of 12
Limnol. Oceanogr.
66, 2021, 3607
3618
© 2021 The Authors.
Limnology and Oceanography
published by Wiley Periodicals LLC on
behalf of Association for the Sciences of Limnology and Oceanography.
doi: 10.1002/lno.11901
Sulfur isotope fractionations constrain the biological cycling of
dimethylsulfoniopropionate in the upper ocean
Daniela Osorio-Rodriguez
,
1
*
Manuel Razo-Mejia,
2
Nathan F. Dalleska,
1
Alex L. Sessions,
1
Victoria J. Orphan,
1
Jess F. Adkins
1
*
1
Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, California
2
Division of Biology and Biological Engineering, California Institute of Technology, Pasadena, California
Abstract
The rapid turnover of dimethylsulfoniopropionate (DMSP), likely the most relevant dissolved organic sulfur
compound in the surface ocean, makes it pivotal to understand the cycling of organic sulfur. Dim-
ethylsulfoniopropionate is mainly synthesized by phytoplankton, and it can be utilized as carbon and sulfur
sources by marine bacteria or cleaved by bacteria or algae to produce the volatile compound dimethylsul
fi
de
(DMS), involved in the formation of sulfate aerosols. The
fl
uxes between the consumption (i.e., demethylation)
and cleavage pathways are thought to depend on community interactions and their sulfur demand. However, a
quantitative assessment of the sulfur partitioning between each of these pathways is still missing. Here, we
report for the
fi
rst time the sulfur isotope fractionations by enzymes involved in DMSP degradation with differ-
ent catalytic mechanisms, expressed heterologously in
Escherichia coli
. We show that the residual DMSP from
the demethylation pathway is 2.7
enriched in
δ
34
S relative to the initial DMSP, and that the fractionation
factor (
34
ε
) of the cleavage pathways varies between

1 and

9
. The incorporation of these fractionation fac-
tors into mass balance calculations constrains the biological fates of DMSP in seawater, supports the notion that
demethylation dominates over cleavage in marine environments, and could be used as a proxy for the domi-
nant pathways of degradation of DMSP by marine microbial communities.
Dissolved organic sulfur is comprised of 432 identi
fi
ed
compounds (Tang 2020), and at l
east 800 compounds predicted
by mass and structure (Ksionzek et al. 2016). Dimethyl-
sulfoniopropionate (DMSP) is the most abundant known and
quanti
fi
able dissolved organic S specie
s, contributing 2.3% of the
minimum estimated marine di
ssolved organic S (Ksionzek
et al. 2016) at an average concentration of 1
2nM(Kieneand
Slezak 2006; Levine et al. 2016). Dimethylsulfoniopropionate
producers can accumulate it intracellularly, generating a particu-
late DMSP pool with concentrations up to nearly 500 mM inside
the cell (Mcparland and Levine 2019). The production of DMSP
is mainly attributed to marine phytoplankton (Keller 1989),
although bacteria and corals are low DMSP producers (Mcparland
and Levine 2019). It has a turnover of hours to days (Zubkov
et al. 2002; Galí and Sim

o 2015; Levine et al. 2016), and has
been hypothesized to be involved in different physiological func-
tions, including protection fro
m cold (Kirst et al. 1991; Karsten
et al. 1996), osmotic (Dickson and Kirst 1987), and oxidative
stresses (Sunda et al. 2002). One of the biological degradation
products of DMSP, dimethylsul
fi
de (DMS), gathered atmospheric
chemistry research attention on volatile dissolved organic S more
than 30 years ago, when its potential to in
fl
uence global climate
by means of aerosol formation was
fi
rst pointed out (Charlson
et al. 1987). The so-called CLAW hypothesis predicted a negative
climate feedback where Earth
s temperature would be regulated
by the interaction between sulfu
r emissions from phytoplankton
and cloud formation. Although DMS emissions might actually
not be signi
fi
cant for global climate regulation under a warming
scenario (i.e., Quinn and Bates 2011), they are expected to alter
the regional formation of sulfate aerosols (Sanchez et al. 2018),
with potential impacts on the weather at high latitudes (Wang
et al. 2018).
Beyond DMS, dissolved organic S is usually disregarded by
sulfur biogeochemists because its abundance in seawater is
exceeded by six orders of magnitude by that of sulfate
(Ksionzek et al. 2016), which has a concentration of 28 mM
(Morris and Riley 1966). The fact that sulfate is an important
electron acceptor, responsible for the remineralization of up
to half of the organic matter in coastal sediments
*Correspondence: dosorior@caltech.edu, jess@gps.caltech.edu
This is an open access article under the terms of the Creative Commons
Attribution-NonCommercial License, which permits use, distribution and
reproduction in any medium, provided the original work is properly cited
and is not used for commercial purposes.
Additional Supporting Information may be found in the online version of
this article.
3607
(Jorgensen 1982), and that it leaves a
fi
ngerprint in the rock
record as sulfate or sul
fi
de (its reduction product) minerals,
has made it central in the study of the geologic sulfur cycle
(Garrels and Lerman 1981). Nonetheless, the slow turnover of
sulfate, which has a residence time of

10
7
years in the ocean
(Holland 1973), highlights the relevance of addressing the
dynamics of highly labile dissolved organic S, such as DMSP
and DMS, toward a comprehensive understanding of the
short-scale processes that may affect the sulfur cycle. Reduced
sulfur, and in particular dissolved organic S, might also be fun-
damental to understanding the sulfur cycle during the
Archaean, where oxygen and sulfate concentrations were neg-
ligible (Fakhraee and Katsev 2019).
Recently, microbial ecologists have rekindled interest in
DMSP as the role of dissolved organic S in ecosystem dynamics
has been highlighted (Levine 2016). Both DMSP and DMS have
been demonstrated to be strong chemoattractants for marine
bacteria and zooplankton (i.e., Seymour et al. 2010). Further-
more, DMSP induces the production of quorum sensing mole-
cules (Johnson et al. 2016), is a mediator of bacterial virulence
toward DMSP-producing algae (Barak-Gavish et al. 2018), and
its cleavage to DMS may generate acrylate as a byproduct, con-
sidered to be a potential predator deterrent (i.e., Wolfe
et al. 1997). Dimethylsulfoniopropionate has also been proven
to satisfy up to 13% of the carbon (Levine et al. 2016), and
100% of the sulfur demand of marine heterotrophic bacteria
(Kiene and Linn 2000). In fact, the most abundant marine bac-
teria (SAR11 clade) are not able to assimilate sulfate and rely
exclusively on DMSP and other reduced sulfur compounds to
satisfy their sulfur requirements (Tripp et al. 2008). Getting to
know the relative routing of DMSP between the demethylation
(i.e., consumption) pathway vs. the cleavage (DMS generating)
pathways (Fig. 1) is then critical to understand the importance
of DMSP and DMS in the marine trophic webs, both as nutrient
sources and as ecologically relevant molecules.
Insight into the relative contributions that each of the
demethylation and cleavage genes/pathways may have to
the fate of DMSP in seawater has been gained from studies
with
35
S

labeled DMSP (Kiene and Linn 2000), stable sulfur
isotopes, and ocean expeditions data. The Global Ocean (GOS;
Rusch et al. 2007) and Tara Oceans (Pesant et al. 2015) expedi-
tions have collected both chemical (concentrations and sulfur
isotopic compositions) and biological (genomic, trans-
criptomic and proteomic) information relevant to the dynam-
ics of DMSP and DMS. Sulfur isotopes are deemed the most
direct and precise geochemical proxies to trace the transforma-
tions of sulfur as it moves through different reservoirs
(reviewed by Fike et al. 2015). Sulfur isotope measurements of
DMSP and DMS in marine surface waters (Amrani et al. 2013;
Carnat et al. 2018), as well as the fractionation in the DMS
produced from DMSP in marine algae (Oduro et al. 2012) were
the
fi
rst attempts to identify a S isotopic signature during
DMSP transformations. Although these fractionation factors
constitute a proxy for the eukaryotic cleavage pathway, spe-
ci
fi
c
δ
34
S signatures for the demethylation pathway and the
bacterial cleavage pathways have not been determined.
Here, we constrained for the
fi
rst time the S isotope frac-
tionations of individual enzymes involved in the cleavage
(DMSP lyases) and demethylation (DMSP demethylase) path-
ways of DMSP. We performed mass balance calculations to
model how the relative contributions of the different biologi-
cal processes that act on DMSP account for the
δ
34
S values of
S
O
-
O
THF
DmdA
5-methyl-THF
S
O
-
O
DMSP
Demethylation
Cleavage
Cleavage
A
Acetyl-S-CoA
OH
O
CO-S-CoA
S
H
3
CO
-
DMS
3-Hydroxypropionyl-
Methyl mercaptopropionate
CoA (3-HP-CoA)
Acetate
S
S
DMS
S
Other DMSP Lyases
(Alma1, DddP/Q/L/W/Y/K)
DddD
O
-
O
Acrylate
+
+
+
+
B
C
MMPA
Fig 1.
Biological fates of DMSP (modi
fi
ed from Lei et al. 2018a). (
A
) Demethylation pathway. The
fi
rst step (shown here) is catalyzed by DmdA, which
uses tetrahydrofolate (THF) as a cofactor (Reisch et al. 2008). This pathway is utilized for the consumption of DMSP as a carbon and sulfur source by
marine bacteria. (
B
) Cleavage pathway with production of 3-Hydroxypropionyl CoA, catalyzed by DddD. This pathway requires acetyl-CoA as cofactor
(Alcolombri et al. 2014b). (
C
) Cleavage pathway with production of acrylate, catalyzed by DMSP lyases other than DddD. Both (
B
) and (
C
) produce
DMS, a volatile S species that is usually released to the water column.
Osorio-Rodriguez et al.
Use of sulfur isotopes to constrain DMSP cycling
3608
total (particulate
+
dissolved) DMSP in seawater. Our data sug-
gest that the fractionation imparted by the DMSP degrading
enzymes is small, but nonetheless useful as a proxy for the
main biological degradation pathways of DMSP in natural
samples. Thus, the fractionation factors reported here provide
a way to establish the relevance of the different cleavage path-
ways of DMSP in natural environments.
Methods
Cultures and extraction of cell lysates
We obtained cell lysates from
Escherichia coli
BL21 cells
independently transformed with DmdA, the only DMSP
demethylase described to date (kindly provided by Will Whit-
man from the University of Georgia, Athens), and different
DMSP lyases (kindly provided by Dan Taw
fi
k from the Weiz-
mann Institute of Science, Israel), under control of the lac pro-
moter. These clones were used to individually express each
enzyme and had the advantage that their transcription could
be regulated with the incorporation of the lac operator
inducer. We do not have a reason to expect that the expres-
sion of the enzymes in
E. coli
would affect their S isotope frac-
tionations. The clones are listed in Table 1, and the natural
taxonomic distributions of each enzyme are reviewed by Lei
et al. (2018a).
Each of the
E. coli
transformants was separately grown in
Luria-Bertani (LB) agar plates incubated overnight at 37

C.
Individual colonies retrieved from the solid media were used
to inoculate 5 mL liquid LB medium for overnight incubations
at 37

C, and 1 mL of these cultures was added to
fl
asks with
1 L of sterile liquid LB media, that were kept at 37

C until
they reached an OD
600
of 0.6
0.8 (late exponential phase).
Solid and liquid culture media were supplemented with the
corresponding antibiotic (50
μ
g/mL ampicillin or kanamycin),
which inhibits the growth of cells that do not possess the
corresponding cloning vectors. The enzyme induction and
the extraction of cell lysates were performed following a modi-
fi
cation of a previously described protocol (Lei et al. 2018b).
Brie
fl
y, the growth temperature was reduced to 16

C, and
enzyme expression was induced overnight with 0.1 mM of the
lac operator inducer isopropyl
β
-d-1-thiogalactopyranoside
(IPTG). The cells were harvested by centrifugation at 4

C and
resuspended in lysis buffer (5mM Tris
HCl pH 8.0, 0.2
g
/L
lysozyme), followed by sonication (10s on, 10s off for 4 min)
and subsequent centrifugation at 10,000

g
for 1 h. Total pro-
tein concentration in the crude extracts was determined using
the Bio-Rad Bradford reagent with bovine serum albumin as
the standard. The average concentration of each heterologous
enzyme was calculated by order of magnitude approximations
following So et al. (2011).
Dimethylsulfoniopropionate biodegradation experiments
Dimethylsulfoniopropionate degradation by cell lysates
with each one of the enzymes in Table 1 was individually
assayed in a similar way as previously described (Lei
et al. 2018b). A DMSP stock solution was prepared by mixing
solid DMSP (Sigma-Aldrich) with reaction buffer (5 mM Tris
HCl pH 8.0), utilized to provide a suitable pH buffering for the
enzymatic reactions. Reaction assays were set up in triplicates
by mixing crude cell extracts, DMSP stock, and reaction buffer
in plastic vials with a total volume between 1 and 5 mL at
28

C. The cell lysates were added at an estimated total protein
concentration of 5.6 mg/mL for DmdA, 1
μ
g/mL for Alma1,
8.2 mg/mL for DddP, 14
μ
g/mL for DddY, 0.9 mg/mL for
DddK, 3 mg/mL for DddQ, and 0.6 mg/mL for DddD, which
allowed to observe DMSP degradation over similar periods of
time. The reaction mixture for the DddD assay was sup-
plemented with 10
μ
M acetyl-CoA, which takes the methyl
group removed from DMSP, and that for the DmdA assay was
set up in anaerobic conditions with 0.685 mM tetrahydro-
folate (THF), required as a cofactor. At de
fi
nite time intervals,
the reaction was quenched by
fi
ltering through Amicon
Ultra-4 or Ultra-15 30 K (Millipore Sigma) with centrifugation
at 5000

g
for 10 min and 4

C for DMSP quanti
fi
cation and S
isotope measurements.
Table 1.
List of
E. coli
BL21 clones expressing different genes involved in DMSP degrading pathways utilized in this study.
Gene
Pathway
Cloned in
E. coli
from
Reference
dmdA
Demethylation
Ruegeria pomeroyi
DSS-3 (Roseobacter;
α
-Proteobacteria)
Reisch et al. (2008)
alma1
Cleavage (DMS
+
acrylate)
Emiliania huxleyi
(Coccolithophore)
Alcolombri et al. (2015)
dddP
Cleavage (DMS
+
acrylate)
Ruegeria pomeroyi
DSS-3 (Roseobacter;
α
-Proteobacteria)
Todd et al. (2011)
dddY
Cleavage (DMS
+
acrylate)
Desulfovibrio acrylicus
(
δ
-Proteobacteria)
Lei et al. (2018b)
dddK
Cleavage (DMS
+
acrylate)
Pelagibacter ubique
(SAR11; a-Proteobacteria)
Lei et al. (2018b)
dddQ
Cleavage (DMS
+
acrylate)
Ruegeria pomeroyi
DSS-3 (Roseobacter;
α
-Proteobacteria)
Todd et al. (2011)
dddD
Cleavage (DMS
+
3-HP-CoA)
Marinomonas
MWYL1 (
γ
-Proteobacteria)
Alcolombri et al. (2014b)
Osorio-Rodriguez et al.
Use of sulfur isotopes to constrain DMSP cycling
3609
Dimethylsulfoniopropionate quanti
fi
cation
For DMSP separation and quanti
fi
cation, an Acquity
ultra-performance liquid chromatograph (Waters, Milford,
Massachusetts) coupled to a Xevo G2-S electrospray ionization
quadrupole time-of-
fl
ight mass spectrometer (Waters Micro-
mass, Manchester, England) operated in positive ion mode
[UPLC/(
+
) ESI-Q-TOF-MS] was used. Samples were prepared by
diluting the stopped DMSP reactions 1:100 in acetonitrile to
fall into the linear detection range of 1.5
30
μ
M. The UPLC
separation was carried out with an Acquity UPLC
BEH
HILIC column (1.7
μ
m, 2.1 mm

100 mm) kept at 27

C with
water (solvent A) and acetonitrile (solvent B) following the
same gradient used by Spielmeyer and Pohnert (2010). The
separation started with 10% A at a
fl
ow rate of 0.25 mL/min
for 0.4 min. The gradient was linearly increased to 60% A
until 1.7 min. At 1.9 min, the
fl
ow rate was increased to
0.6 mL/min. At 2.7 min, the
fl
ow rate and gradient were set
back to 0.25 mL/min and 10% A. Finally, the column was
equilibrated for 1.3 min, resulting in a total analysis time of
4 min. Calibration was performed with standards from 1 to
30
μ
M DMSP, and each sample was diluted for quanti
fi
cation
within this range. The retention time of DMSP was 2.4 min.
Acetonitrile containing 5% v/v water was used as the working
fl
uid in the autosampler syringe in order to maintain the low
water content of the sample solution and initial mobile phase
concentration that is critical for successful HILIC chromatog-
raphy. The mass range from 50 to 300
m/z
using a scan rate of
0.3 s was recorded. The injection volume was 1
μ
L and the
sample was kept at 4

C. The optimized ESI parameters used
were 3 kV capillary voltage, 40 V sampling cone, 80 V source
offset, 120

C source temperature, 450

C desolvation tempera-
ture, and 6 V collision energy.
MS
MS mode data were also acquired to eliminate the pos-
sibility of isobaric interferences. For this measurement, the
quadrupole was set to pass a range of

1 m/z around the mass
of the parent ion (134 m/z). The collision energy was
increased to 30 V, and the signal between 50 and 300 m/z was
recorded at high resolution in the time-of-
fl
ight analyzer. The
product ion at 73 m/z was used for quantitation. Instrumental
stability (i.e., chromatographic and mass spectral reproducibil-
ity) was veri
fi
ed within 5% using a standard solution of DMSP
(Sigma-Aldrich) run periodically (one standard every 10 sam-
ples) during routine analysis. Data were acquired and
processed using MassLynx v4.1 software. The enzymatic rates
of DMSP consumption over time were
fi
tted to Michaelis
Menten kinetics using the previously reported Michaelis
Menten constant (
K
M
) for each heterologously expressed
enzyme (same references as those in Table 1 except for DddP
(Kirkwood et al. 2010), DddK (Peng et al. 2019), and DddQ
(Burkhardt et al. 2017).
Sulfur isotope analysis
To determine the isotopic composition of the residual
DMSP (leftover DMSP after the enzymatic reaction) over the
course of the enzyme assays, a volume of supernatant from
the quenched reactions containing 3
5
μ
g of S was freeze
dried, resuspended in distilled water, and allowed to evaporate
in tin capsules heated at 60

C to concentrate DMSP and
remove volatile sulfur. Sulfur was measured as SO
2
by EA-IRMS
(Carlo Erba NC 2500 Elemental Analyzer connected to a Delta
+
XL, ThermoQuest, via the Thermo Con
fl
o III interface). We
report sulfur isotope ratios using the conventional delta nota-
tion relative to the international standard Vienna-Canyon
Diablo Troilite (VCDT)
δ
34
S
¼
34
R
sample
=
34
R
VCTD


1,
ð
1
Þ
where
34
R refers to the
34
S
=
32
S
ratio. The values of each sample
were corrected by subtracting the blank and using a linear
interpolation between two in-house working standards (sulfa-
nilamide and seawater), with an analytical repeatability better
than 0.26
.
The sulfur isotope fractionation factors for each enzyme
(
34
ε
enz
) were calculated from the slope of the linear regression
analysis of the most accurate approximate solution to the Ray-
leigh distillation equation (Mariotti et al. 1981; Scott
et al. 2004):
ln 1
þ
δ
34
S
DMSP

¼
ln 1
þ
δ
34
S
DMSP,0


34
ε
enz

ln f
R
ðÞð
2
Þ
where enz can be replaced by any of the enzymes studied,
f
R
is the fraction of remaining DMSP in the assay vials, and
δ
34
S
DMSP,0
and
δ
34
S
DMSP
are the sulfur isotopic compositions of
the initial and remaining DMSP at the time of the measure-
ment, respectively. The details of the corrections performed
are included in the Supporting Information.
Data analysis
For the analysis of the data
both the substrate degradation
kinetics and the inference of the fractionation factor
we took
a Bayesian analysis approach. A full description of how this
analysis was performed, including both the theoretical back-
ground and the assumptions behind the statistical analysis,
can be found in the Supporting Information.
Data and code availability
All data and custom scripts were collected and stored using
Git version control. Code for raw data processing, analysis,
modeling, and
fi
gure generation is available in the GitHub
repository (https://github.com/daniosro/DMSP).
Results
The DMSP concentration in the reaction vials at different
timepoints in the enzyme assays with DmdA (DMSP
demethylase) or various DMSP lyases are shown in Fig. 1.
Their
fi
ts to Michaelis
Menten reaction kinetics were good for
DmdA, Alma1 and DddK, and satisfactory for DddD and DddY
Osorio-Rodriguez et al.
Use of sulfur isotopes to constrain DMSP cycling
3610
(Fig. S1). However, we noticed that in the case of DddP and
DddQ the curves seem to
fl
atten out before the reactions were
completed. This was not surprising, since a low catalytic activ-
ity for both enzymes has been recognized (Alcolombri
et al. 2014a; Wang et al. 2015). To investigate if a deviation
from Michaelis
Menten kinetics in these enzymes could be
explained by a loss of their activity during the course of the
reactions, we repeated the Michaelis
Menten
fi
t incorporating
a
fi
rst order decay rate for the enzymes. If an enzyme loses
activity over the course of the reaction, its concentration in
the assay vial (
E
) is assumed to decrease following a
fi
rst order
rate (
k
):
dE
dt
¼
k
E
ð
3
Þ
In turn, the change in the concentration of DMSP over
time will be affected by the decrease in the amount of
enzyme:
dDMSP
dt
¼

v
max
EDMSP
K
M
þ
DMSP
ð
4
Þ
Where
K
M
is the previously reported Michaelis
Menten
constant (see the Methods section) and
v
max
is the
fi
tted maxi-
mum velocity per enzyme (units of nM substrate/min), such
that
V
max
, the maximum catalytic activity when the enzyme
is saturated, is equal to
v
max
E
. The results of this modeling
show very good agreement with the data (Fig. 1). We per-
formed additional enzyme assays for Alma1, the enzyme with
the highest activity, in order to get experimental support for
this hypothesis (Supporting Information). When the enzyme
is in a very low concentration, the reaction stalls early and
when most of the DMSP is still remaining in the reaction.
Adding higher concentrations of enzyme proportionally
increases the fraction of DMSP degraded. Similarly, when addi-
tional DMSP is added to reaction vials where most of it was
already consumed, the added DMSP was degraded very slowly
(Fig. S2). Thus, the experimental results validate the modeled
prediction of a loss of enzymatic activity over the course of
the DMSP degradation assays.
Despite reacting DMSP with large concentrations of DddP
and DddQ, the enzymes that seem to exhibit the larger loss of
activity, we still could not detect complete degradation
of DMSP by these enzymes. To establish if this could affect
the fractionation factors calculated below from the isotopic
compositions of DMSP over the course of the enzymatic deg-
radation experiments, we performed modeling for DddP. We
integrated Eqs. 3 and 4 for
34
DMSP and
32
DMSP (more details
are provided in the Supporting Information), and used their
values to compute the
δ
34
S
of DMSP for
k
=
0 and
k
=
0.08.
We determined the theoretic fractionation factors (
34
ε
) from
DmdA (DMSP demethylase)
DddD (Bacterial DMSP lyase)
Alma1 (Eukaryotic DMSP lyase)
DddP (Bacterial DMSP lyase)
DddQ (Bacterial DMSP lyase)
DddY (Bacterial DMSP lyase)
DddK (Bacterial DMSP lyase)
0183654
Time (min)
0
58
116
174
231
[DMSP] ( M)
0193756
Time (min)
0
27
54
81
108
[DMSP] ( M)
0
9
18
26
Time (min)
0
23
46
69
91
[DMSP] ( M)
0 183654
Time (min)
0
56
112
167
223
[DMSP] ( M)
0122437
Time (min)
0
38
75
113
150
[DMSP] ( M)
0193756
Time (min)
0
27
54
81
108
[DMSP] ( M)
0142842
Time (min)
0
58
115
173
231
[DMSP] ( M)
v
max
/K
M
= 0.029±0.01
v
max
/K
M
= 0.082±0.01
v
max
/K
M
= 0.125±0.01
v
max
/K
M
= 0.072±0.02
v
max
/K
M
= 0.101±0.03
v
max
/K
M
= 0.084±0.02
v
max
/K
M
= 0.015±0.004
A
B
C
D
E
F
G
Fig 2.
Degradation of DMSP by the enzymes depicted in Fig. 1: (
A
) DmdA, which catalyzes the demethylation pathway; (
B
) DddD, which catalyzes the
cleavage pathway with the production of 3-HP-CoA; (
C
) Alma1, which is the only eukaryotic DMSP lyase described; (
D
) DddP, the most abundant and
expressed bacterial DMSP lyase; and (
E
) DddQ, (
F
) DddY, and (
G
) DddK, other DMSP lyases. The points are combined data from triplicate measurements
for each enzyme and the lines represent a
fi
t of the reaction rate based on the Michaelis
Menten kinetics, assuming that the enzymes lose activity over
time with a
fi
rst order degradation rate.
v
max
=
K
M
is the effective catalytic rate of each enzyme, in min

1
. Shaded regions represent the 95% credible inter-
vals from the Bayesian inference of Michaelis
Menten kinetics with enzyme degradation.
Osorio-Rodriguez et al.
Use of sulfur isotopes to constrain DMSP cycling
3611
the slope of the regression of ln (
δ
34
S
DMSP
+
1) vs. ln
f
R
(frac-
tion of DMSP remaining) in both cases, as described in the
Methods section. We found that the apparent loss of enzyme
activity should not affect the enrichment factors by more than
0.01
(Fig. S3A,B). If the enzyme activity was kept constant,
the
δ
34
S values of DMSP that we measured would be larger,
because they would be driven to a greater extent of reaction
(more DMSP degradation) than under a loss of enzyme activ-
ity. In the two cases, there would be a substantial difference in
the ln (
δ
34
S
DMSP
+
1) as a function of time, but not as a func-
tion of ln
f
R
(Fig. 3C).
Since there is no way to discern if the isotope effects of the
DMSP degrading enzymes impact
V
max
only,
K
M
only or both,
the previous model had to make assumptions about them for
both
34
S and
32
S, as well as about the rate of loss of enzyme
activity. To guarantee the reliability of the fractionation fac-
tors that we determined, we performed sensitivity tests to
determine how much the fractionation factors at steady-state
would change for different enzyme degradation rates,
32
V
max
and
32
V
max
=
32
K
M
. We found that changing
32
V
max
(or
34
V
max
)
would have a negligible impact on the fractionation factor,
whereas changing
32
V
max
=
32
K
M
(or
34
V
max
=
34
K
M
) would only
change it by 0.02
(Fig. S4). These analyses are fully
described in the Supporting Information and allowed us to
further con
fi
rm that the fractionation factors determined from
the
δ
34
S
values of DMSP that we measured are reliable.
Having established that a loss of enzyme activity should
not affect the measured
δ
34
S
values, we used them together
with the fractions of DMSP remaining in each enzymatic reac-
tion at each data point to calculate the fractionation factors
(
34
ε
, Fig. 2). All of the enzymes evaluated were found to have
normal kinetic isotope effects (i.e., negative fractionation fac-
tors), that range between

1.2 and

9.1
.
Discussion
The fractionation factors (
34
ε
) determined here are negative
(normal isotope effects), span a range of

8
, and are not
correlated with the effective catalytic rate (
v
max
=
K
M
) or the
Michaelis
Menten constant (
K
M
) of each enzyme. These
34
ε
values are small compared to those of other biological sulfur
transformations such as sulfate reduction (Sim et al. 2011) and
sulfur disproportionation (Can
fi
eld and Thamdrup 1994). The
most plausible reason is that the cleavage of the C
S bond is
not the rate-limiting step in the reaction, and therefore, there
is little sensitivity to sulfur isotopes once this step takes place
(i.e., Goldstein 1966). Speci
fi
cally, in the case of the DMSP
lyases that cleave DMSP to DMS and acrylate, the sulfur cleav-
age reaction happens near the end of a cascade that is initiated
by removing the hydrogen from the alpha carbon position
(Fig. S5). The larger the reversibility of this H abstraction step,
the larger the size of the kinetic isotope effect on the S
0.0
0.4
0.7
1.1
-ln(f
R
)
13.8
14.8
15.8
16.8
17.8
ln(
34
S
VCDT
+1) ( )
DmdA (DMSP demethylase)
0.0
0.7
1.5
2.2
-ln(f
R
)
9.5
14.3
19.2
24.0
28.9
34
= -5.97 ± 0.19
DddD (Bacterial DMSP lyase)
0.0
1.2
2.3
3.5
-ln(f
R
)
14.1
15.4
16.7
18.0
19.3
Alma1 (Eukaryotic DMSP lyase)
0.0
0.4
0.7
1.1
-ln(f
R
)
13.8
15.1
16.4
17.7
19.0
DddP (Bacterial DMSP lyase)
0.0
0.2
0.5
0.7
-ln(f
R
)
13.4
14.7
16.1
17.4
18.8
34
=-5.26± 0.41
DddQ (Bacterial DMSP lyase)
0.0
0.7
1.3
2.0
-ln(f
R
)
11.2
14.9
18.5
22.1
25.7
34
=-9.09± 1.46
DddY (Bacterial DMSP lyase)
0.0
0.2
0.5
0.7
-ln(f
R
)
12.9
15.1
17.3
19.5
21.7
DddK (Bacterial DMSP lyase)
34
= -2.71 ± 0.17
34
= -1.17 ± 0.02
34
= -3.98 ± 0.12
34
= -5.09 ± 0.19
ln(
34
S
VCDT
+1) ( )
ln(
34
S
VCDT
+1) ( )
ln(
34
S
VCDT
+1) ( )
ln(
34
S
VCDT
+1) ( )
ln(
34
S
VCDT
+1) ( )
ln(
34
S
VCDT
+1) ( )
A
A
B
C
D
E
F
G
Fig 3.
Evolution of the
δ
34
S
values of DMSP as it is degraded by (
A
) DmdA, which catalyzes the demethylation pathway; (
B
) DddD, which catalyzes the
cleavage pathway with the production of 3-HP-CoA; (
C
) Alma1, which is the only eukaryotic DMSP lyase described; (
D
) DddP, the most abundant and
expressed bacterial DMSP lyase, and (
E
) DddQ, (
F
) DddY, and (
G
) DddK, other DMSP lyases. Measurements were made at the same points as the con-
centrations depicted in Fig. 1. The points are combined data from triplicate measurements for each enzyme. Values of ln (
δ
34
S
DMSP
+
1) are plotted
against the negative ln of the fraction of DMSP remaining (
f
R
). The lines represent a linear
fi
t of the Rayleigh distillation equation, where the slope was
taken as a measurement of the fractionation factor,
34
ε
. Shaded regions represent the 95% credible intervals from the Bayesian inference of the linear
regression.
Osorio-Rodriguez et al.
Use of sulfur isotopes to constrain DMSP cycling
3612
isotopes of the residual (remaining) DMSP (Kaplan and
Rittenberg 1964). This expected trend matches our observa-
tions, since the magnitude of the fractionation factors (Alma1
< DddP < DddY

DddQ < DddK) is correlated with the
reversibility of the reactions. Alma1 uses cysteine as a nucleo-
phile (Alcolombri et al. 2015), whereas DddP uses aspartate
and coordination to a Fe atom (Wang et al. 2015), DddY and
DddQ have DMSP coordinated to a Zn (sometimes Fe) atom
and use tyrosine as nucleophile (Li et al. 2014), and DddK uses
tyrosine as well but coordinates DMSP to a Mn or Ni atom
(Schnicker et al. 2017; Peng et al. 2019, Fig. S5). Thus, the
DMSP cleavage reactions with acrylate as a byproduct in
which a stronger nucleophile is involved (i.e., cysteine) are
less reversible than those where a weaker nucleophile is
involved (i.e., tyrosine), and consequently have smaller iso-
tope effects.
Despite the diversity of DMSP degrading enzymes, the Tara
Oceans expedition found that more than 90% of the
expressed bacterial DMSP lyases (fraction <3
μ
m) are DddP
homologs (Curson et al. 2018; Fig. S6). Thus, we modeled the
expected
δ
34
S values of total DMSP in seawater assuming that
DMSP is either demethylated or cleaved by only Alma1
(eukaryotic DMSP lyase) or DddP, incorporating our fraction-
ation factors in the calculations. The model is described in
detail in the Supporting Information, and the results for differ-
ent activities of each enzyme are shown in Fig. 4. We consid-
ered the ocean as a single box with a constant inward
fl
ux of
DMSP from a single process (biosynthesis) and two possible
outward
fl
uxes, cleavage, and demethylation. The mixing ratio
between the three possible consumption pathways for DMSP
that our model considers
demethylation, bacterial cleavage
by DddP, and eukaryotic cleavage by Alma1
will determine
the
δ
34
S of total DMSP in seawater. It would be expected that
the fractionation by Alma1 would be primarily imparted
in the particulate DMSP pool, and that the fractionation by
bacterial enzymes would be imparted in the dissolved DMSP
pool, although our model does not differentiate between these
two. If the production of DMSP is balanced by consumption
(a reasonable assumption in the ocean due to the rapid turn-
over of DMSP), mass balance constrains the isotopic composi-
tion of total DMSP to be different from the input by the
isotope effect (Hayes 2001). In other words, the isotopic com-
position inherited by DMSP from its biosynthetic pathway is
subsequently altered by consumption, and the enzyme with a
higher concentration (due to differences in community com-
position and/or gene expression) or activity (faster reaction
rates) will drive the
δ
34
S of total DMSP toward its fractionation
value. As a consequence, when eukaryotic cleavage is the
dominant process (Alma1;
34
ε

1
), total seawater DMSP
will have the lowest
δ
34
S, when demethylation dominates
(DmdA;
34
ε

3
), total seawater DMSP will have an inter-
mediate
δ
34
S value, and when bacterial cleavage dominates,
total seawater DMSP will be able to reach the heaviest possible
δ
34
S values, assuming that the input of DMSP to the ocean has
an approximately constant
δ
34
S value. Since DmdA has an
intermediate fractionation factor between those of Alma1 and
DddP, the
δ
34
S value of DMSP of an environmental sample
would not be enough to establish whether demethylation or
cleavage processes are dominant, and other biological analysis
tools could be handy in these cases, as further explained
below.
The value of

1.18

0.06
for the
34
ε
of Alma1 agrees
with an
34
ε
of

1to

1.5
reported for DMSP cleavage in
culturing experiments with the macroalgae
Ulva lactuca
and
Ulva linza
(Oduro et al. 2012). To our knowledge, no other
absolute fractionation factors for DMSP degrading enzymes
had been reported before. Our modeling approach demon-
strates their usefulness to infer the DMSP degradation pro-
cesses that dominate over a seawater sample with a particular
δ
34
S
DMSP
. We predict values of total seawater
δ
34
S of DMSP
that range from 18.2 to 21.1
. These values fall within the
range of
δ
34
S of total DMSP measured in seawater to date dur-
ing normal (nonbloom) conditions (Amrani et al. 2013; Car-
nat et al. 2018), spanning 17.8
20.5
at depths up to 140m,
and 18.9
20.3
in surface waters (0
5 m). The model
assumed an isotopic composition of 17
for newly synthe-
sized DMSP (before it is affected by any degradation process)
in order to capture the entire range of nonbloom total seawa-
ter
δ
34
S
DMSP
measurements from those two studies. This value
is lower than the data of intracellular
δ
34
S
DMSP
reported by
Oduro et al. (2012) in macroalgae (18.2

0.6
) and phyto-
plankton (19.6

0.3
), and by Gutierrez-Rodriguez
et al. (2017) in
Phaeocystis
and foraminifera (

20
), which
correspond to particulate DMSP. This indicates that there
must be a normal isotope effect in the synthesis of DMSP from
marine sulfate (21
), and that the intracellular particulate
δ
34
S
DMSP
measured by Oduro et al. (2012) and Gutierrez-
Rodriguez et al. (2017) might already have been affected by
cleavage by Alma1, which would leave that DMSP pool
enriched in
34
S.
On the other hand, the values measured by Amrani
et al. (2013) for the
δ
34
S of DMS in seawater were found to be
consistently higher relative to the
δ
34
S of total DMSP in the
same samples by an average of 0.6
throughout the water
column. The same study found a

0.5
fractionation factor
associated to the volatilization of DMS. The fractionation fac-
tors for DMSP cleavage reported here and the fractionation
factor for DMS volatilization reported by Amrani et al. (2013)
alone would not be able to explain the seawater
δ
34
S
DMS
under
normal conditions. However, under nonsteady-state condi-
tions, such as at the end of a bloom, fast recycling of organic
sulfur compounds might increase the
δ
34
S
DMS
values. The only
other processes that could cause fractionation of S isotopes in
DMS are consumption by organisms and photooxidation. If
photooxidation was responsible for the enrichment of
34
Sin
DMS relative to DMSP, Gutierrez-Rodriguez et al. (2017)
should have found
34
ε
values different than those reported by
Oduro et al. (2012), since these studies performed incubations
Osorio-Rodriguez et al.
Use of sulfur isotopes to constrain DMSP cycling
3613
under light and dark conditions, respectively. Therefore, we
propose that there must be a normal isotope effect of about

2.5 to

7.5
associated with microbial DMS consumption,
which drives the residual DMS back to a
δ
34
S close to that of
seawater sulfate (Fig. 5). This was previously hypothesized by
Amrani et al. (2013) and it is consistent with the observations
from Gutierrez-Rodriguez et al. (2017) in seawater incubations.
Alternatively, the inputs of DMSP to the ocean might have
different
δ
34
S values, which would increase the range of possi-
ble
δ
34
S
DMS
in seawater samples.
The presence of DMSP lyases in marine bacterial genomes
is variable. In particular, DddK is much more abundant in
high southern latitudes (Landa et al. 2019). On the other
hand, it has been established that DMSP and DMS productiv-
ity is high in coastal and marine sediments, and that bacteria
are important DMSP producers in these environments
(Williams et al. 2019). Some DMSP lyases that do not have a
high representation in the global ocean metatranscriptomes,
such as DddD and DddY (Fig. S6), have been isolated from
coastal and intertidal settings (De Souza and Yoch 1995; Todd
Transcripts/L
18.2
18.9
19.6
20.3
21.1
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.0
0.2
0.4
0.6
0.8
1.0
0.2
f
DddP
f
Alma1
f
DmdA
0.4
0.6
0.8
1.0
34
~-1
)
(
34
~-4
)
(
34
~-3
)
(
A
32
V
max
/
32
K
M
18.2
18.9
19.6
20.3
21.1
0.0 0.2 0.4 0.6 0.8 1.0
0
.0
0.0
0.2
0.4
0.6
0.8
1.0
0.2
f
DddP
f
A
lma1
f
DmdA
0.4
0.6
0.8
1.0
34
~-
1
)
(
3
4
~-4 )
(
34
~-
3
)
(
B
34
SDMSP
SW
()
34
S DMSP
SW
()
Fig 4.
Prediction of the
δ
34
S values of total DMSP in seawater (SW, colorbars), assuming that DMSP is degraded only by Alma1 (eukaryotic DMSP lyase),
DddP (most abundant bacterial DMSP lyase), and DmdA (DMSP demethylase). The isotopic mass balance calculation assumed a
δ
34
Sof17
for the
incoming
fl
ux of DMSP. The ternary plots show the expected S isotopic composition of total DMSP when Alma1, DddP and DmdA fractionally contribute
to DMSP degradation amounting to a total of 1 (or 100%), when that contribution is considered in terms of (
A
) the enzyme concentrations in transcripts
per liter or (
B
) the enzyme catalytic rate for the light and most abundant isotope of sulfur (
32
S,
32
V
max
=
32
K
M
) of each enzyme.
Fig 5.
Schematic representation of the predicted and determined isotope fractionations associated with DMSP and DMS synthesis and degradation. The
values with * were reported by Amrani et al. (2013) and the values surrounded by a dashed box are predicted based on the fractionation factors found in
this study. The subscript 0 refers to just synthesized DMSP or DMS, and the subscript SW refers to the range of values of DMSP or DMS in seawater.
Osorio-Rodriguez et al.
Use of sulfur isotopes to constrain DMSP cycling
3614
et al. 2007; Curson et al. 2011), in association with plant roots
and microaerobic environments. These differences could be
responsible for local variability in the
δ
34
S of environmental
DMSP. Studies based on the tracing of
35
S-labeled DMSP
(Kiene and Linn 2000) and the quanti
fi
cation of sulfur species
(Bates 1994) determined that usually less than 30% of DMSP
is cleaved in natural waters. This would imply that demethyla-
tion is the dominant DMSP-degrading process over most of
the ocean, and it is consistent with the presence of DmdA in
19% of the Tara Oceans
surveyed bacterial genomes from all
depths, vs. that of all of the bacterial DMSP lyases combined in
only 9% of them (Landa et al. 2019). It has also been pointed
out that in global surface waters
dmdA
homologs are present in
more than 50% of free living bacterioplankton, whereas genes
that encode DMSP lyases are up to two orders of magnitude less
abundant (Moran et al. 2012). The average value of total seawa-
ter
δ
34
S
DMSP
from Amrani et al. (2013) (19.6
) and those of

19
20
predicted by our model are reasonable if demethyl-
ation is the dominant DMSP degradation process. It has also
been established that the transcripts of DmdA are about one
order of magnitude more abundant than those of DddP in the
open ocean (Levine et al. 2012; Varaljay et al. 2015), and that
both increase during algal blooms (Varaljay et al. 2015). Trans-
criptomic data for Alma1, the eukaryotic DMSP lyase (Vorobev
et al. 2020, Fig. S6), also indicate higher expression levels of
this enzyme in the Southern Ocean and the North Atlantic
Ocean, where high DMSP producers like
Phaeocystis
and
coccolithophores thrive (Yoch 2002). However, no Alma1
transcripts were found in most of the stations sampled by the
Tara Oceans expedition in these and other ocean basins
(Fig. S6). This suggests that the expression of Alma1 is mostly
limited to localized spots in the ocean, possibly associated
with eukaryotic blooms. If that was the case, and its expres-
sion dominated over that of bacterial DMSP lyases during
those blooms, the particulate (and total)
δ
34
S
DMSP
would be
driven closer to that of the source, because of its small normal
isotope effect and the shift in community composition.
Slightly higher values of total
δ
34
S
DMSP
were measured by
Amrani et al. (2013) in a Greenland bloom in samples kept at
25

C (19.5
22.1
) relative to nonbloom conditions. This
bloom was dominated by the high DMS/P producer
Emiliania
huxleyi
, and the wide range of total
δ
34
S
DMSP
, both above and
below the
δ
34
S of seawater, might re
fl
ect a rapid recycling of
organic sulfur during blooms dominated by high DMS/P pro-
ducing algae. In the same study, the values of total
δ
34
S
DMSP
of
a Mediterranean Sea bloom, dominated by small eukaryotes
and the cyanobacteria
Synechococcus
, were not different from
those of nonbloom conditions, which is consistent with an
interplay between bacterial and eukaryotic DMSP degradation
processes when blooms retain a mixed community composi-
tion. More studies on ecosystem and community composition
changes during blooms and across different oceanic regimes
are required to address these differences.
Conclusions
The fractionation factors reported here provide an indica-
tion of the biological fates of DMSP in the ocean. For most of
the ocean, the total
δ
34
S
DMSP
values result from a mixed contri-
bution from demethylation, bacterial cleavage and eukaryotic
cleavage. Our data can be useful to address whether bacterial
or eukaryotic DMSP degrading processes dominate at a local
scale. Since the bacterial DMSP lyases have higher fraction-
ation factors (approx.

4to

9
), samples with heavier-
than-average
δ
34
S
DMSP
measurements might provide key clues
to address why the switch from demethylation to cleavage
happens in bacteria (Sim

o 2001). More measurements of dis-
solved and total seawater
δ
34
S
DMSP
, tied to microbial meta-
genomics and metatranscriptomics, are needed to fully
understand how seawater
δ
34
S
DMSP
values are affected by eco-
logical dynamics, and could also be critical to understanding
the role and evolution of the DMSP degrading enzymes. Addi-
tionally, identifying the S isotope fractionations imparted by
DMSP biosynthesis and biological DMS consumption will be
critical to improve our understanding of relative importance
of bacteria and algae in the cycling of organic S and the impli-
cations of these processes for marine microbial communities.
References
Alcolombri, U., S. Ben-Dor, E. Feldmesser, Y. Levin, D. S.
Taw
fi
k, and A. Vardi. 2015. Identi
fi
cation of the algal
dimethyl sul
fi
de-releasing enzyme: A missing link in the
marine sulfur cycle. Science
348
: 1466
1469.
Alcolombri, U., M. Elias, A. Vardi, and D. S. Taw
fi
k. 2014
a
.
Ambiguous evidence for assigning DddQ as a dim-
ethylsulfoniopropionate lyase and oceanic dimethylsul
fi
de
producer. Proc. Natl. Acad. Sci.
111
: 2078
2079. doi:
10.
1073/pnas.1401685111
Alcolombri, U., P. Laurino, P. Lara-Astiaso, A. Vardi, and D. S.
Taw
fi
k. 2014
b
. DddD is a CoA-transferase/lyase producing
dimethyl sul
fi
de in the marine environment. Biochemistry
53
: 5473
5475. doi:
10.1021/bi500853s
Amrani, A., W. Said-Ahmad, Y. Shaked, and R. P. Kiene. 2013.
Sulfur isotope homogeneity of oceanic DMSP and DMS.
Proc. Natl. Acad. Sci.
110
: 18413
18418.
Barak-Gavish, N., M. J. Frada, C. Ku, and others. 2018. Bacte-
rial virulence against an oceanic bloom-forming phyto-
plankter is mediated by algal DMSP. Sci. Adv.
4
: eaau5716.
doi:
10.1126/sciadv.aau5716
Bates, T. S. 1994. The cycling of sulfur in surface seawater of
the Northeast Paci
fi
c. J. Geophys. Res.
99
: 7835
7843. doi:
10.1029/93JC02782
Burkhardt, I., L. Lauterbach, N. L. Brock, and J. S. Dickschat.
2017. Chemical differentiation of three DMSP lyases from
the marine: Roseobacter group. Org. Biomol. Chem.
15
:
4432
4439. doi:
10.1039/c7ob00913e
Osorio-Rodriguez et al.
Use of sulfur isotopes to constrain DMSP cycling
3615