Plant Science 287 (2019) 110167

Contents lists available at ScienceDirect

Plant Science
journal homepage: www.elsevier.com/locate/plantsci

Identification of the aquaporin gene family in Cannabis sativa and evidence T

for the accumulation of silicon in its tissues
Gea Guerrieroa, , Rupesh Deshmukhb, Humira Sonahb, Kjell Sergeanta, Jean-Francois Hausmana,
⁎

Esther Lentzena, Nathalie Vallea, Khawar Sohail Siddiquic,1, Christopher Exleyd

a
Research and Innovation Department, Luxembourg Institute of Science and Technology, 5 Avenue des Hauts-Fourneaux, L-4362, Esch/Alzette, Luxembourg
b
National Agri-Food Biotechnology Institute (NABI), Sector-81 (Knowledge City), P.O. Manauli, S.A.S. Nagar, Mohali, 140306, Punjab, India
c
Life Sciences Department, King Fahd University of Petroleum and Minerals (KFUPM), 31261, Dhahran, Saudi Arabia
d
The Birchall Centre, Lennard Jones Laboratories, Keele University, Keele, Staffordshire, ST5 5BG, UK

ARTICLE INFO ABSTRACT

Keywords: Cannabis sativa is an economically important crop providing bast fibres for the textile and biocomposite sector.
Cannabis sativa Length is a fundamental characteristic determining the properties of bast fibres. Aquaporins, channel-forming
Aquaporin proteins facilitating the passage of water, urea, as well as elements such as boron and silicon, are known to play a
Silicon role in the control of fibre length in other species, like cotton. By mining the available genome, we here identify,
RT-qPCR
for the first time, the aquaporin gene family of C. sativa. The analysis of published RNA-Seq data and targeted
NanoSIMS
qPCR on a textile variety reveal an organ-specific expression of aquaporin genes. Computational analyses, in-
Biogenic silica
cluding homology-based search, phylogeny and protein modelling, identify two NOD26-like intrinsic proteins
harbouring the Gly-Ser-Gly-Arg (GSGR) aromatic/Arg selectivity filter and 108 amino acid NPA (Asn-Pro-Ala)
spacing, features reported to be associated with silicon permeability. SIMS nano-analysis and silica extraction
coupled to fluorescence microscopy performed on hemp plantlets reveal the presence of silicon in the bast fibres
of the hypocotyl and in leaves. The accumulation of silica in the distal cell walls of bast fibres and in the basal
cells of leaf trichomes is indicative of a mechanical role.

1. Introduction is thought to be regulated by aquaporins [8]. Aquaporins are indeed

Cannabis sativa is an annual herbaceous plant that has witnessed a
renewed interest because of its multiple applications in different in-
dustrial sectors. It is a source of woody and cellulosic fibres; the former
(referred to as hurds or shivs) are used to manufacture a concrete-like
material known as Hempcrete® and the latter are employed in bio-
composites to substitute for glass fibres [1]. Bast fibres are extraxylary
cells that mechanically support the phloem, they are very long and
contain little lignin [2]. The mechanisms involved in bast fibre for-
mation are still not fully unveiled [3], however recent high-throughput
studies have contributed to identify key genes involved in their se-
quential developmental phases and cell wall thickening [4,5].
Bast fibres are very long cells which can attain a length of > 50 mm
for primary bast fibres [2]. The mechanism of bast fibre elongation is of
the diffuse anisotropic type and during growth they invade the middle
lamellas of the neighboring cells (intrusive mechanism) [6,7]. During
the phase of intrusive growth, turgor pressure ensures extension which
membrane proteins involved in the control of plant hydraulic para-
meters at the tissue and cell levels [9]. The role of aquaporins in cell
elongation is confirmed by studies on cotton fibres, where PIP2 genes
(encoding plasma membrane intrinsic proteins belonging to the aqua-
porin family) are specifically induced during fibre development [10].
Recently, aquaporins have been identified and classified in the fibre
crop flax, where genes coding for TIPs (tonoplast intrinsic proteins) and
PIPs are differentially expressed in inner and outer stem tissues [11].
The study of aquaporins in fibre crops is important to understand how
turgor pressure is regulated during the phase of elongation and cell wall
thickening. The stem of fibre crops is, in this respect, a valuable model
system, since it shows a gradient of bast fibre developmental stages
going from the top to the bottom. At the top, the fibres elongate mas-
sively, while after a specific point known as snap point [12] they start
to thicken and this translates into a change in the overall mechanical
properties of the stem.
Specific aquaporins belonging to the NIP-III group have been

⁎
Corresponding author.
E-mail address: gea.guerriero@list.lu (G. Guerriero).
1
Present address: 8 Archibald Crescent, Rosemeadow, NSW2560, Australia.

https://doi.org/10.1016/j.plantsci.2019.110167
Received 26 April 2019; Received in revised form 3 June 2019; Accepted 7 June 2019
Available online 11 June 2019
0168-9452/ © 2019 Elsevier B.V. All rights reserved.
G. Guerriero, et al. Plant Science 287 (2019) 110167

Table 1
Details of the primers used in this study.
Name Sequence Amplicon size (bp) Efficiency (%) Tm (°C) R2
(5'→3')

CsaLsi2-3 Fwd ATTCCCAGCACTTTGTGGAC 74 103.6 79.8 0.981

CsaLsi2-3 Rev GCTAGAACAGCTATCCCACCAG
CsaLsi2-2 Fwd GGCTTCAAATGTCCCAACTG 143 90.8 84.7 0.97
CsaLsi2-2 Rev ATCCCAGCAATGACAGGTTC
CsaLsi2-1 Fwd ATGTTCCACCCCATCCTTTC 77 91.6 83.8 0.993
CsaLsi2-1 Rev ATTGCCGATAGGAGTTGCAG
CsaNIP1-2 Fwd CTCGGTGGTGGTAAATTTGG 124 110 79.8 0.987
CsaNIP1-2 Rev GATTGAAATGGGCACCAGAG
CsaNIP3-2 Fwd TGCTGCCCTAAAACACTTCC 112 97.2 81.8 0.99
CsaNIP3-2 Rev CGTCAAGGGATGAAACACTG
CsaNIP5-1 Fwd CTGCTTCACCGATTTTACCG 75 107.7 80.3 0.993
CsaNIP5-1 Rev AAAGTTCCCACGAACTCTGC

characterized as channels mediating the entry of silicon in plants [13]. to demonstrate the already proven intimate association of biosilicifi-
With respect to silicon, plants can be accumulators, excluders or in- cation with plant cell development, we examine silica skeletons from
termediate-type; the monocot rice is an emblematic example of a silicon other hemp tissues, notably leaves, which are organs characterized by
accumulator, while members of the nightshade family (tomato, to- heterogeneous cell populations (epidermal cells, stomata, trichomes).
bacco) are excluders [14,15]. In rice, a cooperative silicon uptake Fluorescence microscope observations indeed confirm the presence of
system composed of a NIP-III member (Lsi1) and an efflux transporter fine silicified cellular structures which are linked to different develop-
(Lsi2) has been described [16], together with an additional NIP-III mental stages of leaf cells.
channel, Lsi6, responsible for silicon distribution in rice shoots [17].
A recent study has shown that in NIP-III aquaporins a specific spa-
2. Materials and methods
cing of 108 amino acids between the NPA (Asn-Pro-Ala) domains is
conserved among channels reported to be permeable to silicon [18].
2.1. Genome-wide identification of aquaporins
Among fibre crops, C. sativa is known to silicify. Hemp trichomes
were shown to accumulate silicon (as amorphous opaline silica) [19]
Transcriptome assemblies of Purple Kush and Finola were retrieved
and the hurds are known to contain silica, a feature which makes them
from the Cannabis Genome Browser (http://genome.ccbr.utoronto.ca/
associate well with lime-based binders via the pozzolanic reaction (i.e.
downloads.html). A local transcriptome database was developed using
formation of calcium silicate with cementing properties). The associa-
NCBI local BLAST implemented in the BioEdit software tool. A set of
tion of silica with hemp bast fibres may have an important impact in
terms of plant physiology and industrial applications. Since hemp bast
Table 2
fibres are arranged in bundles interspersed in the cortex around the
Aquaporins in the Purple Kush (PK) and Finola (FN) genome assemblies and
central vascular tissues, a ring strengthening the whole stem forms,
proposed final nomenclature.
thereby increasing the overall resistance to mechanical stress. A clear
evidence of the beneficial role of silicon in plants is via its association Purple Kush AQPs Finola AQPs Proposed final nomenclature
with the cell wall [20].
Name Transcript Name Transcript Name
Concerning the cell wall aspect, studies in the literature have
pointed to hemicelluloses [21,22], pectins [23] and cellulose [24] as PKPIP1-1 PK28701.1 FNPIP1-1 FN09398.1 CsaPIP1-1
macromolecules favoring silica precipitation. However, robust data PKPIP1-2 PK19617.1 FNPIP1-2 FN03298.1 CsaPIP1-2
PKPIP1-3 PK28206.1 FNPIP1-3 FN30203.1 CsaPIP1-3
concerning promotion of silica deposition both in vivo and in vitro are
PKPIP1-4 PK28630.1 FNPIP1-4 FN20092.1 CsaPIP1-4
only available for the amorphous macromolecule callose [25]. Callose PKPIP2-1 PK01882.1 FNPIP2-1 FN00769.1 CsaPIP2-1
is an ideal template for silicification, since its amorphous structure PKPIP2-2 PK18043.1 FNPIP2-2 FN36356.1 CsaPIP2-2
provides microenvironments favoring the condensation of silicic acid PKPIP2-3 PK10459.1 FNPIP2-3 FN28470.1 CsaPIP2-3
and the precipitation of silica [26]. PKPIP2-4 PK02995.1 FNPIP2-4 FN07100.1 CsaPIP2-4
FNPIP2-5 FN10254.1 CsaPIP2-5
When silica is extracted from plant organs/tissues and is visualized PKTIP1-1 PK28735.1 FNTIP1-1 FN34985.1 CsaTIP1-1
under the microscope with the fluor PDMPO, very detailed structures PKTIP1-2 PK08469.1 FNTIP1-2 FN17866.1 CsaTIP1-2
can be discerned: stomata, trichomes, epidermal papillae-like projec- FNTIP5-1 FN12886.1 CsaTIP5-1
tions [25]. These results suggest that the various plant cell structures PKTIP2-2 PK11931.1 FNTIP2-2 FN03413.1 CsaTIP2-2
PKTIP2-1 PK10655.1 FNTIP2-1 FN27560.1 CsaTIP2-1
(i.e. stomata, trichomes, cuticular projections) have different propen-
PKTIP4-1 PK19836.1 FNTIP4-1 FN11658.1 CsaTIP4-1
sities to silicify and this propensity must be established by the local PKNIP1-1 PK09307.1 FNNIP1-1 FN18987.1 CsaNIP1-1
abundance and structural properties of the macromolecules (i.e. PKNIP1-2 PK06869.1 FNNIP1-2 FN33566.1 CsaNIP1-2
amorphous cell wall components, like callose) templating silica pre- FNNIP1-3 FN26914.1 CsaNIP1-3
cipitation. PKNIP2-1 PK09456.1 FNNIP2-1 FN14156.1 CsaNIP2-1
PKNIP2-2 PK05483.1 FNNIP2-2 FN31611.1 CsaNIP2-2
We here analyze the aquaporin gene family of hemp and identify
PKNIP3-1 PK21844.1 FNNIP3-1 FN02974.1 CsaNIP3-1
their expression pattern in different organs, as well as during the stages PKNIP5-1 PK29364.1 FNNIP5-1 FN18289.1 CsaNIP5-1
of bast fibre elongation and cell wall thickening. Two systems are used, PKSIP1-1 PK16813.1 FNSIP1-1 FN36397.1 CsaSIP1-1
which have been previously studied by us, i.e. the hypocotyl and adult PKSIP2-1 PK29764.1 FNSIP2-1 FN21234.1 CsaSIP2-1
PKXIP1-1 PK24453.1 FNXIP1-1 FN35077.1 CsaXIP1-1
plants [5,27]. Given the fundamental and application-oriented interest
PKTIP2-3 PK23141.1 CsaTIP2-3
of studying silicon accumulation in C. sativa, we identify two putative PKNIP3-2 PK23987.1 CsaNIP3-2
NIP-III family members that contain the GSGR aromatic/Arg (ar/R) PKNIP5-2 PK11676.1 CsaNIP5-2
selectivity filter reported in silicon channels. Subsequently, we use PKNIP5-3 PK22011.1 CsaNIP5-3
NanoSIMS to map the metalloid distribution in the bast fibres. Finally, PKXIP1-2 PK28369.1 CsaXIP1-2

2
G. Guerriero, et al.
141 known aquaporins from soybean, Arabidopsis thaliana and rice were
used as a query sequence to perform BLAST searches against the C.
sativa transcriptome database. To claim a significant match, E-value
cut-off < 0.00001 and bitscore > 100 were used. Non-redundant top
hits were used for the rest of the analysis. Similarly, to look for hemp
orthologs of OsLsi2, the rice gene was used as query sequence. Flax
aquaporin orthologs were used to construct a phylogenetic tree.
ClustalW was used for protein sequence alignment and the maximum
likelihood method (MLM) for phylogenetic tree construction. The no-
menclature of aquaporins identified in the C. sativa genome was as-
signed based on the phylogenetic tree and the homology with known
aquaporins.
2.2. Protein tertiary structure and pore characterization
The 3D homology models of both hemp aquaporins were generated
with the I-TASSER Suite (http://zhanglab.ccmb.med.umich.edu/I-
TASSER/) [28] utilizing LOMETS, SPICKER, and TM-align. The
models were constrained on A. thaliana aquaporin AtTIP2;1 (PDB 5i32),
they were then refined using REMO by optimizing the backbone hy-
drogen-bonding networks and FG-MD by removing the steric clashes
and improving the torsion angles. The quality of the models was
checked by RAMPAGE [29]. The pore analysis was carried out with the
ChExVis online tool (available at http://vgl.serc.iisc.ernet.in/chexvis/),
using default values [30]. Tetramers of C. sativa NIP2-1 aquaporins
were generated using GalaxyHomomer [31].
2.3. Plant growth, RNA extraction and real-time PCR
Plants of textile hemp (Santhica 27) were grown under controlled
conditions (16 h light, 8 h dark) as previously reported [27]. For SIMS
nano-analysis and silica extraction, hemp plantlets were grown for 6
days, then sodium metasilicate (Na2SiO3) 1 mM (pH 5.6) was added
twice a week (controls were watered with 2 mM NaCl to provide the
same concentration of sodium as the + silicon treatment), until the
3
Plant Science 287 (2019) 110167
plants were 15 days old. Plant tissue sampling, RNA extraction, quality
control/quantification and retrotranscription were performed as pre-
viously described [27]. The primers used for RT-qPCR are indicated in
Table 1, while those used for reference gene amplifications have been
previously described [27]. Tubulin and TIP41 were used as reference
genes. A one-way ANOVA with a Tukey’s post-hoc test was performed
with IBM SPSS Statistics v19, after converting the NRQs (Normalized
Relative Quantities) to log2 values.
2.4. High resolution SIMS imaging
Secondary Ion Mass Spectrometry (SIMS) imaging was performed
on a CAMECA NanoSIMS50 instrument. A Cs+ primary ion beam of
about 100 nm with an impact energy of 16 keV and a current of 1.5 pA
was used for studying the distribution of 12C14N−, 28Si− and 31P− si-
multaneously in multicollection mode. The instrument was tuned for a
mass resolution (Mass/ΔMass) of about 5000. Images of (40 × 40) μm2
were recorded in a pixel format of 256 × 256 image points and a
counting time of 30 ms/pixel. Particular attention was paid to sample
preparation in order to remove water from biological samples, whilst
maintaining their morphological and physiochemical properties as
close as possible to the in vivo state. The hemp samples were prepared
by cryo fixation in liquid propane on a Leica EM CPC and freeze-dried
with a Leica CFD System. Freeze-drying has the advantage that there is
no interaction between cellular components and a liquid organic sol-
vent. The latter may indeed cause dissolution or extraction of compo-
nents, which would induce a change in the physicochemical properties
of the sample [32].
After the freeze-drying step, samples were embedded in acrylic resin
(Unicryl) at - 20 °C and polymerized under ultraviolet light. The em-
bedded samples were adjusted crosswise by ultramicrotomy (Leica
Ultracut UCT) and a thin gold layer (˜10 nm) (Baltec MED020) was
deposited at their surface in order to avoid charging effects during
measurements.
Fig. 1. Maximum likelihood phylogenetic tree of C. sativa and flax
aquaporins. Bootstraps = 1000 (circles indicate bootstraps > 0.7;
the bigger the circle, the higher the bootstrap value). The
Physcomitrella patens glycerol uptake facilitator (NCBI Reference
Sequence: XP_024396316.1) was used as outgroup to root the
tree. Different colours indicate different aquaporin families (SIPs
red, XIPs lilac, NIPs dark blue, TIPs dark green, PIPs pink).
G. Guerriero, et al. Plant Science 287 (2019) 110167

2.5. Extraction of silica skeletons from hemp leaves and PDMPO
fluorescence microscopy
Hemp leaves were collected from plantlets that were 15 days old
and dried until constant weight at 37 °C for 3 days. Sample processing
for silica extraction and visualization with PDMPO at the fluorescence
microscope were carried out as previously described [20].
3. Results and discussion
3.1. Identification of aquaporins in the C. sativa genome
We identified Cannabis aquaporins in different organs by mining
previously published datasets [33,34]. A total of 27 aquaporin ortho-
logs were identified in the C. sativa genome assembly of Purple Kush.
The set of C. sativa aquaporins includes 9 NIPs, 8 PIPs, 6 TIPs, 2 SIPs,

Fig. 2. Characteristic features of the pore predicted using the homology-based protein tertiary structure of C. sativa NIP2-1 aquaporins and alignment with known
silicon channels from rice, soybean and poplar. A) Transmembrane pore of PKNIP2-1 (PK09456.1) with heatmaps of flexibility and hydrophobicity; B) tetramer of
PKNIP2-1; C) transmembrane pore of FNNIP2-1 (FN14156.1) with heatmaps of flexibility and hydrophobicity; D) tetramer of FNNIP2-1; E, F) pores through the
protein ribbon structures; G) Protein sequence alignment showing conserved G-S-G-R Ar/R selectivity filter (denoted with green letters on the top of alignment) and
NPA domain (denoted with black letters).

4
G. Guerriero, et al.
and 2 XIPs (Table 2; Supplementary Table S1). The genome assembly of
an industrial variety, Finola, was also explored to identify aquaporins.
A total of 25 aquaporins (without counting the duplicates; Supple-
mentary Table S2), including 7 NIP, 9 PIP, 6 TIP, 2 SIP and 1 XIP, were
identified (Table 2).
All aquaporins from Finola show homologs in Purple Kush, except
for FNPIP2-5, FNTIP5-1 and FNNIP1-3. The total non-redundant number
of aquaporins identified in these two assemblies of C. sativa is therefore
30 (Table 2). The proposed nomenclature of Table 2 will be used
hereafter to designate the corresponding genes/proteins.
Since a recent publication analysed the aquaporins of another bast
fibre-producing crop, i.e. flax (Linum usitatissimum L.) [11], a phylo-
genetic analysis of the C. sativa aquaporins and L. usitatissimum aqua-
porins was performed (Fig. 1). The combined phylogenetic tree shows
high level of similarity among the aquaporin from these two bast fibre-
producing plants. A characteristic clustering of aquaporins into five
distinct subfamilies like NIP, PIP, TIP (tonoplast intrinsic protein), SIP
(small intrinsic protein) and XIP (uncharacterized intrinsic protein) was
observed.
Based on the ar/R selectivity filters in aquaporins and phylogenetic
position, a corresponding NIP2-1 gene ortholog to the rice silicon
channel OsLsi1, which we named CsaNIP2-1 (Table 2), was identified in
Purple Kush (PK09456.1) and Finola (FN14156.1) (Table 2). The
CsaNIP2-1 protein has all the signature attributes like two conserved
NPA motifs, G-S-G-R ar/R selectivity filters and 108 amino acid spacing
between the NPA motifs reported to be associated with silicon perme-
ability of NIP-IIIs.
Three Cannabis genes corresponding to the rice efflux transporter
OsLsi2 [35] were also retrieved, i.e. CsaLsi2-1 (PK18630.1/FN08935.1),
CsaLsi2-2 (PK08860.1/FN07382.1) and CsaLsi2-3 (PK00413.1/
FN13891.1).
3.2. Protein tertiary structure of C. sativa NIP2-1 aquaporins
Homology-based tertiary structures developed for FN14156.1 (C-
score -0.15 and TM-score 0.12 ± 0.69) and PK09456.1 (C-score value
is -2.46 and the TM-score 0.43 ± 0.14) were used to analyze pore
morphology (indicated respectively FNNIP2-1 and PKNIP2-1).
Transmembrane pore analyses, tetramers of both C. sativa NIP2-1 and
5
Plant Science 287 (2019) 110167
pore structure through the protein ribbon structures are shown in
Fig. 2A–F. The minimum (bottleneck) radius determined by ChExVis
ranges from 1.770-1.644 Å which are smaller than the estimated max-
imum radius of 4 Å for silicic acid (based on quantum calculations
taking into account the electron density) [36]. FNNIP2-1 shows the
largest radius (1.770 Å) for the transmembrane pore. Proteins are dy-
namic and “breathe” in vivo, as well exemplified by the dynamics of
protein binding pockets [37]; therefore, the pore diameter may increase
to accommodate silicic acid passage. A heat map of the pores based on
average flexibility of residues shows indeed that in areas corresponding
to the bottleneck, residues with higher flexibility (Fig. 2A and C, top
half) and less hydrophobicity residues (Fig. 2A and C, bottom half) line
the pores. For example, PKNIP2-1 has a minimum radius of 1.644 Å, but
shows very high flexibility with concomitant low hydrophobicity
(Fig. 2A). Based on these in silico studies, we cannot conclude that the
modelled channels mediate silicon passage without functional proof.
CsaNIP2-1 (protein) has the typical GSGR ar/R selectivity filter
found in all plant aquaporins that have been suggested/shown to be
permeable to silicic acid. In addition, a 108 amino acids spacing be-
tween NPA domains was also observed (Fig. 2G), which was shown to
be a feature shared by all the aquaporins currently reported to be
permeable to silicon [18].
3.3. Gene expression dynamics of C. sativa aquaporins
As a first step towards the determination of the aquaporin expres-
sion in different C. sativa tissues, publicly available transcriptomic da-
tasets were mined [33,34]. The heatmap hierarchical clustering of the
C. sativa aquaporin genes shows heterogeneous expression patterns in
the different tissues analyzed (Fig. 3). The CsaNIP2-2 and CsaTIP1-2
show a preferential expression in roots, others, like CsaXIP1-2 are al-
most exclusively expressed in the aerial organs. The cluster of genes
composed of CsaSIP1-1, CsaPIP2-4, CsaPIP1-3, CsaPIP1-2, CsaTIP1-1 is
expressed constitutively in all the tissues considered. CsaNIP2-1, the
aquaporin putatively involved in silicon channeling, has slightly higher
expression in aerial organs, namely the flowers and the shoot.
We also looked for the expression pattern of the hemp Lsi2 ortholog,
since this belongs to a cooperative system of uptake in rice [16]. The
three Lsi2 genes cluster in different branches; among them, CsaLsi2-2 is
Fig. 3. Heatmap and hierarchical clustering of the ex-
pression values relative to the C. sativa (Purple Kush)
aquaporins and Lsi2 orthologs. The clustering was per-
formed with Pearson distance metrics in complete linkage.
The hierarchical clustering was generated with Cluster 3.0
and visualized with Java TreeView (after log2 transfor-
mation of the values), as reported in Material and
Methods. The expression values are taken from [34].
G. Guerriero, et al.
expressed at extremely low levels in the roots.
We have previously published two sets of RNA-Seq data on the
hypocotyl and on isolated bast fibres collected from adult plants of the
industrial fibre variety Santhica 27 [5,27]. In these studies we aimed at
analyzing the development of the hemp hypocotyl’s tissues from a cell
wall perspective and the processes underlying bast fibre elongation and
thickening. Both datasets were analyzed using the Finola genome an-
notation, since it is another industrial variety used for oil. We looked
specifically for aquaporin genes expressed at statistically significant
varying levels (FDR corrected ANOVA p-values < 0.05) in the condi-
tions studied. Representatives of the NIP, PIP, TIP and XIP subfamilies
were retrieved from our two published datasets (Fig. 4).
In the hypocotyl, the aquaporins expressed at the highest levels in
young developmental stages (6 and 9 days after sowing, i.e. H6 and H9)
are CsaPIP1-3 and CsaTIP2-2; their expression subsequently decreases
Fig. 5. Normalized relative expressions of a subset of aquaporins in different hemp tissues. Error bars are relative to the standard deviations of 4 independent
biological replicates (n = 4). Different letters indicate statistically significant values at the one-way ANOVA test (p < 0.05) after a Tukey’s post-hoc test.
6
Plant Science 287 (2019) 110167
Fig. 4. Gene expression profiles of some aquaporins in
hemp hypocotyls and internodes. A) Expression values
(RPKM) of an aquaporin gene subset obtained from the
previously published RNA-Seq data on the developing
hemp hypocotyl [27]. Error bars are relative to the stan-
dard deviations of 3 independent biological replicates
(n = 3). Details of the RPKM for each single biological
replicate, statistical values, means are provided in Sup-
plementary File 1. B) Expression values (RPKM) of an
aquaporin gene subset obtained from the previously pub-
lished RNA-Seq data on the isolated hemp bast fibres
sampled along different stem regions [5]. Error bars are
relative to the standard deviations of 4 independent bio-
logical replicates (n = 4). Details of the RPKM for each
single biological replicate, statistical values, means are
provided in Supplementary File 1.
at older phases of development (15 and 20 days after sowing, i.e. H15
and H20) (Fig. 4A). In particular, CsaTIP2-2 shows an increased ex-
pression at H9, a phase characterized by rapid growth and therefore
represents an interesting candidate potentially partaking in hypocotyl
elongation, in a manner analogous to what was discussed for the TIP1-1
gene of castorbean [38]. The PIP gene CsaPIP1-4 shows an increase as
the hypocotyl ages, attaining the highest level at H20. Since at H15 and
H20 the hypocotyl develops bast fibres with thick cellulosic cell walls
and since PIPs are known to regulate fibre differentiation in cotton
[10], it will be interesting to investigate the role of CsaPIP1-4 in sec-
ondary cell wall biosynthesis in hemp bast fibres. Studies involving
cellulose biosynthesis inhibitors could be performed to analyze how
perturbations in cellulose deposition affect CsaPIP1-4 expression.
CsaNIP2-1 is expressed at low levels and shows a peak in expression
at H15. CsaNIP5-1 is also expressed at low levels: its expression is
G. Guerriero, et al. Plant Science 287 (2019) 110167

Fig. 6. A) Optical image showing the locations of the SIMS nano-analysis (white
circles) in the xylem and the bast fibres. The scale bar is 200 μm; B) and C)
NanoSIMS images of hemp hypocotyl. Distribution of 12C14N− (upper row),
28
Si- (middle row) and 31P− (bottom row) in the xylem fibres (B) and in the bast
fibres (C) of untreated sample (Control) and silicon-treated sample (1 mM so-
dium metasilicate). The colour scale goes from black to red with increasing
intensity. The scale bars in panels (B) and (C) are 10 μm.

substantially stable at H6, H9 and H15 and decreases at H20.

From the RNA-Seq data of isolated hemp bast fibres, gene expres-
sion data for the NIP, PIP, TIP and XIP members were retrieved
(Fig. 4B). CsaPIP1-1, CsaXIP1-1 and CsaTIP2-2 show an increased ex-
pression along the stem: the bast fibres collected at the internode
containing the snap point and at the bottom show indeed a higher ex-
pression of the genes, as compared to the fibres sampled at the top.
Among the genes displaying low expression, CsaTIP4-1 and CsaNIP3-2
are more expressed at the top and gradually decrease at the bottom.
CsaNIP1-2 increases in expression at the snap point and bottom.
CsaTIP1-2 and CsaXIP1-1 are the only genes in our dataset showing
higher expression at the snap point.
We decided to perform a gene expression analysis on different tis-
sues of adult hemp plants and targeting a subset of aquaporins, notably
the three putative rice Lsi2 orthologs and genes showing a low ex-
pression in our RNA-Seq datasets (i.e. CsaNIP3-2, CsaNIP1-2 and
CsaNIP5-1). The rationale was to determine whether these low ex-
pressed genes showed higher expressions in tissues/organs other than
the hypocotyl or the isolated bast fibres. The genes CsaLsi2-1 and
CsaLsi2-2 show higher expressions in the isolated hemp fibres than the
core tissues (Fig. 5); CsaLsi2-3 shows comparable levels of expression in
the core and the bast fibres, with the exception of the bast fibres
sampled at the bottom, where the expression increases. The higher
expression of Lsi2 genes in the fibres may imply an accumulation of
silicon (as opaline silica) in the apoplast of phloem fibres (as shown in
the next section), a feature that has important consequences for their
physical properties. The accumulation of silica in the walls of fibres
could for example affect the mechanical properties, e.g. the density of
cellulose microfibrils, similarly to what demonstrated in rice cells [22].
It should be noted that Lsi2-1 and Lsi2-2 are expressed at equal levels in
female and male flowers; CsaLsi2-3 shows higher expression in male
flowers as compared to female ones.
The gene CsaNIP3-2 is expressed at higher levels in the leaves than
roots. CsaNIP1-2 is expressed at higher levels in the bast fibres at the
snap point and at the bottom (confirming the RNA-Seq data, Fig. 4B)
and shows the same expression values in the other organs. Finally,
CsaNIP5-1 is expressed at the highest levels in roots.

3.4. High resolution SIMS imaging on stems and silica extraction from
leaves

The SIMS nano-analysis reveals the presence of silicon in the hy-

pocotyls of hemp (Fig. 6). In the xylem vessels, an evident enrichment
in silicon is visible when the plants were supplemented with 1 mM
sodium metasilicate (Fig. 6B). In the bast fibres and in the presence of
sodium metasilicate, the presence of silicon is more evident in the distal
side of the cell walls (i.e. facing the stem cortex) (Fig. 6C). The signal is
continuous on the distal side and covers the whole cell wall depth
(Supplementary Fig. S1). This preferential accumulation can be inter-
preted by invoking a purely mechanical principle: the impregnation of
the distal cell walls of bast fibres with silica results in the formation of a
ring strengthening the whole stem. It is known that plants grown in the
presence of silicon are more vigorous, but above all more resistant to
mechanical stresses, like lodging [39].
PDMPO staining of the silica extracted from the hemp plantlets did
not reveal any silicified structures in bast fibres (not shown). This could
be due to the silica extraction procedure, which employs a step of

7
G. Guerriero, et al. Plant Science 287 (2019) 110167

Fig. 7. Biogenic silica deposition in hemp leaves after acid extraction and fluorescence imaging with PDMPO. A) Isolated trichomes, some of which with basal cells;
B) Isolated trichomes and two adjacent trichomes with basal and epidermal cells; C) Trichomes with epidermal cells showing replicas of waxy “wrinkles” (boxed
region and bottom magnified image thereof); D) Compound trichome base (arrow) interspersed with single-celled trichome bases; E) Reticulated structure possibly
representing a developmental stage of trichomes with compound bases; F) strongly labeled structures inside epidermal cells possibly representing vacuoles/large
vesicles. The scale bars in panels A–C are 200 μm, in D–E 100 μm and in F 50 μm.

filtration through 0.2 μm filters prior to collecting the silica particles.
Any bast fibre-derived silica smaller than 200 nm may thus have been
lost.
The multitude of shapes observed in silica extracted from silicifiers
like rice and horsetail [20,40] indicates an intimate association of si-
licification with cellular processes. Such a versatility can be easily
shown if an organ comprising different cell types is analyzed. Leaves are
excellent organs to study the different types of silicified cells, since it
comprises epidermal jigsaw puzzle-like cells, trichomes, stomata. Since
PDMPO was shown to be very useful in mapping the fine silicified
structures of rice and horsetail leaves [20,25,40], we decided to extract
and analyze the silica obtained from hemp leaves. PDMPO labeling and
fluorescence microscopy reveals fine cellular details, such as non-se-
creting trichomes and associated basal cells (Fig. 7). It is possible to
observe detailed cellular features, notably the ring of basal cells con-
touring the base of the hollow non-secretory trichome shaft (Figs. 7A
and B), “wrinkles” from the waxy replicas of leaf epidermal cells
(Fig. 7C), a compound trichome base interspersed with single-celled
trichome bases (Fig. 7D). Reticulated structures (maybe representing a
developmental stage of trichomes with compound bases) associated
with the trichomes are also present (Fig. 7E); strongly labeled structures
are observed inside the epidermal cells which may represent vacuoles/
large vesicles (Fig. 7F).
The presence of silicified trichomes was previously reported in
Cannabis [19]; here we provide evidence that the finest details of the
cells composing the base of the trichome are perfectly replicated in the
silica residues extracted. This feature was also observed in thale cress
[41]. Silica residues extracted from rice, horsetail and fern [26] also
show fine cell details, with stomata providing an emblematic example
[20,25]. The silicification of the trichome base is a clear demonstration
of the mechanical role played by silica, which in this case provides the
required mechanical requisites to support the trichome shaft.
4. Conclusions
We have here identified 30 aquaporins of the important multi-
purpose plant C. sativa. We have analyzed the expression of some
aquaporins in different tissues of a textile hemp variety and provided
for the first time evidence for the presence of genes with homology to
the rice OsLsi1 and OsLsi2. Modelling of the pores suggests that the
diameter of these aquaporins is too small for the computed maximum
radius of silicic acid, despite the presence of more flexible amino acids
at the pore bottlenecks. By using NanoSIMS and fluorescence micro-
scopy coupled to PDMPO staining, we show that silicon is found in
hemp tissues.
Overall, our study shows not only that silicon is present in hemp
tissues, but also that it accumulates preferentially on the distal side of
the bast fibre cell walls. The results provided have relevance in terms of
application-oriented follow up studies, since silica is known to be fire-
retardant and to increase the durability of the biomaterial (i.e. re-
sistance to natural decay).
Author contributions
GG, RD, J-FH, NV, KSS and CE conceived the experimental work;
GG, RD, HS, EL, NV, KSS and CE carried out the experiments; GG, RD,
HS, EL, NV, KSS and CE collected, analyzed and interpreted the data.
GG, RD, NV and KSS wrote the initial draft. All authors contributed to
editing and to the finalization of the manuscript.
Funding
GG, KS and J-FH thank The Fonds National de la Recherche,
Luxembourg, (Project CANCAN C13/SR/5774202) for financial sup-
port.
Acknowledgements
The personal assistance of KSS by KFUPM is acknowledged. Laurent
Solinhac and Aude Corvisy are thanked for their technical assistance.
Appendix A. Supplementary data
Supplementary material related to this article can be found, in the
online version, at doi:https://doi.org/10.1016/j.plantsci.2019.110167.

8
G. Guerriero, et al.
References
[1] C.M. Andre, J.-F. Hausman, G. Guerriero, Cannabis sativa: the plant of the thousand
and one molecules, Front. Plant Sci. 7 (2016) 19, https://doi.org/10.3389/fpls.
2016.00019.
[2] G. Guerriero, K. Sergeant, J.-F. Hausman, Integrated -Omics: a powerful approach to
understanding the heterogeneous lignification of fibre crops, Int. J. Mol. Sci. 14
(2013) 10958–10978, https://doi.org/10.3390/ijms140610958.
[3] T. Gorshkova, N. Brutch, B. Chabbert, M. Deyholos, T. Hayashi, S. Lev-Yadun,
E.J. Mellerowicz, C. Morvan, G. Neutelings, G. Pilate, Plant fiber formation: state of
the art, recent and expected progress, and open questions, Crit. Rev. Plant Sci. 31
(2012) 201–228, https://doi.org/10.1080/07352689.2011.616096.
[4] O. Gorshkov, N. Mokshina, V. Gorshkov, S. Chemikosova, Y. Gogolev, T. Gorshkova,
Transcriptome portrait of cellulose-enriched flax fibres at advanced stage of spe-
cialization, Plant Mol. Biol. (2016) 1–19, https://doi.org/10.1007/s11103-016-
0571-7.
[5] G. Guerriero, M. Behr, S. Legay, L. Mangeot-Peter, S. Zorzan, M. Ghoniem, J.-
F. Hausman, Transcriptomic profiling of hemp bast fibres at different develop-
mental stages, Sci. Rep. 7 (2017) 4961, https://doi.org/10.1038/s41598-017-
05200-8.
[6] M.V. Ageeva, B. Petrovská, H. Kieft, V.V. Sal’nikov, A.V. Snegireva, J.E.G. van Dam,
W.L.H. van Veenendaal, A.M.C. Emons, T.A. Gorshkova, A.A.M. van Lammeren,
Intrusive growth of flax phloem fibers is of intercalary type, Planta 222 (2005)
565–574, https://doi.org/10.1007/s00425-005-1536-2.
[7] G. Guerriero, J.-F. Hausman, G. Cai, No stress! relax! mechanisms governing growth
and shape in plant cells, Int. J. Mol. Sci. 15 (2014) 5094–5114, https://doi.org/10.
3390/ijms15035094.
[8] A.V. Snegireva, M.V. Ageeva, S.I. Amenitskii, T.E. Chernova, M. Ebskamp,
T.A. Gorshkova, Intrusive growth of sclerenchyma fibers, Russ. J. Plant Physiol. 57
(2010) 342–355, https://doi.org/10.1134/S1021443710030052.
[9] F. Chaumont, S.D. Tyerman, Aquaporins: highly regulated channels controlling
plant water relations1, Plant Physiol. 164 (2014) 1600–1618, https://doi.org/10.
1104/pp.113.233791.
[10] D.-D. Li, X.-M. Ruan, J. Zhang, Y.-J. Wu, X.-L. Wang, X.-B. Li, Cotton plasma
membrane intrinsic protein 2s (PIP2s) selectively interact to regulate their water
channel activities and are required for fibre development, New Phytol. 199 (2013)
695–707, https://doi.org/10.1111/nph.12309.
[11] S.M. Shivaraj, R.K. Deshmukh, R. Rai, R. Bélanger, P.K. Agrawal, P.K. Dash,
Genome-wide identification, characterization, and expression profile of aquaporin
gene family in flax (Linum usitatissimum), Sci. Rep. 7 (2017) 46137, https://doi.org/
10.1038/srep46137.
[12] T.A. Gorshkova, V.V. Sal’nikov, S.B. Chemikosova, M.V. Ageeva, N.V. Pavlencheva,
J.E.G. van Dam, The snap point: a transition point in Linum usitatissimum bast fiber
development, Ind. Crops Prod. 18 (2003) 213–221, https://doi.org/10.1016/
S0926-6690(03)00043-8.
[13] J.F. Ma, K. Tamai, N. Yamaji, N. Mitani, S. Konishi, M. Katsuhara, M. Ishiguro,
Y. Murata, M. Yano, A silicon transporter in rice, Nature 440 (2006) 688–691,
https://doi.org/10.1038/nature04590.
[14] G. Heine, G. Tikum, W.J. Horst, Silicon nutrition of tomato and bitter gourd with
special emphasis on silicon distribution in root fractions, J. Plant Nutr. Soil Sci. 168
(2005) 600–606, https://doi.org/10.1002/jpln.200420508.
[15] W. Zellner, J. Frantz, S. Leisner, Silicon delays tobacco ringspot virus systemic
symptoms in Nicotiana tabacum, J. Plant Physiol. 168 (2011) 1866–1869, https://
doi.org/10.1016/j.jplph.2011.04.002.
[16] J.F. Ma, N. Yamaji, A cooperative system of silicon transport in plants, Trends Plant
Sci. 20 (2015) 435–442, https://doi.org/10.1016/j.tplants.2015.04.007.
[17] N. Yamaji, N. Mitatni, J.F. Ma, A transporter regulating silicon distribution in rice
shoots, Plant Cell 20 (2008) 1381–1389, https://doi.org/10.1105/tpc.108.059311.
[18] R.K. Deshmukh, J. Vivancos, G. Ramakrishnan, V. Guérin, G. Carpentier, H. Sonah,
C. Labbé, P. Isenring, F.J. Belzile, R.R. Bélanger, A precise spacing between the NPA
domains of aquaporins is essential for silicon permeability in plants, Plant J. 83
(2015) 489–500, https://doi.org/10.1111/tpj.12904.
[19] P. Dayanandan, P.B. Kaufman, Trichomes of Cannabis sativa L. (Cannabaceae), Am.
J. Bot. 63 (1976) 578–591, https://doi.org/10.2307/2441821.
Plant Science 287 (2019) 110167
[20] G. Guerriero, C. Law, I. Stokes, K.L. Moore, C. Exley, Rough and tough. How does
silicic acid protect horsetail from fungal infection? J. Trace Elem. Med. Biol. Organ
Soc. Miner. Trace Elem. GMS 47 (2018) 45–52, https://doi.org/10.1016/j.jtemb.
2018.01.015.
[21] S.C. Fry, B.H.W.A. Nesselrode, J.G. Miller, B.R. Mewburn, Mixed-linkage
(1 > 3,1 > 4)-beta-D-glucan is a major hemicellulose of Equisetum (horsetail) cell
walls, New Phytol. 179 (2008) 104–115, https://doi.org/10.1111/j.1469-8137.
2008.02435.x.
[22] C. He, J. Ma, L. Wang, A hemicellulose-bound form of silicon with potential to
improve the mechanical properties and regeneration of the cell wall of rice, New
Phytol. 206 (2015) 1051–1062, https://doi.org/10.1111/nph.13282.
[23] O. Leroux, F. Leroux, A.A. Mastroberti, F. Santos-Silva, D. Van Loo, A. Bagniewska-
Zadworna, L. Van Hoorebeke, S. Bals, Z.A. Popper, J.E. de Araujo Mariath,
Heterogeneity of silica and glycan-epitope distribution in epidermal idioblast cell
walls in Adiantum raddianum laminae, Planta 237 (2013) 1453–1464, https://doi.
org/10.1007/s00425-013-1856-6.
[24] C.C. Perry, Y. Lu, Preparation of silicas from silicon complexes: role of cellulose in
polymerisation and aggregation control, J. Chem. Soc. Faraday Trans. 88 (1992)
2915–2921, https://doi.org/10.1039/FT9928802915.
[25] C. Law, C. Exley, New insight into silica deposition in horsetail (Equisetum arvense),
BMC Plant Biol. 11 (2011) 112, https://doi.org/10.1186/1471-2229-11-112.
[26] C. Exley, A possible mechanism of biological silicification in plants, Front. Plant Sci.
6 (2015), https://doi.org/10.3389/fpls.2015.00853.
[27] M. Behr, S. Legay, E. Žižková, V. Motyka, P.I. Dobrev, J.-F. Hausman, S. Lutts,
G. Guerriero, Studying secondary growth and bast fiber development: the hemp
hypocotyl peeks behind the wall, Front. Plant Sci. 7 (2016) 1733.
[28] J. Yang, R. Yan, A. Roy, D. Xu, J. Poisson, Y. Zhang, The I-TASSER suite: protein
structure and function prediction, Nat. Methods 12 (2015) 7–8, https://doi.org/10.
1038/nmeth.3213.
[29] S.C. Lovell, I.W. Davis, W.B. Arendall, P.I.W. de Bakker, J.M. Word, M.G. Prisant,
J.S. Richardson, D.C. Richardson, Structure validation by Calpha geometry: phi, psi
and Cbeta deviation, Proteins 50 (2003) 437–450, https://doi.org/10.1002/prot.
10286.
[30] T.B. Masood, S. Sandhya, N. Chandra, V. Natarajan, CHEXVIS: a tool for molecular
channel extraction and visualization, BMC Bioinformatics 16 (2015) 119, https://
doi.org/10.1186/s12859-015-0545-9.
[31] M. Baek, T. Park, L. Heo, C. Park, C. Seok, GalaxyHomomer: a web server for
protein homo-oligomer structure prediction from a monomer sequence or structure,
Nucleic Acids Res. 45 (2017) W320–W324, https://doi.org/10.1093/nar/gkx246.
[32] L. Edelmann, Freeze-dried and resin-embedded biological material is well suited for
ultrastructure research, J. Microsc. 207 (2002) 5–26.
[33] H. van Bakel, J.M. Stout, A.G. Cote, C.M. Tallon, A.G. Sharpe, T.R. Hughes,
J.E. Page, The draft genome and transcriptome of Cannabis sativa, Genome Biol. 12
(2011) R102, https://doi.org/10.1186/gb-2011-12-10-r102.
[34] L. Massimino, In silico gene expression profiling in Cannabis sativa, F1000Research 6
(2017), https://doi.org/10.12688/f1000research.10631.1.
[35] J.F. Ma, N. Yamaji, N. Mitani, K. Tamai, S. Konishi, T. Fujiwara, M. Katsuhara,
M. Yano, An efflux transporter of silicon in rice, Nature 448 (2007) 209–212,
https://doi.org/10.1038/nature05964.
[36] C. Exley, G. Guerriero, X. Lopez, Silicic acid: the omniscient molecule, Sci. Total
Environ. 665 (2019) 432–437, https://doi.org/10.1016/j.scitotenv.2019.02.197.
[37] A. Stank, D.B. Kokh, J.C. Fuller, R.C. Wade, Protein binding pocket dynamics, Acc.
Chem. Res. 49 (2016) 809–815, https://doi.org/10.1021/acs.accounts.5b00516.
[38] D.A. Eisenbarth, A.R. Weig, Dynamics of aquaporins and water relations during
hypocotyl elongation in Ricinus communis L. seedlings, J. Exp. Bot. 56 (2005)
1831–1842, https://doi.org/10.1093/jxb/eri173.
[39] D. Dorairaj, M.R. Ismail, U.R. Sinniah, T.K. Ban, Influence of silicon on growth,
yield, and lodging resistance of MR219, a lowland rice of Malaysia, J. Plant Nutr. 40
(2017) 1111–1124, https://doi.org/10.1080/01904167.2016.1264420.
[40] G. Guerriero, I. Stokes, C. Exley, Is callose required for silicification in plants? Biol.
Lett. 14 (2018), https://doi.org/10.1098/rsbl.2018.0338.
[41] T. Brugiére, C. Exley, Callose-associated silica deposition in Arabidopsis, J. Trace
Elem. Med. Biol. Organ Soc. Miner. Trace Elem. GMS 39 (2017) 86–90, https://doi.
org/10.1016/j.jtemb.2016.08.005.