved in lignin pathways and higher proportions of

proteinogenic aromatic amino acids suggest a rerouting of redundant carbon atoms under low-

nitrogen conditions

Under low-nitrogen conditions (Cn and cn), we observed higher proportions of aromatic amino acids

such as Phe and Tyr (Figure 1e). Because the synthesis and degradation of aromatic amino acids are

closely connected to lignin metabolism (Cong et al., 2013), the higher proportions of aromatic amino

acids that we observed could be indicative of changes in lignin metabolism. Lignin biosynthesis

involves the degradation of phenylalanine to 4-coumarate (a monolignol), the conversion between

4-coumarate and caffeoyl-CoA, the synthesis of sinapate (a monolignol) and the assembly of lignin

from monolignols (MetaCyc, Karp et al., 2002). Our transcriptome analysis indicated that the

transcript abundance of many genes encoding enzymes involved in lignin pathways increased under

low-nitrogen conditions. This included (i) phenylalanine ammonia lyase (PAL) which converts Phe to

trans-cinnamate; (ii) cinnamoyl-CoA reductase which converts coumaryl-CoA to coumaraldehyde

(CCAR); and (iii) caffeic acid 3-O-methyltransferase (CA3MT) involved in the conversion of caffeoyl-

CoA and feruloyl-CoA. In addition to changes in gene expression related to lignin biosynthesis, we

also observed that two genes encoding arogenate/prephenate dehydratase (APDH) were

upregulated under the Cn condition. Since arogenate and prephenate are intermediate precursors of

This article is protected by copyright. All rights reserved.


phenylalanine, this observation suggests increased activity of the Phe biosynthetic pathway. The

upregulation of lignin pathway related genes combined with the increased expression of genes
Accepted Article
encoding enzymes involved in Phe metabolism suggest that excess carbon atoms were redirected to

these pathways from central carbon metabolism under the low-nitrogen conditions. This could help

to alleviate changes in the balance of carbon and nitrogen uptake (as illustrated in Figure 1b).

3.4 Some flux changes including oxPPP and anaplerotic reactions are likely regulated at the

transcriptional level

Under the low-nitrogen conditions, 13C MFA estimated reduced flux through the plastidic oxPPP

pathway, increased flux of anaplerotic reactions and fructose bisphosphate aldolase reactions. These

results are similar to the flux responses reported by Allen and Young (2013) for soybean embryos fed

different ratios of carbon and nitrogen. However, we did observe some notable exceptions in flux

results between poplar cell suspensions and soybean embryos, including increased activity of the

upstream components of the TCA cycle under low-nitrogen supply. We also found that the plastidic

oxPPP flux spanned almost half of the allowable stoichiometric range across the four conditions (e.g.

CN: 6.37 ± 0.89, 84% of allowable stoichiometric range; Cn: 1.88 ± 0.65, 45% of allowable

stoichiometric range). Such flexibility in oxPPP flux is higher than what has previously been

measured in plants via isotope labeling. For example, Masakapalli et al. (Masakapalli et al., 2013)

found that the plastidic oxPPP flux in Arabidopsis cells growing in ammonium-free media was about

50% higher than the plastidic oxPPP flux in cells growing in MS media. Allen and Young (2013)

observed a one-fold increase in oxPPP flux in developing soybean embryos after increasing the

carbon to nitrogen ratio from 13:1 to 37:1 in growth medium. Despite the substantial difference in

fluxes between the four environmental conditions used in this study, the majority of the observed

flux changes agreed with the microarray data (as mentioned in Section 2.6), suggesting that a

significant number of these metabolic reactions might be controlled at the transcriptional level.

This article is protected by copyright. All rights reserved.


However, there are exceptions for some reactions, such as GAPDH and PDC, where gene expression

and the flux estimations deviated or showed opposite trends (Figure 6). In these cases the enzymes
Accepted Article
encoded by these genes are likely to be regulated more by post-transcriptional or post-translational

intracellular processes as opposed to transcription (Tovar-Mendez et al., 2003; Fan et al., 2014). In

fact, post-translational regulation of plant GAPDH through oxidative modifications has been

extensively reported (reviewed by Zaffagnini et al., 2013). Additionally, previous studies with yeast

PDC revealed that its activity can be regulated by phosphorylation (de Assis et al., 2013) or by

allosteric binding (Kutter et al., 2009).

3.5 Lower NADPH production through oxPPP is consistent with lower NADPH requirement under

low-nitrogen conditions

It has been shown that altering nutrient conditions can significantly redirect carbon flow from

glycolysis to PPP (Ayar-Kayali, 2010; Cadière et al., 2011; Masakapalli et al., 2013; Masakapalli,

Ritala, et al., 2014). Interestingly, in the current study, we observed substantially reversed flux

through plastid phosphoglucose isomerase (PGIFP, Section 2.3), as a consequence of the large flux

associated with oxidative PPP (G6PDHP, Section 2.3). This indicates that the poplar cells produced

pyruvate, the substrate of the downstream TCA cycle, primarily through the PPP and not by

glycolysis. Moreover, our flux estimations also showed that poplar cells under high-nitrogen

conditions exhibited a higher oxidative PPP flux, and therefore, a greater reversed PGI flux. The

three reactions in oxPPP, catalyzed by G6PDH, gluconolactonase and 6-phosphogluconate, together

produce two NADPH at the cost of losing one carbon atom in the form of CO2. One major consumer

of NADPH or NADH is nitrate reduction. The conversion of one mole of nitrate to ammonia costs one

mole of NADPH or NADH and six moles of reduced ferredoxin (i.e. four moles of NADPH equivalent).

The activity of the plastidic oxidative PPP is directly linked to the assimilation of inorganic nitrogen

since a knockdown of the plastidic oxidative PPP enzyme 6-phosphogluconolactonase in Arabidopsis

This article is protected by copyright. All rights reserved.


resulted in nitrogen-starved phenotype characterized by reduced amino acid contents (Bussell et al.,

2013). Another major NADPH sink is the fatty acid synthesis pathway. The addition of every two-
Accepted Article
carbon unit to a growing fatty acid chain is accompanied by two reductive steps – 3-ketoacyl-ACP

reductase and enoyl-ACP reductase – each of which converts an NADPH molecule to NADP+. Our

estimated fluxes showed that upon nitrogen deficiency, poplar cells redirected carbon flow from

oxPPP to downstream components of the TCA cycle, which was accompanied by a decrease in the

rate of NADPH production. This is consistent with the similar CO2 out-flux associated with the

glucose uptake rate across the four conditions (Figure 5b), and also explained the similar biomass

yields irrespective of carbon levels (shown in Figure 1c). Quantitatively, lower NADPH production via

plastidic oxPPP corresponded well to lower NADPH consumption involved in nitrate reduction, lipid

synthesis and amino acid synthesis under the low-nitrogen conditions (Figure S3, Table S8). It

appears that for the poplar cell suspensions studied here, the reduced NADPH production via the

oxPPP cannot be offset by increased flux of malate decarboxylation. Our observation, however, is in

contrast to a recent study on alga Chlorella protothecoides (Gopalakrishnan et al., 2015), which

found that upon nitrogen limitation the oxPPP flux increased by about 6-fold at the expense of

decreased PGI flux. This difference might be explained by the increased lipid production requirement

in algae under nitrogen limitation (Fakhry and El Maghraby, 2015), which was not observed in our

poplar cells (Table 1).

3.6 Photosynthetic and light-harvesting proteins may serve as nitrogen sinks

The gene expression data generated from these experiments contained substantial information

related to the operation of metabolic processes beyond central carbon metabolism under low-

nitrogen availability. One interesting observation is the lower expression levels of photosystem and

light-harvesting genes under low-nitrogen availability. Previous reports with various species have

shown that the expression levels of many photosynthesis related genes can be affected by limited

This article is protected by copyright. All rights reserved.


nitrogen availability. For example, Juergens, et al. (2015) found that nitrogen starvation in

Chlamydomonas reinhardtii resulted in reduced transcript levels and abundance of proteins involved
Accepted Article
in photosystems I and II, likely to avoid energy overflow caused by photosynthesis. Similar results

have also been observed for maize in which many photosynthesis related genes were repressed (e.g.

GRMZM2G461279 and GRMZM2G429955 that encode Photosystem II type I chlorophyll a/b-binding

protein show -1.1 log2-fold change under nitrogen limitation) in nitrogen deficient leaves (Schluter et

al., 2012). Martin et al. (2002) also studied the effects of varying carbon and nitrogen availability on

the growth of Arabidopsis seedlings, and found that the expression levels of photosynthetic genes

dropped significantly under nitrogen starvation conditions. We have previously demonstrated that

the poplar cells used in this study did not fix atmospheric CO2 (Nargund et al., 2014). Even if

photosynthetic electron transport was active, it would have affected the electron or energy balance,

but not the carbon balance in our metabolic network. This is because the metabolites in our

metabolic network are not involved in the photosynthetic electron transport chain. Interestingly, in

these heterotrophic tissues we observed significant downregulation of photosystem genes under

low-nitrogen supply, suggesting that these photosystem and light-harvesting proteins served as

nitrogen sinks in poplar cells when nitrogen supply is sufficient. This hypothesis was supported by

the substantial larger reflux of acetyl-CoA under the Cn condition (0.35 ± 0.04 μmol day-1 [mg

biomass]-1), compared to the CN condition (0.14 ± 0.03 μmol day-1 [mg biomass]-1), since acetyl-CoA

can be produced from the degradation of lipids and various amino acids including Glu, Thr, Val and

Tyr. Moreover, our previous proteomic study of poplar bark tissue also supports this hypothesis

(Islam et al., 2015). The poplar bark tissue employed in that study was also heterotrophic and

derived from the same general part of the plant as the cell suspensions. Multidimensional protein

identification technology (MudPIT) analysis of this tissue revealed the synthesis of RuBisCO and

several photosystem II proteins such as photosystem II subunit P-1 during spring shoot growth and

nitrogen remobilization. This validates that RuBisCO and photosystem proteins can act in nitrogen

storage and mobilization in a heterotrophic tissue from poplar.

This article is protected by copyright. All rights reserved.


Based on comparisons of cell growth, intracellular fluxes and transcriptomics data across the four

environmental conditions, we observed that the poplar cell suspensions enhance their NUE in
Accepted Article
respond to low nitrogen availability (Figure 8). When nitrogen supply is abundant, the oxPPP is more

active and produces more reducing equivalents (e.g. NADPH), while inorganic nitrogen is likely to be

partitioned into various storage proteins including photosystem proteins. In nitrogen-limited

conditions, the oxPPP flux is reduced, consistent with the lower demand of reducing power to

assimilate and reduce NO3-, and storage proteins including photosystems proteins are synthesized to

a lesser degree. In addition, a larger proportion of the carbon flux is redirected to the TCA cycle,

maintaining a robust ATP production for major cellular activities including biomass synthesis.

4. Conclusion and Outlook

In this study, we applied a fully compartmented model of central carbon metabolism for MFA. This

model was justified not only by the measured isotopomer abundance data, but also by the

expression levels and compartment assignment of the genes encoding enzymes involved in pentose

phosphate pathways. Estimated fluxes explained cell growth characteristics and indicated which

fluxes could be transcriptionally regulated. However, the flux results alone were insufficient to

directly reveal how the poplar cells improve NUE at low nitrogen supply. This is because the

understanding of cellular nitrogen recycling mechanism requires information including protein

abundance and/or expression levels of genes which cannot be obtained from MFA. However,

predicting fluxes from gene expression data alone is not as straightforward as using isotope-assisted

MFA, since gene expression may not necessarily be consistent with enzyme levels, enzyme activities

or flux values (Arvas et al., 2011; Vogel and Marcotte, 2012; Kumar et al., 2015). In fact, Vogel and

Marcotte (2012) found that the correlation between gene expression levels and protein levels is only

about 40% in various organisms. Nevertheless, transcriptomic data covers a wide range of cellular

This article is protected by copyright. All rights reserved.


activities and thus provides valuable insight to how cells responded to nitrogen deficiency beyond

central carbon metabolism. Here we successfully connected the fluxomics and transcriptomics
Accepted Article
observations to identify possible mechanisms of how poplar cells improve NUE under low-nitrogen

supply. We also showed that there are consistencies between fluxes and gene expression levels for

some pathways (including G6PDH and PEPC) indicating these pathways could be regulated at the

transcriptional level.

To fully understand the cellular basis of nitrogen cycling in poplar cells and unequivocally identify

which proteins are recruited and used in this cycling, additional approaches such as proteomics will

be required. For example, approaches similar to those used in Arabidopsis thaliana to identify

peptides via 15N isotopic labeling (Nelson et al., 2007) and compartmental information of specific

proteins using MFA and 13C peptide labelling experiments (Mandy et al., 2014) could help to better

understand cellular nitrogen cycling in poplar cells. The application of genome-scale models

(Poolman et al., 2009; Simons et al., 2014) may also be of value in resolving where and how carbon

or nitrogen flow is redirected. The application of these approaches to the poplar cell suspension

systems would advance our understanding of NUE response in poplar cells at the metabolic and

genetic levels and allow for the development of potential strategies for the improvement of NUE.

5. Experimental Procedures

5.1 Poplar cell suspension culture and cell growth measurements

Throughout this work we used poplar (Populus. tremula × Populus. alba clone 717-1B4) cell

suspensions maintained in 125 mL Erlenmeyer flasks shaken at 125 rpm. The cell suspensions were

initiated from internode sections of in-vitro shoot cultures and represented mesophyll parenchyma

cells. In stems, it is the parenchyma cells that function in storage, metabolism as well as a critical role

in transport. Therefore, the metabolic response reported in this article can be understood to reflect

the nitrogen recycling machinery in poplar stem tissues.

This article is protected by copyright. All rights reserved.


The cells were grown and maintained in MS media supplemented with 20 g L-1 glucose and 1.0 mg L-1

2,4D, 0.1 mg L-1 NAA, and 0.01 mg L-1 BA. The cell suspensions were transferred weekly to fresh
Accepted Article
media. Culture flask containing the cells were placed at 20 °C under continuous light from cool-white

fluorescent light (Ecolux® Technology Plant and Aquarium F40T12 bulbs, photosynthetically active

radiation of 200-300 μmol m-2 s-1, incident radiation on flasks 28.3 ± 2.6 μmol m-2 s-1). The major

reason for growing the cells under continuous light is to minimize the possible influence of diurnal

fluctuations on cellular metabolism (Urbanczyk-Wochniak et al., 2005; Gibon et al., 2006; Lee et al.,

2010). We subcultured the cell suspensions every 7 d by transferring 0.6 g fresh cells from the

previous growth cycle into 30 mL Murashige and Skoog medium (Phytotechnology Laboratories,

Shawnee Mission, KS) containing 20 g L-1 glucose and 1.0 mg L-1 2,4D, 0.1 mg L-1 NAA, and 0.01 mg L-1

BA. The four carbon and nitrogen levels used in this study consisted of (i) high carbon-high nitrogen

(CN: 20 g L-1 Glucose, 1,650 mg L-1 NH4NO3 and 1,900 mg L-1 KNO3 ), (ii) high carbon-low nitrogen (Cn :

20 g L-1 Glucose, 264 mg L-1 NH4NO3 and 304 mg L-1 KNO3), (iii) low carbon-high nitrogen (cN: 10 g L-1

Glucose, 1,650 mg L-1 NH4NO3 and 1,900 mg L-1 KNO3) and (iv) low carbon-low nitrogen (cn : 10 g L-1

Glucose, 264 mg L-1 NH4NO3 and 304 mg L-1 KNO3). The carbon and nitrogen source concentrations

were initial concentrations in the fresh media, and thus they gradually decreased over a 7-d period

before they were replenished every 7 d. Nevertheless, a comparison of the residual nutrient amount

between the four environmental conditions (Figure 1c and 1d) shows that at any time during the

growth cycle, cells under the high-carbon or high-nitrogen conditions were always exposed to higher

carbon or nitrogen concentrations as compared to the cells under the low-carbon or low-nitrogen

conditions. In all the environmental conditions, the ratio of NH4NO3 to KNO3 was constant. We chose

the carbon and nitrogen levels based on preliminary experiments that showed under these nutrient

concentrations the suspension cultures grew continuously from 0 d to 7 d, irrespective of the

treatments but also exhibited difference in growth rates and biomass production relative to the CN

condition.

This article is protected by copyright. All rights reserved.


Growth measurements of cell suspensions were made at 2 d, 4 d, 6 d and 7 d, after transfer to fresh

media for each environmental condition. At each sampling interval, cells from three individual shake
Accepted Article
flasks (representing biological replicates), were harvested by vacuum filtration through glass

microfiber filter paper (Whatman, Piscataway, NJ). Fresh weights of the harvested cells were

obtained by weighing them immediately after the filtration; dry weights were obtained by weighing

the harvested cells after lyophilization. The media collected by vacuum filtration was used to

measure residual glucose using a glucose meter (YSI Incorporated, Yellow Springs, OH), and residual

nitrate and ammonium using an assay kit (Sigma-Aldrich, St. Louis, MO). The residual nutrient

concentrations were also performed on three biological replicates for each environmental condition.

Biomass yield relative to carbon or nitrogen was estimated from the slopes of the fitted lines in

Figure 1c and 1d. 95% confidence intervals of the growth rate μ and the slopes in (c) – (d) were

calculated from statistical bootstrapping based on the mean and the standard deviations of the

measured data. Specifically, from the mean and standard deviations of the cell dry weight and

nutrient concentration measurements, we obtained 2000 synthetic datasets, fit exponential curves

(Figure 1a) or straight lines (Figures 1b, 1c, 1d) through them and calculated either the growth rate μ

(Figure 1a) or the slope of the fitted line (Figures 1b, 1c, 1d). The results from the 2000 fits were

listed at ascending order, from which the top and bottom 2.5% data points were discarded to obtain

95% confidence intervals.

5.2 Three-week 13C isotope labeling experiments

For 13C labeling, three replicate flasks of cells for each environmental condition were grown in media

where the carbon was supplied as either (i) 98% U-13C glucose, (ii) 98% 1-13C glucose or (iii) a mixture

of 28% U-13C and 72% naturally abundant glucose (Cambridge Isotope Laboratories, Andover, MA).

All replicates were grown for 3 weeks, with weekly subculturing to the appropriate 13C-labeled

This article is protected by copyright. All rights reserved.


medium. At each subculture, 0.6 g of fresh cells were transferred to fresh media. This three-week

cycle labeling strategy prevented the depletion of carbon and nitrogen and helped to establish the
Accepted Article
metabolic steady state (Figure 2). Cells were harvested after 7 d of growth following the third weekly

sub-culturing using vacuum filtration and immediately frozen in liquid nitrogen, and stored at –80 °C

until used for analysis, In addition, three replicate flasks were also grown at the same time using

unlabeled glucose and were used for microarray analysis.

5.3 Extraction and quantification methods of biomass components

The frozen cells from the LEs were lyophilized and ground to fine powder with mortar and pestle.

Lipids were extracted with the Folch method (Folch et al., 1957). Briefly, ground cells were placed in

2 mL of a methanol:water (4/3, v/v) mixture and heated at 70 °C for 15 min. Chloroform was then

added to give a mixture of chloroform/methanol/water (8/4/3, v/v/v). After heating at 70 °C for

another 5 min the extractions were centrifuged to separate the layers and the lipid layer (lower

layer) was collected and dried under a stream of nitrogen gas. Total dry lipids were weighed.

Transesterification of lipids was performed by adding 1 mL 3 N HCl in methanol to the dried lipids

and heating at 70 °C for 1 hour. Following transesterification, hexane was added and the fatty acid

methyl esters (FAME) were collected from the upper phase after centrifugation and analyzed by GC-

MS. After baseline correction, the mass spectrum corresponding to each fatty acid component was

integrated, which was then used to obtain its relative abundance. Heptadecanoic acid was used as

an external standard.

Starch was extracted from ground cells by re-suspending in 1.5 mL of 0.1 M sodium acetate buffer

(pH 5.0) and heating to 90 °C for 20 min. Soluble starch was digested with amylase (50 μg/mL

overnight at 37 °C and the digestion was terminated by heating at 100 °C for 10 min. The digestion

product, maltose, was collected by removing the supernatant after centrifugation, dried under a

This article is protected by copyright. All rights reserved.


stream of nitrogen gas, and weighed. The dry weight of starch was estimated by subtracting water

added during hydrolysis (1 molecule water or 18 g mol-1 per maltose molecule, 342 g mol-1).
Accepted Article
Protocols for the extraction, derivatization and quantification of soluble metabolites and

proteinogenic amino acids, parameters used for GC-MS analysis, and correction of the raw mass

isotopomer abundance data for the natural abundances of hydrogen, nitrogen, oxygen, sulfur,

silicon and non-metabolic carbon atoms were used as previously described (Nargund et al., 2014).

5.4 Flux evaluation from isotopomer data

Our established metabolic network model consisted of 34 free fluxes to fit and 14 measured fluxes.

Of these 34 free fluxes, there were nine net fluxes among which one net flux is related to metabolite

reflux (Nargund et al., 2014), 24 reversibility extents and one scrambling extent, as defined in

Supplementary Material of Sriram et al. (Sriram et al., 2004). The number of independent

isotopomer measurements from three labeling experiments was in the range 363 and 373 for the

four environmental conditions (Supporting note 2), which is much larger than the number of free

fluxes in the metabolic model. We calculated glucose influx from its concentration in spent media

after cell collection at 7 d and the cell growth rate (Fig. 1a and 1c). Biosynthetic fluxes of amino acids

and soluble metabolites were determined from compositions of proteinogenic amino acids,

compositions of biomass components and cell growth rate (Figure 1, Table 1 and Table S1).

We employed our 13C MFA program NMR2Flux+ for evaluation of metabolic fluxes from the mass

isotopomer and biosynthetic flux data. Details about NMR2Flux+ have been previously described

(Sriram et al., 2004; Sriram et al., 2008; Nargund and Sriram, 2013). Briefly, this program is supplied

with information on the metabolic network and LE(s) including metabolite names, chemical reaction

This article is protected by copyright. All rights reserved.


stoichiometries, carbon atom rearrangements, isotopically labeled forms of the carbon source(s),

mass isotopomer measurements and relationship of the measured mass isotopomer fragments to
Accepted Article
central carbon precursors. It uses a flux-isotopomer model, nonlinear optimization and Monte Carlo

statistical analysis to calculate fluxes and confidence intervals from a given set of amino acid mass

isotopomer measurements. To assess how well the evaluated fluxes account for the supplied

isotopomer measurements, the program uses a sum of squared residuals (SSR) metric, calculated as:

2
Ij IJx
SSR
j j

where Ix represents measured (mass) isotopomer abundances, I represents calculated isotopomer

abundances corresponding to the evaluated set of fluxes, σ represents the measured standard

deviations of the isotopomer abundances and the index j cycles through all the isotopomer

abundances. For each of the environmental conditions, the SSR was calculated for approximately

400 isotopomer abundances from three parallel LEs. The statistically acceptable SSR was calculated

via the chi-square test, given a confidence level (95%), the degrees of freedom (DOF_data), which is

equal to the number of independent measurements minus the number of free fluxes (Supporting

note 2).

Subsequently, we carried out a bootstrap Monte-Carlo algorithm to calculate the standard

deviations and 95% confidence intervals of the fluxes. Briefly, for each condition, we used the

measured isotopomer abundances and standard deviations to generate N synthetic, normally

distributed isotopomer abundance datasets (N was approximately 400). Then we independently

evaluated fluxes from each synthetic dataset by using the optimization procedure described above.

From the distributions of the resulting N sets of fluxes, we calculated the standard deviations of the

fluxes. To find confidence intervals, the flux values of each reaction were listed in ascending order,

This article is protected by copyright. All rights reserved.


from which the top and bottom 2.5% data points were discarded. All NMR2Flux+ simulations were

performed on three identical workstations equipped with an Intel Xeon E5620 CPU and 32 GB of
Accepted Article
RAM.

5.5 RNA extraction and microarray data processing

Suspension cells for microarray were collected by vacuum filtration and the collected cells were

immediately frozen in liquid nitrogen. Frozen cells were ground to fine powder using pre-chilled

mortar and pestle and the ground frozen cells were transferred to 2 mL tubes containing 1~1.5 mL

RLT buffer (QIAGEN, Venlo, Netherlands), 0.01 g soluble polyvinylpyrrolidone (Sigma-Aldrich, St.

Louis, MO) and 10 μg/mL of 2-Mercaptoethanol (Sigma-Aldrich, St. Louis, MO). After the cells were

suspended in the buffer, 0.4 mL 5 M potassium acetate (Sigma-Aldrich, St. Louis, MO) at pH 6.5 was

added. After vortexing the extraction was centrifuged at 12, 000 rpm, the supernatant was

transferred to two new 2 mL clear tubes and loaded into the automated QIACube (QIAGEN) using

RNeasy plant mini kit (QIAGEN) with on column DNase digestion. RNA was quantified using a

NanoDrop spectrophotometer (Thermo Scientific, Wilmington, DE).

RNA from three biological replications of each environmental condition was submitted to the

Biopolymer/Genomics Core Laboratory (University of Maryland, Baltimore) for microarray

processing by hybridization to the GeneChip® poplar genome arrays (Affymetrix). The microarray

raw data contained 61,413 probesets and was analyzed using ArrayStar 5.0 (DNASTAR, Inc., Madison,

WI) with the RMA normalization method. In addition, we also performed an unsupervised clustering

analysis [principal components analysis (PCA)] to make an initial comparison of the four

environmental conditions. The correspondence between microarray probesets and gene IDs were

obtained from Affymetrix poplar annotations, release 36. Gene annotations were obtained from

PoplarCyc 5.0, UniProt and PLEXdb.

This article is protected by copyright. All rights reserved.


Author contributions
Accepted Article
XZ, AM, SN, GDC and GS conceived this study; XZ, AM, SN and GS designed it. GDC provided poplar

cell suspension cultures; XZ, AM and SN performed experiments; SN, AM, XZ and GS analyzed data

and drafted the manuscript. XZ, GDC and GS critically edited the manuscript.

Acknowledgments

The authors thank Margaret N. Simons and Whitney Hollinshead for assistance with experiments.

This work was funded by the U.S. National Science Foundation (award IOS-0922650) to GDC and GS.

The authors declare no conflict of interest.

Short Supporting Information Legends

Figure S1. Unsupervised comparison of isotopomer data and clustering analysis of gene expression

between carbon and nitrogen conditions.

Figure S2. Comparison between Ser and glycerol (Glr) MIDs and between calculated isotopomer and

measured isotopomers supports a compartmented model.

Figure S3. Lower NADPH supply under low-nitrogen conditions is consistent with lower NADPH

demand.

Table S1. Abundances of proteinogenic amino acids from poplar suspension cells supplied with

different levels of carbon and nitrogen.

Table S2-S4. Mass isotopomer distribution (MID) measurements for three 13C glucose LEs for each of

the environmental conditions.

This article is protected by copyright. All rights reserved.


Table S5. The compartmented metabolic network model used for 13C MFA.

Table S6. Metabolic fluxes evaluated by 13C MFA for each of the environmental conditions.
Accepted Article
Table S7. Microarray results for genes that were differentially expressed between the four

environmental conditions.

Table S8. NADPH demand and production rate calculated from biomass compositions and fluxes.

Supporting note 1. Higher nitrogen use efficiency (NUE) under low nitrogen conditions was also

supported by analysis of biomass yield relative to nitrogen amount in synthesized proteins.

Supporting note 2. Justification of applying a compartmented metabolic network model in MFA.

References

Allen, A.E., Dupont, C.L., Obornik, M., et al. (2011) Evolution and metabolic significance of the urea
cycle in photosynthetic diatoms. Nature, 473, 203–207.

Allen, D.K. and Young, J.D. (2013) Carbon and Nitrogen Provisions Alter the Metabolic Flux in
Developing Soybean Embryos. Plant Physiology, 161, 1458–1475.

Amtmann, A. and Armengaud, P. (2009) Effects of N, P, K and S on metabolism: new knowledge


gained from multi-level analysis. Current Opinion in Plant Biology, 12, 275–283.

Apweiler, R., Bairoch, A., Wu, C.H., et al. (2004) UniProt: the Universal Protein knowledgebase.
Nucleic Acids Research, 32, D115-119.

Arvas, M., Pakula, T., Smit, B., et al. (2011) Correlation of gene expression and protein production
rate - a system wide study. BMC Genomics, 12.

Assis, L.J. de, Zingali, R.B., Masuda, C.A., Rodrigues, S.P. and Montero-Lomelí, M. (2013) Pyruvate
decarboxylase activity is regulated by the Ser/Thr protein phosphatase Sit4p in the yeast
Saccharomyces cerevisiae. FEMS Yeast Research, 13, 518–528.

Ayar-Kayali, H. (2010) Pentose phosphate pathway flux analysis for glycopeptide antibiotic
vancomycin production during glucose-limited cultivation of Amycolatopsis orientalis.
Preparative Biochemistry and Biotechnology, 41, 94–105.

This article is protected by copyright. All rights reserved.


Bussell, J.D., Keech, O., Fenske, R. and Smith, S.M. (2013) Requirement for the plastidial oxidative
pentose phosphate pathway for nitrate assimilation in Arabidopsis. The Plant Journal, 75,
578–591.
Accepted Article
Cadière, A., Ortiz-Julien, A., Camarasa, C. and Dequin, S. (2011) Evolutionary engineered
Saccharomyces cerevisiae wine yeast strains with increased in vivo flux through the pentose
phosphate pathway. Metabolic Engineering, 13, 263–271.

Canales, J., Moyano, T.C., Villarroel, E. and Gutiérrez, R.A. (2014) Systems analysis of transcriptome
data provides new hypotheses about Arabidopsis root response to nitrate treatments.
Frontiers in Plant Science, 5.

Cantón, F.R., Suárez, M.F. and Cánovas, F.M. (2005) Molecular aspects of nitrogen mobilization and
recycling in trees. Photosynthesis Research, 83, 265–278.

Chen, M., Wang, C., Bao, H., Chen, H. and Wang, Y. (2016) Genome-wide identification and
characterization of novel lncRNAs in Populus under nitrogen deficiency. Molecular genetics
and genomics, 291, 1663–1680.

Cong, F., Diehl, B.G., Hill, J.L., Brown, N.R. and Tien, M. (2013) Covalent bond formation between
amino acids and lignin: Cross-coupling between proteins and lignin. Phytochemistry, 96,
449–456.

Escobar, M.A., Geisler, D.A. and Rasmusson, A.G. (2006) Reorganization of the alternative pathways
of the Arabidopsis respiratory chain by nitrogen supply: opposing effects of ammonium and
nitrate. The Plant Journal, 45, 775–788.

Fakhry, E.M. and El Maghraby, D.M. (2015) Lipid accumulation in response to nitrogen limitation
and variation of temperature in Nannochloropsis salina. Botanical Studies, 56.

Fan, J., Kang, H.-B., Shan, C., et al. (2014) Tyr-301 Phosphorylation Inhibits Pyruvate Dehydrogenase
by Blocking Substrate Binding and Promotes the Warburg Effect. Journal of Biological
Chemistry, 289, 26533–26541.

Folch, J., Lees, M. and Sloane Stanley, G.H. (1957) A simple method for the isolation and purification
of total lipides from animal tissues. J. Biol. Chem., 226, 497–509.

Gao, P., Xin, Z. and Zheng, Z.-L. (2008) The OSU1/QUA2/TSD2-encoded putative methyltransferase
is a critical modulator of carbon and nitrogen nutrient balance response in Arabidopsis. PLoS
ONE, 3, e1387.

Gibon, Y., Usadel, B., Blaesing, O.E., Kamlage, B., Hoehne, M., Trethewey, R. and Stitt, M. (2006)
Integration of metabolite with transcript and enzyme activity profiling during diurnal cycles
in Arabidopsis rosettes. Genome Biology, 7, R76.

Gopalakrishnan, S., Baker, J., Kristoffersen, L. and Betenbaugh, M.J. (2015) Redistribution of
metabolic fluxes in Chlorella protothecoides by variation of media nitrogen concentration.
Metabolic Engineering Communications, 2, 124–131.

This article is protected by copyright. All rights reserved.


Gutierrez, R.A., Gifford, M.L., Poultney, C., Wang, R., Shasha, D.E., Coruzzi, G.M. and Crawford,
N.M. (2007) Insights into the genomic nitrate response using genetics and the Sungear
Software System. Journal of Experimental Botany, 58, 2359–2367.
Accepted Article
Gutierrez, R.A., Lejay, L.V., Dean, A., Chiaromonte, F., Shasha, D.E. and Coruzzi, G.M. (2007)
Qualitative network models and genome-wide expression data define carbon/nitrogen-
responsive molecular machines in Arabidopsis. Genome Biology, 8.

He, L., Xiao, Y., Gebreselassie, N., Zhang, F., Antoniewiez, M.R., Tang, Y.J. and Peng, L. (2014)
Central metabolic responses to the overproduction of fatty acids in Escherichia coli based on
13
C-metabolic flux analysis. Biotechnology and Bioengineering, 111, 575–585.

Heldt, H.-W. and Piechulla, B. (2010) Plant Biochemistry 4th ed., Academic Press.

Hockin, N.L., Mock, T., Mulholland, F., Kopriva, S. and Malin, G. (2011) The Response of Diatom
Central Carbon Metabolism to Nitrogen Starvation Is Different from That of Green Algae and
Higher Plants. Plant Physiology, 158, 299–312.

Islam, N., Li, G., Garrett, W.M., Lin, R., Sriram, G., Cooper, B. and Coleman, G.D. (2015) Proteomics
of nitrogen remobilization in poplar bark. Journal of Proteome Research, 14, 1112–1126.

Juergens, M.T., Deshpande, R.R., Lucker, B.F., et al. (2015) The regulation of photosynthetic
structure and function during nitrogen deprivation in Chlamydomonas reinhardtii. Plant
Physiology, 167, 558–573.

Junker, B.H., Lonien, J., Heady, L.E., Rogers, A. and Schwender, J. (2007) Parallel determination of
enzyme activities and in vivo fluxes in Brassica napus embryos grown on organic or inorganic
nitrogen source. Phytochemistry, 68, 2232–2242.

Karp, P.D., Riley, M., Paley, S.M. and Pellegrini-Toole, A. (2002) The MetaCyc Database. Nucleic
Acids Res., 30, 59–61.

Kind, S., Becker, J. and Wittmann, C. (2013) Increased lysine production by flux coupling of the
tricarboxylic acid cycle and the lysine biosynthetic pathway—Metabolic engineering of the
availability of succinyl-CoA in Corynebacterium glutamicum. Metabolic Engineering, 15, 184–
195.

Krouk, G., Mirowski, P., LeCun, Y., Shasha, D.E. and Coruzzi, G.M. (2010) Predictive network
modeling of the high-resolution dynamic plant transcriptome in response to nitrate. Genome
Biology, 11, R123.

Krouk, G., Tranchina, D., Lejay, L., Cruikshank, A.A., Shasha, D., Coruzzi, G.M. and Gutiérrez, R.A.
(2009) A Systems Approach Uncovers Restrictions for Signal Interactions Regulating
Genome-wide Responses to Nutritional Cues in Arabidopsis R. Ranganathan, ed. PLoS
Computational Biology, 5, e1000326.

Kruger, N.J. and Schaewen, A. von (2003) The oxidative pentose phosphate pathway: structure and
organisation. Current Opinion in Plant Biology, 6, 236–246.

This article is protected by copyright. All rights reserved.


Kumar, A., Pillay, B. and Olaniran, A.O. (2015) Enzyme activity and gene expression profiles of
Xanthobacter autotrophicus GJ10 during aerobic biodegradation of 1,2-dichloroethane.
World Journal of Microbiology and Biotechnology, 31, 1211–1216.
Accepted Article
Kutter, S., Weiss, M.S., Wille, G., Golbik, R., Spinka, M. and Konig, S. (2009) Covalently Bound
Substrate at the Regulatory Site of Yeast Pyruvate Decarboxylases Triggers Allosteric Enzyme
Activation. Journal of Biological Chemistry, 284, 12136–12144.

Lee, C.P., Eubel, H. and Millar, A.H. (2010) Diurnal Changes in Mitochondrial Function Reveal Daily
Optimization of Light and Dark Respiratory Metabolism in Arabidopsis. Molecular & Cellular
Proteomics, 9, 2125–2139.

Lonien, J. and Schwender, J. (2009) Analysis of metabolic flux phenotypes for two Arabidopsis
mutants with severe impairment in seed storage lipid synthesis. Plant Physiology, 151,
1617–1634.

Lonien, J. and Schwender, J. (2009) Analysis of metabolic flux phenotypes for two Arabidopsis
thaliana mutants with severe impairment in seed storage lipid synthesis. Plant Physiology,
151, 1617–1634.

Ma, F., Jazmin, L.J., Young, J.D. and Allen, D.K. (2014) Isotopically nonstationary 13 C flux analysis of
changes in Arabidopsis thaliana leaf metabolism due to high light acclimation. Proceedings
of the National Academy of Sciences, 111, 16967–16972.

Mandy, D.E., Goldford, J.E., Yang, H., Allen, D.K. and Libourel, I.G.L. (2014) Metabolic flux analysis
using 13 C peptide label measurements. The Plant Journal, 77, 476–486.

Martin, T., Oswald, O. and Graham, I.A. (2002) Arabidopsis seedling growth, storage lipid
mobilization, and photosynthetic gene expression are regulated by carbon:nitrogen
availability. Plant Physiology, 128, 472–481.

Masakapalli, S.K., Bryant, F.M., Kruger, N.J. and Ratcliffe, R.G. (2014) The metabolic flux phenotype
of heterotrophic Arabidopsis cells reveals a flexible balance between the cytosolic and
plastidic contributions to carbohydrate oxidation in response to phosphate limitation. The
Plant Journal, 78, 964–977.

Masakapalli, S.K., Kruger, N.J. and Ratcliffe, R.G. (2013) The metabolic flux phenotype of
heterotrophic Arabidopsis cells reveals a complex response to changes in nitrogen supply.
The Plant Journal, 74, 569–582.

Masakapalli, S.K., Lay, P.L., Huddleston, J.E., Pollock, N.L., Kruger, N.J. and Ratcliffe, R.G. (2010)
Subcellular flux analysis of central metabolism in a heterotrophic Arabidopsis thaliana cell
suspension using steady-state stable isotope labeling. Plant Physiology, 152, 602–609.

Masakapalli, S.K., Ritala, A., Dong, L., Krol, A.R. van der, Oksman-Caldentey, K.-M., Ratcliffe, R.G.
and Sweetlove, L.J. (2014) Metabolic flux phenotype of tobacco hairy roots engineered for
increased geraniol production. Phytochemistry, 99, 73–85.

This article is protected by copyright. All rights reserved.


Nargund, S., Misra, A., Zhang, X., Coleman, G.D. and Sriram, G. (2014) Flux and reflux: metabolite
reflux in plant suspension cells and its implications for isotope-assisted metabolic flux
analysis. Molecular BioSystems, 10, 1496–1508.
Accepted Article
Nargund, S. and Sriram, G. (2013) Designer labels for plant metabolism: statistical design of isotope
labeling experiments for improved quantification of flux in complex plant metabolic
networks. Molecular BioSystems, 9, 99.

Nelson, C.J., Huttlin, E.L., Hegeman, A.D., Harms, A.C. and Sussman, M.R. (2007) Implications
of15N-metabolic labeling for automated peptide identification inArabidopsis thaliana.
Proteomics, 7, 1279–1292.

Noorhana (2011) Effect of Nitrogen on Growth and Lipid Content of Chlorella pyrenoidosa. American
Journal of Biochemistry and Biotechnology, 7, 124–129.

Nunes-Nesi, A., Fernie, A.R. and Stitt, M. (2010) Metabolic and signaling aspects underpinning the
regulation of plant carbon nitrogen interactions. Molecular Plant, 3, 973–996.

Poolman, M.G., Miguet, L., Sweetlove, L.J. and Fell, D.A. (2009) A genome-scale metabolic model of
Arabidopsis thaliana and some of its properties. Plant Physiology, 151, 1570–1581.

Pregitzer, K.S., Dickmann, D.I., Hendrick, R. and Nguyen, P.V. (1990) Whole-tree carbon and
nitrogen partitioning in young hybrid poplars. Tree Physiology, 7, 79–93.

Sahrawy, M., Avila, C., Chueca, A., Cánovas, F.M. and López-Gorgé, J. (2004) Increased sucrose level
and altered nitrogen metabolism in Arabidopsis thaliana transgenic plants expressing
antisense chloroplastic fructose-1,6-bisphosphatase. Journal of Experimental Botany, 55,
2495–2503.

Schluter, U., Mascher, M., Colmsee, C., Scholz, U., Brautigam, A., Fahnenstich, H. and Sonnewald,
U. (2012) Maize source leaf adaptation to nitrogen deficiency affects not only nitrogen and
carbon metabolism nut also control of phosphate homeostasis. Plant Physiology, 160, 1384–
1406.

Simons, M., Misra, A. and Sriram, G. (2014) Genome-Scale Models of Plant Metabolism. In G.
Sriram, ed. Plant Metabolism. Methods in Molecular Biology. Humana Press, pp. 213–230.
Available at: http://link.springer.com/protocol/10.1007/978-1-62703-661-0_13 [Accessed
December 23, 2014].

Sriram, G., Fulton, D.B., Iyer, V.V., Peterson, J.M., Zhou, R., Westgate, M.E., Spalding, M.H. and
Shanks, J.V. (2004) Quantification of compartmented metabolic fluxes in developing
soybean embryos by employing biosynthetically directed fractional 13C labeling, two-
dimensional [13C, 1H] nuclear magnetic resonance, and comprehensive isotopomer
balancing. Plant Physiology, 136, 3043–3057.

Sriram, G., Iyer, V.V., Fulton, D.B. and Shanks, J.V. (2007) Identification of hexose hydrolysis
products in metabolic flux analytes: A case study of levulinic acid in plant protein

This article is protected by copyright. All rights reserved.


hydrolysate. Metabolic Engineering, 9, 442–451.

Sriram, G., Rahib, L., He, J.-S., Campos, A.E., Parr, L.S., Liao, J.C. and Dipple, K.M. (2008) Global
metabolic effects of glycerol kinase overexpression in rat hepatoma cells. Molecular Genetics
Accepted Article
and Metabolism, 93, 145–159.

Tovar-Mendez, A., Miernyk, J.A. and Randall, D.D. (2003) Regulation of pyruvate dehydrogenase
complex activity in plant cells. European Journal of Biochemistry, 270, 1043–1049.

Truong, Q., Koch, K., Yoon, J.M., Everard, J.D. and Shanks, J.V. (2013) Influence of carbon to
nitrogen ratios on soybean somatic embryo (cv. Jack) growth and composition. Journal of
Experimental Botany, 64, 2985–2995.

Urbanczyk-Wochniak, E., Baxter, C., Kolbe, A., Kopka, J., Sweetlove, L.J. and Fernie, A.R. (2005)
Profiling of diurnal patterns of metabolite and transcript abundance in potato (Solanum
tuberosum) leaves. Planta, 221, 891–903.

Villadsen, J., Nielsen, J. and Lidén, G. (2011) Bioreaction engineering principles 3. ed., New York:
Springer.

Vogel, C. and Marcotte, E.M. (2012) Insights into the regulation of protein abundance from
proteomic and transcriptomic analyses. Nature Reviews Genetics, 13, 227–232.

Wasylenko, T.M. and Stephanopoulos, G. (2014) Metabolomic and 13C-Metabolic Flux Analysis of a
Xylose-Consuming Saccharomyces cerevisiae Strain Expressing Xylose Isomerase: Xylose
Metabolic Flux Analysis. Biotechnology and Bioengineering, 112, 470–483.

Wise, R.P., Caldo, R.A., Hong, L., Shen, L., Cannon, E. and Dickerson, J.A. (2007) BarleyBase/PLEXdb.
Methods in Molecular Biology, 406, 347–363.

Zaffagnini, M., Fermani, S., Costa, A., Lemaire, S.D. and Trost, P. (2013) Plant cytoplasmic GAPDH:
redox post-translational modifications and moonlighting properties. Frontiers in Plant
Science, 4. Available at:
http://journal.frontiersin.org/article/10.3389/fpls.2013.00450/abstract [Accessed April 25,
2017].

Zamboni, N., Fendt, S.-M., Rühl, M. and Sauer, U. (2009) 13C-based metabolic flux analysis. Nature
Protocols, 4, 878–892.

This article is protected by copyright. All rights reserved.


Tables
Accepted Article
Table 1. Biomass composition of cells growing under the four environmental conditions.

Extraction and quantification of proteins, lipids, starch and soluble metabolites are as described in

Section 5.3 of the Experimental Procedures. After extraction was completed, the remaining insoluble

cell pellets were lyophilized and weighed. Standard deviations (SD) were calculated from three

biological replicates. The mass balance did not close at 100% for all the four environmental

conditions likely because: (i) There was soluble mineral ash that was extracted together with soluble

metabolites; (ii) there were nucleotides and nucleotides-derived products that were not analyzed as

they were not included in our metabolite network model; (iii) The recovery of amino acids by

protein hydrolysis usually cannot reach 100%; and (iv) small amount of biomass might be lost during

sample transfer.

% of cell CN cN Cn cn
dry wt Mean SD Mean SD Mean SD Mean SD
Proteins 12.73 0.86 8.72 0.51 7.91 0.69 7.72 0.35
Lipids 6.50 0.16 6.62 0.83 4.95 0.58 5.34 1.63
Starch 3.00 0.50 2.02 0.10 2.88 0.46 2.56 0.40
Insoluble
51.30 6.02 50.80 4.38 52.60 6.85 51.50 5.23
cell pellets

Figure captions

Figure 1. Effect of carbon and nitrogen levels on growth, NUE and proteinogenic amino acid

proportions of poplar cell suspensions.

Poplar cell suspension cultures were grown under four different carbon and nitrogen levels that

included high carbon-high nitrogen (CN; represented in black), low carbon-high nitrogen (cN;

represented in green), high carbon-low nitrogen (Cn; represented in blue) and low carbon-low

This article is protected by copyright. All rights reserved.


nitrogen (cn; represented in red). All values are based on three biological replicates. (a) Cell growth

was measured as dry biomass weight (mg) versus time (d) and is presented as μ (d-1) which is an
Accepted Article
t
index obtained by fitting the cell growth to the exponential curve ( X  X 0  e , X (mg) – cell dry

weight at time t (day); X0 (mg) – initial cell dry weight). The data point at 7 d for the cN condition was

not fitted to calculate μ since cell growth had apparently started to slow down. (b) Correlation

between residual nitrogen in media (mg) and residual carbon in media (mg). (c) Correlation between

dry biomass weight (mg) and residual carbon in media (mg). (d) Correlation between dry biomass

weight (mg) and residual nitrogen in media (mg). The slopes of the linear trendlines in (c) and (d),

represent the biomass yield relative to carbon or nitrogen, respectively, and are tabulated adjacent

to the plots. 95% confidence intervals of the slopes in (a) – (d) were calculated from Monte Carlo

simulations, based on the mean and the standard deviations of the measured data. (e) Protein

proportion of biomass and amino acid proportions in protein under four different environmental

conditions. The patterns in the stacked bar plot represent different amino acids, which are organized

based on their metabolic precursors as indicated to the right of the panel. Glu represents glutamate

+ glutamine, while Asp represents aspartate + asparagine.

Figure 2. Verification of isotopic steady state attainment via 13C enrichments and MIDs of soluble

amino acids from the cells harvested after three consecutive weeks of growth under the CN

condition with 28% U-13C labeled glucose.

(a) 13C enrichments of all the amino acid fragments from the cells harvested after 7 d, 14 d and 21 d.

(b) Principal component analysis of the entire MID dataset obtained from cells harvested after each

week of 13C treatment. This comparison shows that the 13C enrichments and MIDs of soluble amino

acids from cells of two-week old and three-week cell suspensions are similar, indicating the

approximate achievement of isotopic steady state. Error bars represent the standard deviations

between the three biological replicates.

This article is protected by copyright. All rights reserved.


Figure 3. MIDs of certain amino acids from cells grown under the different carbon and nitrogen

levels.
Accepted Article
MIDs of fragments of three amino acids (a) His (from a 98% 1-13C glucose LE), (b) Thr (from a 98% U-
13
C glucose LE) and (c) Glu (from a 98% U-13C glucose LE) are shown for the four environmental

conditions (CN, cN, Cn and cn). The different conditions are designated by four colors: black

represents CN; green represents cN, blue represents Cn and red represents cn. Different MIDs from

M+0 to M+n are distinguished by patterned filled bars, from bottom to top.

Figure 4. Metabolic flux map: low-nitrogen conditions yield lower plastidic oxPPP flux and higher

net anaplerotic flux.

(a) Flux map for a compartmented (cytosol, mitochondrion and plastid) metabolic network model,

with glycolysis and PPP duplicated in both cytosol and plastid. Circles represent metabolites and

arrows represent chemical reactions. Flux values for the four conditions CN (black), cN (green), Cn

(blue) and cn (red) are labeled in boxes adjacent to the arrows. The lengths of the grey bars

overlaying the numbers represent the scales of the flux values. For bidirectional reactions, net fluxes

are shown. (b) Flux map for an uncompartmented metabolic network model, where fluxes of

identical reactions in the cytosol and in the plastid from the compartmented model are

consolidated.

Figure 5. Metabolic flux map highlights: flux phenotype space and carbon partitioning.

(a) The four polyhedrons with different colored borders represent the flux spaces of the four

nutrient conditions used in this study: black, CN; green, cN; blue, Cn, red, cn. The fluxes (μmol day-1

[mg biomass]-1) represented are the oxPPP flux, the net anaplerotic flux (neither is easily measurable

This article is protected by copyright. All rights reserved.


from extracellular measurements) and the lipid synthesis flux (measurable from biomass

composition). Each flux space was constructed from constraints imposed by stoichiometry, reaction
Accepted Article
irreversibility and measured extracellular fluxes for the four carbon and nitrogen conditions.

Therefore, these surfaces are the loci of all possible flux values for the respective conditions. The

four ellipsoids denote the flux values under the four carbon and nitrogen conditions (same colors as

above), as determined experimentally by 13C MFA, with each ellipsoid encompassing mean ±

standard deviation of the three fluxes respectively. The four conditions are separated mainly along

the g6pdh flux axis. (b) Overall partitioning (μg day-1 [mg biomass]-1) of carbon from glucose

(principal carbon source supplied) and acetyl-CoA (from glucose and from reflux of proteins and

lipids) between biomass and CO2 (c) Partitioning (μmol day-1 [mg biomass]-1) between glycolysis and

oxPPP at G6P branch point. (d) Partitioning (μmol day-1 [mg biomass]-1) between TCA cycle and

anaplerosis at the consolidated PEP/Pyr branch point. In (c) and (d), arrow thicknesses are

proportional to the relative fluxes.

Figure 6. Comparison between evaluated fluxes and transcript measurements suggests that many

flux changes were transcriptionally regulated.

Flux changes and relative transcript levels of the genes encoding the enzymes catalyzing the

reactions are shown for nine important metabolic reactions. Flux names, chemical reaction

stoichiometry, enzyme names, Gene IDs as well as fold changes of both fluxes and gene expression

levels are listed. Fold changes relative to the CN (high-carbon, high nitrogen) condition are shown as

color scales and separate color scales are used for flux and gene expression.

This article is protected by copyright. All rights reserved.


Figure 7. Relative transcript abundance of selected genes in response to altered carbon and

nitrogen levels.
Accepted Article
Selected examples of relative transcript abundance for genes encoding proteins involved in (a)

central carbon metabolism, (b) amino acid synthesis/degradation, (c) sugar synthesis, (d)

photosynthesis/light capture, (e) lignan- and lignin-related and (f) redox and cytochrome, are shown

here. The log2-fold changes in transcript levels are relative to the CN condition. In each panel, rows

correspond to one of the four conditions CN, cN, Cn and cn, and columns correspond to individual

genes. Heat maps represent gene expression levels (log2-scale) and range from red (positive fold

change) to green (negative fold change). Gene name abbreviations are shown in the figure and they

follow the HUGO Gene Nomenclature Committee guidelines

(http://www.genenames.org/about/guidelines). A complete list of the 892 differentially expressed

genes with log2-fold changes between -300 and 30 and corresponding p-values from Student’s test

are listed in Table S7.

Figure 8. Lower NAPDH production, lower oxPPP flux and higher TCA cycle flux may contribute to

improved NUE under low nitrogen supply.

Summary and proposed scheme of the differences in NAPDH production, oxPPP and TCA cycle flux

and how they may impact poplar cell suspension NUE under high, (a) CN, and low, (b) Cn, levels of

nitrogen availability. Fluxes in both panels are normalized to glucose uptake rates. The thicknesses of

flux arrows represent relative flux for carbon, Ⓒ, nitrogen, Ⓝ, and (NAD[P]H), ⓔ. The areas of the

NH4 and NO3– boxes are proportional to their concentrations in media. Dashed lines represent

hypothesized processes.

This article is protected by copyright. All rights reserved.


Accepted Article

This article is protected by copyright. All rights reserved.


Accepted Article

This article is protected by copyright. All rights reserved.


Accepted Article

This article is protected by copyright. All rights reserved.


Accepted Article

This article is protected by copyright. All rights reserved.


Accepted Article

This article is protected by copyright. All rights reserved.

You might also like