PNAS202209152_proof.pdf - pnas.202209152.pdf

Novel sulfur isotope analyses constrain sulfurized porewater

uxes as a minor component of marine dissolved

organic matter

Alexandra A. Phillips

a,1,2

, Margot E. White

, Michael Seidel

, Fenfang Wu

, Frank F. Pavia

, Preston C. Kemeny

, Audrey C. Ma

Lihini I. Aluwihare

, Thorsten Dittmar

c,e

, and Alex L. Sessions

Edited by Franc ̧ois Morel, Princeton University, Princeton, NJ; received June 19, 2022; accepted September 2, 2022

Marine dissolved organic matter (DOM) is a major reservoir that links global carbon,

nitrogen, and phosphorus. DOM is also important for marine sulfur biogeochemistry

as the largest water column reservoir of organic sulfur. Dissolved organic sulfur (DOS)

can originate from phytoplankton-derived biomolecules in the surface ocean or from

abiotically

“

sulfurized

”

organic matter diffusing from sul

fi

dic sediments. These sources

differ in

S isotope ratios (

δ

S values), with phytoplankton-produced DOS track-

ing marine sulfate (21

‰

) and sulfurized DOS mirroring sedimentary porewater sul

fi

(

∼

0to

–

‰

). We measured the

δ

S values of solid-phase extracted (SPE) DOM

from marine water columns and porewater from sul

fi

dic sediments. Marine DOM

SPE

δ

S values ranged from 14.9

‰

to 19.9

‰

and C:S ratios from 153 to 303, with lower

δ

S values corresponding to higher C:S ratios. Marine DOM

SPE

samples showed con-

sistent trends with depth:

δ

S values decreased, C:S ratios increased, and

δ

C values

were constant. Porewater DOM

SPE

was

S-depleted (

∼

-0.6

‰

) and sulfur-rich (C:S

∼

37) compared with water column samples. We interpret these trends as re

fl

ecting at

most 20% (and on average

∼

8%) contribution of abiotic sulfurized sources to marine

DOS

SPE

and conclude that sulfurized porewater is not a main component of oceanic

DOS and DOM. We hypothesize that heterogeneity in

δ

S values and C:S ratios

fl

ects the combination of sulfurized porewater inputs and preferential microbial scav-

enging of sulfur relative to carbon without isotope fractionation. Our

fi

ndings

strengthen links between oceanic sulfur and carbon cycling, supporting a realization

that organic sulfur, not just sulfate, is important to marine biogeochemistry.

dissolved organic matter

dissolved organic sulfur

stable isotopes

marine sulfur cycle

sulfurization

Dissolved organic matter (DOM) is the largest inventory of

fi

xed carbon in the ocean

(

∼

662 Pg carbon) and is a critical component of marine food webs and nutrient

cycling (1). DOM has been hypothesized to impact climate over geological time scales

via carbon sequestration (1), as a highly recalcitrant subset of DOM persists for many

thousands of years, surviving multiple mixing cycles of the ocean for poorly understood

reasons (2

–

4). Therefore, despite decades of intensive study, fundamental questions

remain regarding DOM sources and sinks in the modern ocean.

Ksionzek et al. demonstrated that solid-phase extracted DOM (DOM

SPE

) contains a

substantial quantity of organic sulfur (5). The marine dissolved organic sulfur (DOS)

pool is estimated to contain

∼

7 Pg sulfur, more than 10 times that in phytoplankton,

bacteria, and particulate organic sulfur (POS) combined (6). Although estimates of DOS

concentration vary (5, 7), it is nevertheless clear that DOS is central to marine sulfur

cycling (6), with growing evidence for important links to carbon cycling. For example,

reduced sulfur compounds within DOS lower trace metal availability by tightly binding

free zinc and copper and potentially limiting primary production (8). Concentrations of

alkyl thiols, such as the amino acid cysteine, have been correlated to chlorophyll concen-

trations, implying further connections to phytoplankton growth (9, 10). DOS metabo-

lites may also limit the growth of some marine heterotrophs, as clades of the ubiquitous

SAR11 and SAR86 bacteria are unable to assimilate sulfate and must rely on scavenging

reduced OS from the water column (11, 12). It remains unknown whether this subset of

S-containing molecules (i.e., DOS) within DOM behaves similarly to the larger dissolved

organic carbon (DOC) pool or has unique origins and/or dynamics.

fi

rst-order question regarding DOS dynamics is the origin of the organic com-

pounds (Fig. 1). Ksionzek et al. proposed a DOS cycle that mirrors DOC, with DOS

produced mainly by phytoplankton in the sunlit ocean. Under this hypothesis, microbial

reworking during the aging of DOS, such as heterotrophic uptake for growth or remi-

neralization back to sulfate, leaves remaining DOS compounds increasingly recalcitrant.

Signi

cance

Marine dissolved organic matter

(DOM) is a vast reservoir of

enigmatically old carbon with

important connections to the

global carbon cycle and climate.

Previous work hypothesized that

highly recalcitrant,

“

sulfurized

”

organic matter from sul

fi

dic

sediments could potentially

account for this long-lived carbon.

To quantify the contribution of

this source, we made sulfur

isotope measurements of marine

DOM, measuring samples across

the Paci

fi

c and Atlantic Oceans as

well as from the porewaters of

sul

fi

dic sediments. We found

values and C:S ratios that were

consistent with

∼

8% contribution

of this abiotic source and conclude

that these sulfurized porewater

fl

uxes are not a main component

of oceanic DOM budgets.

Author contributions: A.A.P. and A.L.S. designed research;

A.A.P., M.E.W., M.S., F.F.P., P.C.K., L.I.A., and T.D.

performed research; A.A.P., F.W., and A.L.S. contributed

new reagents/analytic tools; A.A.P., F.W., A.C.M., and A.L.S.

analyzed data; and A.A.P., M.E.W., M.S., L.I.A., T.D., and

A.L.S. wrote the paper.

The authors declare no competing interest.

This article is a PNAS Direct Submission.

This article is distributed under

Creative Commons

Attribution-NonCommercial-NoDerivatives License 4.0

(CC BY-NC-ND)

Present address: Earth Sciences Department, University

of California Santa Barbara, Santa Barbara, CA 93016.

To whom correspondence may be addressed. Email:

phillips.alexandra.a@gmail.com.

This article contains supporting information online at

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.

2209152119/-/DCSupplemental

Published October 6, 2022.

PNAS

2022 Vol. 119 No. 41 e2209152119

https://doi.org/10.1073/pnas.2209152119

1of8

RESEARCH ARTICLE

EARTH, ATMOSPHERIC, AND PLANETARY SCIENCES

DOS may therefore offer a unique lens to DOM cycling more

broadly (5). A second hypothesis by Pohlabeln et al. posits that a

signi

fi

cant portion of DOS originates in euxinic sediments (13).

Here, sul

fi

de produced via microbial sulfate reduction reacts abi-

otically with organic molecules to

“

sulfurize

”

organic matter

(14

–

16). The resulting covalent C-S bonds are thought to be

resistant to microbial degradation as a result of stable S-S cross-

linking and replace otherwise more labile functional groups

(17

–

19). As such, sulfurization reactions have been recognized as

an important pathway for sedimentary organic matter preserva-

tion (20

–

22) and implicated in the carbon cycles of the ancient

ocean (e.g., during ocean anoxic events (20, 23)). In modern

euxinic sediments, these reactions are a major process: in a study

of sedimentary porewater from the Santa Barbara Basin, 70% of

detected DOS

SPE

formulas were products of sul

fi

de and polysul-

fi

de sulfurization reactions (24). Pohlabeln et al. calculated the

fl

ux of these highly recalcitrant DOS species as a signi

fi

cant

source to benthic DOS

SPE

and by extension to water column

DOM

SPE

(13). While the proportion of DOM molecules that

contain sulfur is not known precisely, estimates from molecular

formulas of SPE DOM imply it could be

∼

–

10% (7). Such

numbers suggest that porewater-sourced, sulfurized molecules

could represent an overlooked subcycle within DOM and might

also help to explain why some components of DOM

SPE

have

surprisingly long lifetimes, clarifying connections between DOC

and DOS cycles in the ocean (25).

The sulfur isotopic composition of DOS

SPE

can distinguish

between organic matter produced by phytoplankton or by pore-

water reactions because these sources differ in

S values by

∼

‰

(Fig. 1). This signal is easily resolved given typical mea-

surement errors on organic sulfur of

0.2

‰

(26). Although

carbon and nitrogen isotopic measurements have previously

proven useful for studying marine DOM

SPE

(e.g., (27)), to our

knowledge, no such measurements have been made for sulfur,

likely for two reasons. First, dissolved inorganic sulfate, at 28

mM, is 4

–

5 orders of magnitude higher in concentration than

DOS

SPE

(

∼

100 nM). To isolate organic sulfur compounds

from this high salt background, studies rely on SPE on a

styrene-divinylbenzene stationary phase (Bond Elute PPL (28))

that is considered to impart negligible isotope fractionations for

Fig. 1.

A simpli

fi

ed marine organic sulfur cycle highlighting processes that may impact the sulfur isotope composition and sulfur content of dissolved

organic sulfur (DOS). Numbers underneath each reservoir give the approximate

S value and molar C:S ratio, based on initial studies of organic sulfur in

phytoplankton (39, 40), water column DOS (5), and particulate organic sulfur (POS) (42

–

44), porewater DOS (13), and porewater sul

fi

des (54). Results from

this study are highlighted in pink. Through lysis, exudation, and senescenc

e, phytoplankton biomass enters the DOS pool. DOS may also originate from

sedi-

ments, where sul

fi

de is incorporated into sedimentary organic matter to form sulfurized DOS. The sulfur isotopic composition of DOS

SPE

indicates the rela-

tive balance of these two sources.

2of8

https://doi.org/10.1073/pnas.2209152119

pnas.org

organic compounds (29

–

31) and should be negligible for sulfur

atoms that do not participate in hydrophobic binding interac-

tions. For example, hydrophobic separation of volatile organic

sulfur compounds by gas chromatography does not yield frac-

tionation of sulfur isotopes across peaks (32). PPL resins isolate

DOM with moderate yields for DOC (

∼

60%), low (

∼

21%)

yields for DON, and unknown yields for DOS (33). However,

given that abiotically sulfurized organic matter typically has sulf-

hydryl groups and hydrophobic hydrocarbon regions (24), and

given the associated mechanism of retention on SPE columns,

PPL should be well suited for retention of sulfurized DOM.

Indeed, in a study of dozens of model organic sulfur com-

pounds, extraction ef

fi

ciency was highest for uncharged, slightly

polar, medium-sized analytes (34). Furthermore, both experimen-

tally sulfurized organic matter (13) and sulfurized organic matter

from euxinic sediments (24) have been shown to be retained on

PPL resins for molecular characterization. We therefore expect

SratiosandC:SratiosofDOS

SPE

to preferentially retain sulfu-

rized components of bulk DOM. Here, we show that SPE also

reduces sulfate contamination to nondetectable levels, allowing

accurate isotopic measurements of DOS

SPE

. The second problem

is that conventional isotope ratio mass spectrometry (IRMS)

methods have required much more sulfur (

∼

g) than carbon

(

∼

g). Given that DOM

SPE

has high C:S ratios (

∼

150

–

300),

traditional techniques would require extraction from hundreds

of liters of seawater per measurement. Recent developments in

combustion elemental analysis IRMS (EA-IRMS) have enabled

much lower sample sizes (1-10

g sulfur (26)) for organic sulfur

compounds. Here, we used the SPE method and adapted our

EA-IRMS measurements to allow simultaneous determination of

Cand

S values, and C:S molar ratios on samples of

∼

350

gDOM

SPE

. We applied this approach to extracts from the

Paci

fi

c and Atlantic coastal zones and ocean basins, generating

pro

fi

les of

SvaluesofmarineDOM

SPE

. Our data indicate

that porewater sulfurization reactions contribute minimally to

the marine DOS

SPE

pool, and by extension, DOS and DOM.

Results

A global sample suite of 90 marine,

fi

ve estuarine, and

fi

ve pore-

water DOM

SPE

samples were analyzed for

Cand

Svalues,

and C:S molar ratios (Fig. 2

and

SI Appendix

,TableS1

). Sam-

ples spanned gyres, shelves, restricted basins, oxygen minimum

zones (OMZs), and coastal oceans.

Porewater and Estuary Samples.

Porewater DOM

SPE

was ana-

lyzed from sul

fi

dic sediments in the back barrier tidal

fl

ats of

the North Sea, yielding a

S value of

–

0.2

‰

, C:S molar ratio

of 18, and

C value of

–

23.3

‰

. Meanwhile, porewater

DOM

SPE

samples from sul

fi

dic sediments in the mangrove

tidal creek within the Caet

e Estuary ranged in

S values from

–

2.7 to 0.7

‰

, with C:S molar ratios between 40 and 45 (Fig.

) and

C values between

–

27.1 and

–

26.8

‰

. DOM

SPE

surface water samples from the Caet

e Estuary were analyzed

across a transect from the coastal ocean to the mangrove-

fringed estuary. C:S ratios ranged from 74 to 117, with

values between 4.2 and 8.7

‰

C values spanned

–

28.6 to

–

24.2

‰

across the transect.

Fig. 2.

(

) Sample locations spanned gyres (N Atlantic: pink triangles, N Paci

fi

c: green triangles, S Paci

fi

c: dark blue triangles), shelves (NE Paci

fi

c Shelf:

yellow circles), mangrove-fringed estuaries (Caet

e Estuary: light blue left-facing triangles), restricted hypoxic basins (San Pedro Basin: turquoise diamonds),

oxygen minimum zones (NE Paci

fi

c OMZ: magenta squares), coastal settings (NE Paci

fi

c: light pink right-facing triangles), and sul

fi

dic porewater (North Sea:

dark gray down-facing triangles). (

)

S values against C:S ratios of all analyzed DOM

SPE

samples (

100). (C) Expanded view of marine DOM

SPE

samples,

showing a negative correlation between DOM

SPE

S values and C:S ratios (

0.24,

0.001). 1

SEs are shown for all samples in (

) and (C), but in

many cases, are within the size of the symbol. VCDT, Vienna Canyon Diablo Troilite.

PNAS

2022 Vol. 119 No. 41 e2209152119

https://doi.org/10.1073/pnas.2209152119

3of8

Marine Samples.

Marine DOM

SPE

ranged in

S values from

15 to 20

‰

and in molar C:S from 150 to 300. We recognized

three distinct behaviors over different depth intervals for

DOM

SPE

measurements, across all sampling locations (Fig. 3 and

Table 1). For the purpose of calculating quantitative statistics, we

chose typical cutoffs for epipelagic (150 m) and bathypelagic

(750 m) zones, while recognizing that such sharp boundaries are

somewhat arbitrary and may vary throughout the world

’

s oceans.

In the epipelagic shallow samples between 0 and 150 m , C:S

ratios of DOM

SPE

increased (

0.23;

0.005), resulting in

DOC (

0.48;

0.001) and apparent (see

Materials and

Methods

)DOS

SPE

concentrations (

0.39;

0.005) that

decreased sharply with depth, and

Sand

Cvaluesdidnot

change signi

fi

cantly. Average

SvaluesandC:Sratioswere

18.6

0.8

‰

and 213

32, respectively, in this interval. At

mesopelagic depths, between 150 and 750 m, the C:S ratios did

not change. Because DOC decreased gradually, apparent DOS

SPE

also decreased gradually.

S values decreased signi

fi

cantly (

0.32;

0.005), while

C values did not change. Average

S values and C:S ratios were 17.8

0.8

‰

and 236

24,

respectively. In the bathypelagic samples, between 750 and

5,000 m, none of the parameters showed statistically signi

fi

cant

Table 1. Properties of marine DOM

SPE

samples binned by depth

Depth range (m)

DOM

SPE

C:S (molar)

DOM

SPE

‰

)

DOM

SPE

‰

)

DOC (

DOS (nM)

–

150

213

18.6

0.8

22.4

0.3

283

150

–

750

236

17.8

0.8

22.2

0.3

192

750

–

5,000

252

17.1

1.1

22.2

0.3

163

Averages are reported with 1

SDs.

Fig. 3.

DOM

SPE

properties plotted against depth for (

) C:S ratios, (

)

S values (

)

C values, (

) DOC concentration, and (

) calculated DOS concentra-

tion. Note the scale breaks at 150 m, 750 m on the

axis. Colors and symbols for stations are the same as in Fig. 2; signi

fi

cant correlations are marked in

gray. 1

SEs are shown for

S values, but error bars for C:S,

C values, DOC, and DOS are within the size of the symbol. DOM

SPE

C:S ratios slightly

increased with depth in the

fi

rst 150 m, but did not change signi

fi

cantly between 150 and 5,000 m.

S values averaged 18.6

‰

in the

fi

rst 150 m and

decreased between 150 and 750 m to an average of 17.1

‰

. DOM

SPE

C values did not change systematically with depth. Both DOC and DOS concentra-

tions decreased within the

fi

rst 150 m, but did not vary further with depth. For station-speci

fi

c plots, see

SI Appendix

(

SI Appendix

, Figs. S8, S9, S11

–

S14

4of8

https://doi.org/10.1073/pnas.2209152119

pnas.org

correlations with depth. Average

SvaluesandC:Sratiosinthe

deepest samples were 17.1

1.1

‰

and 252

27, respectively.

DOM

SPE

C values were constant throughout the water col-

umn, with an average of

–

22.3

0.3

‰

.DOM

SPE

S values

also had statistically signi

fi

cant positive correlations with DOC

concentration (

0.25;

0.001) and, therefore, apparent

DOS

SPE

concentration (

0.38;

0.001) and tempera-

ture (

0.28;

0.001), although all these parameters

covaried with depth (

SI Appendix

,Fig.S7

Discussion

To assess the possible contribution of sulfurized porewater

DOS sources to the open ocean using a two-parameter (

C:S), two-endmember mixing calculation, we

fi

rst constrained

1) the composition of biotic DOS

SPE

in the surface ocean,

which we de

fi

ned as the upper 50 m of the water column, and

2) the composition of abiotic DOS

SPE

in sul

fi

dic porewaters.

The results of 1) and 2) allow us to conclude that 3) there is

limited isotopic evidence for porewater sources to global marine

DOS

SPE

and more broadly, to DOS and DOM. Finally, we

discuss 4) potential mechanisms for observed sulfur isotope het-

erogeneity with depth.

Composition of Biotic DOS in the Surface Ocean.

Primary pro-

ducers in the photic zone invest energy to transform inorganic sul-

fate into biomass through assimilatory sulfate reduction (6). Very

few studies have examined the consequences of these reactions for

the sulfur isotope composition of organic matter in phytoplank-

ton; preliminary measurements of marine algae found that bulk

biomass was minimally (by 0.8

‰

)

S-depleted from marine sul-

fate (35), which is a nearly constant 21

‰

(36). Further work on

phytoplankton dimethylsulfoniopropionate and dimethyl sul

fi

(DMS) found

S-depletions from marine sulfate by

∼

–

‰

(37

–

39). Our dataset of surface DOM

SPE

50 m depth)

values averaged 18.6

0.6

‰

, which aligns with these studies

and supports the idea that phytoplankton-derived organic sulfur

in the surface oceans is only slightly fractionated (by up to

∼

‰

) relative to marine sulfate.

There are few studies that have investigated C:S ratios of

organic matter from primary producers, but most converge on

similar amounts of organic sulfur as phosphorous (i.e., S:P

∼

1).

Laboratory experiments found cultures of

Synechococcus

incor-

porating carbon, nitrogen, and sulfur at a ratio of 95:16:1 (40).

A study of 15 marine eukaryotic phytoplankton found average

stoichiometries for C:N:P:S of 124:16:1:1.3 (41). Meanwhile,

suspended marine POS, which is assumed to derive from phy-

toplankton, has C:N:S ratios of

∼

110

–

187:27:1 in the North

Paci

fi

c (42

–

44). Euphotic zone DOM

SPE

(

50 m depth) C:S

ratios in this study were similar, averaging 192

25 and rang-

ing between 153

–

243. This aligns with previous measurements

of DOM

SPE

C:S ratios, which ranged between 188 and 290 in

the upper 100 m of the Eastern Atlantic and Southern Ocean

(5). Higher C:S ratios in DOM/POM versus phytoplankton

suggest substantial losses of relatively sulfur-rich compounds.

This presumably includes volatile species like DMS (C:S

2),

which accounts for 40% of the atmospheric sulfur

fl

ux from

the ocean (45). In summary, we take our phytoplankton-

derived surface DOS

SPE

endmember to have a

S value of

18.6

0.6

‰

and a C:S ratio between 153 and 243.

Composition of Abiotic DOS in Sulfidic Porewater.

Porewater

sul

fi

de and polysul

fi

des react abiotically with organic functional

groups to form new C-S bonds in a process called sulfurization

(16). The sul

fi

de derives from dissimilatory sulfate reduction

(DSR), which, unlike the assimilatory pathway, strongly frac-

tionates sulfur isotopes and results in porewater sul

fi

values as low as

–

‰

(46

–

48). DOS compounds formed by

this pathway are thus also expected to have very low

S val-

ues. We analyzed porewater samples from tidal

fl

ats with high

rates of sulfate reduction and organic matter sulfurization: the

mangrove-fringed Caet

e Estuary and the North Sea (49

–

51).

Because the magnitude of sulfur isotope fractionation by DSR

is generally inversely correlated with the sulfate reduction rate

(52, 53), these samples should provide an upper bound on

porewater sul

fi

de and thus sulfurized DOS

SPE

S values.

Our data provide direct constraints on porewater DOS

SPE

sulfur isotope compositions and indicate that sulfurization is a

major process in these sediments. The

S values of DOM

SPE

in porewaters ranged from

–

2.7 to 0.7

‰

S-depleted by

almost 20

‰

relative to phytoplankton organic sulfur, but rela-

tively

S-enriched compared with most porewater H

S meas-

urements (46, 54). DOM

SPE

C:S ratios ranged between 18 and

45, with much higher sulfur contents than phytoplankton-

derived DOS in the surface ocean. These values align with both

laboratory studies that incubated DOM

SPE

in sul

fi

dic water,

fi

nding post-sulfurization C:S ratios

∼

15 and previous measure-

ments of North Sea porewaters with a C:S ratio of

∼

27 (13).

Porewater DOM

SPE

C values differed in the Caet

e Estuary

(

–

26.9

0.1

‰

) and the North Sea (

–

23.3

0.2

‰

), likely

fl

ecting higher terrestrial in

fl

uences in the mangrove-fringed

estuary. Notably, while terrestrial OS is poorly constrained for

S ratios, C:S values are generally higher (55); therefore,

while we see in

fl

uences of terrestrial carbon, no such inputs of

terrestrial sulfur are observed. We therefore assume a sul

fi

dic

porewater-derived sedimentary endmember to have a

S value

–

2.7 to 0.7

‰

with C:S ratios between 18 and 45.

Limited Evidence for Porewater Sources to DOS.

Mixing calcu-

lations using the

S and C:S ratios for the two-endmember

sources described above and the measured

S and C:S ratios

of DOM

SPE

indicate that sulfurized porewater sources cannot

contribute more than

∼

20% of the DOS

SPE

in any one sample

from our open-marine dataset, and no more than

∼

8% on

average for the deep-ocean samples (Fig. 4). Given that other

processes could also lead to

S depletion (see below) and that

our porewater samples are, if anything, more

S-enriched than

sulfurized porewater DOS

SPE

from typical marine sediments

(which have more

S-depleted sul

fi

des), this calculation repre-

sents a very conservative maximum estimate (that is, any value

20% is plausible, including 0%). Only samples from the

mangrove-fringed Caet

e Estuary were within a range of signi

fi

cant contributions (50

–

80%) from sulfurized organic matter;

however, this contribution appears short-lived, likely due to

rapid (i.e., weeks to months) photochemical oxidation of and

removal from the DOM

SPE

pool in the estuary before reaching

the open ocean (56). Without more data to examine differences

between ocean basins, we are unable to address the long-term

accumulation of sulfurized porewater DOS

SPE

, which is

assumed to persist in the ocean over multiple mixing cycles.

However, the few deep Atlantic data that we do have do not

support this hypothesis, as samples from the Bermuda Atlantic

Time-Series (BATS) are more

S-depleted (Figs. 2 and 3; pink

triangles) than those from the (older) deep Paci

fi

c Ocean.

Thus, we can conclude that marine DOS has a dominantly

(

92%) biotic origin, that is, presumably produced by micro-

organisms in the sunlit surface ocean.

Sulfurization reactions have also been documented to occur

within anoxic microenvironments in the pelagic ocean, such as

PNAS

2022 Vol. 119 No. 41 e2209152119

https://doi.org/10.1073/pnas.2209152119

5of8

sinking particles in OMZs, where genomic studies found active

transcription of genes for sulfate reduction (57, 58), and incuba-

tion studies with radioactively labeled sulfur con

fi

rmed that partic-

ulate organic matter sulfurization was occurring (59). It remains

uncertain if this process translates from the particulate to the dis-

solved sulfur pool. Our data from the NE Paci

fi

cOMZ(

31;

magenta squares, Figs. 2

–

4) were collected from adjacent stations

to Raven et al. (59) and spanned dissolved O

concentrations

from 0 to 250

M. Yet, we observed no signi

fi

cant changes to

either C:S ratios or

S values with dissolved oxygen (

Appendix

,Fig.S10

). This suggests that sulfurization reactions in

that anoxic water column minimally or negligibly contribute to

marine DOS; globally signi

fi

cant contributions from other, more

sul

fi

dic water bodies, seem unlikely (60).

The conclusion that porewater

fl

uxes of sulfurized organic

matter do not greatly impact the present marine global DOS

inventory was somewhat unexpected, given previous estimates of

large benthic

fl

uxes (30-200 Tg sulfur yr

(13)), which are cal-

culated by dividing DOC

fl

uxes by porewater DOM

SPE

C:S

ratios. There are several potential explanations for this discrep-

ancy, including incorrect estimates of 1) DOM

SPE

C:S ratios,

and/or 2) porewater DOC

fl

uxes. Existing measurements of

porewater DOM

SPE

C:S ratios are highly biased toward pore-

water samples and environments with known sulfurization

—

global average porewater DOM

SPE

C:S ratios are much higher,

then calculated DOS

fl

uxes would decrease. Notably, DOC

fl

uxes are highest in areas with high rates of sulfate reduction

(61), i.e., continental margin sediments (121-233 Tg carbon

(62, 63)) and intertidal sediments (106-416 Tg carbon

, ref (64). Given that

∼

70% of porewater DOC

fl

ux is

estimated to occur from sediments that have active sulfate reduc-

tion, the bias in porewater DOM

SPE

C:S ratios is likely only

minor. Alternatively, calculated DOC

fl

uxes might be overesti-

mated. Our dataset offers a window to sedimentary DOS/DOC

fl

uxes for comparison: Given a conservative maximum of 20%

(Fig. 4) of the

6,700 Tg sulfur DOS inventory that was

initially estimated to derive from porewater

fl

uxes, and an aver-

age lifetime for DOS of

∼

4,000 y (5) assuming that DOS com-

pounds share the same average radiocarbon age as bulk DOC,

the sedimentary

fl

ux of DOS to the global ocean is

0.4 Tg sul-

fur yr

.PorewaterC:Sratios(

∼

–

50) further imply DOC

fl

uxes of 1.6

–

7.6 Tg carbon yr

, orders of magnitude below

those reported from either intertidal or continental margin sedi-

ments. Although it is possible to invoke a DOC

fl

ux with little

concurrent DOS

fl

ux, such a scenario would require C:S ratios

2,500 that are dif

fi

cult to imagine from areas with active sulfur

cycling (and notably, much higher than the highest recorded

value of 1,472 (7)). Either sulfurized DOS suffers extreme pref-

erential loss over DOC or porewaters are orders of magnitude

less signi

fi

cant sources of marine DOM than previously thought.

Finally, if the lifetime of porewater-derived DOS in the water

column is signi

fi

cantly shorter than the average radiocarbon age

of DOC used above (

∼

4,000 y), then sedimentary porewater

fl

uxes could be higher, but this would mean that porewater

DOS does not contribute to the refractory DOM reservoir.

Causes of Spatial

δ

S Heterogeneity in DOM.

We next turn to

the observed decrease in

S and increase in C:S ratios of

DOM

SPE

with depth (Figs. 2

and 3). In shallow (0-150 m

depth) waters, the loss of apparent DOS

SPE

and DOC, increase

in C:S ratios, and near-constant

S and

C values with

depth can be readily explained by the rapid uptake and/or

remineralization of phytoplankton-derived labile DOS com-

pounds without fractionation. The preferential loss of sulfur

relative to carbon is similar to that observed for nitrogen and

phosphorous over the same depth by other studies (1, 65) and

is unsurprising given that a majority of organic sulfur is present

as diverse bioavailable components like thiols, sulfonates, or thio-

phenes (6, 7). The minimal isotopic fractionation also agrees

with studies of both speci

fi

c degradation reactions (38) and bulk

trophic level effects (66, 67). In contrast, DOS

SPE

in the deep

ocean is lower in concentration (163

19 versus 283

66 nM),

has higher C:S ratio (252

27 versus 213

32), and lower

value (17.1

1.1

‰

versus 18.6

0.8

‰

; Table 1). Changes in

DOS

SPE

concentration and C:S ratios can be explained as the

continuous degradation of semilabile and semirefractory com-

pounds. The lower

S values of deep water DOM

SPE

cannot,

however, be easily explained by a simple degradation model (Fig.

4) because the shift is in the opposite direction of that expected

from a normal isotope effect.

Strong intramolecular heterogeneity in

S of DOS

SPE

compounds could potentially yield the observed isotopic pat-

terns if the DOS

SPE

that is removed with depth is signi

fi

cantly

S-enriched, relative to the remaining DOS

SPE

.Forexample,

modeling studies and initial data hint that semilabile, reduced

organic sulfur compounds like amino acids are more

S-enriched

relative to more recalcitrant, oxidized organic sulfur (26, 55).

However, at present there is little direct compound-speci

fi

csulfur

isotope data to either support or reject this hypothesis. Generation

of DOS compounds via the hydrolysis of sinking particles could

leave POS

S-enriched and DOS

SPE

S-depleted, but such a sce-

nario would require large isotope fractionations or active turnover

of a large proportion of the deep DOS pool. Finally, addition of

terrestrial DOS (generally high in C:S and lower in

S) to the

deep ocean could explain lower

Swithdepthbutisinconsis-

tent with our nearly invariant

Cvalues(Fig.3

Alternatively, deep-ocean DOS

SPE

could simply be

S-depleted

due to mixing with porewater sulfurized compounds, as discussed

above. While this should add S-rich compounds that lower the

Fig. 4.

DOM

SPE

samples plotted in the same coordinate space and color

scheme as Fig. 1

, with C:S molar ratios on the

axis and

S values on

the

axis. A linear mixing model is superimposed in the white- to blue-

colored polygon, with dashed lines at 10% intervals. Endmembers for the

mixing space were inferred from marine surface (

50 m) extracts and pore-

water samples and are detailed in the discussion. Marine samples have, on

average,

∼

8% DOS contributions from porewater sources and at maximum

∼

20% (two samples). We hypothesize that heterogeneity in

Svaluesand

C:S ratios is a combination of both mixing from porewater sources that low-

ers

S values, and organic sulfur removal without fractionation that

increases C:S ratios (gray arrows). VCDT, Vienna Canyon Diablo Troilite.

6of8

https://doi.org/10.1073/pnas.2209152119

pnas.org

C:S ratios, the large

S contrast between surface and porewater

DOS

SPE

means that the isotopic composition is much more sensi-

tive to this addition than the elemental ratio (Fig. 4). If this mech-

anistic explanation is correct, the trends of decreasing apparent

DOS, increasing C:S ratios, and decreasing

Svaluesobserved

in intermediate-depth waters can be understood as the superposi-

tion of two processes (Fig. 4, gray arrows). Thus, we hypothesize

that C:S ratios of DOM

SPE

increase with depth (up to

∼

750 m)

due to mixing with older water masses and concurrent slow degra-

dation, but without appreciable isotopic fractionation. Meanwhile,

the

S of deep DOS

SPE

is lower due to the addition of small

amounts of

S-depleted sulfurized DOM

SPE

from porewaters and

possibly other sources.

Conclusions.

We developed an improved (

∼

10x more sensitive

for sulfur) analytical measurement for marine DOM

SPE

enabling concurrent measurements of

Svalues,

Cvalues,

and C:S ratios on

∼

350

gDOM

SPE

. We used this technique

to produce a global survey of marine DOM

SPE

S values

and show that abiotically sulfurized organic matter from sul

fi

dic porewater is on average

8% of the marine DOS

SPE

pool.

As this is a conservative upper bound, we conclude that sulfu-

rized porewater DOS is only a small component of the oceanic

DOM inventory. The accumulation of refractory, sulfurized

porewater is therefore not a main process that could explain

the old radiocarbon ages of oceanic DOM

SPE

.Instead,DOS

apparently derives mostly from biological assimilation of sul-

fate in the sunlit surface ocean. Trends of increasing C:S ratios

and decreasing

S values with depth could re

fl

ect the contin-

uous removal of surface-derived DOM

SPE

superimposed on a

background of this small inventory of sulfurized DOS

SPE

Materials and Methods

Sample Collection.

DOM

SPE

samples were collected between 2016 and 2021

on 10 cruises (see

SI Appendix

for details), including the BATS and the Hawaii

Ocean Time-Series, that covered the following regions: N Paci

fi

c Gyre, N Atlantic

Gyre, S Paci

fi

c Gyre, NE Paci

fi

c, NE Paci

fi

cShelf,NEPaci

fi

c OMZ, San Pedro Basin

(coastal California), and the Caet

e Estuary (Amazonian mangroves). Porewater

DOM

SPE

samples were collected from the intertidal sediments of the mangrove-

fringed Caet

e Estuary, south of the Amazon Estuary in North Brazil, and a North

Sea intertidal

fl

at in Germany. For marine samples, conductivity, temperature,

and density casts for physical parameters (i.e., dissolved oxygen, salinity,

fl

uores-

cence, temperature) were taken at each station and can be found in the

Appendix

(

SI Appendix

,Figs.S1

–

). Seawater (

∼

–

20 L samples) was collected

from Niskin bottles into acid-washed polyethylene containers and

fi

ltered

through a 0.80/0.45

mcapsule

fi

lter (AcroPak 500) prior to acidi

fi

cation to pH

2 with reagent grade 12 N hydrochloric acid. DOC concentrations were measured

prior to SPE via high-temperature combustion on a Shimadzu Total Organic Car-

bon Analyzer (Shimadzu Corp). Nutrients (nitrate, phosphate, and silicate) were

measured via colorimetry on an AutoAnalyzer II (Seal Analytical; see

SI Appendix

for station-speci

fi

c methods details and references).

DOM

SPE

Isolation.

DOM

SPE

was concentrated using SPE with Bond Elute PPL

cartridges (Agilent; 1 g, 6 mL size) following Dittmar et al. (28). DOM

SPE

samples

were eluted in GC-grade methanol (28). Extracts were dried under a stream

of N

gas and transferred to 2 mL GC vials and then to 150

L glass inserts.

Aliquots corresponding to

∼

4.5

gsulfur(

∼

350

g DOC) were transferred in

methanol from 2 mL GC vials into smooth-walled tin EA capsules (6

2.9 mm,

OEA Laboratories). Methanol was evaporated at room temperature (

∼

2h)prior

to folding.

EA-IRMS Measurements.

Tin capsules containing dried DOM

SPE

were folded

closed, loaded into an autosampler, and then combusted and analyzed in a

Thermo Scienti

fi

c EA IsoLink IRMS System for determination of C:S molar ratios,

S values, and

C values. The system comprised a Flash combustion EA

coupled to a Delta V Plus IRMS via a ConFlo IV Universal Interface. Carbon and

sulfur isotope analysis and data processing followed a previously published

method (26). Due to the high C:S molar ratios of DOM

SPE

,CO

was diluted by

88.4% following combustion via the ConFlo. Urea standards were run at the

same settings to account for any possible fractionation of

Cbythisdilution.

Sulfur isotope and concentration standards included a methionine working

standard, seawater sulfate, and silver sul

fi

de reference materials (IAEA S1, S2,

S3). Additionally, a working standard of DOM

SPE

extracted in a large batch

(

∼

200 L) from the Scripps Institution of Oceanography (SIO) pier was run in at

least triplicate with each sample set. Sample

S values were corrected for lin-

earity (peak height) effects and then calibrated to IAEA reference materials. sul-

fur content (

0.10

g sulfur) and

Svalues(

∼

‰

) of tin capsules were also

measured by EA-IRMS and used to correct subsequent analyses for the blank

contribution (68).

S values are reported as permil (

‰

) variations relative to

the Vienna Canyon Diablo Troilite reference frame, while

Cvaluesare

reported relative to Vienna Pee Dee Belemnite. C:S ratios were obtained by

dividing corrected carbon and sulfur amounts (molar ratio), as calculated from

EA peak areas. Extracts supplied by collaborators were often limited by sample

size to single or duplicate analyses, so we could not estimate precision directly

for each sample. Instead, uncertainties in isotopic compositions and C:S ratios

for samples are reported using the SD of our DOM

SPE

SIO pier standard

measurements, divided by the square root of sample replicates analyzed. These

SEs (1

)were

≤

0.2

‰

for

Svaluesand

Cvaluesand

≤

6forC:Sratios.

DOS

SPE

concentrations were calculated as the product of measured total DOC

concentration and DOM

SPE

C:S ratio, and we therefore refer to DOS

SPE

concen-

trations as

“

apparent

”

throughout.

Sulfate Carryover.

We also considered the carryover of seawater sulfate in

DOM

SPE

extracts as a potential source of systematic error. Blank extractions

using 28 mM Na

in deionized water on the PPL SPE cartridges showed

no measurable sulfur above the capsule blank (0.10

g sulfur) via EA-IRMS.

Taking this as the upper limit for sulfate blank, and assuming a

S value

‰

(36), we calculated a worst-case error (for a sample with just 1

sulfur and measured

Sof15

‰

)of0.6

‰

. For a sulfate blank half that

size, and a more typical sample of 4

g sulfur and

17.5

‰

, the effect

would be just 0.05

‰

. There was no correlation between SO

peak size in the

EA and measured

0.0004), so we conclude that sulfate contami-

nation is very unlikely to have caused either the relative

S enrichments or

the depth-related trends in our dataset. DOM

SPE

samples with peak sizes that

corresponded to

g sulfur were not reported to further minimize any

possibility of blank-related artifacts.

Data, Materials, and Software Availability.

All study data are included in

the article and/or supporting information.

ACKNOWLEDGMENTS.

We thank the crew, administrative teams, and science

parties of the

R/V Atlantic Explorer

and the

R/V Kilo Moana

for their assistance

during the Bermuda Atlantic Time-Series and Hawaii Ocean Time-Series cruises,

especially Rod Johnson and Carolina Funkey. We thank Daniela Osorio Rodriguez

and Sijia Dong for their assistance in sample collection aboard the

R/V Sally Ride

andJess Adkins from Caltech. We thank Troy Gunderson from the San Pedro

Ocean Time-Series and the crew of the

R/V Yellow

fi

We acknowledge Mike

Beman for inviting our participation on cruises that provided the NE Paci

fi

cOMZ

samples and Ken Smith for supporting the acquisition of the Station M NE Paci

fi

samples. We are grateful to members and technicians in the Aluwihare Lab at

Scripps Institution of Oceanography, especially Brandon Stephens, Irina Koester,

and Tran Nguyen. We acknowledge Usha Lingappa for drafting Figure 1. We

thank colleagues for early reviews and conversations about the manuscript,

including Tony Wang, Hannah Dion-Kirschner, Morgan Raven, and Ted Present,

and other members of the Adkins and Sessions Labs at Caltech. We thank

University of California Santa Cruz Professor Matthew McCarthy for early conversa-

tions and support for the project. Color-blind-friendly palettes were generated

from Paul Tol

’

s online resource. Funding for this work was provided by NSF

OCE (Division of Ocean Sciences) Grant 2023676 to A.L.S. and A.A.P.; M.S. and

T.D. acknowledge funding by the DFG-FAPERJ (German Research Foundation)

cooperative project (DI 842/6-1) and within the Cluster of Excellence EXC

2077

“

The Ocean Floor - Earth

’

s Uncharted Interface

”

(DFG Project number

PNAS

2022 Vol. 119 No. 41 e2209152119

https://doi.org/10.1073/pnas.2209152119

7of8