Contrasting mechanisms of proteomic nitrogen thrift in Prochlorococcus

Molecular Ecology (2011) 20, 92–104 doi: 10.1111/j.1365-294X.2010.04914.x

Contrasting mechanisms of proteomic nitrogen thrift inProchlorococcus

JAMES D. J . GILBERT*† and WILLIAM F. FAGAN*

*University of Maryland, College Park MD 20742, USA, †University of Cambridge, Downing Street, Cambridge CB2 3EJ, UK

Corresponde

E-mail: james

Abstract

Organisms limited by carbon, nitrogen or sulphur can reduce protein production costs by

transitions to less costly amino acids, or by reducing protein expression. These

alternative mechanisms of nutrient thrift might respond differently to selection, but

this possibility remains untested. We hypothesized that relatively invariant sequence

composition responds to long-term variation in nutrient concentrations, whereas

dynamic expression profiles vary with nutrient predictability. Prolonged nutrient

scarcity favours proteome-wide nutrient reduction. Under stable, nonfluctuating nutrient

availability, reduction of nutrient content typically occurs in proteins upregulated when

nutrient availability is low, e.g. assimilation and catabolism. We suggest that fluctuating

nutrient availability favours mechanisms involving short-term downregulation of

nutrient-rich proteins. We analysed protein nitrogen content in six high-light, low-

nutrient adapted (HL) vs. six low-light, high-nutrient adapted (LL) Prochlorococcus(marine cyanobacteria) strains, alongside expression data under experimental nitrogen

and phosphorus limitation in two strains, MED4 (HL) vs. MIT9313 (LL). HL strains

contained less nitrogen, but DNA GC content confounded this relationship. While

anabolic and catabolic proteins had normal nitrogen content, most strains showed

reduced nitrogen in typical nitrogen stress response proteins. In the experimental data

set, though, proteins upregulated under nitrogen limitation were nitrogen-poor only in

MIT9313, not MED4. MIT9313 responded similarly to nitrogen and phosphorus

limitation, with slow, sustained downregulation of nitrogen-rich ribosomal proteins. In

contrast, under nitrogen but not phosphorus limitation, MED4 rapidly downregulated

ribosomal proteins. MED4’s specific, rapid nitrogen response suggests adaptation to

fluctuating conditions, supporting previous work. Thus, we identify contrasting

proteomic nitrogen thrift mechanisms within Prochlorococcus consistent with different

nutrient regimes.

Keywords: comparative proteomics, ecological stoichiometry, gene expression, molecular evolu-

tion, nutrient dynamics, stress response

Received 24 November 2009; revision received 31 August 2010; accepted 3 September 2010

Introduction

In nutrient-limited organisms, protein production costs

in terms of carbon (C), nitrogen (N) or sulphur (S, hence-

forth ‘nutrients’) can be reduced, by evolutionary transi-

tions to amino acids poorer in that nutrient (Mazel &

Marliere 1989; Baudouin-Cornu et al. 2001; Elser et al.

2006; but see Bragg & Wagner 2007), or by reduction of

nce: James D. J. Gilbert, Fax: (+44) 1223 336676;

[email protected]

protein expression levels (Bragg & Wagner 2009). How-

ever, selective forces affecting protein sequences and

those affecting expression levels can be different.

Within an individual organism, protein sequences are

effectively immutable, whereas expression levels are

highly dynamic. Therefore, we might expect protein

nutrient content to evolve in response to long-term, pre-

dictable variation in nutrient availability experienced by

an organism over its lifetime. By contrast, rapid, unpre-

dictable fluctuations in nutrient availability may select

for expression programs that reduce nutrient demand

� 2010 Blackwell Publishing Ltd

PROCHLO ROCOCCUS PROTEOMIC NITROGEN T HRI FT 93

by reducing production of the most expensive proteins,

thus providing an organism with a range of rapid-reac-

tion strategies for coping with short-term variation in

nutrient availability. To date, no study has addressed

this latter possibility (Bragg & Wagner 2007, 2009).

Here, we used data from conspecific cyanobacterial

strains under contrasting selective regimes of nutrient

availability to develop and evaluate these hypotheses.

Where nutrient availability is typically nonlimiting,

evidence suggests that nutrient reduction occurs in pro-

teins predictably upregulated on rare occasions when

nutrient availability is low, such as nutrient uptake

pathways (Baudouin-Cornu et al. 2001), catabolic

machinery (Acquisti et al. 2009b), and proteins found to

be upregulated under experimental nutrient limitation

(Cuhel et al. 1981; Mazel & Marliere 1989; Naito et al.

1994; Petrucco et al. 1996; Auger et al. 2001; Takahashi

et al. 2001; Fauchon et al. 2002; Tralau et al. 2007).

Nutrient reduction is most strongly evident in highly

expressed proteins (Acquisti et al. 2009b; Li et al. 2009).

In contrast, predictably low nutrient availability has

selected for general nutrient reduction across entire pro-

teomes (Acquisti et al. 2009a), again most strongly in

highly expressed proteins (Elser et al. 2006). No study

has yet addressed the possibility of reducing expression

levels specifically to reduce protein material costs, but

we hypothesize that organisms may alleviate temporary

nutrient limitation via short-term downregulation of

nutrient-rich proteins. Targets for rapid downregulation

may be specific pools of nutrient-rich proteins, e.g. the

ribosomal proteins (Acquisti et al. 2009b), or alterna-

tively, the most nutrient-rich proteins regardless of

function.

Prochlorococcus marinus is a complex of marine cyano-

bacterial strains, superabundant in vast oceanic gyres

and thought to be the most abundant phototrophic

organism (Partensky et al. 1999). Prochlorococcus are

ideal for testing the effects of nutrient availability on

the atomic composition and expression of proteins.

Twelve Prochlorococcus proteomes have been elucidated

(Kettler et al. 2007), with at least two studies of global

gene expression under nutrient limitation (N: Tolonen

et al. 2006; P: Martiny et al. 2006). The 12 strains are

classified into low-light adapted (‘LL’) vs. high-light

adapted ecotypes (‘HL’, Rocap et al. 2002; a distinction

we maintain by convention), but the distributions of

these two ecotypes broadly correspond to different oce-

anic niches with respect not only to light and depth,

but also to temperature and nutrient availability (John-

son et al. 2006). Although these distinctions are by no

means perfect, loosely speaking, HL strains are abundant

in surface waters, with low nutrient availability, and LL

strains predominate in deeper waters where nutrient

concentrations are higher (Johnson et al. 2006) and in this


paper we use ‘HL’ and ‘LL’ as abbreviations for these

broad niches. N availability is particularly important for

the evolutionary dynamics of Prochlorococcus (Garcia-Fer-

nandez et al. 2004). Strain-specific capabilities for N

source utilization arise because strains possess different

complements of enzymes appropriate to their respective

nutrient regimes (Garcia-Fernandez et al. 2004). The

availability of N sources utilized by HL strains of Prochlo-

rococcus (predominantly NH4+ and urea) in surface

waters is low (Lipschultz 2001, Karl 2002). In addition,

despite relatively thorough mixing of surface layers, sur-

face N availability fluctuates widely, both spatially (e.g.

Smith 1986, Varela & Harrison 1999) and temporally (e.g.

Beers & Kelly 1965; Mopper & Lindroth 1982; Brzezinski

1988), and on the scale of the individual cell is highly het-

erogeneous (Fenchel 2002, Church 2009). Consistent with

this, surface-living, HL Prochlorococcus strains like MED4

have genomic architectures adapted for life in low and

fluctuating N conditions (Rocap et al. 2003). For example,

MED4 has a reduced genome with low GC content,

reducing cell volume (thereby increasing surface

area:volume ratio to facilitate uptake) and reducing geno-

mic N and P demand (Dufresne et al. 2005). MED4 also

has a rapid expression response to transient N depriva-

tion (Tolonen et al. 2006). In contrast, the N sources uti-

lized by LL strains (e.g. nitrite) are higher at the base of

the euphotic zone where LL strains are typically found

(see e.g. Johnson et al. 2006).

At least two selection pressures may shape protein

N content in Prochlorococcus. First, Lv et al. (2008)

found that N content was higher in strains isolated

from deeper in the water column, even after matching

pairs of strains with similar N utilization capacities.

The authors related this finding to a depth gradient in

N availability. Second, however, the authors did not

account for the fact that proteomic N content may also

be at least partly determined by genomic GC content,

because the GC content of DNA codons correlates

with the N content of the encoded amino acid (Bragg

& Hyder 2004). GC content is subject to a range of

selection pressures unrelated to nutrient availability

(e.g. Birdsell 2002) and so may drive evolution in pro-

tein N content independently of ecological selection.

Here, we looked further at whether Prochlorococcus

strains have become adapted to their respective

regimes of N availability through three mechanisms:

(i) by wholesale reduction of N in the proteomes of

strains adapted to low-N environments (independently

of reduced GC content); (ii) by reducing the material

N cost of protein production through evolutionary

transitions to low-N amino acids in key proteins of the

N stress response, or (iii) by evolving mechanisms to

reduce expression levels of high-N proteins when N

availability is low.

94 J . D . J . GILBERT and W. F . FAGAN

First, we predicted that generally lower N availability

has led to proteome-wide selection for lower N in HL

strains, which typically exist at shallow depths where N

availability is low, independently of the fact that they

are known to have greatly reduced genomic GC content

compared to LL strains (Rocap et al. 2003). Second, we

predicted that life under typically stable N availability

should have left evolutionary signatures of reduced N

in proteins upregulated in response to limitation in LL

strains; in HL strains, these signatures may be either

evident or obscured by proteome-wide N reduction in

response to generally low N availability. Evolutionary

reduction of N in the sequences of commonly upregu-

lated proteins would require changes in protein N con-

tent over and above changes observed in the whole

proteome (Baudouin-Cornu et al. 2001; Bragg & Wagner

2009), despite constraints on gene function and stability,

and relatively little available variation in N content (0–3

atoms per amino acid). These long-term evolutionary

changes would be most effective if N limitation were

infrequent and predictable, as is the case for LL strains

in deeper waters, where C:N:P levels are closer to a rel-

atively constant ‘Redfield Ratio’ (Redfield 1934) and any

fluctuations encountered are likely to be prolonged.

Third, we tested the hypothesis that fluctuating N avail-

ability in LL strains has selected for expression pro-

grams that yield rapid downregulation of high-N

proteins under experimental N shortage, but not under

P shortage. Although N and P uptake and metabolism

are linked in complex ways (e.g. Barak & van Rijn

2000), we would not necessarily expect to see downre-

gulation of high-N proteins under P shortage. The abil-

ity to rapidly downregulate high-N proteins would

require relatively small, short-term changes to expres-

sion programs within the cell, as opposed to evolution-

ary substitution of N-rich residues for N-poor residues

in proteins whose efficiency is central to the N stress

response. This strategy would be most beneficial if N

limitation is frequent and unpredictable, as is the case

in the shallow surface waters to which HL strains are

adapted. We used two published data sets (Martiny

et al. 2006; Tolonen et al. 2006) on global expression

under N and P shortage in HL vs. LL strains (MIT9313

vs. MED4, respectively). In both studies, the authors

measured global transcriptional responses in both Pro-

chlorococcus strains over a 48-h time-course after resu-

spending cells in N- or P-free medium, respectively.

Reanalysing their data, we looked at expression levels

of all proteins with the top 1% of N content in each

proteome under N and P shortage, and also at expres-

sion patterns of ribosomal proteins, because Acquisti

et al. (2009b) found that the sequences of ribosomal

proteins are relatively N-rich. These authors proposed

that ribosomal proteins may require high proportions of

N-containing positively charged amino acids to bind to

negatively charged ribosomal RNA. Because protein

synthesis is reduced in times of N limitation, ribosomes

may additionally act as a cellular ‘store’ of N, which

can be broken down when the cell suffers N shortage

(Acquisti et al. 2009b).

Methods

We downloaded the 12 completely sequenced Prochloro-

coccus genome and proteome sequences from Cyano-

base (Nakao et al. 2010), plus the database of 1273

orthologs given in Kettler et al. (2007) for analyses

involving only genes shared by all 12 strains. Following

Elser et al. (2006), for each protein we calculated N con-

tent per amino acid side chain as

Nprotein ¼ ð100=LproteinÞ:X

wipi

Where Lprotein is the protein sequence length, wi is the

number of N atoms in the side chain of amino acid i

and pi is the frequency of amino acid i in the protein.

We took supplementary information on gene expression

under N limitation over time from Tolonen et al. (2006;

MED4: n = 1696; MIT9313, n = 2246) and under P limi-

tation over time from Martiny et al. (2006; MED4:

n = 1698; MIT9313, n = 2250). Transcript expression lev-

els were analysed as given by Tolonen et al. (2006) and

by Martiny et al. (2006), as log2 fold changes in tran-

script abundance between two conditions (i.e. N-replete

and N-limited, or P-replete and P-limited, respectively;

henceforth referred to as ‘fold change’). For ease of sta-

tistical analysis, we back-transformed fold changes to

ratios of transcript abundance and analysed the log-

transformed ratios (henceforth ‘expression levels’). To

compare expression under N and P limitation, we nor-

malized expression data by dividing abundance ratios

by their standard deviation before log transforming. All

analyses were carried out in R 2.9.1 (R Development

Core Team 2009).

Adaptation to low N availability: N content of theproteome

First, we asked whether N content was related to GC

content, HL or LL adaptation, or both. Because values

for orthologous proteins between strains were not inde-

pendent, we analysed protein N content using a linear

mixed model (LME; function lme() in R), equivalent to

a repeated measures ANOVA, using only orthologs pres-

ent across all 12 strains. We used ‘N content’ as a

response variable, ‘GC content’ and ‘light adaptation’

(i.e. HL vs. LL) as predictors, and ‘protein ID’ as a ran-

dom effect in the model. We first fitted the full model



(GC content and light adaptation, plus their interaction)

and sequentially dropped terms from the model, retain-

ing terms only if dropping them caused a significant

decrease in the fit of the model to the data. We assessed

model performance using likelihood ratio tests. Follow-

ing Paul et al. (2010), for comparison, we also calculated

the N content and GC content of Escherichia coli K12

and Francisella tularensis, two outgroup species with

similar GC content to LL and HL strains of Prochlorococcus,

respectively. We also asked whether the MED4 prote-

ome was lower in N than that of MIT9313, reflecting

reduced N usage. We compared the entire protein sets

with unpaired Wilcoxon tests, then compared only

shared proteins using a paired Wilcoxon test.

Adaptation to stable N availability: N content ofdifferentially expressed proteins

We next performed several analyses to test whether

upregulated transcripts were particularly low in N.

First, we tested the hypothesis of Acquisti et al. (2009b)

that catabolic and anabolic proteins are, respectively,

depleted and enriched in N compared to the proteomes

of all 12 strains. We used GO terms to identify proteins

involved in catabolic and anabolic processes, plus the

anabolic ‘machinery’ (sites of anabolic activity, i.e. ribo-

somes) and compared their N content using Wilcoxon

tests. For catabolic machinery (i.e. the proteasome), no

proteasome homologue exists in Prochlorococcus’ close

relative Synechococcus sp. WH 8102 (Valas & Bourne

2008) and no GO terms were found corresponding

either to proteasomes or to sites of protein degradation,

so we assumed that these structures or their homo-

logues have been lost in the process of genome reduc-

tion (see Dufresne et al. 2005). Catabolic, anabolic and

ribosomal proteins were identified by searching for

‘proteol* or degrad* or catabol*’, ‘biosynthe* or anabol*’

and ‘ribosom*’, respectively, within GO annotations

(terms listed in Appendix A).

Second, we identified proteins that had been previ-

ously found to be commonly upregulated and ⁄ or main-

tained at high levels as part of N stress responses in

several strains of Prochlorococcus (as opposed to proteins

that formed part of strain-specific responses and were

found not to be upregulated by other strains). These

were transporters for ammonium (amt) urea (urt, ure),

and, where present, cyanate (cyn) and nitrite (e.g.

PMT2240 in MIT9313); proteins of the GS-glutamate

synthase pathway (gln, gls) and high-light inducible

(hli) proteins, which protect photosystems from light

damage (see e.g. Linden et al. 2002; Tolonen et al.

2006), plus the lon protease, which may degrade pro-

teins during nutritional stress (Gomez-Baena et al.

2009). For each strain, we compared the N content of


the proteome to the N content of these proteins using

Wilcoxon tests. Since we were comparing internally,

within each proteome, we used all proteins and did not

restrict our analyses to orthologs.

Because N stress responses are strain-specific, an

approach using proteins that are merely hypothesized

to be commonly upregulated based on work conducted

in other strains can only be an approximation. There-

fore, to examine individual strains’ N stress responses

in detail, we analysed the N content of transcripts iden-

tified by Tolonen et al. (2006) as significantly

upregulated by MIT9313 or MED4 during N limitation

relative to the same transcripts under N-replete condi-

tions (NH4+), and we compared these to transcripts

from the remainder of the proteome of each strain. For

each strain, we counted the number of time points at

which each protein was upregulated during the 48-h

time-course, categorizing proteins into those that were

never upregulated, upregulated at one time point (rep-

resenting transitory upregulation), or upregulated at

more than more time point (representing sustained

upregulation). We also performed the same analysis but

looking only at proteins with orthologs in both strains.

For both analyses, we used a simple two-way ANOVA

with ‘N content’ as the response, and ‘strain’ and ‘num-

ber of points upregulated’ (0, 1, or >1 time points) as

predictor variables. Differences between strains in the

N content of proteins upregulated under N limitation

would be evident as an interaction between ‘strain’ and

‘number of points upregulated’. Additionally, to give a

more realistic estimate of the metabolic impact of

upregulation, we also analysed the same data using

weighted least squares, weighting by median protein

expression level (averaged across time points).

Finally, we analysed N content in the subsets of tran-

scripts that fell into coregulated ‘‘gene clusters’’ form-

ing the regulatory response to imposed N limitation

(MED4: n = 9 clusters, MIT9313: n = 7 clusters). We cal-

culated the median expression level across time points

for each cluster, and the median N content, and fitted a

linear model with ‘median cluster N content’ as a

response variable, and ‘median cluster expression level’

and ‘strain’ as predictor variables.

Adaptation to fluctuating N availability: expressionprofiles of N-rich proteins under N and P limitation

We tested the hypothesis that high-N proteins would

tend to be downregulated under N limitation, but not

necessarily under P limitation, using time-course

expression data from Tolonen et al. (2006; N limitation)

and Martiny et al. (2006; P limitation). First, we analy-

sed the expression levels of transcripts with the highest

1% of N content (MED4: 17 of 1696 proteins; MIT9313,

0.35 0.40 0.45 0.50

0.32

0.33

0.34

0.35

0.36

0.37

0.38

DNA GC content

Pro

tein

N c

onte

nt

HL-adapted ProchlorococcusLL-adapted ProchlorococcusE. coliF. tularensis

Fig. 1 Median ± 95%CI protein N content against median GC

content of all shared genes in LL-adapted and HL-adapted

strains of Prochlorococcus (n = 1273). Regression line is shown

for LL-adapted strains. Escherichia coli and Francisella tularensis

are given as outgroups for comparison.


23 of 2246 proteins) compared to the rest of the

proteome. Second, we analysed the N content of ribo-

somal proteins, with respect to the proteome, to confirm

their high N content. We then analysed the expression

levels of these proteins compared to the proteome of

each strain. To account for the possibility that regula-

tion of protein synthesis in general was driving expres-

sion of ribosomal proteins, we also analysed the N

content and expression of nonribosomal proteins

involved in translation (terms containing ‘translation*’:

GO:0006412, GO:0006418, GO:0006414, GO:0003746,

GO:0006415, GO:0016149, GO:0003747, GO:0006450,

GO:0006413, GO:0003743, GO:0006614, n = 44 in both

strains), and proteins involved in transcription (terms

containing ‘transcription*’: GO:0006355, GO:0006352,

GO:0003700, GO:0006353, GO:0003715, GO:0045449,

GO:0008134, GO:0006350, GO:0031554, GO:0030528,

GO:0006351, GO:0003711, n = 32 in both strains). In all

cases, we first analysed the amalgamated data using

simple Wilcoxon tests, and also looked for temporal

patterns of regulation with a LME model while account-

ing for nonindependence of estimates of the same tran-

script over time. For each model, we used the

categorical variable of interest and ‘time point’ as pre-

dictors, ‘expression level’ as the response variable, and

‘protein ID’ as a random effect.

To look for a more general relationship between N

content and protein expression levels, rather than

downregulation of a specific subset such as the ribo-

somes, we also analysed expression levels across the

entire proteome using N content as a continuous vari-

able. We used a LME model to account for nonindepen-

dence of expression estimates for the same gene at

different times, using ‘expression level’ as a response

variable, ‘N content’ and ‘strain’ as predictor variables,

and ‘protein ID’ within ‘time stage’ as a nested random

effect. Differences between MED4 and MIT9313 in the

way protein expression varied with N content would be

evident as a significant interaction between ‘N content’

and ‘strain’. Finally, we repeated this analysis excluding

the ribosomal proteins.

Results

Adaptation to low N availability: N content of theproteome

As expected, we found a positive relationship between

GC content and N content overall (Fig. 1). Thus, unsur-

prisingly, the low-GC, HL-adapted MED4 proteome

was also lower in N than that of the relatively high-GC,

LL-adapted MIT9313. Proteins encoded by Prochlorococ-

cus MED4 transcripts were lower in N than MIT9313

proteins, in the whole sample (median N content in

MED4, 0.343, IQR 0.086; median in MIT9313, 0.372,

IQR = 0.104; Wilcoxon test, P < 0.001) and when ortho-

logs were paired (median difference 0.028, IQR 0.053, or

a difference of 7.5% in N in amino acid side chains;

Wilcoxon paired test, P < 0.001).

This typified a wider ecological pattern. Protein for

protein, LL Prochlorococcus strains had higher N content

and higher GC content than HL strains (Fig. 1). How-

ever, LL vs. HL adaptation also affected the relationship

between N content and GC content: LL strains had a

positive relationship between N and GC, as found by

Bragg & Hyder (2004), whereas this relationship was

not evident in HL strains (LME model, dropping the

‘GC content’ · ‘light adaptation’ interaction, likelihood

ratio test, v2 = 25.49, P < 0.001). E. coli, included as an

outgroup for comparison, had a mean proteomic N con-

tent typical of the LL strains of Prochlorococcus but

higher than that of the HL strains. F. tularensis, by con-

trast, had much lower proteomic N content than the HL

strains of Prochlorococcus despite a similarly reduced

genome with comparably low GC content.


First, we tested Acquisti et al.’s (2009b) prediction that

anabolic and catabolic proteins should be N-rich and

N-poor, respectively, owing to predictable patterns of

regulation in response to conditions of nutrient stress.

We analysed the N content of proteins involved in

catabolic and anabolic processes, plus the anabolic


0.2

0.4

0.6

0.8

N c

onte

nt

MIT

9215

:HL

MIT

9515

:HL

AS

9601

:HL

MIT

9301

:HL

MIT

9312

:HL

ME

D4:

HL

NAT

L1A

:LL

MIT

9303

:LL

SS

120:

LL

MIT

9211

:LL

NAT

L2A

:LL

MIT

9313

:LL

ProteomeAnabolicCatabolicRibosome

Fig. 2 Median ± interquartile range (boxes) and 90% quantile

(whiskers) of N content for the proteome (white bars) vs. ana-

bolic (grey bars), catabolic (dark grey bars) and ribosomal pro-

teins (red bars) in 12 Prochlorococcus strains.

0.1

0.2

0.3

0.4

0.5

0.6

N c

onte

nt

MIT

9215

:HL

(25)

MIT

9515

:HL

(44)

AS

9601

:HL

(31)

MIT

9301

:HL

(27)

MIT

9312

:HL

(26)

ME

D4:

HL

(19)

NAT

L1A

:LL

(12)

MIT

9303

:LL

(23)

SS

120:

LL (

21)

MIT

9211

:LL

(20)

NAT

L2A

:LL

(19)

MIT

9313

:LL

(13)

ProteomeN stress-related proteins

Fig. 3 Median ± interquartile range (boxes) and 90% quantile

(whiskers) of N content for the proteome (white bars) vs. pro-

teins commonly highly upregulated and ⁄ or maintained at high

levels during Prochlorococcus N stress responses (grey bars) (see

text for genes).


machinery (i.e. ribosomes), in all 12. The ribosomal pro-

teins were substantially N-rich compared to the prote-

ome in all strains (Wilcoxon tests, P < 0.0001, Fig. 2).

Anabolic proteins were in fact lower in N than the pro-

teome in some cases (AS9601, MIT9211, MIT9312,

MIT9313, NATL2A, SS120; P < 0.05) but not in others.

In no case were catabolic proteins different from the

proteome in N content (Wilcoxon tests, all NS).

Second, we looked across the 12 proteomes at pro-

teins that have been previously found to be predictably

upregulated and ⁄ or maintained at high levels by

diverse Prochlorococcus strains as part of N stress

responses. These proteins were significantly lower in N

than their respective proteomes for eight of 12 strains

(Wilcoxon tests, P < 0.01; LL: SS120, MIT9313, NATL2A,

MIT9211; HL: MIT9215, MIT9301, MIT9312, MED4). In 2

of the remaining strains (HL: MIT9515, AS9601), this

difference was again negative, but nonsignificant. In

MIT9303 and NATL1A (both LL), median N content of

these proteins was approximately the same as in the

wider proteome (Fig. 3).

Published expression data for MIT9313 and MED4

allowed us to look in detail at proteins of the individual

N stress responses of a LL-adapted strain vs. a HL-

adapted strain. In the model of N content in proteins

upregulated at varying numbers of time points under N

limitation, the two strains showed different patterns

evidenced by a significant interaction term between

‘strain’ and ‘number of points upregulated’ (F2,3938 =

7.01, P < 0.001). In MIT9313, proteins that underwent


sustained upregulation under experimental N limitation

(i.e. were upregulated at more than one time point)

were significantly lower in N than proteins that were

never upregulated or were upregulated at only one

point in the time-course (Fig. 4, upper bars). However,

this pattern was not evident in MED4, where proteins

that underwent sustained upregulation were no differ-

ent in N content from the remainder of the proteome.

This pattern was similar when only orthologous pro-

teins were analysed (dropping the ‘strain’ · ‘number of

points upregulated’ interaction, F2,2564 = 3.94, P < 0.01;

Fig. 4, lower bars). When weighted by median expres-

sion level to give a more biologically relevant assess-

ment of the metabolic effect of upregulation, proteins

undergoing sustained upregulation under N limitation

were again significantly lower in N content than the

proteome in MIT9313, but not in MED4 (weighted least

squares regression, dropping the ‘strain’ · ‘number of

points upregulated’ interaction, F2,3938 = 113.0, P <

0.001).

Finally, we analysed the N content of co-regulated

gene clusters constituting the N stress response, identi-

fied by Tolonen et al. (2006; MIT9313, n = 9 clusters;

MED4, n = 7 clusters) using a linear model with ‘med-

ian cluster N content’ as a response variable, and ‘med-

ian cluster expression level’ and ‘strain’ as predictor

variables. Confirming the above findings, the strains

showed different patterns in cluster N content with

respect to cluster regulation, evidenced by a significant

0.1

0.2

0.3

0.4

0.5

0.6

0.1

0.2

0.3

0.4

0.5

0.6

N c

onte

nt

MIT9313 MED4

Upregulated at 0 points1 point>1 points

Fig. 4 Median ± 95%CI (notches), interquartile range (boxes)

and 90% quantile (whiskers) of N content in the whole proteo-

mes (upper bars) and in orthologous proteins (lower bars) of

Prochlorococcus strains MIT9313 and MED4, for proteins never

upregulated (white bars) vs. proteins upregulated at one time

point (light grey bars) or more than one time point (dark grey

bars) under experimental N limitation.


interaction between ‘median cluster expression level’

and ‘strain’ (removing this interaction, F1,13 = 7.80,

P < 0.01). In MIT9313, cluster median N content was

negatively related to cluster median expression level; in

MED4, there was no relationship between these two

variables (Fig. 5).

−5 0 5 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7MIT9313

Expression level

N c

onte

nt

1

2

3

4

5

6

7

Fig. 5 Median N content against median expression level for gene c

strains MIT9313 and MED4. Solid error bars show interquartile rang

slopes from linear model. Functional enrichment within clusters (fro

Transport and binding; (3) Regulation; (4) Amino acid synthesis; (5)

and respiration; (9) Photosynthesis and respiration. MIT9313: (1) Tran

and binding; (5) Fatty acid and phospholipid; (6) Photosynthesis and

Adaptation to fluctuating N availability: expressionprofiles of N-rich proteins

We then asked whether high-N proteins tended to be

downregulated under conditions of N limitation com-

pared to the rest of the proteome, and we compared

this with the expression of high-N proteins under P

limitation. We predicted that high-N proteins would be

downregulated under N limitation but not necessarily

under P limitation and that this pattern would be stron-

ger in MED4 because this strain is adapted to temporal

fluctuations in N availability.

Initially, we looked at the proteins with the top 1%

of N content regardless of function. Amalgamated

across time stages, the highest N proteins were in fact

downregulated under N limitation compared to the rest

of the proteome, highly significant in MED4 (Wilcoxon

test, P < 0.001) and weakly significantly in MIT9313

(P = 0.04). Under P limitation, proteins in the top and

bottom 1% of N content were not differently regulated

compared to the proteome (Wilcoxon test, all NS;

Fig. 6). However, in both strains the top 1% of pro-

teins was composed almost exclusively of ribosomal

proteins. We therefore expanded our analysis, looking

at expression patterns of ribosomal proteins compared

to the rest of the proteome. Across all time points com-

bined, the ribosomal proteins were strongly downregu-

lated relative to the proteomes in both strains under N

limitation (Wilcoxon tests, P < 0.001 in both strains)

and under P limitation (P < 0.001 in both strains). By

contrast, under N limitation, nonribosomal proteins

involved in transcription actually had slightly higher

expression than the proteome in both strains (MIT9313,

−5 0 5 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7MED4

Expression level

1

2

3

45

6

7

89

lusters coexpressed in response to N shortage in Prochlorococcus

e, dotted error bars show range. Dashed lines show estimated

m Tolonen et al. 2006): MED4: (1) Transport and binding; (2)

Regulation; (6) Translation; (7) Translation; (8) Photosynthesis

sport and binding; (2) Regulation; (3) Regulation; (4) Transport

respiration; (7) Translation.


−3

−2

−1

0

1

2

3

Nor

mal

ized

exp

ress

ion

leve

l

MIT9313 MED4 MIT9313 MED4P limitation N limitation

N-poor proteinsProteomeN-rich proteins

Fig. 6 Median ± 95%CI (notches), interquartile range (boxes)

and 90% quantile (whiskers) of normalized expression levels

under P and N limitation for N-poor proteins (dark grey bars)

and N-rich proteins (light grey bars), compared to the pro-

teome (white bars), in strains MIT9313 and MED4.


P = 0.03; MED4, P = 0.06), while nonribosomal proteins

involved in translation had lower expression than the

proteome in MIT9313 (P < 0.001) but not in MED4

(P = 0.10). Under P limitation, proteins of transcription

and translation were not expressed differently from the

proteome in either strain.

When we looked specifically for temporal differences

in regulation, the LME model of expression levels

revealed strongly divergent patterns of regulation over

time among the two strains under both N and P-limited

MIT9313

Nor

mal

ized

exp

ress

ion

leve

l

−2−1

012

−2−1

012

0 h 3 h 6 h 12 h 24 h 48 h

ProteomeTranscriptionTranslationRibosome

Fig. 7 Median ± 95%CI (notches), interquartile range (boxes) and 90%

(0, 3, 6, 12, 24, 48h) in ribosomal proteins (red bars) and proteins in

bars), compared to the proteome (white bars), under P limitation (M

et al. 2006; bottom series). Time points in the series have been align

limitation, and at 4 h for P limitation.


conditions (dropping the ‘functional group’ · ‘strain’ ·‘time point’ interaction, P < 0.0001 for both N and P

limitation). Under both N- and P-limited conditions,

this pattern arose mostly from between-strain differ-

ences in temporal expression patterns of the ribosomal

proteins (Fig. 7); transcription proteins and nonriboso-

mal translation proteins showed relatively little change

in expression over time in both strains. In MIT9313,

ribosomal protein expression followed a similar pattern

under both N and P limitation, which suggested a gen-

eralized stress response. Downregulation of ribosomal

proteins began at approximately 6 h following N or P

deprivation and decreased over time such that the

greatest degree of downregulation compared to the pro-

teome was at the last time point, 48 h (N) or 24 h (P)

(Fig. 7). In MED4, ribosomal proteins were rapidly and

strongly downregulated under N limitation from 3h

onwards, with the peak downregulation at 12–24 h.

Under P limitation, however, we found no evidence of

downregulation of ribosomal proteins over time, and

only at the last time point, 48 h, were ribosomal pro-

teins downregulated (Fig. 7).

Finally, we analysed expression levels using N con-

tent as a continuous variable to look for a more general

relationship between N content and protein expression

levels, using a LME model to account for nonindepen-

dence of expression estimates for the same transcript at

different time points. The results were qualitatively sim-

ilar to those described earlier, in that N content was

weakly negatively associated with expression levels in

MIT9313, but was strongly negatively associated with

expression levels in MED4 (dropping ‘strain’ · ‘N con-

tent’ interaction, likelihood ratio test, v2 = 4.0, d.f. = 1,

P < 0.05). However, this relationship disappeared when

MED4

−2−1

012

−2−1

012

quantile (whiskers) of normalized expression levels over time

volved in transcription (light grey bars) and translation (pink

artiny et al. 2006; top series) and under N limitation (Tolonen

ed for clarity, except that the second time point is at 3 h for N


we excluded ribosomal proteins from the analysis

(dropping ‘strain’ · ‘N content’ interaction, likelihood

ratio test, v2 = 1.0, d.f. = 1, P = 0.43). Once ribosomal

proteins were excluded, expression levels were actually

slightly positively related to protein N content (drop-

ping ‘N content’, v2 = 8.0, d.f. = 1, P = 0.01).

Discussion

Adaptation to long-term N availability: N content ofthe proteome

Among Prochlorococcus strains, the positive N con-

tent ⁄ GC content relationship was as expected, given the

underlying positive relationship detected by Bragg &

Hyder (2004). Thus, it is unsurprising that MED4, a

strain with exceptionally low GC content, had proteins

lower in N than MIT9313 proteins, although the

approximately 7.5% difference in N content between

the strains was large given the relatively close related-

ness of these strains (shared genes among Prochlorococcus

represent 73% of the MED4 genome and 56% of the

MIT9313 genome; Dufresne et al. 2005). By comparison,

Elser et al. (2006) found a 7% difference in N content

across the proteomes of animals vs. plants. It is also

unsurprising that the low-GC, HL-adapted strains had

relatively low N content compared to the higher GC

LL-adapted strains. Lv et al. (2008) found an association

between isolation depth and N content and interpreted

this finding in terms of variation in N availability with

depth. Although it should be noted that HL and LL

adaptations in Prochlorococcus cannot be assumed to cor-

respond perfectly with distinct regimes of low and high

N availability, respectively (the environmental factors

determining Prochlorococcus distributions are complex;

Johnson et al. 2006), our findings are nevertheless con-

sistent with general adaptation of the genomic architec-

ture of Prochlorococcus to low N availability in surface

waters (Garcia-Fernandez et al. 2004; Dufresne et al.

2005; Lv et al. 2008; Paul et al. 2010). They are also con-

sistent with reduction of N in macromolecules generally

under N limitation (Elser et al. 2006; Acquisti et al.

2009a). HL strains of Prochlorococcus are known to have

distinctive patterns of atomic composition and amino

acid usage compared to LL strains; this holds true even

when controlling for the effect of genome reduction by

specifically comparing HL Prochlorococcus strains to LL

strains that also have reduced genomes (Paul et al.

2010). However, given the relatively tight positive rela-

tionship between protein N content and genomic GC

content, we cannot rule out the explanation that alterna-

tive types of selection on GC content may be indirectly

causing reduction in proteomic N content. For example,

as a result of adaptive gene loss through genomic

‘streamlining’, MED4 has lost several DNA repair

enzymes (Dufresne et al. 2005), the absence of which

results in a mutational bias toward AT pairs in knock-

out mutants of other species (Mackay et al. 1994; Horst

et al. 1999).

Interestingly, HL vs. LL adaptation also affected the

relationship between proteomic N content and GC con-

tent (Fig. 1). While LL strains showed a positive N vs.

GC relationship, HL strains had no such relationship,

although they varied much less in both N content and

GC content. This suggests one of three explanations.

First, N content in HL strains may evolve indepen-

dently of GC content, while in LL strains, evolution of

proteomic N content may be more closely correlated

with changes in GC content. This would be consistent

with the idea that there is direct selection for N conser-

vation in proteomes of HL strains but not LL strains.

Alternatively, HL strains may have reached a GC con-

tent ‘floor’ beyond which further reduction is maladap-

tive, especially for free-living cyanobacteria. Other

bacteria with similarly low GC content are mostly intra-

cellular parasites such as F. tularensis (Fig. 1 and Paul

et al. 2010). Such a floor would reduce the range of

possible genomic GC content, and hence possibly N

content, for a HL-adapted strains, reducing the detect-

ability of a N vs. GC content relationship. Another

alternative is that HL strains have reached a floor in N

content rather than GC content, beyond which function-

ality is impaired. Evidence suggests that this is unlikely.

F. tularensis, a species with comparable GC content to

HL-adapted Prochlorococcus, had substantially less N in

its proteome (Fig. 1), despite its comparatively N-rich

intracellular environment. Moreover, Bragg & Hyder

(2004) assessed the N content of proteomes of nearly

100 prokaryotes, and the minimum proteomic N con-

tent was <0.31 atoms per amino acid, much lower than

we observed for MED4; in fact, given its GC content,

MED4 actually had comparatively N-rich proteins (their

Fig. 2b, p. S376). Finally, Elser et al. (2006) report N

contents of <0.3 atoms per amino acid in highly

expressed proteins of eukaryotic plants (their Fig. 2b,

p. 1949), implying that if such an N floor exists, it is

well below the 0.343 that we observed in the MED4

proteome.


Another, stronger line of evidence for proteomic nutri-

ent conservation strategies comes from signatures of

limitation within proteomes. Anabolic and catabolic

proteins were not substantially different in N from their

respective proteomes in any strains; only the ribosomal

proteins were uniformly high in N.


PROCHLOROCOCCUS PROTEOMI C NITROGEN T HRIFT 101

Acquisti et al. (2009b) recently detected significant N

enrichment (even ‘N reservoirs’) in the anabolic

machinery in diverse eukaryotic species, most especially

in the ribosomal proteins. Such patterns may arise for

two reasons. First, ribosomal proteins bind to negatively

charged rRNA, which selects for a high density of

amino acids with positively charged NH+ groups. Sec-

ond, ribosomes are required mostly when raw materials

for protein synthesis are abundant, but are broken

down when nutrient concentrations are low. Thus, the

cellular N pool contained in ribosomes can be recycled

in times of N limitation without impairment of cellular

function. This is likely to be especially important in spe-

cies with reduced genomes, such as HL-adapted Pro-

chlorococcus, which possess relatively few proteins that

can serve as such a storage pool.

We also found signatures of N limitation across most

of the 12 strains, in the proteins commonly found to be

upregulated as part of diverse Prochlorococcus N stress

responses (Linden et al. 2002; Tolonen et al. 2006;

Gomez-Baena et al. 2009). These were broadly lower in

N than their respective proteomes, although with excep-

tions in LL strains MIT9301 and NATL1A. This sug-

gests that there has been selection for N conservation

across Prochlorococcus strains, consistent with the central

role of N dynamics in their ecology (e.g. Moore et al.

2002). However, our results also suggested that detailed

expression data are necessary to form robust conclu-

sions about nutrient-sparing strategies. Focusing on

those two strains for which detailed, global expression

data under short-term N limitation were available,

MIT9313 and MED4 (Tolonen et al. 2006), proteins

upregulated in response to N limitation were particu-

larly low in N only in strain MIT9313, not in strain

MED4. Two lines of evidence suggested this: (i) pro-

teins that underwent sustained upregulation under N

limitation (i.e. proteins that were found by Tolonen

et al. (2006) to be upregulated at more than one time

point) were N-poor compared to the proteome in

MIT9313, but not in MED4, (ii) co-regulated protein

clusters that responded in synchrony to N limitation

showed a negative relationship between N content and

expression level in MIT9313, but not in MED4. To our

knowledge, this is the first demonstration of a N spar-

ing response among upregulated proteins in an experi-

mental context, although other authors have pointed

out strongly suggestive correlations from comparative-

phylogenetic evidence (Elser et al. 2006) or from com-

parisons between different components within proteo-

mes (Acquisti et al. 2009b; Li et al. 2009). Our findings

complement existing evidence for evolutionary conser-

vation of S in proteins upregulated under experimental

S limitation (Mazel & Marliere 1989; Takahashi et al.

2001; Fauchon et al. 2002).


Why were upregulated proteins not relatively N-poor

in strain MED4? For reasons discussed earlier, it is unli-

kely that the entire MED4 proteome has reached an N

‘floor’ which obscures any further reduction in N con-

tent in N stress proteins. Hence, this lack of proteomic

thrift remains an open question. As one possibility,

Bragg & Wagner (2007) suggested that long-term adap-

tation to nutrient limitation should result in the gradual

evolution of physiological mechanisms to increase nutri-

ent capture affinity. This would actually relax selection

for nutrient thrift in the proteome and may explain why

the authors did not observe C thrift in upregulated pro-

teins relative to the proteome even in yeast strains

adapted to long-term C limitation. This may also be the

case in MED4, although substantial further work would

be required to test such a hypothesis. For example,

while the global N response regulator NtcA plays a cen-

tral role in the N response in most Prochlorococcus

strains, including MIT9313 and MED4, there are impor-

tant strain-specific differences in the way N metabolism

is integrated with other physiological systems, princi-

pally C metabolism (Tolonen et al. 2006). Unlike

MIT9313, strain MED4 shuts down many metabolic

activities rapidly in response to N limitation, including

the C fixation machinery. Thus, in response to transient

N shortages, MED4 may be able to reduce its metabolic

demand for N by reducing protein expression wholesale

very rapidly and efficiently (see e.g. Fig. 5), and this

reduction in total protein production may possibly relax

selection for N reduction in the N stress machinery.

Adaptation to fluctuating N availability: expressionprofiles of N-rich proteins

Tolonen et al. (2006) found that MED4’s N stress

response was generally rapid and transient compared

to a slower, more sustained response in MIT9313, and

suggested that this may represent adaptation to fluctu-

ating nutrient availability in surface waters. Here, we

have hypothesized that one function of this rapid

response is to bring about a transient reduction in cellu-

lar N demand by reducing production of the most

N-rich proteins, the ribosomal proteins, although data

from more than two strains would be required to test

this hypothesis rigorously. Ribosomal proteins were

among the most strongly downregulated proteins in

both strains (Tolonen et al. 2006; their Fig. 3, p. 3), but

downregulation of ribosomal transcripts under N limi-

tation began much more rapidly in MED4 than in

MIT9313 (i.e. downregulation was strongly in evidence

as early as 3 h after the onset of N limitation). Downre-

gulation of ribosomes usually occurs as part of a gener-

alized stress response, reducing cellular nutrient

demand wholesale and reducing growth rate (Gourse


et al. 1996). This is consistent with our results for

MIT9313: ribosomal proteins were downregulated grad-

ually under both N limitation and P limitation. In

MED4, however, patterns of ribosomal protein expres-

sion were different under N vs. P limitation. Ribosomal

proteins were rapidly downregulated under N limita-

tion, when they would be relatively very costly for the

cell to produce, but were not downregulated under P

limitation, when N availability was not limiting. Thus,

we can cautiously suggest that MED4 may have

evolved rapid downregulation of ribosomes as a tran-

sient N conservation response, because of adaptation to

frequent and unpredictable fluctuations in environmen-

tal N availability. This confirms and extends Acquisti

et al. (2009b) finding of N enrichment in ribosomal pro-

teins to an ecologically relevant context. While environ-

ments with typically stable N may select for reduction

of the N content of upregulated machinery, in environ-

ments with fluctuating N levels such as that of MED4,

it may be more advantageous to have a response based

on rapid downregulation of N-rich ribosomal tran-

scripts.

If, as suggested by Acquisti et al. (2009b), ribosomal

proteins are primarily N-rich because their role in bind-

ing to rRNA selects for high charge density, then envi-

ronments that tend to destabilize these electrostatic

interactions should also select for ribosomal proteins

with a higher charge density and hence higher N cost.

It would now be very interesting to compare the rela-

tive charge density and N content of ribosomal proteins

and RNA across the prokaryotes, with respect to selec-

tion on ribosomal stability, e.g. in species living at high

temperatures (Hickey & Singer 2004; Bragg et al. 2006).

The nucleotide composition of ribosomal RNA in ther-

mophiles is under strong selection for stability (Hickey

& Singer 2004). GC content of ribosomal RNA, which

directly determines N content, is also positively related

to molecular stability, and the GC content of rRNA

increases with growth temperature (Galtier & Lobry

1997). Strong selection on ribosome stability may there-

fore indirectly select for both high-N ribosomal RNA

and high-N ribosomal proteins, with a corresponding

increase in the organism’s demand for N. Such a sce-

nario may help determine an organism’s best strategy

for responding to N limitation; for example, hyperther-

mophiles may be forced to downregulate their ribo-

somal proteins when N is scarce.

Conclusion

By synthesizing published data on proteome N content

in 12 strains of Prochlorococcus and expression profiles

for two strains under short-term N and P limitation, we

have identified three different N thrift mechanisms in

closely related proteomes consistent with different eco-

logical regimes of N availability. This allows us to move

toward broad hypotheses relating environmental

dynamics to the evolution of protein nutrient content

and expression levels. These findings shed light on the

heretofore open question of under what conditions it is

advantageous for any organism to reduce protein mate-

rial costs through evolutionary reduction of sequence

nutrient content (Baudouin-Cornu et al. 2001), vs. sim-

ply producing less of the protein by reducing expres-

sion levels, as suggested by Bragg & Wagner (2009).

Our results suggest that (i) relatively stable nutrient

availability selects for evolutionary reduction of nutrient

content in the proteins commonly upregulated under

nutrient limitation, but that (ii) generally low nutrient

availability (e.g. in HL strains) may also select for

wholesale, proteome-wide nutrient reduction in protein

sequences. In addition to these, (iii) fluctuating or

unpredictable nutrient availability may also select for a

rapid, transient response involving downregulation of

the most nutrient-rich proteins. The three mechanisms

are of course not expected to be mutually exclusive.

Other constraints are likely to influence which mecha-

nism is most important for reducing protein production

costs. For example, certain proteins are likely to be con-

strained either to have high expression under nutrient

shortage (e.g. high-affinity uptake proteins, Baudouin-

Cornu et al. 2001) or to have high N content (e.g. posi-

tively charged ribosomal proteins, Acquisti et al.

2009b).

The relative feasibility of each mechanism for achiev-

ing proteomic thrift probably depends upon which of

these constraints is most important to cellular function.

In the case of MED4, which rapidly shuts down photo-

synthesis under N limitation (Tolonen et al. 2006), the

machinery of protein synthesis (e.g. ribosomal proteins)

may be ‘free’ to be downregulated if necessary, follow-

ing this shutdown. In contrast, owing to low light levels

MIT9313 cannot afford to shut down its metabolism

until an advanced stage of starvation (Tolonen et al.

2006), so N-rich proteins may be required for longer

periods than in MED4. Further studies of proteomic

thrift in other taxa should therefore consider how the

relationship between nutrient content and expression

levels is constrained by protein function under different

environmental conditions. Similar constraints may

apply, for example, in proteins constrained by selection

for chemical stability to have high numbers of disul-

phide bonds (and hence high S content, Bragg et al.

2006). In such proteins, different regimes of environ-

mental S availability may influence the optimal strategy

to reduce the costs of protein production. This may be

relevant, for example, in bacteria inhabiting different

ecological niches surrounding hydrothermal vents.


PROCHLOROCOCCUS PROTEOMI C NITROGEN T HRIFT 103

Acknowledgements

This work was supported by the National Science Foundation.

Special thanks to C. Acquisti, S. Kumar, J.J. Elser, A. Manica,

F.S. Gilbert, J. Keagy and three anonymous referees for useful

discussions and comments on the manuscript.

References

Acquisti CA, Elser JJ, Kumar S (2009a) Ecological nitrogen

limitation shapes the DNA composition of plant genomes.

Molecular Biology and Evolution, 26, 953–956.

Acquisti CA, Kumar S, Elser JJ (2009b) Signatures of nitrogen

limitation in the elemental composition of the proteins

involved in the metabolic apparatus. Proceedings Royal Society

B, 276, 2605–2610.

Auger S, Danchin A, Martin-Verstraete I (2001) Global expres-

sion profile of Bacillus subtilis grown in the presence of

sulfate or methionine. Journal of Bacteriology, 184, 5179–5186.

Barak Y, van Rijn J (2000) Relationship between nitrite

reduction and active phosphate uptake in the phosphate-

accumulating denitrifier Pseudomonas sp. Strain JR 12. Applied

and Environmental Microbiology, 66, 5236–5240.

Baudouin-Cornu P, Surdin-Kerjan Y, Marliere P, Thomas D

(2001) Molecular evolution of protein atomic composition.

Science, 293, 297–300.

Beers JR, Kelly AC (1965) Short-term variation of ammonia in

the Sargasso Sea off Bermuda. Deep-Sea Research, 12, 21–25.

Birdsell JA (2002) Integrating genomics, bioinformatics, and

classical genetics to study the effects of recombination on

genome evolution. Molecular Biology and Evolution, 19, 1181–

1197.

Bragg JG, Hyder CL (2004) Nitrogen versus carbon use in

prokaryotic genomes and proteomes. Proceedings Royal

Society B, 271(Suppl 5), S374–S377.

Bragg JG, Wagner A (2007) Protein carbon content evolves in

response to carbon availability and may influence the fate of

duplicated genes. Proceedings Royal Society B, 274, 1063–1070.

Bragg JG, Wagner A (2009) Protein material costs: single atoms

can make an evolutionary difference. Trends in Genetics, 25,

5–8.

Bragg JG, Thomas D, Baudouin-Cornu P (2006) Variation

among species in proteomic sulphur content is related to

environmental conditions. Proceedings Royal Society B, 273,

1293–1300.

Brzezinski MA (1988) Vertical distribution of ammonium in

stratified oligotrophic waters. Limnology and Oceanography,

33, 1176–1182.

Church MJ (2009) The trophic tapestry of the sea. Proceedings

of the National Academy of Sciences of the USA, 106, 15519–

15520.

Cuhel RL, Taylor CD, Jannasch HW (1981) Assimilatory sulfur

metabolism in marine microorganisms: sulfur metabolism,

growth and protein synthesis of Pseudomonas halodurans and

Alteromonas luteo-violaceus during sulfate limitation. Archives

of Microbiology, 130, 1–7.

Dufresne A, Garczarek L, Partensky P (2005) Accelerated

evolution associated with genome reduction in a free-living

prokaryote. Genome Biology, 6, R14.

Elser JJ, Fagan WF, Subramanian S, Kumar S (2006)

Signatures of ecological resource availability in the animal


and plant proteomes. Molecular Biology and Evolution, 23,

1946–1951.

Fauchon M, Lagniel G, Aude J-C, Lombardia L, Soularue P,

Petat C, Marguerie G et al. (2002) Sulfur Sparing in the Yeast

Proteome in Response to Sulfur Demand. Molecular Cell, 9,

713–723.

Fenchel T (2009) Microbial behavior in a heterogeneous world.

Science, 296, 1068–1071.

Galtier N, Lobry JR (1997) Relationships between genomic

G+C content, RNA secondary structures, and optimal

growth temperature in prokaryotes. Journal of Molecular

Evolution, 44, 632–636.

Garcia-Fernandez JM, Tandeau de Marsac N, Diez J (2004)

Streamlined regulation and gene loss as adaptive mechanisms

in Prochlorococcus for optimized nitrogen utilization in

oligotrophic environments. Microbiology and Molecular Biology

Reviews, 68, 630–638.

Gomez-Baena G, Rangel OA, Lopez-Lozano A, Garcıa-

Fernandez JM, Diez J (2009) Stress responses in Prochlorococcus

MIT9313 vs SS120 involve differential expression of genes

encoding proteases ClpP, FtsH and Lon. Research in Micro-

biology, 160, 567–575.

Hickey DA, Singer GAC (2004) Genomic and proteomic adap-

tations to growth at high temperature. Genome Biology, 5,

1171–1177.

Horst JP, Wu TH, Marinus MG (1999) Escherichia coli mutator

genes. Trends in Microbiology, 7, 29–36.

Johnson ZI, Zinser ER, Coe A, McNulty NP, Woodward EMS,

Chisholm SW (2006) Niche partitioning among Prochlorococcus

ecotypes along ocean-scale environmental gradients. Science,

311, 1737–1740.

Karl DM (2002) Nutrient dynamics in the deep blue sea. Trends

in Microbiology, 10, 410–418.

Kettler GC, Martiny AC, Huang K, Zucker J, Coleman ML,

Rodrigue S, Chen F et al. (2007) Patterns and implications of

gene gain and loss in the evolution of Prochlorococcus. PLoS

Genetics, 3, e231.

Li N, Lv J, Niu D-K (2009) Low contents of carbon and

nitrogen in highly abundant proteins: evidence of selection

for the economy of atomic composition. Journal of Molecular

Evolution, 68, 248–255.

Linden D, Erdner E, Marie D, Prasil O, Koblızek M, Le Gall F,

Rippka R et al. (2002) Nitrogen stress response of

Prochlorococcus strain PCC 9511 (oxyphotobacteria) involves

contrasting regulation of ntcA and amt1. Journal of Phycology, 38,

1113–1124.

Lipschultz F (2001) A time-series assessment of the nitrogen

cycle at BATS. Deep Sea Research Part II: Topical Studies in

Oceanography, 48, 1897–1924.

Lv J, Ning L, Deng-Ke N (2008) Association between the

availability of environmental resources and the atomic

composition of organismal proteomes: Evidence from

Prochlorococcus strains living at different depths. Biochemical

and Biophysical Research Communications, 375, 241–246.

Mackay WJ, Han S, Samson LD (1994) DNA alkylation repair

limits spontaneous base substitution mutations in Escherichia

coli. Journal of Bacteriology, 176, 3224–3230.

Martiny AC, Coleman ML, Chisholm S (2006) Phosphate

acquisition genes in Prochlorococcus ecotypes: evidence for

genome-wide adaptation. Proceedings of the National Academy

of Sciences of the United States of America, 103, 12552–12557.


Mazel D, Marliere P (1989) Adaptive eradication of methionine

and cysteine from cyanobacterial light-harvesting proteins.

Nature, 341, 245–248.

Moore LR, Post AF, Rocap G, Chisholm SW (2002) Utilization of

different nitrogen sources by the marine cyanobacteria Pro-

chlorococcus and Synechococcus. Limnology and Oceanography, 47,

989–996.

Mopper K, Lindroth P (1982) Diel and depth variations in

dissolved free amino acids and ammonium in the Baltic Sea

determined by shipboard HPLC analysis. Limnology and

Oceanography, 27, 336–347.

Naito S, Hirai MY, Chino M, Komeda Y (1994) Expression of a

Soybean (Glycine max [L.] Merr.) Seed storage protein gene

in transgenic Arabidopsis thaliana and its response to

nutritional stress and to abscisic acid mutations. Plant

Physiology, 104, 497–503.

Nakao M, Okamoto S, Kohara M, Fujishiro T, Fujisawa T, Sato

S, Tabata S et al. (2010) Cyanobase: the cyanobacterial

genome database update 2010. Nucleic Acids Research, 38,

D379–D381. URL: http://genome.kazusa.or.jp/cyanobase.

Partensky F, Hess WR, Vaulot D (1999) Prochlorococcus, a

marine photosynthetic prokaryote of global significance.

Microbiology and Molecular Biology Reviews, 63, 106–127.

Paul S, Dutta A, Bag SK, Das S, Dutta C (2010) Distinct, ecotype-

specific genome and proteome signatures in the marine

cyanobacteria Prochlorococcus. BMC Genomics, 11, 103.

Petrucco S, Bolchi A, Foroni C, Percudani R, Rossi GL, Ottonello

S (1996) A maize gene encoding an NADPH binding enzyme

highly homologous to isoflavone reductases is activated in

response to sulfur starvation. Plant Cell, 8, 69–80.

R Development Core Team (2009) R: A Language and Environment

for Statistical Computing. R Foundation for Statistical Computing,

Vienna, Austria. ISBN 3-900051-07-0, URL: http://www.

R-project.org.

Redfield AC (1934) On the proportions of organic derivatives

in sea water and their relation to the composition of

plankton. In: James Johnstone Memorial Volume (ed. Daniel

RJ). pp. 176–192, University Press of Liverpool, Liverpool,

UK.

Rocap G, Distel DL, Waterbury JB, Chisholm S (2002) Resolution

of Prochlorococcus and Synechococcus ecotypes using 16S-23S

rDNA internal transcribed spacer (ITS) sequences. Applied and

Environmental Microbiology, 68, 1180.

Rocap G, Larimer FW, Lamerdin J, Malfatti S, Chain P,

Ahlgren NA et al. (2003) Genome divergence in two Prochlo-

rococcus ecotypes reflects oceanic niche differentiation.

Nature, 424, 1042–1047.

Smith DF (1986) Small-Scale Spatial Heterogeneity in Dissolved

Nutrient Concentrations. Limnology and Oceanography, 31,

167–171.

Takahashi H, Braby CE, Grossman AR (2001) Sulfur economy

and cell wall biosynthesis during sulfur limitation of

Chlamydomonas reinhardtii. Plant Physiology, 127, 665–673.

Tolonen AC, Aach J, Lindell D, Johnson ZI, Rector T, Steen R,

Church GM et al. (2006) Global gene expression of Prochlo-

rococcus ecotypes in response to changes in nitrogen availability.

Molecular Systems Biology, 2, 1–15.

Tralau T, Vuilleumier S, Thibault C, Campbell BJ, Hart CA,

Kertesz MA (2007) Transcriptomic analysis of the sulfate

starvation response of Pseudomonas aeruginosa. Journal of

Bacteriology, 189, 6743–6750.

Valas RE, Bourne PE (2008) Rethinking proteasome evolution:

two novel bacterial proteasomes. Journal of Molecular

Evolution, 66, 494–504.

Varela DE, Harrison PJ (1999). Seasonal variability in

nitrogenous nutrition of phytoplankton assemblages in the

northeastern subarctic Pacific Ocean. Deep Sea Research Part

II: Topical Studies in Oceanography, 46, 2505–2538.

JDJG is a postdoc working in WFF’s laboratory. He is inter-

ested in the impact of nutritional tradeoffs on evolutionary

biology at multiple levels, ranging from molecular evolution to

behavioural ecology. Research by WFF meshes field research,

theoretical models and bioinformatics to address critical ques-

tions in ecology, evolution, and conservation biology.

Appendix A

Catabolic proteins: GO:0006508, GO:0030163, GO:0009253,

GO:0016998, GO:0006546, GO:0016075, GO:0006527, GO:0006308,

GO:0006007, GO:0030655, GO:0009264, GO:0006510, GO:0006402;

Anabolic proteins: GO:0009236, GO:0009228, GO:0009231,

GO:0015995, GO:0033014,GO:0000105, GO:0009097, GO:0016114,

GO:0008615, GO:0008652,GO:0019877, GO:0017000, GO:0009252,

GO:0030494, GO:0009435,GO:0010024, GO:0008299, GO:0019295,

GO:0009058, GO:0009073,GO:0019277, GO:0044249, GO:0006207,

GO:0006596, GO:0006221,GO:0006779, GO:0009103, GO:0006171,

GO:0006750, GO:0008295,GO:0006529, GO:0009245, GO:0006229,

GO:0016117, GO:0006571,GO:0008610, GO:0006633, GO:0006164,

GO:0009113, GO:0006526,GO:0000271, GO:0009059, GO:0006189,

GO:0006799, GO:0008654,GO:0009396, GO:0009190, GO:0006241,

GO:0006183, GO:0006228,GO:0006542, GO:0009098, GO:0006777,

GO:0009089, GO:0019438,GO:0006564, GO:0009250, GO:0009234,

GO:0016051, GO:0006783,GO:0019288, GO:0006188, GO:0009082,

GO:0006561, GO:0015937,GO:0009086, GO:0015940, GO:0008616,

GO:0006231, GO:0009107,GO:0009088, GO:0006754, GO:0045226,

GO:0006535, GO:0009168,GO:0042450, GO:0019363, GO:0006537,

GO:0009102, GO:0009094,GO:0005978, GO:0006233, GO:0006235,

GO:0000162, GO:0006222,GO:0009152, GO:0009165, GO:0009156,

GO:0006177, GO:0019281,GO:0019856;

Ribosomal proteins: GO:0005840, GO:0003735, GO:0015935,

GO:0015934, GO:0043022, GO:0042254


Contrasting mechanisms of proteomic nitrogen thrift in Prochlorococcus

Documents

Transcript of Contrasting mechanisms of proteomic nitrogen thrift in Prochlorococcus