BIOGEOCHEMISTRY
Plants sustain the terrestrial silicon cycle during
ecosystem retrogression
F. de Tombeur^1 *, B. L. Turner^2 , E. Laliberté3,4, H. Lambers^4 , G. Mahy^1 , M.-P. Faucon^5 ,
G. Zemunik^4 , J.-T. Cornelis^1
The biogeochemical silicon cycle influences global primary productivity and carbon cycling, yet changes
in silicon sources and cycling during long-term development of terrestrial ecosystems remain poorly
understood. Here, we show that terrestrial silicon cycling shifts from pedological to biological control
during long-term ecosystem development along 2-million-year soil chronosequences in Western
Australia. Silicon availability is determined by pedogenic silicon in young soils and recycling of plant-
derived silicon in old soils as pedogenic pools become depleted. Unlike concentrations of major
nutrients, which decline markedly in strongly weathered soils, foliar silicon concentrations increase
continuously as soils age. Our findings show that the retention of silicon by plants during ecosystem
retrogression sustains its terrestrial cycling, suggesting important plant benefits associated with this
element in nutrient-poor environments.
S
ilicon (Si) is widely recognized as an im-
portant regulator of the global carbon
cycle via its effect on diatom productivity
in oceans ( 1 ) and the weathering of sili-
cate minerals on continents ( 2 ). Si is also
a beneficial plant nutrient ( 3 , 4 ), improving
resistance to herbivory and pathogens ( 5 )and
mitigating the negative effects of several abiotic
stresses ( 6 ), including nutrient limitation ( 7 , 8 ).
As a result, Si improves plant performance and
can contribute to the functioning of terrestrial
ecosystems ( 5 , 9 , 10 ). Detailed information on
long-term controls on Si cycling therefore un-
derpins our understanding of Si-related func-
tions in plants, fluxes to aquatic ecosystems,
and ultimately the fixation of atmospheric
carbon in terrestrial and oceanic ecosystems.
The release of Si into the soil solution reg-
ulates its availability to plants and its transfer
from land to oceans. While the concentration
of dissolved Si in the soil solution has long
been understood to be driven primarily by
geochemical processes (i.e., mineral dissolu-
tion), it is now recognized that Si mobility is
influenced strongly by plant biocycling ( 11 – 13 ).
The polymerization of amorphous silica in leaf
tissues (i.e., the formation of phytoliths) and
its return to topsoil after leaf shedding builds
a pool of reactive silicate in soil ( 14 ). However,
the magnitude of geochemical versus bio-
logical processes in controlling the release
of Si to the soil solution remains debated.
Although soil scientists often assume that
geochemical processes control dissolved Si
concentrations ( 15 ), mass-balance calculations
suggest a strong imprint of biological pro-
cesses (i.e., phytolith formation in plants and
dissolution in soils) on the Si cycle ( 11 , 16 , 17 ),
driven by the order-of-magnitude greater disso-
lution rate of phytoliths compared with clay
minerals ( 14 ). As soil Si is derived ultimately
from the parent rock, plant-available Si con-
centrations are expected to decrease with soil
age through desilication(i.e., Si leaching during
pedogenesis) ( 18 ), thus increasing the impor-
tance of biological processes as soils and
ecosystems develop ( 12 , 13 ). However, the
emergence of biological control of terrestrial
Si cycling as soils age is still poorly under-
stood, in part because of the limited number
of study systems spanning sufficiently long
time scales.
To quantify changes in pedological and
biological controls of Si cycling during long-
term ecosystem development, we studied Si in
soils and plants along a pair of 2-million-year
coastal dune chronosequences in southwestern
Australia ( 19 ). Suchlong-term chronosequences
that have not been directly affected by Pleisto-
cene glaciations are rare worldwide ( 20 ). The
Jurien Bay and Guilderton chronosequences
include the end-members of soil formation
( 18 , 21 ), providing a rare opportunity to study
long-term shifts in biogeochemical cycles. Soil
development along these chronosequences
includes carbonate leaching from Holocene
soils [<6.5 thousand years (ka); stages 1 to 3],
formation of secondary Si-bearing minerals
in young Middle Pleistocene soils (~120 ka;
stage 4) followed by their loss via dissolution
in medium-aged and old Middle Pleistocene
soils (~250 to 500 ka; stage 5), to yield quartz-
rich soils of Early Pleistocene age (~2000 ka;
stage 6) ( 18 , 21 ).
Along each chronosequence, we quantified
the pools of reactive Si-bearing phases and
plant-available Si in the soils and physicallyextracted phytoliths ( 22 ). In addition, we quan-
tified Si and major nutrients in mature leaves
of the most abundant plants growing along
the best-studied of the two chronosequences
(Jurien Bay) ( 21 , 23 , 24 ). We used the con-
centrations of Si and nutrients in leaves to
indicate the degree of elemental biocycling.
We hypothesized that, as soils aged, the pools
of reactive and plant-available Si would be in-
creasingly determined by recycling from phy-
toliths. We also hypothesized that plant foliar
Si concentrations would decrease with soil age
owing to the loss of Si-bearing minerals and
quartz enrichment, as do the concentrations of
major rock-derived nutrients, such as phos-
phorus, during long-term pedogenesis ( 24 ).
ThereactivepedogenicSipool(poorlycrys-
talline aluminosilicatesof nonbiogenic origin,
estimated by oxalate extraction) increased
markedly from Holocene soils (stages 1 to 3)
to young Middle Pleistocene soils (stage 4),
associated with the formation of clay minerals
( 18 ) (Fig. 1A; from≤250 kg ha−^1 to≥2000 kg
ha−^1 ). However, desilication during prolonged
soil weathering resulted in the complete loss
of the reactive pedogenic Si pool in the oldest
stage of the chronosequences.
The Si:Al ratio of alkali-reactive Si (mea-
sured in hot 1% Na 2 CO 3 ) indicates the origin
of this pool: values >5 suggest a biogenic
origin ( 22 ). Alkali-reactive Si stocks were
lowest in the three first stages (≤1200 kg ha−^1 )
and had a mostly nonbiogenic origin (Si:Al
0.5 to 1.8) (Fig. 1B), indicating a contribution of
lithogenic and/or pedogenic minerals. Alkali-
reactive Si increased strongly in stage 4 (3800
to 6100 kg ha−^1 ), but the Si:Al ratio remained
typical of lithogenic and pedogenic minerals
(1.4 to 2.1). In contrast to the reactive pedo-
genic Si pool, however, alkali-reactive Si did
not disappear during long-term pedogenesis,
varying between 2500 and 6300 kg ha−^1 in the
most advanced stage of soil weathering and
having Si:Al ratios >5, which indicates a bio-
genic origin ( 25 ).
Plant-available Si quantified by extraction in
0.01 M CaCl 2 followed a pattern similar to the
reactive pedogenic Si pool, increasing to a maxi-
mum by stage 4 and then decreasing toward
the oldest stage (Fig. 1C). The stocks of plant-
available and reactive pedogenic Si were
significantly correlated [coefficient of deter-
mination (R^2 ) = 0.68;P< 0.01;n= 9 soil
profiles] along both chronosequences.
Soil phytoliths extracted physically by gravi-
metric separation were concentrated in the
surface soil horizon, where plant-available Si
concentrations were also highest ( 22 ). The
concentration of soil phytoliths was positively
correlated with that of plant-available Si in soil
horizons dominated by quartz minerals (Fig. 2).
The contribution of phytoliths to plant-available
Si was supported by dissolution features on
phytoliths, which increased with depth in theRESEARCH
de Tombeuret al.,Science 369 , 1245–1248 (2020) 4 September 2020 1of4
(^1) TERRA Teaching and Research Centre, Gembloux Agro-Bio
Tech, University of Liege, 5030 Gembloux, Belgium.
(^2) Smithsonian Tropical Research Institute, Balboa, Ancon,
Panama.^3 Institut de Recherche en Biologie Végétale,
Département de Sciences Biologiques, Université de
Montréal, Montréal, QC H1X 2B2, Canada.^4 School of
Biological Sciences, The University of Western Australia,
Perth, WA 6009, Australia.^5 AGHYLE, UniLaSalle, 60026
Beauvais, France.
*Corresponding author. Email: [email protected]