Environmental Pollution 236 (2018) 416e424

Contents lists available at ScienceDirect

Environmental Pollution
journal homepage: www.elsevier.com/locate/envpol

What is the aquatic toxicity of saponin-rich plant extracts used as

biopesticides?*
Xiaogang Jiang*, Hans Chr Bruun Hansen, Bjarne W. Strobel, Nina Cedergreen
Department of Plant and Environmental Sciences, University of Copenhagen, Thorvaldsensvej 40, 1871 Frederiksberg, Denmark

a r t i c l e i n f o a b s t r a c t

Article history: Saponin-rich extracts from Quillaja saponaria and Chenopodium quinoa have been registered by US EPA as
Received 7 May 2017 active ingredients in biopesticides, and extract from tea seed powder, Camellia oleifera has been proposed
Received in revised form for biocidal use. If saponin-rich biopesticides are efficient against pests, they are most likely also
10 January 2018
bioactive in the aquatic environment against non-target organisms. The aim of this study was to conduct
Accepted 17 January 2018
an effect assessment of saponin-rich plant extracts by using species sensitivity distributions based on
acute toxicity tests. The maximal concentrations protecting 95% of the aquatic species (HC5) of saponins
extracted from quillaja bark, tea seed coat and quinoa seed coat were 2.91 ± 1.00, 0.22 ± 0.11 and
Keywords:
Saponins
22.9 ± 5.84 mg/L, respectively. The 100-fold difference in toxicity between the saponin-rich extracts from
Biopesticides different plant species, indicate that saponin toxicity depends on the species it origins from, making
Species sensitivity distribution “read-across” between saponins a dubious exercise. In addition, the predicted environmental concen-
Aquatic toxicity trations of different saponins are close to or higher than their water quality standard, which means that
the extracts might pose a risk to the aquatic environment if not used cautiously.
© 2018 Elsevier Ltd. All rights reserved.

1. Introduction
Saponins are a class of natural compounds, found abundantly in
more than 100 families of plants (Güçlü-Üstündag
and Mazza,
2007). They are characterized by their structure containing a tri-
terpene or steroid aglycone bound to a sugar moiety (Fig. 1). Sa-
ponins have antiviral (Güçlü-Üstündag
and Mazza, 2007),
antiprotozoal and antibacterial activities (Potter et al., 2010) and are
regarded as wormicides (Szakiel et al., 2005), fungicides (Rubio-
Moraga et al., 2011) and piscicides (Cannon et al., 2004). These
uses indicate that saponins are bioactive and may be used as bio-
pesticide ingredients. The definition of biopesticides by the U.S.
Environmental Protection Agency (EPA) is: “Biopesticides include
naturally occurring substances that control pests (biochemical
pesticides), microorganisms that control pests (microbial pesti-
cides), and pesticidal substances produced by plants containing
added genetic material (plant-incorporated protectants).” (U.S.
Environmental Protection Agency, 2017). Saponin-rich plant
*
This paper has been recommended for acceptance by Dr. Harmon Sarah
Michele.
* Corresponding author.
E-mail addresses: jiang@plen.ku.dk (X. Jiang), haha@plen.ku.dk (H.C.B. Hansen),
bjwe@plen.ku.dk (B.W. Strobel), ncf@plen.ku.dk (N. Cedergreen).
extracts belong to the first group, and saponins from Quillaja sap-
onaria and Chenopodium quinoa have already registered by U. S. EPA
as biopesticides (U.S. Environmental Protection Agency, 2007,
2005).
Saponins from Q. saponaria can be used for control/suppression
of pathogenic fungi or plant parasitic nematodes in vegetable or
fruit fields using a dose of 1.4e3.7 g liquid extract/m2 (~8.6% of
saponins) (U.S. Environmental Protection Agency, 2007). Saponins
from tea seed powder or pellets (Camellia oleifera) have been used
as soil additives with documented growth enhancing effects with
1.5 g powder/m2 for pot-grown beet, oat, and barley plants
(Andresen and Cedergreen, 2010), or have been used to expel
earthworms from golf courses and sport fields applying 30 g pel-
lets/m2 once a month (Potter et al., 2010). Saponin extracts from
C. quinoa can be used for control of pathogenic fungi on tubers,
legume, and cereal seeds, with 1 g liquid/m2 (U.S. Environmental
Protection Agency, 2005). They also showed potential use against
snails, with an application dose of 0.6 g coat/m2 (10 mg saponin/L in
the water of rice fields) (Joshi et al., 2008). The saponin-rich ma-
terials are often biowaste products. The tea seed pellets are, for
example produced during oil production (Andresen and
Cedergreen, 2010), while quinoa coat is generated during clean-
ing of the quinoa seed for consumption (San Martín et al., 2008).
There is a great interest in creating a rational reuse of the biowastes,

https://doi.org/10.1016/j.envpol.2018.01.058
0269-7491/© 2018 Elsevier Ltd. All rights reserved.
 X. Jiang et al. / Environmental Pollution 236 (2018) 416e424
Fig. 1. Structure of quillaja saponin. R may be either -H or substituted by one or more sugar moieties to form different saponins. The lipophilic part (shown in red) located in the
middle of the structure is the aglycone. Modified after Fernandez-Tejada et al. (2014). (For interpretation of the references to colour in this figure legend, the reader is referred to the
Web version of this article.)
and the saponin-rich bioproducts seem to have a considerable
potential. The saponins in the three plant species contain a tri-
terpene aglycone, which means that all saponins have a similar
aglycone with different functional groups and sugar chains
attached. It is, however, unknown to what extent the naturally
produced but different saponins have a similar biological activity.
Also, information on their specific toxicity to a wide range of spe-
cies is lacking. Information on toxic concentration ranges and
source-specific saponin toxicity are important, if we are going to
use saponin-rich plant extracts in large quantities as biopesticides
in an environmentally sustainable way.
The very low octanol/water partition coefficient (kow) of sapo-
nins and hence poor bonding to organic matter make saponins
prone to leaching from soils (U.S. Environmental Protection Agency,
2007). Hence, to know their toxicity to aquatic organisms is
important. In Europe, there are two regulatory frameworks giving
guidance on how to produce water quality standards. One is
Environmental Quality Standards (EQS) for chemicals which are
used for chemicals in general and used retrospectively described in
the Water Framework Directive (WFD) (European Commission,
2011). Another one is Regulatory Acceptable Concentrations
(RAC), which is used prospectively for Plant Protection Products
and their Residues (EFSA PPR, 2013). Within both frameworks,
several options are possible to derive water quality standard for
saponins depending on the data available. One of the more
comprehensive and preferred methods is to use Species Sensitivity
Distributions (SSDs). SSDs describe the variation among a set of
species in the sensitivity towards a certain compound (or a mixture
of compounds) by applying a statistical distribution to the data
(Posthuma et al., 2002). Based on the distribution, a concentration
protecting a certain percentage of the species in the community can
be determined. Hence, if we want to protect at least 95% of the
species, the environmental concentration must be below the con-
centration that is not hazardous to 5% of the species (HC5). The SSD
based water quality standards underlying the WFD and Plant Pro-
tection Products Regulation both use the median HC5 derived from
the SSD curve and an additional assessment factor (AF) for
extrapolation to the field, but the methodology used is slightly
different as well as the terminology. The WFD makes a distinction
in AA-EQS (based on chronic toxicity data) and an MAC-EQS (based
on acute toxicity data) and for both AA-EQS and MAC-EQS deriva-
tion based on SSDs at least 10 valid toxicity data for at least 8
different taxonomic groups are required. For MAC-EQS derivation
the median HC5 from the SSD constructed with acute toxicity data
(50% effect or lethal concentrations, LC50 and/or EC50 values) and an
AF of 10 is selected (European Commission, 2011). In the aquatic
417
effect assessment underlying the PPP Regulation, a distinction is
made between an acute (based on acute LC50/EC50 data) and a
chronic (based on chronic No Observable Effect Concentrations
NOEC or EC10 values) Regulatory Acceptable Concentration (RACs-
w;ac and RACsw;ch, respectively). In deriving the RACsw;ac and
RACsw;ch by means of the SSD approach at least 8 valid toxicity data
for the potentially sensitive taxonomic groups should be provided
(European Commission, 2011). For saponins with a more or less
biocidal mode-of-action toxicity data of aquatic vertebrates, in-
vertebrates, plants and microorganisms may be used to construct
the SSD if the toxicity data comprise at least 6 different taxonomical
orders/families (EFSA PPR, 2013). Note that in pesticide risk
assessment the functions of microorganisms need to be protected
and not necessarily their structural properties. For RACsw;ac deri-
vation the median HC5 from an SSD constructed with acute toxicity
data (LC50 and/or EC50 values) and an AF of 3e6 is selected (EFSA
PPR, 2013). The size of the selected AF depends on several
criteria, amongst others, on the position of the toxicity data in the
tail of the SSD, the goodness-of-fit of the SSD curve, the steepness of
the SSD curve, the toxicity data used and the 95% confidence limit
around the HC5 estimate. This study will use both the SSD derives
MAC-EQS and RACsw;ac estimates to assess the potential risks of
saponin-rich plant extracts to the aquatic environment.
The aim of this study was, therefore, threefold: first, to test the
sensitivity of non-target aquatic species to saponin-rich plant ex-
tracts to find the more sensitive species among plants, algae,
worms, mollusks, arthropods, fish embryo and bacteria. Secondly,
to test whether the bioactivity of the saponin-rich extracts of
different botanical origin based on triterpene sapogenins was
similar, making read across between saponins possible and thirdly,
to determine the MAC-EQS and RACsw,ac for the three different
saponin-rich plant extracts and to evaluate them in relation to the
Predicted Environmental Concentration (PEC).
2. Materials and methods
2.1. The tested saponin-rich plant extracts
Three kinds of saponin crude extracts from quillaja bark, tea
seed coat, and quinoa seed coat were tested in this study, together
with one pure quillaja saponin mixture, which was included as a
reference and used as a standard for tentative quantification of
saponin content of the extracts. The sources of the three saponin
crude extracts were: quillaja extract: S7900® from Q. saponaria bark
(Sigma-Aldrich), tea extract: water extract of tea seed coat powder
from C. oleifera (NOR-ADD A/S, Denmark), quinoa extract: water
418 X. Jiang et al. / Environmental Pollution 236 (2018) 416e424

extract of quinoa seed coat powder, a Real-type variety originating
from Bolivia (Ruiz et al., 2017).
Water extraction was conducted by adding 5 g tea or quinoa
seed coat powder to 50 mL MilliQ water. After 20 min stirring
(using a 10 mm magnetic stirrer at 100 rpm/min), the supernatant
was passed through a filter paper (Whatman® quantitative filter
paper, ashless, Grade 41, Sigma-Aldrich). The crude extracts were
frozen and freeze-dried (Christ® LOC-1m, Germany) for 24 h to
obtain a storable powder that could be used throughout the study.
The reason for using water extract in this study was to mimic the
scenario in the field, where the saponin-rich plant products are
often directly applied instead of the pure saponins.
The saponins from three extracts were marked as quillaja sa-
ponins (saponins from quillaja bark), tea saponins (saponins from
tea seed coat powder) and quinoa saponins (saponins from quinoa
seed coat powder), respectively. The reference saponin (Quil-A®)
was a purified commercial Purified quillaja saponins, from Brenn-
tag Biosector A/S, containing >95% saponins, <5% residue water,
and it was marked as Quil-A in this study.
2.2. Organisms and acute toxicity tests methods
Twelve organisms were cultured and tested based on the ‘‘OECD
guideline for Testing of Chemicals’’ or “International Standard
(ISO)”, including plants, algae, worms, mollusks, arthropods, and
microbes. An overview of the endpoints and test conditions used in
the twelve tests is given in Table 1.
The concentrations of tests were firstly set at 12.5, 25, 50, 100,
and 200 mg/L (saponin extracts’ concentration). If the 50% effect
was not reached with these concentrations, lower or higher con-
centrations would be chosen in the adjusted tests. Apart from
adjusting the concentration range, smaller intervals between con-
centrations could be implemented as well in order to get a better-
described concentration-response curve in the cases where the
slope was steep. Generally, the highest concentration used in the
tests was 200 mg/L of saponins.
2.3. Semi-quantification of total saponin content of the extracts
To be able to compare the toxicity of the different plant extracts
relative to their saponin content, a semi-quantitative method was
used for determination of saponin. Quil-A was used as a standard,
assuming 100% pure saponin, and relative saponin contents of
quillaja extract, tea extract and quinoa extract were expressed as a
weight percentage relative to Quil-A, w/w % (Quil-A/sample %).
The method uses a vanillin-sulfuric acid assay and exploits the
colour reaction of oxidized triterpene saponin backbones (agly-
cones) with vanillin, which was measured as the absorbance at
480 nm using a UV/VIS spectrometer (Lambda 25, PerkinElmer)
(Cheok et al., 2014; Goel et al., 2012). For the vanillin-sulfuric acid
assay, 100 mL samples were added to a 20 mL glass vial containing
0.3 ml of vanillin reagent (8% vanillin, w/v in 99.9% ethanol) and
3 ml of 72% (v/v) sulfuric acid placed in an ice water bath. The
samples were then heated in a 60
C water bath for 10 min and
cooled in the ice water bath before measuring absorbance. In order
to reduce the effect of the matrix in the extract on the absorbance,
standard addition was used as an analytical method (Harris, 2010),
which is further described in Supplementary Materials.
2.4. Statistics
Concentration-response curves of the 12 acute toxicity tests
were performed and plotted using the open source software R,
version 3.3.2 (R Team, 2014). Data was described with a three
parameter log-logistic regression model using the drc-package
(Ritz and Streibig, 2005), assuming normal or binary distribution

Table 1
Experimental conditions of the individual toxicity tests and the guidelines followed. Climate chamber conditions are given as temperature (
C) and light:dark period in hours
(L:D). Irradiance is given in mmol/m2/s Photosynthetic Active Radiation (PAR) for the photosynthesizing organisms.

Organisms Taxonomic Age/Size Duration Endpoints Medium Climate chamber Replicates per concentration References
group conditions

Vibrio fischeri Proteobacteria e 30 min Luminescence 2% NaCl 10 ± 1
C in dark Three replicates (ISO, 2007)
inhibition


Pseudokirchneriella Chlorophyta e 48 h Relative growth Algae growth 20 ± 1 C, 24:0 h, Three replicates (OECD,2006a)
subcapitata rate medium 90
Lemna minor Tracheophyta Two weeks old 7d Relative growth K-medium 24 ± 1
C, 16:8 h, Six replicates with one frond (OECD, 2006b)
rate 100 each
Daphnia magna Arthropoda 24 h old 48 h Immobilisation M7 medium 20 ± 1
C, 16:8 h Four replicates with five (OECD, 2004)
neonates in each


Chironomus riparius Arthropoda Fourth instar 48 h Immobilisation M7 medium 20 ± 1 C, 16:8 h Four replicates with five larvae (OECD, 2011)
larvae in each
Chaoborus Arthropoda Around 10 mm 48 h Immobilisation M7 medium 20 ± 1
C, 16:8 h Four replicates with five larvae (OECD, 2011)
crystallinus in each
Gammarus pulex Arthropoda Random size 48 h Mortality Artificial 10 ± 1
C, 12:12 h Twelve replicates with one (Rasmussen
(>5 mm) freshwater gammerus in each et al., 2016)
Tubifex tubifex Annelida Random size 48 h Mortality Artificial 20 ± 1
C, in dark Four replicates with five worms (OECD, 2007)
(>10 mm) freshwater in each
Lumbriculus Annelida Random size 48 h Mortality Artificial 20 ± 1
C, 16:8 h Four replicates with five worms (OECD, 2007)
variegatus (>10 mm) freshwater in each
Lymnaea stagnalis Mollusca 12 weeks old (5 48 h Mortality Artificial 20 ± 1
C, 16:8 h Four replicates with five snails (Ducrot et al.,
e10 mm) freshwater in each 2014)
Lymnaea stagnalis Mollusca 24 h old 7d Mortality Artificial 20 ± 1
C, 16:8 h Sixteen replicates with one (Hallett et al.,
embryo freshwater embryo in each 2016)
Danio rerio embryo Chordata 24 h old 4d Mortality Artificial 26 ± 1
C in dark Five replicates with two (OECD, 2013)
freshwater embryos in each

V. fischeri was a BioTox™ kit with lyophilized bacteria, from Aboatox Oy, Finland. C. crystallinus and T. tubifex were bought from an aquarium store (Neonfisken) in Copenhagen,
Denmark. L. variegatus was originally from Department of Science and Environment, Roskilde University, Denmark. G. pulex was collected from a small stream in the north of
Copenhagen, coordinates: 55
480 5800 N 12
180 4500 E (d
m's’’). L. stagnalis were originally from Department of Biology, University of Southern Denmark. All acquired organisms
were acclimatized to growth conditions for at least two days prior to testing. D. rerio embryo was provided by Department of Veterinary Disease Biology, University of
Copenhagen, Denmark. Other organisms were kept and cultured in our own lab. All used mediums such as K-medium, M7 medium and Artificial freshwater are described in
the previous papers (Jessing et al., 2014; OECD, 2011; Rasmussen et al., 2016).
 X. Jiang et al. / Environmental Pollution 236 (2018) 416e424
of data depending on the endpoint monitored (Equation (1)):
d
y¼
(1)
1þ x b
e
where y is the measured endpoint (survival/growth rate/inhibited
rate); d corresponds to the untreated controls; e is the concentra-
tion causing 50% lethality or effect, depending on the endpoint
measured; b is proportional to the slope around e, and x is the
treatment concentration.
E(L)C50 (the concentrations causing 50% lethality/effect) was
retrieved and used for the SSDs. SSDs are generated using a log-
logistic distribution (Equation (1)) (Xu et al., 2015), using open
source software R version 3.3.2 based on the drc-package. In the
SSD model, y in Equation (1) is the proportion of affected species, x
is the saponin concentration of different E(L)C50, and e is the
saponin concentration of 50% species affected. The proportion of
affected species was calculated as shown in Equation (2) (Posthuma
et al., 2002).
Ri
yi ¼
nþ1
where yi is the proportion of affected species; Ri is the rank of
species and n is the number of E(L)C50 values used in the model.
From the SSDs, a hazardous concentration (HCp) is identified at
which a certain percentage (p) of all species is assumed to be
affected. HC5 is derived from the SSD model using the software R.
Species not reaching and E(L)C50 value at the highest saponin
concentration tested (200 mg/L) were included in the rank but not
in the SSD-fit. The EFSA Aquatic Guidance Document suggests that
a larger than toxicity value can be used in the SSD if this value is
higher than the highest non-censored value (EFSA PPR, 2013).
In principle, the PEC of saponins should be estimated using
implemented exposure scenarios and models (e.g. FOCUS). As little
guidance to the application is given and the input required to model
surface runoff and leaching is not available, we choose to use a
simple “worst-case loading” scenario based on spray drift (Equa-
tion (3)) (EFSA PPR, 2013). The biopesticide environmental con-
centration, expressed as the mass that enters the water per surface
area of water, can be obtained with knowledge of water depth, field
dose per area, and percent spray drift. In this case, we assume a
field dose on 30 cm depth water with 30% spray drift deposition.
Dsd App
c¼
h regression parameters of SSDs are shown in Table 3. Generally, all
where c is the saponin concentration in surface water (g/m3); Dsd is
the spray drift deposition as fraction of the application dose; App is
the application dose given as mass per surface area (h/m2) and h is
the surface water depth (m).
3. Results
3.1. Semi-quantification of total saponins
The total saponin content in the different plant extracts was
estimated using a standard addition method, and the standard
addition curves of quillaja extract, tea extract, quinoa extract are
shown in Supplementary Materials, Fig. S1. The resulting total
saponin contents of the quillaja, tea and quinoa extracts were
estimated to be 67.2 ± 1.2%, 57.3 ± 1.3%, and 67.9 ± 1.4% of saponins
in the dried extracts, respectively (w/w % Quil-A/sample %). The
E(L)C50 and HC5 of the different saponin extracts shown in this
study are expressed as total nominal saponin concentrations
419
calculated from the weight of the dried extracts dissolved in the
stock solutions multiplied with the saponin percentage.
3.2. Acute toxicity results
The E(L)C50 of the different saponin-rich plant extracts are
shown in Table 2. The concentration-response curves are shown in
Supplementary Materials (Figs. S2eS13). The full concentration-
response curves of the acute tests were conducted on quillaja
extract and the reference saponin, Quil-A. Some organisms were
not sensitive to saponins and 50% effect was not reached even at
200 mg/L of saponins. This was the case for Vibrio fischeri, Chiro-
nomus riparius, Chaoborus crystallinus and Gammarus pulex. Those
organisms were considered as insensitive species, and the whole
acute tests were therefore not conducted on tea extract and quinoa
extract, but the highest concentration (200 mg/L of saponins) was
tested to show no effect. Therefore, E(L)C50 of those insensitive
species is expressed as above 200 mg/L.
To the reference Quil-A, there is more than 20-fold difference in
toxicity between the most sensitive and most insensitive species,
with the E(L)C50 estimates varying from 11.9 mg/L to more than
(2)
200 mg/L. The most sensitive species was the Danio rerio (fish)
embryos. A similar pattern was shown for quillaja saponins, the
crude extract from the same plant species. Tea saponins showed
higher selectivity between species, with an approximately 100-fold
difference in species sensitivity, varying from the lowest LC50 of
~2 mg/L for the fish embryo, followed by the snail, L. stagnalis, the
two worm species Tubifex tubifex and Lumbriculus variegatus to over
200 mg/L for the arthropod C. crystallinus. Quinoa saponins seem to
be the least toxic of the three plant extracts, when given on a
relative saponin content basis, with a concentration of 49.5 mg/L
for the snail embryo test being the lowest LC50 value.
3.3. Species sensitivity distributions
The SSDs are constructed using the E(L)C50 values in Table 2 and
are shown in Fig. 2. The data of Lymnaea stagnalis is not included in
the SSD approaches since the more sensitive life-stage (L. stagnalis
embryo) is used. In addition, in order to make all dataset compa-
rable and consistent, the data of L. stagnalis is also not included in
tea saponin SSD approach, even though the snail adult is more
sensitive than the embryo in this case. Therefore, there are 11 E(L)
C50 values applied to construct the SSD giving n ¼ 11 for Equation
(2). The species that has an E(L)C50 value > 200 mg/L are considered
(3) in the rank, but data are not included in the SSD-fit. Detailed
species display the same overall pattern in terms of sensitivity. The
most sensitive species are worms, fish and snail embryos, while the
tested microbe and arthropods are relatively insensitive.
Quillaja saponins and the standard saponin Quil-A, have similar
SSD curves (Fig. 2 A, D), which was expected as they originate from
the same plant species. Due to the higher variance in species
sensitivity towards tea extracts, the tea saponins SSD curve is
relatively shallow with a b parameter of e 0.58 (Fig. 2 B). Hence, the
retrieved HC5-value is almost 10-times lower than the LC50 of the
most sensitive species and the coefficient of variation (CV) of the
parameter gets relatively high (50%, as opposed to the 34, 26 and
30% for quillaja saponins, quinoa saponins and Quil-A, respec-
tively). The quinoa saponins are the least toxic and the SSD curve is
relatively steep (1.21) according to the PPP Regulation, defining
steep as there is less than 100-fold difference between the least and
most sensitive species (EFSA PPR, 2013) (Fig. 2 C). The differences in
the species sensitivity towards the different plant extracts has the
consequence that the saponin concentration derived from quinoa
saponins protecting 95% of the species (to a 50% effect level) could
420 X. Jiang et al. / Environmental Pollution 236 (2018) 416e424

Table 2
E(L)C50 of saponin-rich plant extracts towards different organisms. All values are given ±standard error with 95% confidence limits around E(L)C50 given in brackets.

Organisms E(L)C50 (mg/L) towards different organisms

Quillaja saponins Tea saponins Quinoa saponins Quil-A

Vibrio fischeri >200 >200 >200 >200

Pseudokirchneriella subcapitata 146 ± 18.7 63.4 ± 3.91 >200 >200
(109, 183) (55.7, 71.1)
Lemna minor 103 ± 11.7 60.8 ± 4.68 >200 92.8 ± 11.9
(80.1, 126) (51.6, 70.0) (69.5, 116)
Daphnia magna 18.3 ± 1.53 15.1 ± 2.49 190 ± 60.4 22.9 ± 0.73
(15.3, 21.3) (10.2, 20.0) (71.6, 308) (21.5, 24.3)
Chironomus riparius 313 ± 79.5 104 ± 17.2 283 ± 25.3 108 ± 8.17
(157, 469) (70.3, 138) (233, 333) (92.0, 124)
Chaoborus crystallinus >200 >200 >200 >200
Gammarus pulex >200 148 ± 79.2 >200 115 ± 32.0
(0, 303) (52.3, 178)
Tubifex tubifex 47.6 ± 3.74 2.77 ± 0.25 127 ± 11.5 37.3 ± 2.18
(40.3, 54.9) (2.28, 3.26) (105, 150) (33.0, 41.6)
Lumbriculus variegatus 40.1 ± 4.01 1.95 ± 0.16 85.8 ± 8.63 39.0 ± 40.5
(32.2, 48.0) (1.64, 2.26) (68.9, 103) (0, 118)
Lymnaea stagnalis 52.1 ± 96.5 2.23 ± 0.36 65.6 ± 113 31.5 ± 28.3
(0, 241) (1.52, 2.94) (0, 287) (0, 87.0)
Lymnaea stagnalis embryo 37.0 ± 3.76 10.7 ± 1.35 49.5 ± 7.12 18.7 ± 2.81
(29.6, 44.4) (8.05, 13.3) (35.5, 63.5) (13.2, 24.2)
Danio rerio embryo 9.02 ± 1.30 1.93 ± 0.27 89.6 ± 13.6 11.9 ± 1.80
(6.47, 11.6) (1.40, 2.46) (62.9, 116) (8.37, 15.4)

All concentrations expressed in saponin concentration relative to Quil-A.

affect nearly 50% of the species if they had been derived from tea
saponins, as the HC5-value of quinoa saponins is close to HC50-
value of tea saponins (Table 3).
The median HC5 derived from the SSD model, giving the highest
concentration protecting 95% of species to a 50% effect level, are
shown in Table 3. There is approximately 100-fold difference be-
tween HC5-values of tea saponins and quinoa saponins, which are
the most and least toxic saponins in this study. However, there is
only around 8-fold difference between tea saponins and quinoa
saponins when comparing the HC50-values. This demonstrates
clearly the effect the slope of the SSD has on HC5-values, as also
mentioned above.
The HC5-values were used to calculate quality criteria for surface
waters according to the WFD and PPP Regulation, as MAC-EQS and
RACsw,ac, respectively (Table 4). MAC-EQS is derived from median
HC5-values divided by an assessment factor of 10, and RACsw,ac
using an assessment factor between 3 and 6. Here we use the
higher value of 6, as the factor between the lowest and highest E(L)
C50 values used to construct the SSD curve is less than 100 (EFSA
PPR, 2013). For tea saponins, it could, however, be discussed if the
assessment factor should be lower, due to the relatively low slope of
the curve and the consequent conservative HC5-value (Fig. 2 B). To
get an estimate of potential surface water concentrations, we used
Equation (3) assuming a field dose with 30% spray drift deposition
on 30 cm depth water and a 10% saponin content in the solid
product (10% saponins content is estimated by U.S. Environmental
Protection Agency, 2007) and no degradation or sorption (EFSA
PPR, 2013). As official recommended doses are rarely given, we
used doses given in the literature together with what they were
used for and the references. The PECs in the surface water of quillaja
saponins, tea saponins and quinoa saponins are given in Table 4.
4. Discussion
The present study confirms earlier observations of “soft skin”
organisms, such as worms, snails and fish embryos, being the most
sensitive to saponins. This is believed to be due to saponins
detergent-like structures, causing damages to cell membranes by
creating pores, vesiculation and membrane domain disruption,
ultimately killing the organisms (Augustin et al., 2012). Surpris-
ingly, the microbe, V. fischeri, is extremely insensitive to all three
kinds of saponins. This is unexpected since saponins are supposed
to have a high cytotoxicity to single-cell organisms, because of their
membrane disrupting properties, and because of their documented
antibacterial activities (Dinda et al., 2010). Study on rumen mi-
crobes, however, has shown 100% degradation of saponins within
24 h (Makkar and Becker, 1997). Hence, some microbes may be
resistant to the soap-like properties of saponins and might rather
use them as substrates. This possibility cannot be excluded to occur
for V. fischeri. Moreover, a recent study shows an EC50 of quillaja
saponin towards V. fischeri of 578 mg/L, supporting our V. fischeri
test has an EC50 > 200 mg/L (De Oliveira et al., 2017). The study also
shows that organisms in their early stage are more sensitive to
saponins than more grown individuals. This is evident for
L. stagnalis where the embryo is more vulnerable than 5e10 mm
grown individuals for three of the four extracts. The similarity in
which species are the most sensitive across the different saponin-
rich extracts illustrates that the saponins most likely have a
similar mode of action.
The activity of the different saponin extracts may depend both
on the composition of the saponins and on the non-saponin con-
tent of the extracts, such as tannins and polyphenol. Comparing
Quil-A and quillaja saponins, however, they have very similar SSD
patterns and similar HC5-values (Fig. 2, Table 3). This suggests that
it is the saponins that contribute most to the bioactivity, at least in
this species, while the non-saponin content contributes less. For
the much more toxic tea saponins, we do, however not know if
other components do also contribute to toxicity. Studies of the
molluscicidal activity of tea extract and tea saponins do, however,
show similar responses indicating that the saponins in the tea
extract are the main toxic compounds (LC50 of tea saponin is
0.66 mg/L, while it is 6.79 mg/L for the extract) (Kijprayoon et al.,
2014). The much lower toxicity of saponins from quinoa saponins
compared to the standard quillaja saponins (Quil-A) emphasizes
that the difference in the structure and composition of the saponins
from different plant species does affect their bioactivity and
resulting toxicity. We, therefore, recommend not using surrogate or
read-across data during registration of saponin-rich plant extracts
 X. Jiang et al. / Environmental Pollution 236 (2018) 416e424 421

Fig. 2. Species sensitivity distribution curves. A: quillaja saponins, B: tea saponins, C: quinoa saponins, D: reference quillaja saponins (Quil-A). All concentrations are expressed in
total saponin concentrations. The dots are input data, E(L)C50 values of different species, while the line it the regression line based on Equation (2).

Table 3
Summary of regression parameters of the SSD model for quillaja saponins, tea saponins, quinoa saponins and reference quillaja saponins (Quil-A), given together with the HC5
of different saponins. All values are given ±standard error with 95% confidence limits around E(L)C50 given in brackets.

Parameters Stressors

Quillaja saponins Tea saponins Quinoa saponins Quil-A

b (proportional to the slope around e) 0.82 ± 0.10 0.58 ± 0.06 1.21 ± 0.20 1.06 ± 0.12
(-1.02, 0.62) (-0.70, 0.46) (-1.60, 0.82) (-1.30, 0.82)
e (HC50, mg/L) 105 ± 11.0 34.4 ± 5.17 260 ± 25.7 72.0 ± 6.93
(83.4, 127) (24.3, 44.5) (210, 310) (58.4, 85.6)
HC5 (mg/L) 2.91 ± 1.00 0.22 ± 0.11 22.9 ± 5.84 4.42 ± 1.33
(0.95, 4.87) (0, 0.44) (11.5, 34.4) (1.81, 7.03)

for the use as pesticides or for other purposes, due to the strikingly attempt to make an overview of bioactivity of the different sapo-
different bioactivities. nins (Fig. 3). There are many types of aglycones in saponins
Assuming that only the aglycone is the causative agent of depending on the plant species it derives from, and often the
bioactivity, neglecting the effect of the sugar moiety, we may functional groups differ from each other as well. Although quillaja
422 X. Jiang et al. / Environmental Pollution 236 (2018) 416e424

Table 4
Predicted environmental concentration (PEC) in the aquatic environment based on doses given in the literature together with the water quality criteria MAC-EQS and RACsw,ac.
All concentrations are given in mg/L.

Saponin-rich Objectives Dosing PECa MAC- RACsw,ac References

plant products EQS

Liquid quillaja For control/suppression of pathogenic fungi or plant parasitic nematodes in 1.4e3.7 g 0.1 0.29 0.49 (U.S. Environmental Protection
extract vegetable or fruit fields liquid/m2 e0.4 Agency, 2007)
Tea seed coat Expel earthworms in turfgrass 30.0 g 3 0.02 0.04 (Potter et al., 2010)
powder/
m2
Quinoa seed For control/suppression of pathogenic fungi or bacteria in vegetable fields and 0.6e2.0 g 0.1 2.29 3.82 (Joshi et al., 2008; U.S. Environmental
coat Golden apple snail (Pomacea canaliculata) in rice water coat/m2 e0.2 Protection Agency,  2005)

g
dose 30%10% 
a m2
Assuming a field dose with 30% spray drift deposition on 30 cm depth water and a 10% saponin content in the solid product, which is PEC ¼ 30cm .

Fig. 3. Structures of different aglycones depending on the functional groups attached to the R1 and R2 positions.

saponins, tea saponins and quinoa saponins may include many
types of aglycones, the most abundant ones being reported for the
three species are shown in Fig. 3 (Güçlü-Üstündag
and Mazza,
2007). There are basically three types of very abundant aglycone
structures present, which are defined by their functional groups at
the C-4 position. Quillaic acid has an aldehyde (eCHO) group,
hederaginin has a hydroxymethyl (eCH2OH) group and oleanolic
acid has a hydrogen atom (eH) at the C-4 position (Fig. 3). While
aglycones of quillaja saponins mainly consist of quillaic acid, tea
saponins mainly consist of quillaic acid and hederagenin (Kitagawa
et al., 1998), while quinoa saponins contain approximately 50% of
hederagenin and 25% of oleanolic acid while rest are unknown
aglycones (Ruiz et al., 2017). As we observe tea saponins to be up to
10-fold more toxic to some organisms than quillaja saponins, it
suggests that the oleanolic acid type of aglycone may be relatively
inactive. Hence, the additional functional groups at C-4 of quillaic
acid and hederagenin seem to be essential to the bioactivity, no
matter if it is an aldehyde or a hydroxymethyl group. A study of
saponins from Barbarea vulgaris which reported that the type of
saponins with oleanolic acid had a less antifeedant effect on her-
bivorous insects than the ones with hederagenin supports this
observation of the importance of the aglycone structure (Augustin
et al., 2012). Other studies show that the functional group of
aldehyde at C-4 is essential for quillaja saponin bioactivities (anti-
body stimulation and induction of an immune response), as de-
rivatives modified at this function (without aldehyde group) did
not show adjuvant activity (Moses et al., 2014). Despite this, there
are studies showing that aldehyde is a more important functional
group for adjuvant effects than the other two groups (eCH2OH, eH)
(Greatrex et al., 2015), as oleanolic acid has a higher cytotoxic ac-
tivity when measured on cancer and normal cells than quillaic acid
and hederagenin (Barthomeuf et al., 2002; Gauthier et al., 2009;
Podolak et al., 2010). Hence, the relative bioactivity of aglycones
may differ depending on the endpoint measured.
The effects of the sugar moieties on bioactivity can, however, not
be ignored. Numerous studies confirm that the bioactivity of sa-
ponins is influenced both by the aglycone and the sugar moieties
(Podolak et al., 2010), especially the sugar attached at the C-3 po-
sition. The study on Barbarea vulgaris saponins reports that ole-
anolic acid and hederagenin with a sugar chain attached in C-3
position increased the antifeedant effect, compared to those that
have no sugar attached (Augustin et al., 2012). Pilot studies in our
 X. Jiang et al. / Environmental Pollution 236 (2018) 416e424
own lab have shown a large decrease in toxicity of both quillaja
saponins and tea saponins towards daphnids, when cleaving off all
sugar chains (only aglycone left) (Unpublished data). However, the
sugar moieties are extremely different and complex among quillaja
saponins, tea saponins and quinoa saponins, hence, more re-
searches are needed to explain the effect of different sugar chains
on different types of bioactivity.
The difference in bioactivity between the saponins of different
plant origin was also reflected in the tentative risk assessment done
on the plant extracts. The SSD-based water quality criteria MAC-
EQS and RACsw, ac exceeded the worst case predicted environ-
mental concentration 100-fold in the case of tea saponins, while the
quality criteria were not exceeded for quinoa saponins and were on
the borderline of being so for quillaja saponins (Table 4). Hence,
saponin-rich plant extracts may pose a risk, if not used cautiously.
More knowledge on use and fate in the environment is, however,
needed to be able to do a proper risk assessment. Likewise,
analytical techniques to measure saponins in the environment
would be an advantage, though this may be a big challenge due to
the structural diversity of plant-produced saponins.
Acknowledgements
We wish to thank Assoc. Prof. Sven-Erik Jacobsen, University of
Copenhagen, for providing Quinoa seed coat powder, NOR-ADD A/
S, Denmark, for providing tea seed powder, Assoc. Prof. Henrik
Holbech, University of Southern Denmark, for providing the culture
of Lymnaea stagnalis, Assoc. Prof. Annemette Palmqvist, University
of Roskilde, for providing the culture of Lumbriculus variegatus, and
Assist. Prof. Louise von Gersdorff Jørgensen, University of Copen-
hagen, for providing Danio rerio embryos. We also want to give our
gratitude to the financial support of University of Copenhagen and
China Scholarship Council.
Appendix A. Supplementary data
Supplementary data related to this article can be found at
https://doi.org/10.1016/j.envpol.2018.01.058.
References
Andresen, M., Cedergreen, N., 2010. Plant growth is stimulated by tea-seed extract:
a new natural growth regulator? Hortscience 45, 1848e1853.
Augustin, J.M., Drok, S., Shinoda, T., Sanmiya, K., Nielsen, J.K., Khakimov, B.,
Olsen, C.E., Hansen, E.H., Kuzina, V., Ekstrom, C.T., Hauser, T.P., Bak, S., 2012.
UDP-glycosyltransferases from the UGT73C subfamily in Barbarea vulgaris ca-
talyse sapogenin 3-O-glucosylation in saponin-mediated insect resistance.
Plant Physiol. 160, 1881e1895. https://doi.org/10.1104/pp.112.202747.
Barthomeuf, C., Debiton, E., Mshvildadze, V., Kemertelidze, E., Balansard, G., 2002.
In vitro activity of hederacolchisid A1 compared with other saponins from
hedera colchica against proliferation of human carcinoma and melanoma cells.
Planta Med. 68, 672e675. https://doi.org/10.1055/s-2002-33807.
Cannon, J.G., Burton, R.A., Wood, S.G., Owen, N.L., 2004. Naturally occurring fish
poisons from plants. J. Chem. Educ. 81, 1457. https://doi.org/10.1021/
ed081p1457.
Cheok, C.Y., Salman, H.A.K., Sulaiman, R., 2014. Extraction and quantification of
saponins: a review. Food Res. Int. 59, 16e40. https://doi.org/10.1016/
j.foodres.2014.01.057.
De Oliveira, D.W.F., Cara, A.B., Lechuga-Villena, M., García-Rom an, M., Melo, V.M.M.,
Gonçalves, L.R.B., Vaz, D.A., 2017. Aquatic toxicity and biodegradability of a
surfactant produced by Bacillus subtilis ICA56. J. Environ. Sci. Heal. - Part A
Toxic/Hazardous Subst. Environ. Eng 52, 174e181. https://doi.org/10.1080/
10934529.2016.1240491.
Dinda, B., Debnath, S., Mohanta, B.C., Harigaya, Y., 2010. Naturally occurring tri-
terpenoid saponins. Chem. Biodivers. 7, 2327e2580. https://doi.org/10.1002/
cbdv.200800070.
Ducrot, V., Askem, C., Azam, D., Brettschneider, D., Brown, R., Charles, S., Coke, M.,
Collinet, M., Delignette-Muller, M.L., Forfait-Dubuc, C., Holbech, H.,
Hutchinson, T., Jach, A., Kinnberg, K.L., Lacoste, C., Le Page, G., Matthiessen, P.,
Oehlmann, J., Rice, L., Roberts, E., Ruppert, K., Davis, J.E., Veauvy, C., Weltje, L.,
Wortham, R., Lagadic, L., 2014. Development and validation of an OECD
reproductive toxicity test guideline with the pond snail Lymnaea stagnalis
423
(Mollusca, Gastropoda). Regul. Toxicol. Pharmacol. 70, 605e614. https://doi.org/
10.1016/j.yrtph.2014.09.004.
EFSA PPR, 2013. Guidance on tiered risk assessment for plant protection products
for aquatic organisms in edge-of-field surface waters. EFSA J 11, 3290. https://
doi.org/10.2903/j.efsa.2013.3290.
European Commission, 2011. Common implementation strategy for the water
framework directive (2000/60/EC). Guidance document No. 27. Technical
Guidance for Deriving Environmental Quality Standards. https://doi.org/
10.2779/43816.
Fernandez-Tejada, A., Chea, E.K., George, C., Pillarsetty, N., Gardner, J.R.,
Livingston, P.O., Ragupathi, G., Lewis, J.S., Tan, D.S., Gin, D.Y., 2014. Development
of a minimal saponin vaccine adjuvant based on QS-21. Nat. Chem. 6, 635e643.
https://doi.org/10.1038/nchem.1963.
Gauthier, C., Legault, J., Girard-Lalancette, K., Mshvildadze, V., Pichette, A., 2009.
Haemolytic activity, cytotoxicity and membrane cell permeabilization of semi-
synthetic and natural lupane- and oleanane-type saponins. Bioorg. Med. Chem.
17, 2002e2008. https://doi.org/10.1016/j.bmc.2009.01.022.
Goel, N., Sirohi, S.K., Dwivedi, J., 2012. Estimation of total saponins and evaluate
their effect on in vitro methanogenesis and rumen fermentation pattern in
wheat straw based diet. J. Adv. Vet. Res. 2, 120e126.
Greatrex, B.W., Daines, A.M., Hook, S., Lenz, D.H., McBurney, W., Rades, T.,
Rendle, P.M., 2015. Synthesis, formulation, and adjuvanticity of mono-
desmosidic saponins with olenanolic acid, hederagenin and gypsogenin agly-
cones, and some C-28 ester derivatives. ChemistryOpen 740e755. https://
doi.org/10.1002/open.201500149.
Güçlü-Üstündag € Mazza, G., 2007. Saponins: properties, applications and pro-

, O.,
cessing. Crit. Rev. Food Sci. Nutr. 47, 231e258. https://doi.org/10.1080/
10408390600698197.
Hallett, K.C., Atfield, A., Comber, S., Hutchinson, T.H., 2016. Developmental toxicity
of metaldehyde in the embryos of Lymnaea stagnalis (Gastropoda: pulmonata)
co-exposed to the synergist piperonyl butoxide. Sci. Total Environ. 543, 37e43.
https://doi.org/10.1016/j.scitotenv.2015.11.040.
Harris, D.C., 2010. Quantitative Chemical Analysis, eighth ed. W. H. Freeman and
Company, New York.
ISO (International Organization for Standardization), 2007. Water Quality-
Determination of the Inhibitory Effect of Water Samples on Te Light Emission
of vibrio Fischeri (Luminescent Bacteria Test)- Part 3: Method Using Freeze-
dried Bacteria. ISO 11348e32007 1e21.
Jessing, K.K., Andresen, M., Cedergreen, N., 2014. Temperature-dependent toxicity
of artemisinin toward the macrophyte Lemna minor and the algae Pseudo-
kirchneriella subcapitata. Water, Air, Soil Pollut. 225 https://doi.org/10.1007/
s11270-014-2010-1.
Joshi, R.C., San Martín, R., Saez-Navarrete, C., Alarcon, J., Sainz, J., Antolin, M.M.,
Martin, A.R., Sebastian, L.S., 2008. Efficacy of quinoa (Chenopodium quinoa)
saponins against golden apple snail (Pomacea canaliculata) in the Philippines
under laboratory conditions. Crop Protect. 27, 553e557. https://doi.org/10.1016/
j.cropro.2007.08.010.
Kijprayoon, S., Tolieng, V., Petsom, A., Chaicharoenpong, C., 2014. Molluscicidal
activity of Camellia oleifera seed meal. Sci. Asia 40, 393e399. https://doi.org/
10.2306/scienceasia1513-1874.2014.40.393.
Kitagawa, I., Hori, K., Motozawa, T., Murakami, T., Yoshikawa, M., 1998. Structures of
new acylated oleanene-type triterpene oligoglycosides, theasaponins E1 and
E2, from the seeds of tea plant, Camellia sinensis (L.) O. KUNTZE. Chem. Pharm.
Bull. (Tokyo) 46, 1901e1906. https://doi.org/10.1248/cpb.46.1901.
Makkar, H.P., Becker, K., 1997. Degradation of quillaja saponins by mixed culture of
rumen microbes. Lett. Appl. Microbiol. 25, 243e245. https://doi.org/10.1046/
j.1472-765X.1997.00207.x.
Moses, T., Papadopoulou, K.K., Osbourn, A., 2014. Metabolic and functional diversity
of saponins, biosynthetic intermediates and semi-synthetic derivatives. Crit.
Rev. Biochem. Mol. Biol. 49, 439e462. https://doi.org/10.3109/
10409238.2014.953628.
OECD, 2013. Test No. 236: Fish Embryo Acute Toxicity (FET) Test, OECD, OECD
Guidelines for the Testing of Chemicals, Section 2. OECD Publishing. 10.1787/
9789264203709-en.
OECD, 2011. Test No. 235: Chironomus sp., Acute Immobilisation Test, OECD, OECD
Guidelines for the Testing of Chemicals, Section 2. OECD Publishing. 10.1787/
9789264122383-en.
OECD, 2007. Test No. 225: Sediment-water Lumbriculus Toxicity Test Using Spiked
Sediment, OECD, OECD Guidelines for the Testing of Chemicals, Section 2. OECD
Publishing. 10.1787/9789264067356-en.
OECD, 2006a. Test No. 201: Alga, Growth Inhibition Test, OECD, OECD Guidelines for
the Testing of Chemicals, Section 2: Effects on Biotic Systems. OECD Publishing.
10.1787/9789264069923-en.
OECD, 2006b. Test No. 221: Lemna Sp. Growth Inhibition Test, OECD, OECD
Guidelines for the Testing of Chemicals, Section 2. OECD Publishing. 10.1787/
9789264016194-en.
OECD, 2004. Test No. 202: Daphnia Sp. Acute Immobilisation Test, OECD, OECD
Guidelines for the Testing of Chemicals, Section 2. OECD Publishing. 10.1787/
9789264069947-en.
Podolak, I., Galanty, A., Sobolewska, D., 2010. Saponins as cytotoxic agents: a review.
Phytochem. Rev. 9, 425e474. https://doi.org/10.1007/s11101-010-9183-z.
Posthuma II, Leo, Glenn, W., Suter Traas, T.P., 2002. Species Sensitivity Distributions
in Ecotoxicology. Lewis Publishers, Boca Raton, Florida.
Potter, D.A., Redmond, C.T., Meepagala, K.M., Williams, D.W., 2010. Managing
earthworm casts (Oligochaeta: lumbricidae) in turfgrass using a natural
424 X. Jiang et al. / Environmental Pollution 236 (2018) 416e424
byproduct of tea oil (Camellia sp.) manufacture. Pest Manag. Sci. 66, 439e446.
https://doi.org/10.1002/ps.1896.
Rasmussen, J.J., Cedergreen, N., Kronvang, B., Andersen, M.B.B., Nørum, U.,
Kretschmann, A., Strobel, B.W., Hansen, H.C.B., 2016. Suspended particles only
marginally reduce pyrethroid toxicity to the freshwater invertebrate Gammarus
pulex (L.) during pulse exposure. Ecotoxicology 25, 510e520. https://doi.org/
10.1007/s10646-016-1609-y.
Ritz, C., Streibig, J.C., 2005. Bioassay analysis using R. J. Stat. Software 12, 1e22.
https://doi.org/10.18637/jss.v012.i05.
Rubio-Moraga, A.,  Gerwig, G.J., Castro-Díaz, N., Jimeno, M.L., Escribano, J.,
Ferna ndez, J.-A., Kamerling, J.P., 2011. Triterpenoid saponins from corms of
Crocus sativus: Localization, extraction and characterization. Ind. Crops Prod 34,
1401e1409. https://doi.org/10.1016/j.indcrop.2011.04.013.
Ruiz, K.B., Khakimov, B., Engelsen, S.B., Bak, S., Biondi, S., Jacobsen, S.-E., 2017.
Quinoa seed coats as an expanding and sustainable source of bioactive com-
pounds: an investigation of genotypic diversity in saponin profiles. Ind. Crop.
Prod. 104, 156e163. https://doi.org/10.1016/j.indcrop.2017.04.007.
San Martín, R., Ndjoko, K., Hostettmann, K., San Martin, R., Ndjoko, K.,
Hostettmann, K., San Martín, R., Ndjoko, K., Hostettmann, K., 2008. Novel
molluscicide against Pomacea canaliculata based on quinoa (Chenopodium
quinoa) saponins. Crop Protect. 27, 310e319. https://doi.org/10.1016/
j.cropro.2007.03.015.
Szakiel, A., Ruszkowski, D., Janiszowska, W., 2005. Saponins in Calendula officinalis
L. e structure, biosynthesis, transport and biological activity. Phytochem. Rev. 4,
151e158. https://doi.org/10.1007/s11101-005-4053-9.
Team, R.D.C., 2014. R: a Language and Environment for Statistical Computing.
Version: 0.98.1091. Available at: http://www.R-project.org.
U.S. Environmental Protection Agency, 2017. What Are Biopesticides? [WWW
Document]. URL. https://www.epa.gov/pesticides/biopesticides (accessed
4.16.17).
U.S. Environmental Protection Agency, 2007. Saponins of quillaja saponaria (PC code
097095). Biopestic. Regist. Action Doc 1e33.
U.S. Environmental Protection Agency, 2005. Saponins of Chenopodium quinoa (PC
code 097094). Biopestic. Regist. Action Doc 1e39.
Xu, F., Li, Y., Wang, Y., He, W., Kong, X., Qin, N., Liu, W., Wu, W., Jorgensen, S.E., 2015.
Key issues for the development and application of the species sensitivity dis-
tribution (SSD) model for ecological risk assessment. Ecol. Indicat. 54, 227e237.
https://doi.org/10.1016/j.ecolind.2015.02.001.