The biochemical and elemental compositions of marine plankton:

A NMR perspective

J.I. Hedges a,*, J.A. Baldock b, Y. Gelinas a,1, C. Lee c, M.L. Peterson a, S.G. Wakeham d

aSchool of Oceanography, University of Washington, Box 355351, Seattle, WA 98195-5351, USAbCSIRO Land and Water, PMB #2, Glen Osmond, SA 5064, Australia

cMarine Sciences Research Center, Stony Brook University, Stony Brook, NY, 11794-5000, USAdSkidaway Institute of Oceanography, 10 Ocean Science Circle, Savannah, GA 31411, USA

Received 11 July 2001; received in revised form 2 January 2002; accepted 30 January 2002

Abstract

The traditional Redfield–Ketchum–Richards (1963) equation for the production (or respiration) of ‘‘average marine

plankton’’

106 CO2 þ 16 HNO3 þ H3PO4 þ 122 H2O ¼ ðCH2OÞ106ðNH3Þ16ðH3PO4Þ þ 138 O2

has long been a useful guideline for establishing the ratios and reaction extents of the bioactive elements in ocean systems. The

empirical formula on the right of the above equation for marine plankton biomass adequately represents the C/N/P of mixed

marine plankton collected in towed nets, but includes an impossibly elevated hydrogen content and a questionably high level of

organic oxygen. An elevated estimate of oxygen content is particularly critical because it would lead to an underestimate of the

amount of O2 required for complete respiration of plankton biomass. Although direct biochemical measurements have been

used previously to constrain the compositions, and hence the reaction stoichiometries, of marine plankton and their remains,

such analyses can be prone to error and analytical bias. To cast a new light on the chemical composition of marine plankton, we

determined the major functional group distribution of organic carbon in mixed plankton tows from five contrasting ocean sites

using cross-polarization, magic-angle spinning carbon-13 nuclear magnetic resonance (CP/MAS 13C NMR). Using a mixing

model that relates NMR spectral data to biochemical composition, we estimate an average major biochemical composition

(weight basis) for these plankton samples of 65% protein, 19% lipid and 16% carbohydrate. This biochemical composition

corresponds to an average elemental formula for plankton biomass of C106H177O37N17S0.4, whose complete oxidation requires

154 moles of O2. Although preliminary, this 13C NMR-based estimate indicates elemental compositions and respiratory oxygen

demands that are widely different from those indicated by the RKR composition (C106H260O106N16 and 138 O2, respectively)

and those determined in many previous field studies. D 2002 Published by Elsevier Science B.V.

Keywords: Marine plankton; NMR; Biochemical; Elemental; Redfield ratios; Photosynthesis

0304-4203/02/$ - see front matter D 2002 Published by Elsevier Science B.V.

PII: S0304 -4203 (02 )00009 -9

* Corresponding author. Tel.: +1-206-543-0744; fax: +1-206-543-0275.

E-mail address: jihedges@u.washington.edu (J.I. Hedges).1 Now at: Concordia University, Chemistry and Biochemistry Department, 1455 de Maisonneuve Blvd. West, Montreal, Canada, H3G

1M8.

www.elsevier.com/locate/marchem

Marine Chemistry 78 (2002) 47–63

1. Introduction

A representative average elemental composition for

marine plankton biomass can be an extremely useful

oceanographic tool. Such a formula helps identify

limiting dissolved nutrients for phytoplankton produc-

tion (Redfield et al., 1963), allows calculations of

‘‘preformed’’ nutrients as water mass tracers (e.g.

Broecker, 1974), and provides a means for estimating

the uptake and regeneration fluxes ofmultiple bioactive

elements based on the direct measurement of only one

element (Broecker and Peng, 1982; Takahashi et al.,

1985). Dissolved molecular oxygen is particularly

useful for estimating cumulative organic matter respi-

ration because the initial O2 concentration of a sub-

merged water mass is closely constrained by its

temperature and salinity, assuming saturation with

atmospheric O2 prior to downwelling. The difference

between the calculated (initial) andmeasured (sampled)

O2 contents of the subsurface water, its ‘‘apparent

oxygen utilization’’ (AOU), can be used to determine

the cumulative input of dissolved elements (e.g. C, N,

P) released from in situ respiration of organic matter

(Broecker and Peng, 1982).

The atomic C/N/P ratio of 106:16:1 in the biomass

of ‘‘average marine plankton’’ was first published by

Redfield (1934) and formalized by Redfield et al.

(1963). This ‘‘Redfield’’ ratio was based on direct

analysis of these three elements in zooplankton and

phytoplankton collected by towing nets at numerous

sites in the upper ocean. This fundamental Redfield

stoichiometry has withstood the test of time, as

indicated by observed C/N/P regeneration ratios near

106:16:1 in the deep ocean (Takahashi et al., 1985;

Anderson and Sarmiento, 1994; Shaffer et al., 1999).

Redfield’s fundamental stoichiometry was later

‘‘fleshed out’’ by Richards (1965) in the form of the

following equation

106 CO2 þ 16 HNO3 þ H3PO4 þ 122 H2O

¼ ðCH2OÞ106ðNH3Þ16ðH3PO4Þ þ 138 O2 ð1Þ

In this ‘‘RKR’’ equation, each reactant element is

depicted as occurring in uncharged molecules of the

predominant seawater component, whereas (CH2O)106(NH3)16(H3PO4) represents a P-normalized unit of

‘‘phosphorylated amino-carbohydrate’’ in mixed

marine net plankton. The indicated number of oxygen

molecules released by photosynthesis was theoreti-

cally calculated assuming that one mole of O2 is ge-

nerated for every CO2 reduced to CH2O, and 2 moles

of O2 are generated from every HNO3 reduced to NH3

(Redfield et al., 1963). Respiration was taken as the

reverse of Eq. (1) and assumed to be complete to the

level of nitrate production. Thus, complete oxic deg-

radation theoretically would require 138 moles of dis-

solved O2/106 moles of organic carbon (C), and hence

a molar respiration ratio (O2/C) of 1.30.

However, as has been pointed out previously by

Vollenweider (1985) and Anderson (1995), the RKR

equation embodies elevated hydrogen and oxygen

contents for marine plankton versus values that would

be expected based on their published major biochem-

ical compositions. Hydrogen content is elevated in

part because water loss by dehydration that would

result from combining the nominal structural units

(CH2O and NH3) into polysaccharides and proteins

has not been taken into account. This H elevation does

not affect the calculated respiration demand because it

is nullified mathematically by adding more H2O to

the left of Eq. (1). The oxygen content of (CH2O)106(NH3)16(H3PO4), however, is also elevated (Vollen-

weider, 1985; Anderson, 1995) because the major

organic structural unit is formulated as carbohydrate,

which contains appreciably more oxygen than the

other major biochemical components of marine

plankton such as lipids and proteins. Because more

carbohydrate-rich organic substances require less net

oxygen for respiration, the O2/C ratio of 1.30 for

RKR plankton represents a minimal theoretical esti-

mate of O2 demand during respiration, and of O2

production via photosynthesis (Laws, 1991).

Direct analysis of marine plankton could provide

the needed major element information, but necessi-

tates measurements that are difficult for organic

hydrogen, oxygen and sulfur. Measurement of organic

hydrogen content typically involves combustion to

H2O in a CHN analyzer and thus is subject to errors

from other water sources. It is challenging, however,

to completely dry salt-rich plankton samples, espe-

cially if hygroscopic CaCl2 residues are formed when

inorganic carbon is removed by HCl without rinsing

(Froelich, 1980). Additional water of inorganic origin

can be released during CHN analysis (typically at

> 1000 jC) due to thermal breakdown of opal (SiO2�nH2O) and hydrous aluminosilicates. Organic oxygen

J.I. Hedges et al. / Marine Chemistry 78 (2002) 47–6348

often is estimated by mass difference, once the weight

percentages of all the other organic and inorganic

elements are measured by CHN and ash analysis.

This approach is intrinsically insensitive and subject

to cumulative errors in all the contributing determi-

nations, including inorganic mass changes during

heating to ashing temperatures (Hedges and Stern,

1984). Pyrolytic analysis of organic oxygen as CO in

specially adapted CHN analyzers (Chen et al., 1996)

is a more dependable analytical approach, but often

necessitates demineralization and is not always fea-

sible or accurate (later discussion). Quantitative sulfur

measurement, typically performed by determining the

amount of SO2 liberated during sample combustion

(e.g. Chen et al., 1996), can be particularly difficult

because of the low ratio of organic sulfur to seawater-

derived sulfate ion in marine plankton.

The needed elemental data also can be obtained by

a comprehensive quantification of all the major bio-

chemical components of plankton (Vollenweider,

1985; Laws, 1991; Anderson, 1995). This general

approach requires two conditions: (1) that the ele-

mental composition within each major type of bio-

chemical composing marine plankton be known, and

(2) that the relative amounts of each of these major

biochemical types also be determined. The usual

sources of both pieces of information are molecular-

level analyses of the individual biomonomers (e.g.

amino acids, sugars, fatty acids and sterols) that make

up the major biopolymers (e.g. proteins, polysacchar-

ides and lipids). With respect to the first condition, to

meaningfully constrain the elemental compositions

within these three biochemical families necessitates

that at least the major protein amino acids (f 20 in

number), neutral sugars (f 10) and neutral and acidic

lipids (>50) be chromatographically quantified as

individual compounds. This minimal accounting

requires at least three individual preparations and

chromatographic analyses of each sample. The most

comprehensive molecular-level analysis to date for net

plankton material accounted for about 80% of total

carbon in the form of chromatographically resolved

biochemicals (Wakeham et al., 1997a,b). Because the

unrepresented organic carbon could occur in a wide

variety of compositionally dissimilar forms, including

nucleotides (Anderson, 1995), acidic and basic sugars

(e.g. Bergamaschi et al., 1999) and hydrolysis-resist-

ant macromolecules (De Leeuw and Largeau, 1993;

Zegouagh et al., 1999), molecular analysis alone does

not necessarily provide definitive elemental composi-

tions (or overall amounts) for each major biochemical

type.

The second condition of estimating the relative

amounts of each major biochemical type in marine

plankton is usually approached by comparing summed

biomonomer yields among different biochemical

classes, although colorimetric measurements or spec-

tral analyses can also be used. A major challenge in

comparing chemically based estimates of relative

abundances is that each of the several major biochem-

ical types is typically measured by different methods

in different samples—and usually by different labor-

atories at different times. Moreover, it is generally

more challenging to chemically measure absolute

amounts (e.g. yields) of biomonomers than relative

amount (e.g. within-class mole percentages) because

absolute quantification requires detailed knowledge of

efficiencies with which each measured component is

released from a particular sample matrix and recovered

(e.g. Cowie and Hedges, 1984; Pakulski and Benner,

1992). Thus the percentages of the major biochemical

types in marine plankton are at least as uncertain as the

elemental compositions within these mixture endmem-

bers.

In a seminal paper, Anderson (1995) used chem-

ically based literature values for the elemental

compositions and relative abundance of four major

biochemical types to estimate a representative formula

for marine plankton. He calculated an average formula

of C106H185O38N16PO4, which corresponds to an

atomic O2/P of f 150. This formulation represented

a major step toward correcting the elevated oxygen

and hydrogen contents in the RKR equation (C106

H260O106N16, 138 O2), and in bringing the associated

respiratory oxygen demand toward more realistic

higher values. However, this literature-based assess-

ment necessarily drew on a variety of analyses from

disperse laboratories on different biochemical classes

in disparate samples (Laws, 1991; Anderson, 1995).

In addition, these chemical measurements are subject

to the previously discussed uncertainties in composi-

tion and absolute yield. Given especially the challenge

of estimating the relative abundances of the major

biochemical types within complex samples, an inde-

pendent perspective on the elemental composition of

average marine plankton would be useful.

J.I. Hedges et al. / Marine Chemistry 78 (2002) 47–63 49

We describe here the application of cross-polar-

ization, magic-angle spinning 13C nuclear magnetic

resonance (CP/MAS 13C NMR) to estimate the major

biochemical content of marine net plankton from five

contrasting open-ocean sites. These contents are used

in conjunction with recent molecular-level amino acid

and sugar analyses to estimate the elemental compo-

sition of five marine plankton samples. Our NMR-

based results are consistent with elemental composi-

tions estimated by Anderson (1995) using chemical

data drawn from different literature sources. Together,

these independent assessments indicate that the RKR

formula should be supplanted with a compositionally

more realistic stoichiometry. The revised formula for

average marine plankton contains less hydrogen and

oxygen and corresponds to a respiratory oxygen de-

mand intermediate between the low extreme of the

RKR equation and the high values often calculated

from field measurements.

2. Materials and methods

2.1. Samples and elemental analysis

The five studied plankton materials were collected

by oblique vertical tows of a 26-Am net from 100 to 0

m water depth at the locations indicated in Table 1.

Like essentially all net tows, including the samples of

Redfield et al. (1963), these field samples comprise

mixtures of phytoplankton and zooplankton at varying

unknown ratios. Our mixed plankton collections rep-

resent a range of ocean settings including a Pacific

equatorial upwelling zone (E), a monsoonal region of

the Arabian Sea (A) and a transect of three sites (1–3)

across a strong nutrient gradient in the Southern

Ocean. The collected samples were passed through

an 850-Am stainless steel sieve to remove large organ-

isms. The plankton samples were collected onto 90-

mm diam. GF/A glass fiber filters, then scraped off,

freeze-dried and ground to pass a 351-Am-mesh sieve.

Organic carbon (C), total hydrogen (H) and total

nitrogen (N) were determined (Table 1) with a Carlo-

Erba 1001 CHN analyzer (Hedges and Stern, 1984).

Organic carbon was measured after removing carbo-

nates with HCl vapor, whereas H and N were measured

in untreated samples. Total sulfur (S) was determined

by combustion to SO2 in a Carlo-Erba NA-1500

elemental analyzer (Pella and Colombo, 1978). Ash

content was determined by heating individual pre-

weighed samples to 400 jC for 4 h. The final ash

mass was corrected by subtracting the small amount of

remaining C and N (Gelinas et al., 2001a). Oxygen

content was estimated by the difference of total mass

minus the combined masses of C, N, H, S and ash.

2.2. NMR analyses

Solid-state CP/MAS 13C NMR spectra were ac-

quired at a 13C frequency of 50.3 MHz on a Varian

Unity 200 spectrometer. Samples were packed into a

7-mm-diameter cylindrical zirconia rotor with Kel-F

end caps and spun at the magic angle at a rate of

5000F 100 Hz in a Doty Scientific MAS probe. A

conventional cross-polarization pulse sequence was

used with a 1.0-ms contact time. An inversion recov-

ery pulse sequence was employed to estimate a relax-

ation time (T1H value) for each sample. These

Table 1

Sources and measured weight percentages of individual elements in the plankton tow samples

Sample Sym. Lat./Long. Date %C %H %Oa %N %Sb %Ash

Equatorial Pacific E 0jN, 140jW 8/31/92 21.5 3.19 20.3 4.08 1.21 49.7

Arabian Sea A 17jN, 60jW 1/5/96 14.0 2.06 25.0 2.38 1.04 55.5

Southern Ocean 1 1 74jS, 177jE 11/16/98 14.9 2.09 23.2 3.11 1.12 55.5

Southern Ocean 2 2 66jS, 169jW 11/21/98 15.8 2.22 22.3 3.28 0.78 55.6

Southern Ocean 3 3 57jS, 170jW 11/24/98 14.9 2.04 17.2 2.95 0.16 62.8

Mean m f f 16.2 2.32 21.6 3.16 0.86 55.8

Sym. = symbol, Lat. = latitude, Long. = longitude.a Oxygen percentages were calculated by difference.b Sulfur percentages were corrected for sea salt contribution based on mass loss on drying and the concentration of dissolved sulfate ion in

35 ppt seawater.

J.I. Hedges et al. / Marine Chemistry 78 (2002) 47–6350

relaxation times were used to set the duration of the

recycle delay between pulses to >7 times the longest

T1H value measured for each sample (Wilson, 1987).

For all samples, a relaxation delay of 0.3 s was

sufficient to satisfy this requirement. The number of

transients collected for individual samples ranged

between 22,000 and 262,000 depending on the

amount of carbon that could be placed in the rotor.

Spectra were processed with a 50-Hz Lorentzian line

broadening and 0.010 s Gaussian broadening. Spectral

distributions for all samples were determined by

integrating signal intensities within each of the seven

chemical shift regions given in Table 2. The ppm

bounds of these regions were selected based on

previous experience with quantification of 13C NMR

mass spectral data (Hedges et al., 2001; Baldock and

Smernik, submitted for publication; Nelson and Bal-

dock, submitted for publication). Areas were divided

into individual spectral regions by dropping vertical

lines to a baseline defined by the Varian VNMR oper-

ating software between 300 and � 100 ppm. More

intricate spectral deconvolution schemes (e.g. Mao et

al., 2000) were not attempted because insufficient

information presently exists on individual resonance

positions and shapes to adequately differentiate con-

tributions among all the overlapping carbon types in

these plankton samples and their hypothetical bio-

chemical components. Signal intensities associated

with spinning side bands of all major resonances

above 110 ppm consistently accounted for < 1% of

the parent peak intensity, and were numerically allo-

cated back to the spectral region from which they

were derived (Baldock and Smernik, submitted for

publication). Triplicate CP/MAS 13C NMR analyses

were made for each of the three Southern Ocean

samples by repacking aliquots of the same material

into a rotor and acquiring spectra under otherwise

identical conditions. The overall reproducibility with

which major areas (>2 area%) could be determined by

triplicate analyses of individual spectral regions aver-

aged F 4% (percent sample mean deviation) of the

measured value.

2.3. Mixing model

The measured spectral data were entered into a

three-endmember mixing model to estimate the major

biochemical makeup of each plankton tow sample,

and the corresponding elemental composition (Nelson

et al., 1999; Hedges et al., 2001). The model involved

only ‘‘protein,’’ ‘‘carbohydrate’’ and ‘‘lipid’’ end-

members, which were used to represent the more

diverse biochemical types and forms actually occur-

ring in the analyzed plankton samples. Average spec-

tral and elemental characteristics were calculated for

each of the three nominal biochemical endmembers

based on literature values for the relative abundances

and carbon functionalities of their chromatographi-

cally resolved structural units (see below). The three

endmembers were then numerically mixed in the

Table 2

Carbon area percentages measured in spectral regions (ppm) of individual plankton tow samples and calculated for the protein, lipid and

carbohydrate endmembers described in the text

Sample 0–45 (I) 45–60 (II) 60–95 (III) 95–110 (IV) 110–145 (V) 145–165 (VI) 165–215 (VII)

Equatorial Pacific 44.9 14.0 19.0 1.8 6.0 0.0 14.3

Arabian Sea 45.7 12.4 19.4 2.9 6.5 1.1 12.0

Southern Ocean 1a 38.4 16.2 16.4 1.3 8.3 1.3 18.1

Southern Ocean 2a 37.4 15.3 16.9 1.6 8.7 1.7 18.4

Southern Ocean 3a 41.9 15.3 15.1 1.2 6.9 1.6 18.0

Averageb 41.6F 3.3 14.6F 1.3 17.4F 1.6 1.8F 0.6 7.3F 1.0 1.1F 0.6 16.2F 2.6

Protein 36.3 21.0 7.4 0.0 7.8 1.5 26.0

Lipid 83.3 0.0 0.0 0.0 11.1 0.0 5.6

Carbohydrate 0.0 0.0 83.3 16.7 0.0 0.0 0.0

Roman numerals correspond to the following spectral intervals in Fig. 1: I = alkyl, II =N-alkyl, III =O-alkyl, IV = di-O-alkyl, V=C=C* –H

(or –C), VI =C=C* –O (or –N), and VII = carbonyl (primarily in carboxyl, ester and amide).a Carbon percentages are averages of triplicate independent NMR analyses.b Indicated intervals are F one standard deviation of the mean for the five plankton net tow samples.

J.I. Hedges et al. / Marine Chemistry 78 (2002) 47–63 51

model to determine the percentage of each biochem-

ical that gave the best overall agreement between the

calculated and measured spectral abundances for each

plankton tow sample. Fixing the endmember percen-

tages for any sample with this model also fixes the

corresponding elemental composition.

Relative intensities and corresponding elemental

compositions, expected for average plankton protein,

were calculated based on the average mole percen-

tages of individual amino acids (except proline, cys-

tine and tryptophan) measured in five different marine

phytoplankton by Cowie and Hedges (1992). Com-

plementary contributions by proline, cystine and tryp-

tophan were added based on the average amino acid

compositions of 25 different marine plankton pub-

lished by Chau et al. (1967). Although the average

amino acid compositions reported for marine plankton

by Cowie and Hedges (1992) and Chau et al. (1967)

were similar, the more recent compositions were used

here because they were measured with charge-matched

recovery standards. Inspection of other measurements

for the amino acid contents of marine plankton (Chue-

cas and Riley, 1969; Hecky et al., 1973) and upper

ocean particles (Cowey and Corner, 1963, 1966; Lee

and Cronin, 1984; Lee et al., 2000) also indicate mi-

nimal compositional variability.

Once average mole percentages of amino acids

were determined from the literature, individual car-

bons in each amino acid were assigned proportion-

ately to one of the seven spectral regions in Table 2.

These assignments were based on well-established

relationships between carbon structural characteristics

and spectral characteristics (e.g. Wilson, 1987). The

corresponding elemental composition of average

plankton ‘‘protein’’ was calculated using the same

mole percentages and known elemental compositions

of each amino acid (Lehninger, 1975). The fact that

essentially all amino acids in plankton actually occur

in proteins and other combined forms was taken into

account by subtracting one water molecule per amino

acid structural unit. When normalized to 106 carbons,

as in the RKR format, our calculated elemental

composition for average plankton protein was C106

H168O34N28S1.

The elemental composition of the ‘‘carbohydrate’’

endmember was assigned a nominal value of C6H10O5

based on the individual neutral sugar compositions

measured by Hernes et al. (1996) for 11 plankton net

tow samples collected from the equatorial Pacific

Ocean. This formula corresponds to a hexose polymer

(such as cellulose or starch) in which each sugar unit

loses one water molecule in forming a glycosidic

bond. The neutral sugar compositions measured by

Hernes et al. (1996) correspond to an average ele-

mental formula (with all aldoses in dehydrated form)

of C6.0H9.9O4.8. Although data are limited, acidic

sugars appear to compose roughly 5.5 wt.% of the

total carbohydrate in marine phytoplankton and zoo-

plankton and occur almost exclusively as glucuronic

and galacturonic acids (Bergamaschi et al., 1999).

When the above elemental composition derived from

Hernes et al. (1996) is recalculated with an additional

5.5 wt.% of these two 6-carbon uronic acids (also in

dehydrated form), the resulting average formula

becomes C6.0H9.6O4.9. Because neither of these liter-

ature-based average compositions is appreciably dif-

ferent from that of a pure hexose polymer, the use of

C6H10O5 as the carbohydrate endmember in the mix-

ing model seems entirely warranted. Anderson (1995)

assigned the same formula based on different liter-

ature data. This compositional assignment is also

consistent with observed area ratios near a value of

six for anomeric (110 ppm) to total (60–110 ppm)

carbohydrate carbon resonances in the samples ana-

lyzed in the present study (e.g. Fig. 1). The RKR form

for this hypothetical carbohydrate component is

C106H177O88.

The ‘‘lipid’’ component was assigned the structure

of oleic acid, C18H34O2, without assuming any loss of

water upon incorporation into biomass. Oleic acid was

chosen because fatty acids are major components of

lipids in phytoplankton and sinking marine particles

(Wakeham et al., 1997a,b) and because other lipids

and algaenans also contain small amounts of oxygen

(De Leeuw and Largeau, 1993). This monounsatu-

rated fatty acid also helps to account for hydrogen loss

due to the presence of double bonds and rings in many

plankton lipids (e.g. Wakeham et al., 1997a,b). We

explored the average goodness of fit obtained with

stearic (C18:0), oleic (C18:1) and linoleic (C18:2) acid

analogs for the ‘‘lipid’’ endmember in the mixing

model (see next paragraph) and found that oleic acid

fit best, followed closely by stearic acid and distantly

by linoleic acid. Although lipids are by definition

solvent extractable and include a variety of cyclic,

CMC, nitrogen and sulfur-containing examples (Eglin-

J.I. Hedges et al. / Marine Chemistry 78 (2002) 47–6352

ton and Murphy, 1969), a better widely recognized

name for such alkyl-rich material is not evident.

The above distributions of assigned spectral inten-

sities for the protein, carbohydrate and lipid endmem-

bers were used in a simple spreadsheet to calculate the

mole percentages of carbon in each of these three

biochemical classes that are indicated by NMR char-

acterization. Fig. 2 illustrates such a spreadsheet

calculation, in this case as applied to estimating the

percentages of each endmember contributing toward

the seven spectral intensities calculated for ‘‘average’’

plankton in Table 2. The three bars to the left of each

box in Fig. 2 indicate the relative contributions within

a given spectral region made by each of the three

endmembers. The relative magnitudes of these dis-

tributed contributions were determined by the calcu-

lated percentage of total sample carbon that each

biochemical composed. The mole percentages of

carbon in each of the three hypothetical endmembers

were then used as variables to obtain ‘‘calculated’’

total intensities (from protein + carbohydrate + lipid)

in each region that best fit the corresponding inten-

sities measured by 13C NMR (white boxes). The best

fit was obtained by minimizing the sum of the squared

deviations between the calculated and measured inten-

sities in each of the seven spectral regions, and thus

gives most weight to close matching of major inten-

sities. The only additional constraint in the minimiza-

tion routine was that the sum of the mole percentages

of carbon in each sample was forced to a value of 100.

The corresponding weight percentages of carbon in

protein, carbohydrate and lipid are listed in Table 3

along with the resultant elemental formulas. The moles

of oxygen required to completely respire the individual

biochemical mixtures in Table 3 to nitrate, carbon

Fig. 2. Modeled contributions made by the hypothetical protein,

carbohydrate and lipid endmembers (leftmost three bars in each

spectral region) to the corresponding percentage of total area

represented by that carbon type in the ‘‘average’’ plankton sample

(see Table 2). The black ‘‘calculated’’ bars in each of the seven

spectral regions represent the average contribution of carbon made

by the sum of the individual endmembers and the white ‘‘NMR’’

bars present the average measured distribution of signal intensity

from the 13C NMR analyses. A comparison of the black and white

bars provides an indication of the goodness of fit between the

modeled and measured 13C NMR data.

Fig. 1. CP/MAS 13C NMR spectra of the compositionally extreme

Southern Ocean #1 (top) and Arabian Sea (bottom) plankton tow

samples. Also indicated are the seven resonance ranges over which

all spectra were integrated. The structures and terminologies

corresponding to each spectral region are given in Table 2.

J.I. Hedges et al. / Marine Chemistry 78 (2002) 47–63 53

dioxide, water and sulfur trioxide were calculated by

the equation

CaHbOcNdSr þ xO2

¼ aCO2 þ bH2Oþ dHNO3 þ rSO3 ð2Þ

where

x ¼ 1:00a þ 0:25b þ 1:25d þ 1:5r � 0:5c ð3Þ

These formulas are from Anderson (1995), except

that terms for the oxidation of organic sulfur to sulfur

trioxide are included in Eqs. (2) and (3).

3. Results and discussion

The following discussion will focus sequentially

on the compositional information that can be drawn

from the measured elemental compositions, NMR

analyses and model calculations. Selected examples

of plankton compositions determined by other means

will be given from the literature for comparison.

3.1. Measured elemental compositions

The five plankton tow samples contained 50–63

wt.% ash, which is typical for mixed marine plankton

having carbonate and opal tests (Bishop et al., 1977;

Broecker and Peng, 1982; Hernes et al., 1999). These

high mineral contents decreased the directly measured

weight percentages of the organic components by

factors of 2 or more (Table 1). Even after ash correc-

tion, the average organic carbon and oxygen contents

determined for these five plankton samples are 37F 6

and 49F 6 wt.%, respectively. In comparison, the

protein, carbohydrate and lipid endmembers contain

53, 44 and 77 wt.% carbon and 23, 49 and 11 wt.%

oxygen (Table 3). Because the high content of nitrogen

(7.2F 1.3 wt.%, ash-free basis) directly measured in

these plankton tow samples rules out a carbohydrate-

rich composition, their calculated oxygen contents

appear to be erroneously high (see later discussion).

Given that these organic oxygen contents were deter-

mined by subtracting the combined measured masses

of ash, C, H, N and S from the corresponding total

sample mass, the most likely source of error is an un-

derestimate of the inorganic (ash) component of the

Table 3

Modeled biochemical and elemental compositions of the plankton net tow samples

Sample Protein (wt.%) Carbohydrate (wt.%) Lipid (wt.%) Elemental formulaa DO2b RR

Equatorial Pacific 57 21 22 C106H179O37N15S0.3 152 1.43

Arabian Sea 51 24 25 C106H181O37N13S0.3 150 1.41

Southern Ocean 1 73 16 11 C106H174O38N20S0.5 156 1.47

Southern Ocean 2 73 17 10 C106H174O39N20S0.5 156 1.47

Southern Ocean 3 70 15 15 C106H176O36N19S0.4 156 1.47

Average of above 65F 9 19F 4 16F 6 C106H177O37N17S0.4 154 1.45

Anderson, 1995 58by 26 16 C106H175O38N16 149 1.40

Redfield et al., 1963 f f f C106H260O106N16 138 1.30

Parsons and Takahashi, 1973 f f f C106H195O70N15 139 1.31

Dyrssen, 1977 f f f C106H191O86N20 136 1.28

Honjo, 1980 f f f C106H165O28N14 151 1.42

Martin et al., 1987 f f f C106H175O?N15 175 1.65

Chen et al., 1996 f f f C106H121O57N14 129 1.22

Our CHN data f f f C106H182O106N18 125 1.18

(a) wt.%=weight percentage, (b) RR= respiration ratio of O2/C on a molar basis. The weight percentages of protein, carbohydrate and lipid

listed in the first five lines of this table were calculated using the biochemical mixing model described in the text. The corresponding elemental

formulas for the five plankton samples were directly calculated from these model-derived weight percentages using the hypothetical elemental

weight percentages assigned to the protein (53.1% C, 7.0% H, 22.7% O, 16.3% N, 0.9% S), carbohydrate (44.4% C, 6.2% H, 49.4% O) and

lipid (76.6% C, 12.1% H, 11.3% O) endmembers (see text). The intervals indicated for the average calculated weight percentages of protein,

carbohydrate and lipid are F 1 standard deviation for the corresponding mean.a Does not include P in phosphate or the associated four oxygen atoms.b Includes O2 needed to respire organic sulfur when S is given in the biomass formula.

J.I. Hedges et al. / Marine Chemistry 78 (2002) 47–6354

initial sample mass. This discrepancy may result in

part from thermal decomposition of opal and carbo-

nates at ashing temperatures with subsequent loss of

volatiles such as H2O and CO2. Oxygen overestima-

tion also is indicated by the high average atomic O/C

ratio (1.00) that we measured for the five plankton

samples, which greatly exceeds previous literature

values. In contrast, the corresponding average H/C

(1.72) and N/C (0.17) ratios fall within previously

reported ranges (e.g. Parsons and Takahashi, 1973;

Anderson, 1995).

3.2. 13C NMR measurements

The CP/MAS 13C NMR spectra of the Arabian Sea

and Southern Ocean #1 plankton tow mixtures (Fig. 1)

illustrate the compositional extremes observed among

the characterized samples (Tables 1–3). Whereas the

Arabian Sea spectrum is dominated by resonances

derived from alkyl carbons (0–45 ppm), the Southern

Ocean #1 counterpart has in addition comparably

predominant N-alkyl (45–60 ppm) and carbonyl

(165–215 ppm) resonances. The latter resonance is

centered at approximately 175 ppm in both spectra and

thus corresponds to carbon in amide, carboxyl and/or

ester groups (Wilson, 1987). The low-field resonance

at f 20 ppm is due to methyl carbon (e.g. extensive

alkyl branching) and is especially pronounced in the

Southern Ocean #1 sample (Fig. 1). In contrast, O-

alkyl (60–95 ppm) and di-O-alkyl (95–110 ppm) re-

sonances corresponding to carbohydrates are more

evident in the Arabian Sea plankton sample. Weaker

resonances are also evident in both spectra at approx-

imately 130 and 155 ppm for unsaturated carbon subs-

tituted by hydrogen/carbon and by oxygen/nitrogen,

respectively. Overall, the spectrum of the Southern

Ocean #1 plankton tow material closely resembles that

expected for protein-rich material (e.g. Knicker, 2000),

except that some carbohydrate is evident (Table 2).

The Arabian Sea plankton tow sample appears to con-

tain appreciably more carbohydrate and lipid carbon,

with the latter contributing primarily to linear (f 30

ppm) versus methyl-branched (f 15 ppm) alkyl struc-

tures. The higher nitrogen content of the Southern

Ocean #1 (3.11 wt.% N) versus the Arabian Sea (2.38

wt.% N) plankton sample is consistent with this in-

terpretation (Table 1). The few published plankton

spectra available for comparison are for freshwater

samples. These limited examples indicate that green

freshwater algae may be more carbohydrate rich

(Hatcher et al., 1983; Zelibor et al., 1988) than our

mixed marine samples. However, a mixed freshwater

algal culture from a benthic mat in the Everglades

(Knicker et al., 1996) exhibited major spectral features

similar to the Arabian Sea sample in Fig. 1.

3.3. Mixture analysis

Our use of CP/MAS 13C NMR spectra to estimate

the elemental compositions of the five plankton sam-

ples involves the three key assumptions that: (1) each

type of carbon contained in the samples is measured

with the same efficiency by the applied NMR proce-

dure, (2) the three biochemical types chosen as mixture

components account for effectively all the composi-

tional variability among the samples and (3) repre-

sentative elemental and spectral compositions can be

assigned to each biochemical endmember. These as-

sumptions will be discussed in the following para-

graphs, pointing out (when appropriate) how our treat-

ment varies from that of Anderson (1995).

To address the assumption that the various types of

carbon are measured with an equivalent efficiency by

our CP/MAS 13C methodology, preliminary NMR

analyses were conducted to insure that sufficiently

long recycle delay times between pulses were em-

ployed (Section 2.2). Ideally, direct (Bloch-decay)

spectral acquisition would be used in preference to

the CP/MAS method (Mao et al., 2000), which can

underestimate carbon occurring in hydrogen-poor

surroundings. However, Bloch decay spectra require

much longer time periods to acquire than CP/MAS

counterparts and are often not feasible and/or afford-

able. This is especially true when multiple analyses of

individual samples are collected, as was done here, to

establish analytical reproducibility. In addition, the

plankton tow samples examined in this study are all

hydrogen rich and thus should contain only small

amounts of unsaturated carbon devoid of protons,

which are prone to underestimation by CP/MAS

methodology (Golchin et al., 1997; Skjemstad et al.,

1999). The high organic carbon to paramagnetic metal

ratios characteristic of open-ocean plankton also favor

acquisition of representative NMR spectra (Gelinas et

al., 2001b). Essentially, the same acquisition parame-

ters and spectral treatments as were used here have

J.I. Hedges et al. / Marine Chemistry 78 (2002) 47–63 55

been successfully applied to a variety of solid sample

types derived from marine environments (Hedges et

al., 2001; Gelinas et al., 2001b,c).

Our second assumption that marine plankton are

composed only of protein, carbohydrate and lipid is

clearly an oversimplification. In particular, nucleic

acids, which reportedly are composed of 1–7 wt.%

of plankton carbon (Parsons et al., 1961) and are on

average unusually nitrogen rich (N/Cc 0.40) and

hydrogen poor (H/Cc 1.25), have not been taken

into account (see Anderson, 1995). However, given

that the composition of the three endmembers was

derived from experimental data collected from plank-

ton samples (amino acid, lipid and carbohydrate

analyses), the results are expected to be indicative of

the major fraction (>90%) of the plankton carbon.

More molecular-level data on the concentrations of

nucleic acid, and other biochemicals, in marine plank-

ton will be necessary to independently assess the

validity of the three-endmember model.

As for the third assumption of representative ele-

mental compositions, the formulas that we indepen-

dently estimated for the protein, carbohydrate and lipid

endmembers of marine plankton are very similar to

those previously assigned by Anderson (1995) based

on separate literature sources. For example, the for-

mula for plankton protein (C106H168O34N28S1) that we

derived from the data of Cowie and Hedges (1992) is

almost identical to the formula (C106H167O35N28) that

Anderson drew from the earlier compilation of Laws

(1991). In the case of carbohydrate, we have been able

to support Anderson’s endmember of C6H10O5 based

on detailed molecular-level analyses of marine plank-

ton samples (previous discussion). Given the structural

diversity, widely varying abundance, and differing

extraction efficiencies of algal lipids (Volkman et al.,

1998); it is difficult at present to closely constrain the

lipid endmember based on molecular-level analyses.

The oleic acid structural model we selected to repre-

sent algal lipid has atomic H/C and O/C ratios (1.89,

0.111) that are very similar to those of the C40H74O5

endmember (1.85, 0.125) used by both Laws (1991)

and Anderson (1995). Because the elemental compo-

sitions of all three endmembers are internally consis-

tent between our mixing model and Anderson’s

(1995), any major offset in the elemental composition

estimated by these two methods should be largely

traceable to differences in major biochemical abun-

dances as estimated from NMR analysis versus liter-

ature sources, respectively. This clear contrast is

appropriate because direct estimation of major bio-

chemical abundances by NMR analysis is the novel

aspect of our study.

Fig. 2 illustrates the best fit, on a carbon mole

percentage basis, of the mixing calculation to the

average spectral distributions of carbon types for the

five plankton tow samples (Table 2). The relative

variability (percent standard deviation) for the four

major (>10 area%) resonances (I, II, III and VII) in the

overall average spectral distribution was F 10% of the

measured value, with reproducibility dropping off as

area% values approach the average analytical preci-

sion (F 4%) for this data set. This variation among the

five plankton tow samples appears to be due mainly to

systematically higher protein contents of the three

Southern Ocean samples. The pattern of fit of the cal-

culated to the measured spectral areas obtained for the

average sample (Fig. 2) is typical of the entire sample

set. Most of the variability is traceable to an overes-

timation by the mixing model of carbonyl carbon

(region VII, 165–215 ppm) and an underestimate of

unsaturated carbon linked to H or C (region V, 110–

145 ppm). The underestimate of CMC may be due in

part to the exclusion of pigments, and nucleic acids

from the oversimplified three-endmember mixing mo-

del (Laws, 1991; Anderson, 1995).

The apparent overestimation of carbonyl carbon

has several possible sources. First, the N-alkyl (II,

45–60 ppm) resonance, which is a major determinant

for estimating protein carbon in the mixing model, sits

in a crowded spectral region (Fig. 1) and therefore

might be overestimated in area due to overlap with

large flanking alkyl and O-alkyl resonances. Such an

overlap would drive up all estimates of protein car-

bon, and especially that of the comparatively large

carbonyl component. However, the previously dis-

cussed consistent shortfall in the calculated abundan-

ces of unsaturated carbon (110–165 ppm), which also

has a major protein source in the mixing model (Fig.

2), argues against a major overestimation of protein

(see also later comments on N/C). Second, the hypo-

thetical protein endmember may be richer in amide

plus carboxyl carbon (both resonating near 175 ppm)

than the natural counterparts in Table 2. However, the

measured total acidic amino acid contents of the

plankton protein endmember (Cowie and Hedges,

J.I. Hedges et al. / Marine Chemistry 78 (2002) 47–6356

1992) and average protein (Lehninger, 1975) are both

20 mol%. Removing carboxyl (largely ester carbon)

from the lipid endmember would help substantially to

minimize the combined excess in the carbonyl region,

but is difficult to rationalize (Anderson, 1995). Given

the simplicity of the mixing calculation, the overall

fits of the calculated and measured spectral areas are

surprisingly good (Fig. 2), with an average total

squared difference of 9% per sample (F 3% if not

squared).

Application of the previously described mixing

model to the five plankton tow materials gives calcu-

lated protein contents of 51–73 wt.%, with an average

of 65F 9 wt.% (Table 3). The corresponding calcu-

lated carbohydrate 15–24 wt.% (average = 19F 4

wt.%) and lipid 10–25 wt.% (average = 16F 6

wt.%) values range by about a factor of 2. Of the

three endmembers, the weight ratio of carbohydrate to

lipid is least variable, with an average near 1.2. The

average RKR-equivalent formula obtained from the

mixing calculations is C106H177O37N17S0.4 (Table 3).

The mean sulfur content of these samples calculated

from the NMR data is 0.4 mol% (0.6 wt.%), with all

the calculated contribution coming from the protein

endmember (0.6 mol%). Nitrogen is the most variable

elemental component (F 16% coefficient of varia-

tion) in these samples. For the other elements, coef-

ficients of variation were < 3%. By scaling up the

elemental weight percentages experimentally meas-

ured for each of these samples (Table 1) to match the

corresponding weight percent of carbon derived from

the modeled NMR data (Table 3), it is possible to

estimate how well the measured and NMR-derived

compositions agree. This comparison indicates close

agreement for C (forced to match), H and N (Fig. 3),

but points toward erroneously high measured versus

calculated values of O, and especially S.

A convenient way to evaluate these mixing results

is to convert the calculated mixture solutions into

atomic ratios that can then be compared directly with

the measured elemental data (Table 1) and corre-

sponding literature values (Table 3). Borrowing from

the coal petrology (van Krevelen, 1961) and kerogen

(Durand, 1980) literature, these comparisons can be

presented in the form of a van Krevelen plot (atomic

H/C versus O/C) and variants thereof. Such graphical

representations have the advantages that data are

easily compared with the mixing ranges for major

biochemicals and that directional offsets (trajectories)

can be interpreted in terms of specific additions or

removals of known volatile substances from the

measured solid materials (Reuter and Perdue, 1984).

The van Krevelen plot for the five analyzed plank-

ton tow samples (Fig. 4) illustrates several important

patterns. First, the lightly shaded triangle bounded by

the letters L, P and C represents the possible composi-

tional space within which mixtures of the three hypo-

thetical endmembers can exist. Samples that plot

outside this shaded region must contain appreciable

quantities of molecular species with elemental com-

positions different from that of the endmembers and/

or include erroneous elemental data. Fig. 4 indicates

that the elemental compositions directly measured for

our five samples (enclosed within the white rectangle)

are too oxygen-rich. All the measured atomic H/C

ratios are within the range (f 1.6–1.9) expected for

protein/carbohydrate/lipid mixtures. However, all but

one of the corresponding measured O/C ratios exceed

Fig. 3. Ratios of the average measured elemental compositions of

the five plankton tow samples versus the corresponding modeled

compositions derived from the average 13C NMR spectrum in Table

2. Values in substantial excess of 1 indicate elevated measured

concentrations.

J.I. Hedges et al. / Marine Chemistry 78 (2002) 47–63 57

that of pure carbohydrate (f 0.83), the most oxygen-

rich of the major biochemicals, including nucleic

acids (O/Cc 0.7; Anderson, 1995). Second, the ele-

mental compositions calculated for the same five

plankton samples by applying the mixing model to

the measured 13C NMR data all cluster in a narrow

range (enclosed within the shaded rectangle) near the

protein endmember (Fig. 4). Because the protein

contents are more variable versus a relatively constant

carbohydrate/lipid ratio, these compositional points

trend upward and away from the protein endmember.

Finally, in comparison to literature values, the com-

positional point of Anderson (1995) falls nearby the

NMR-based range (Fig. 4). Anderson used published

weight percentages of biochemicals in marine phyto-

plankton (54% protein, 26% carbohydrate, 16% lipid

and 4% nucleic acid) to estimate an average elemental

composition of C106H175O42N16. The major difference

between Anderson’s mean biochemical composition

versus that calculated here is that our samples are on

average somewhat more protein rich and carbohydrate

poor, largely due to the influence of the three nitro-

gen-rich Southern Ocean samples (Table 3). In con-

trast, the elemental compositions reported for RKR

plankton (Fig. 4), average marine phytoplankton (Par-

sons and Takahashi, 1973), plankton (Honjo, 1980),

suspended surface seawater particles (Chen et al.,

1996) and regenerated upper ocean particles (Dyrssen,

1977) are very different (Table 3). In fact, none of

these published elemental compositions lies within the

theoretical mixing range of the major biochemical

endmembers. Thus, the oceanographic literature pres-

ently contains highly variable and often questionable

assessments of the elemental makeup of surface ocean

particles.

Because the elemental composition for Redfield

plankton is so widely employed, it is useful to consider

why the RKR composition point is so far removed

from the biochemical mixing range in a van Krevelen

plot (Fig. 4). One reason for this large offset is that the

RKR formulation for plankton biomass [(CH2O)106(NH3)16(H3PO4)] represents both carbohydrate and

amine structural units in their fully hydrated forms,

without allowing for water loss upon condensation of

individual structural units into polysaccharides and

proteins. In the case of hydrogen, comparison of the

NMR-based elemental composition for net-plankton

(C106H177O37N17) versus the RKR-equivalent formula

(C106H260O106N16) indicates an excess of approxi-

mately 80 H atoms in the RKR formulation. The

average mole percentages of total carbon that occur

in each of the major biochemical types calculated from

the NMR data in Table 2 are 62% protein, 15%

carbohydrate and 23% lipid. Combining this informa-

tion with an average of roughly five carbons in a

‘‘typical’’ amino acid structural unit, six carbons per

individual sugar structural unit, and 18 carbons per

lipid structural unit, the 106 carbons in a RKR-type

formulation can be apportioned approximately into 14

amino acid, three sugar and one lipid structural units.

Given an average of one water molecule loss per non-

lipid structural unit combined into biopolymer, a total

of approximately 17 water molecules (34 hydrogens)

must be lost from the RKR formula by dehydrative

coupling alone. A second large source of excess H in

the RKR formula results from the representation of

nitrogen as (NH3)16. Because every nitrogen in an

amide bond retains only one hydrogen atom, each of

the 16 ammonia units should lose two hydrogens,

equivalent to f 32 hydrogens total. The remaining

excess hydrogens result primarily from surplus H in

carbohydrate (RKR) versus protein structural units.

Fig. 4. Directly measured (unshaded rectangle) and NMR-based

elemental (shaded rectangle) compositions for the five plankton tow

samples in the form of a van Krevelen (1961) plot of atomic H/C

versus O/C ratios. Symbols: E = Equatorial Pacific, A=Arabian Sea,

1 = Southern Ocean sample 1, 2 = Southern Ocean sample 2,

3 = Southern Ocean sample 3. The circled letters (R) and (la)

correspond respectively to the RKR and Anderson compositions in

Table 3. The compositions assigned to the three hypothetical bio-

chemical endmembers described in the text are C for carbohydrate,

L for lipid and P for protein.

J.I. Hedges et al. / Marine Chemistry 78 (2002) 47–6358

The elemental trajectory (slope =H/O = 2) resulting

from the combined removal of approximately 40 water

molecules (80 hydrogens) from RKR plankton (see

Fig. 4) is extensive, but explicable from the NMR-

based biochemical composition.

The dehydration trajectory in Fig. 4, however,

passes through the oxygen-rich portion of the bio-

chemical mixing triangle, and thus corresponds to

much higher carbohydrate compositions than are esti-

mated by NMR. If the NMR-derived composition

range for net plankton is correct, then an additional

downward correction in oxygen content of RKR

plankton is necessary. After removal of the equivalent

of f 40 water molecules, the carbohydrate-based

RKR formula (C106H176O64N16) still contains f 27

more oxygen atoms than the NMR-based formula,

necessitating the removal of about 14 moles of O2.

However, most of this difference can be attributed to

the presence of 21 moles of lipid carbon, which con-

tain approximately 20 fewer moles of O2 (accounting

for the carboxyl) than occur in the equivalent amount

of carbohydrate. The remaining excess RKR oxygen

may result from the lower oxygen content of protein

versus polysaccharide. According to Eqs. (2) and (3),

the calculated amount of oxygen needed to completely

respire the average NMR-based plankton composition

(C106H177O37N17) is 154F 3 moles O2, which cor-

responds to a respiration ratio of 1.45F 0.03 moles

O2/mole C. Corresponding values calculated for the

RKR formulation [Eq. (1)] are 138 moles of dissolved

O2/mole C, and thus a molar respiration ratio of 1.30.

The calculated sulfur content of these samples is so

small that its effect on respiration demand is compara-

ble to the oxygen rounding error (F 1/150 O2). Phos-

phorus and associated oxygen have been excluded

from consideration because phosphate undergoes no

change in redox state during photosynthesis and remi-

neralization and thus do not affect respiration demand.

Fig. 5 illustrates a plot of the carbon-normalized

atomic ratios of total nitrogen versus respiratory O2

demand in the same general format as used for Fig.

4. From this comparison, it is evident that the

measured (N/C = 0.167F 0.012) and modeled (N/C =

0.164F 0.026) nitrogen contents of the five net tow

samples are essentially the same, even though their

mole percentages of protein were calculated from

NMR data with no constraint by measured elemental

compositions. All the N/C results for the literature

sources in Table 3 correspond closely to the values

(measured and modeled) for our net tow samples,

and are within the vertical extent of the biochemical

mixing range. Thus it appears in general that the

carbon and nitrogen contents of marine plankton

samples are being measured consistently by the

oceanographic community. Although the five plank-

ton tow samples we analyzed are on average nitrogen

rich versus the RKR value of N/C = 0.151, they fall

within the range of natural variability reported in the

literature (Table 3). Our NMR-based mixture calcu-

lations therefore are supported (within the range of

natural variability) by N/C ratios that were independ-

ently measured for these same samples.

In contrast, the respiration ratios that were calcu-

lated from direct CHN analysis are disperse and not

confined uniformly to the biochemical mixing zone

(Fig. 5). The measured O2/C ratios for the five plank-

ton tow samples are all below the hypothetical mixing

range because their organic oxygen contents (esti-

mated by difference) are erroneously high (Fig. 3).

RKR plankton exhibit the minimal respiratory oxygen

demand (O2/C = 1.30) possible for a biochemical mix-

ture (Fig. 5) that exhibits a N/C of 0.15 (C/N = 6.6).

This shortfall occurs because all carbon is treated as

occurring in carbohydrate, which is the most oxygen-

Fig. 5. Directly measured and modeled elemental compositions for

the five plankton tow samples in a plot of atomic N/C versus O2/C

ratios. The respiration ratio corresponds to the calculated number of

O2 moles needed to completely respire 1 mole of organic carbon in

plankton tow material. Symbols are as in Fig. 4.

J.I. Hedges et al. / Marine Chemistry 78 (2002) 47–63 59

rich major biochemical. The ascribed overall O2/C for

RKR plankton is 1.30 (versus 1.00) owing to the

included 16 amine groups [(Eq. (1))], which were

assumed to require 1 O2 each for complete respiration

to nitrate (Redfield et al., 1963). Protein exhibits the

highest respiration ratio (1.57) of the major biochem-

icals, with corresponding lipid and carbohydrate values

of 1.42 and 1.00. None of the O2/C ratios calculated

from the NMR data exceed 1.50 (Fig. 5). Their average

respiration ratio (1.45F 0.03) is just slightly higher

than the value (1.42) that Anderson (1995) estimated

from literature values for the major biochemical com-

position of marine phytoplankton (Fig. 5). Martin et al.

(1987) calculated an average respiration ratio of 1.65

for particulate organic carbon recycled in the upper 100

m of the northeast Pacific ocean, but did not correct for

the organic oxygen content of these materials (Ander-

son and Sarmiento, 1994). Most O2/C values derived

from dissolved oxygen and nutrient profiles in the

ocean may be overestimates because recent anthropo-

genic CO2 inputs result in an underestimate of in situ

carbon respiration (Kortzinger et al., 2001). Even after

correction, published O2/C ratios (Table 3) span the

entire biochemical mixing region (1.30–1.70) over the

corresponding likely range of N/C values from 0.15 to

0.20 (see Kortzinger et al., 2001). Clearly, many

published and widely cited estimates for the elemental

compositions and respiratory O2 demands of marine

plankton (and their remains) fail to satisfy the funda-

mental constraint of biogeochemical consistency.

Inconsistencies, however, also can occur in direct

analyses of biochemical compositions. For example,

the biochemical compositions modeled using NMR

data from the Equatorial Pacific plankton tow sample

can be contrasted to those directly measured for the

same sample (Wakeham et al., 1997a,b) by chromato-

graphic analyses. This comparison (Fig. 6) shows that

the mole percentage of amino acid carbon is appreci-

ably higher for this sample set when measured at the

molecular level versus by NMR. More strikingly,

however, only about one-half and one-quarter, respec-

tively, of the lipid (largely fatty acids and sterols) and

carbohydrate (aldoses only) carbon indicated by 13C

NMR are directly measured. Although some of this

analytical discrepancy may derive from forcing all

plankton carbon into only three biochemical categor-

ies, the internal consistency of the spectral (Fig. 2) and

modeled elemental data (Figs. 4 and 5) suggests that

lipid-and carbohydrate-like components of marine

plankton are being incompletely measured. This infer-

ence is consistent with the observation that 18% of the

total carbon in this sample is unaccounted for at the

molecular level (Wakeham et al., 1997a,b). It remains

to be seen whether these shortfalls result from incom-

plete chromatographic measurements of the targeted

hydrolyzable neutral sugars and solvent-extractable

lipids, or from the presence of compositionally similar

materials (e.g. amino sugars versus aldoses; or algae-

nans versus lipids) that require different analytical

procedures (Hedges et al., 2000). Nevertheless, 13C

NMR measurements hold great promise for ground-

truthing analytical procedures for marine materials,

both by identifying incomplete measurements among

solubilized products and for detecting unextracted

residues in the treated solids.

4. Overview

Clearly, CP/MAS 13C NMR analysis of only five

plankton tow samples is insufficient to provide a new

elemental formulation to globally represent the ele-

mental composition of ‘‘average marine plankton.’’ In

addition, these preliminary results should be tested by

other spectroscopic techniques and independent meth-

ods of analysis. The present study, however, demon-

strates that CP/MAS 13C NMR can rapidly and

sensitively provide an overall assessment of the major

biochemical and elemental compositions of marine

Fig. 6. Percentages of total carbon in the Equatorial Pacific plankton

tow sample that can be measured by 13C NMR versus by direct

chromatographic methods (Wakeham et al., 1997a,b).

J.I. Hedges et al. / Marine Chemistry 78 (2002) 47–6360

plankton with minimal sample preparation and loss.

Quantitatively, the 13C NMR spectra and mixing

model results reported here are consistent with pub-

lished biochemical compositions (Anderson, 1995)

and supported by C, N and H compositions measured

directly for the same samples. These results indicate

that direct measurements of the oxygen (by differ-

ence) and sulfur contents of marine plankton are prone

to large errors, as are estimates of respiratory O2

demands based on water column profiles (Kortzinger

et al., 2001). Moreover, only a fraction of the O-alkyl

(f carbohydrate) and polymethylenic (f lipid) car-

bon in marine plankton appears to be measured by

molecular methods in common use today.

These direct 13C NMR measurements support

previous indications (Vollenweider, 1985; Anderson,

1995) that biopolymer compositions constrain the

elemental formulas for marine plankton to values that

are much less hydrogen and oxygen rich than indi-

cated by the RKR formula. Based on the variability

exhibited by our small sample set, formulas in the

range of C106H175–180O35–40N15–20S0.3 –0.5, which

require 150–155 moles of O2 for complete respira-

tion, appear to be more realistic and useful (see also

Anderson, 1995). This natural range in elemental

composition is greater than can be calculated to result

from analytical variability within the 13C CP/MAS

method, although the quantitative effects of the many

associated assumptions are difficult to assess. Overall,

CP/MAS 13C NMR provides an independent and

highly revealing new perspective on the elemental

compositions and reaction stoichiometries of marine

plankton, with the potential for parallel applications to

sinking (Hedges et al., 2001) and sedimentary (Gel-

inas et al., 2001c) marine organic matter.

Acknowledgements

We thank P. Hernes, B. Bergamaschi, J. Murray, J.

Dymond and the captain and crew of the R/V Thomas

G. Thompson and the RVIB Nathaniel B. Palmer for

cruise support. F. Prahl kindly provided analytical

assistance. This manuscript benefited greatly from

detailed comments by Ellery Ingall, an anonymous

reviewer, Anthony Aufdenkampe, Angie Dickens and

the UW Marine Organic Geochemistry (MOG) read-

ing group. This research was supported by grants from

the National Science Foundation to J.H., C.L. and

S.W. and a Canadian Natural Sciences and Engineer-

ing Research Council (NSERC) Post-doctoral Fellow-

ship to Y.G. Development of the data modeling

procedure was a direct result of a fellowship provided

to JAB by the International Scientific Collaborations

Program of the Australian Academy of Science.

References

Anderson, L.A., 1995. On the hydrogen and oxygen content of

marine plankton. Deep-Sea Res. 42, 1675–1680.

Anderson, L.A., Sarmiento, J.L., 1994. Redfield ratios on reminer-

alization determined by nutrient data analysis. Global Biogeo-

chem. Cycles 8, 65–80.

Baldock, J.A., Smernik, R.J., submitted for publication. Chemical

composition and bioavailability of thermally altered Pinus resin-

osa (red pine) wood. Org. Geochem.

Bergamaschi, B.A., Walters, J.S., Hedges, J.I., 1999. Distribution of

uronic acids and O-methyl sugars in sinking and sedimentary

particles in two coastal marine environments. Geochim. Cosmo-

chim. Acta 63, 413–425.

Bishop, J.K.B., Edmond, J.M., Ketten, D.R., Bacon, M.P., Silker,

W.B., 1977. The chemistry, biology, and vertical flux of partic-

ulate matter from the upper 400 m of the equatorial Atlantic

Ocean. Deep-Sea Res. 24, 511–548.

Broecker, W.S., 1974. ‘‘NO’’ a conservative water-mass tracer.

Earth Planet. Sci. Lett. 23, 100–107.

Broecker, W.S., Peng, T.-H., 1982. Tracers in the Sea. Eldigo Press,

New York.

Chau, Y.K., Chuecas, L., Riley, J.P., 1967. The component com-

bined amino acids of some marine phytoplankton species. J.

Mar. Biol. U. K. 47, 543–554.

Chen, C.-T.A., Gong, G.-C., Wang, S.-L., Bychkov, A.S., 1996.

Redfield ratios and regeneration rates of particulate matter in

the Sea of Japan as a model of closed system. Geophys. Res.

Lett. 23, 1785–1788.

Chuecas, L., Riley, J.P., 1969. The component combined amino

acids of some marine diatoms. J. Mar. Biol. U. K. 49, 117–120.

Cowey, C.B., Corner, E.D.S., 1963. On the nutrition and metabo-

lism of zooplankton. J. Mar. Biol. U. K. 43, 495–511.

Cowey, C.B., Corner, E.D.S., 1966. The amino-acid composition of

certain unicellular algae, and the faecal pellets produced by

Calanus finmarchicus when feeding on them. In: Barnes, H.

(Ed.), Some Contemporary Studies In Marine Science. Allen

and Utwin, London, pp. 225–231.

Cowie, G.L., Hedges, J.I., 1984. Determination of neutral sugars in

plankton, sediments, and wood by capillary gas chromatography

of equilibrated isomeric mixtures. Anal. Chem. 56, 497–504.

Cowie, G.L., Hedges, J.I., 1992. Sources and reactivities of amino

acids in a coastal marine environment. Limnol. Oceanogr. 37,

703–724.

De Leeuw, J.W., Largeau, C., 1993. A review of macromolecular

organic compounds that comprise living organisms and their

J.I. Hedges et al. / Marine Chemistry 78 (2002) 47–63 61

role in kerogen, coal and petroleum formation. In: Engel, M.,

Macko, S.A. (Eds.), Organic Geochemistry. Plenum, New York,

pp. 23–72.

Dyrssen, D., 1977. The chemistry of plankton production and de-

composition in seawater. In: Anderson, N.R., Zahuranec, B.J.

(Eds.), Ocean Sound Scattering Prediction. Plenum, New York,

pp. 65–84.

Durand, B., 1980. Kerogen. Edition Technip, Paris.

Eglinton, G., Murphy, M.T.J., 1969. Organic Geochemistry. Spring-

er-Verlag, New York.

Froelich, P.N., 1980. Analysis of organic carbon in marine sedi-

ments. Limnol. Oceanogr. 25, 564–572.

Gelinas, Y., Prentice, K., Baldock, J.A., Hedges, J.I., 2001a. An

improved thermal oxidation method for quantifying soot-phase

black carbon in complex sediment and soil matrices. Environ.

Sci. Technol. 35, 3519–3525.

Gelinas, Y., Baldock, J.A., Hedges, J.I., 2001b. Demineralization of

marine and freshwater sediments for CP/MAS 13C NMR anal-

ysis. Org. Geochem. 32, 677–693.

Gelinas, Y., Baldock, J.A., Hedges, J.I., 2001c. Carbon composition

of immature organic matter from marine sediments: effect of

oxygen exposure on oil generation potential. Science 294,

145–148.

Golchin, A., Clarke, P., Baldock, J.A., Higashi, T., Skjemstad, J.O.,

Oades, J.M., 1997. The effects of vegetation and burning on the

chemical composition of soil organic matter in a volcanic ash

soil as shown by 13C NMR spectroscopy: I. Whole soil and

humic acid fraction. Geoderma 76, 155–174.

Hatcher, P.G., Spiker, E.C., Szeverenyi, N.M., Maciel, G.E., 1983.

Selective preservation and origin of petroleum-forming aquatic

kerogen. Nature 305, 498–501.

Hecky, R.E., Mopper, K., Kilham, P., Degens, E.T., 1973. The

amino acid and sugar composition of diatom cell walls. Mar.

Biol. 19, 323–331.

Hedges, J.I., Stern, J.H., 1984. Carbon and nitrogen determinations

of carbonate-containing solids. Limnol. Oceanogr. 29, 657–

663.

Hedges, J.I., Eglinton, G., Hatcher, P.G., Kirchman, D.L., Arnosti, C.,

Derenne, S., Evershed, R.P., Kogel-Knabner, I., de Leeuw, J.W.,

Littke, R., Michaelis, W., Rullkotter, J., 2000. The molecularly

uncharacterized component of nonliving organic matter in natural

environments. Org. Geochem. 31, 945–958.

Hedges, J.I., Baldock, J.A., Gelinas, Y., Lee, C., Peterson, M.L.,

Wakeham, S.G., 2001. Evidence for non-selective preservation

of organic matter in sinking marine particles. Nature 409, 801–

804.

Hernes, P.J., Hedges, J.I., Peterson, M.L., Wakeham, S.G., Lee, C.,

1996. Neutral carbohydrate geochemistry of particulate material

in the central equatorial Pacific. Deep-Sea Res. II 43, 1181–

1204.

Hernes, P.J., Peterson, M.L., Murray, J.W., Wakeham, S.G., Lee, C.,

Hedges, J.I., 2001. Particulate carbon and nitrogen fluxes and

compositions in the central Equatorial Pacific. Deep-Sea Res. I

48, 1999–2023.

Honjo, S., 1980. Material fluxes and modes of sedimentation in the

mesopelagic and bathypelagic zones. J. Mar. Res. 38, 53–97.

Knicker, H., 2000. Biogenic nitrogen in soils as revealed by solid-

state carbon-13 and nitrogen-15 nuclear magnetic resonance

spectroscopy. J. Environ. Qual. 29, 715–723.

Knicker, H., Scaroni, A.W., Hatcher, P.G., 1996. 13C and 15N NMR

spectroscopic investigation on the formation of fossil algal res-

idues. Org. Geochem. 24, 661–669.

Kortzinger, A., Hedges, J.I., Quay, P.D., 2001. Redfield ratios re-

visited, removing the biasing effect of anthropogenic CO2. Lim-

nol. Oceanogr. 46, 964–970.

Laws, E.A., 1991. Photosynthetic quotients, new production and net

community production in the open ocean. Deep-Sea Res. 38,

143–167.

Lee, C., Cronin, C., 1984. Particulate amino acids in the sea: effects

of primary productivity and biological decomposition. J. Mar.

Res. 42, 1075–1097.

Lee, C., Wakeham, S.G., Hedges, J.I., 2000. Composition and flux

of particulate amino acids and chloropigments in equatorial

Pacific seawater and sediment. Deep-Sea Res. I 47, 1535–

1568.

Lehninger, A.L., 1975. Biochemistry Worth, New York.

Mao, J.-D., Hu, W.-G., Schmidt-Rohr, K., Davies, G., Ghabbour,

E.A., Xing, B., 2000. Quantitative characterization of humic

substances by solid-state carbon-13 nuclear magnetic resonance.

Soil Sci. Soc. Am. J. 64, 873–884.

Martin, J.H., Knauer, G.A., Karl, D.M., Broenkow, W.W., 1987.

VERTEX: carbon cycling in the northeast Pacific. Deep-Sea

Res. 34, 267–285.

Nelson, P.N., Baldock, J.A., submitted for publication. Common

biomolecules can explain the chemical composition of natural

organic materials from diverse sources. Biogeochemistry.

Nelson, P.N., Baldock, J.A., Oades, J.M., Churchman, G.J., 1999.

Dispersed clay and organic matter in soil: their nature and asso-

ciations. Aust. J. Soil Res. 37, 289–315.

Pakulski, J.D., Benner, R., 1992. An improved method for the hy-

drolysis and MBTH analysis of dissolved and particulate carbo-

hydrates in seawater. Mar. Chem. 40, 143–160.

Parsons, T.R., Takahashi, M., 1973. Biological Oceanographic Pro-

cesses Pergamon, Oxford.

Parsons, T.R., Stephens, K., Strickland, J.D.H., 1961. On the chem-

ical composition of eleven species of marine phytoplankters. J.

Fish. Res. Board Can. 18, 1001–1015.

Pella, E., Colombo, B., 1978. Simultaneous C–H–N and S micro-

determination by combustion and gas chromatography. Mikro-

chim. Acta 1, 271–286.

Redfield, A.C., 1934. On the proportions of derivatives in sea water

and their relation to the composition of plankton. James Johnson

Memorial Volume. University of Liverpool, Liverpool, UK, pp.

176–192.

Redfield, A.C., Ketchum, B.H., Richards, F.A., 1963. The influence

of organisms on the composition of seawater. In: Hill, M.N.

(Ed.), The Sea, vol. 2. Interscience, New York, pp. 26–77.

Reuter, J.H., Perdue, E.M., 1984. A chemical structural model of

early diagenesis of sedimentary humus/proto-kerogens. Mitt.

Geol.-Palaeont. Inst. Univ. Hamburg. 56, 249–262.

Richards, F.A., 1965. Anoxic basins and fjords. In: Riley, J.P.,

Skirrow, D. (Eds.), Chemical Oceanography, vol. 1. Academic

Press, New York, pp. 611–645.

Shaffer, G., Bendtsen, J., Ulloa, O., 1999. Fractionation during

J.I. Hedges et al. / Marine Chemistry 78 (2002) 47–6362

remineralization of organic matter in the ocean. Deep-Sea Res. I

46, 185–204.

Skjemstad, J.O., Taylor, J.A., Smernik, R.J., 1999. Estimation of

charcoal (char) in soils. Commun. Soil Sci. Plant Anal. 30,

2283–2298.

Takahashi, T., Broecker, W.S., Langer, S., 1985. Redfield ratios

based on chemical data from isopycnal surfaces. J. Geophys.

Res. 90, 6907–6924.

Van Krevelen, D.W., 1961. Coal. Elsevier, Amsterdam.

Volkman, J.K., Barrett, S.M., Blackburn, S.I., Mansour, M.P.,

Sikes, E.L., Gelin, F., 1998. Microalgal biomarkers, a review

of recent research developments. Org. Geochem. 29, 1163–

1179.

Vollenweider, R.A., 1985. Elemental and biochemical composition

of plankton biomass: some comments and explorations. Arch.

Hydrobiol. 105, 11–29.

Wakeham, S.G., Lee, C., Hedges, J.I., Hernes, P.J., Peterson, M.L.,

1997a. Molecular indicators of diagenetic status in marine or-

ganic matter. Geochim. Cosmochim. Acta 61, 5363–5369.

Wakeham, S.G., Hedges, J.I., Lee, C., Peterson, M.L., Hernes, P.J.,

1997b. Compositions and transport of lipid biomarkers through

the water column and surficial sediments of the Equatorial Pa-

cific Ocean. Deep-Sea Res. II 44, 2131–2162.

Wilson, M.A., 1987. NMR Techniques and Applications in Geo-

chemistry and Soil Chemistry. Pergamon, Oxford.

Zegouagh, Y., Derenne, S., Largeau, C., Bertrand, P., Sicre, M.-A.,

Saliot, A., 1999. Refractory organic matter in sediments from

the North–West African upwelling system: abundance, chem-

ical structure and origin. Org. Geochem. 30, 101–117.

Zelibor, J.L., Romankiw, L., Hatcher, P.G., Colwell, R.R., 1988.

Comparative analysis of the chemical composition of mixed

and pure cultures of green algae and their decomposed residues

by 13C nuclear magnetic resonance spectroscopy. Appl. Envi-

ron. Microbiol. 54, 1051–1060.

J.I. Hedges et al. / Marine Chemistry 78 (2002) 47–63 63