In plants, conjugation and compartmentalization appears as major steps in the detoxification process of xenobiotics. Once transported into the cells, xenobiotic compounds are mainly metabolized by xenome enzymes, reducing their toxicity in the form of conjugates before compartmentalization or excretion. In the present work, we performed both cellular localization of phe and oxidative stress markers in Arabidopsis and Spartina to assess the putative role of specialized cells in xenobiotic detoxification strategies.

Arabidopsis plantlets were grown under control and moderate phe (25 µm) conditions, and leaves were filtered under vacuum with Singlet Oxygen Sensor Green (SOSG). Compared to the control ( Figure 1A ), trichomes in treated plants were found to be highly reactive to phe ( Figure 1B ), as they presented pronounced fluorescent patches, indicating a severe oxidative stress response through singlet oxygen production. Moreover, phe detected by epifluorescence microscopy pointed out a significant role of these structures in xenobiotic sequestration and detoxification ( Figure 1C ). We observed that the intensity of the phe-induced fluorescence was amplified as the concentration was increased. Indeed, the visualization of the leaf surface of plantlets grown in the presence of 200 µm phe revealed additional high fluorescent spots ( Figure 1D ). Close cellular screening showed that these spots were found to be localized specifically in the pavement cells ( Figure 1E ) from the epidermal layer. Here, we noticed that cells involved in phe metabolization/storage depend on phe concentration. In fact, in Arabidopsis phe was only detected in trichomes under moderate-induced stress, whereas under sublethal concentration (200 µm phe, Dumas et al., 2016), it was detected in both trichomes and pavement cells, indicating different detoxification strategies depending on the physicochemical conditions. When exceeding these concentrations, Arabidopsis show morphological and cellular symptoms of xenobiotic stress such as chlorosis, necrotic lesions, perturbations in signaling and metabolic pathways that regulate ROS and responses related to pathogen defense and cell death (Weisman et al., 2010; Dumas et al., 2016).

Similar experiments were performed in species from Spartina, described to be highly tolerant to PAHs (Alvarez et al., 2018; Cavé-Radet et al., 2019). Here, we focused on the two hexaploid parents S. maritima and S. alterniflora, and the allododecaploid S. anglica which resulted from genome doubling of their interspecific hybrid. Leaves of these three Spartina species were grown under high phe concentration (400 µm), followed by Nitroblue Tetrazolium (NBT) staining. UV light microscope observations revealed that superoxide radical production was mostly restricted to salt glands, mainly distributed along the adaxial leaf surface ( Figure 2A ). Basically, salt glands (also called hydathodes) are specialized structures involved in salt accumulation and excretion in Spartina and other Poaceae species (Skelding and Winterbotham, 1939; Barhoumi et al., 2008). In order to understand the structure of these particular cells and to verify if there is any co-localization of phe with this excess production of superoxide radicals, we investigated semi-thin leaf sections ( Figure 2B ). As first described by Skelding and Winterbotham (1939), we observed that salt glands include two distinct cells, a small cap cell ( Figure 2B-1 ) inserted into the epidermis and a much larger basal cell ( Figure 2B-2 ). Confocal microscopy analysis of thin sections of phe-treated leaf sections did not detect any phe signal in the cap cells, in contrast to the basal cells where strong intracellular fluorescence patches associated to phe sequestration were found ( Figure 2C ).

According to our results, we assume that trichomes, pavement cells and the basal cells of salt glands are potentially specialized cells involved in the accumulation and/or dissipation of organic xenobiotics in higher plants. Such observations are corroborated by the inorganic detoxification potentials previously reported in trichomes (Hall, 2002; Wienkoop et al., 2004) and salt glands (Thomson, 1975; Kraus et al., 1986; Burke et al., 2000). In response to organic xenobiotics, our results are consistent regarding the recent reports by Alves et al. (2017) about PAH cellular localization in Medicago sativa, where authors revealed phe and other PAHs accumulated in glandular secreting trichomes. However, to our knowledge the putative impact of pavement cells in xenobiotic detoxification is still missing in the literature.

At the molecular level, several studies provided evidences that trichomes although being non-secreting unicellular structures in Arabidopsis, are central accumulating and/or detoxifying structures. In fact, Gutierrez-Alcala et al. (2000) reported up to 300-fold more glutathione (GSH) content in trichomes as compared to other epidermal cells. As GSH are expected to be involved in ROS scavenging and xenobiotic detoxification (Das and Roychoudhury, 2014), such feature supports the high conjugation potential of trichomes. Moreover, the enrichment in peroxidases and CYPs involved in xenobiotic hydroxylation (Aziz et al., 2005), and the identification of detoxifying molecules under sulfur-induced stress (Wienkoop et al., 2004) in Arabidopsis trichomes are consistent with our observations. Complementary results were also reported in trichomes of tobacco, where expression profiling identified genes involved in responses to abiotic and biotic stresses, specifically expressed under cadmium exposure (Harada et al., 2010).

In a previous study, we reported enhanced tolerance to PAHs in S. anglica compared to the parental species S. alterniflora and S. maritima (Cavé-Radet et al., 2019). In this paper, we used comparative analyses of phe and ROS cellular compartmentalization, and photosynthetic indices, to highlight enhanced tolerance to phe in S. anglica following allopolyploidization. Other supplementary results in accordance with these reports are provided in this manuscript as supplementary data. On leaves grown one month under 100 µm phe (as described in the materials & methods section), we observed that S. maritima and S. alterniflora turned senescent indicating a high degradation and cell death phenotype, unlike S. anglica ( Supplementary Figure S1 ). Even cultivated under 400 µm phe for 10 days, we did not measure by spectrometry a significant reduction in photosynthetic pigment contents (chlorophylls and carotenoids) in S. anglica in contrast to the parental species ( Supplementary Table S1 ).

In order to find out if there is any relationship between salt gland features and phe tolerance abilities, we compared salt gland densities and cell sizes between species. Using both UV light and scanning electron microscopy, we performed the measures and provided high quality images of salt glands along the adaxial leaf surface ( Figure 3A ), and cap cells of salt glands in the leaf ridges ( Figure 3B ). No significant difference in salt gland density was observed between species, with average densities of 37.5 ± 1.6 gland.mm^-2 for S. alterniflora, 34.0 ± 1.6 for S. maritima and 33.1 ± 1.7 for S. anglica ( Figure 3C ). Similarly, no significant difference in the size of the cap cell of the salt gland was observed (average size of cap salt gland cells of 91.6 ± 3.4 µm² in S. alterniflora, 80.0 ± 6.0 µm² in S. maritima and 81.5 ± 5.1 µm² in S. anglica; Figure 3D ). By contrast, significant differences were found for basal cells, with larger basal cells found in S. anglica (174.2 ± 8.0 µm²) as compared to the parental species (143.3 ± 5.5 in S. alterniflora and 137.6 ± 4.4 in S. maritima; Figure 3E ).

Our results provide a positive correlation between the size of basal salt gland cells and phe tolerance abilities previously reported following allopolyploidization in Spartina (Cavé-Radet et al., 2019). One can assume that genome doubling in S. anglica may increase cell size, as we observed on basal salt gland cells compared to the parental species. Surprisingly, our investigations did not show any differences between cap cell sizes from hexaploid and dodecaploid Spartina species. Hence, we assume that larger basal cells in S. anglica may be related with enhanced tolerance to phe (among allopolyploidy evolutionary novelties), for example through increased metabolism and storage capacities while no differences in salt gland densities were reported between species. In combination with previous observations, these data are consistent with the function of salt glands as central PAH detoxifying structures (e.g. storage/metabolization). Therefore, complementary analyses are needed to validate the role of basal salt gland cells in response to organic xenobiotic-induced stress. In Spartina, cell specific transcriptome profiling would be of high interest to conduct gene set enrichment analysis of basal salt gland cells between species. This would provide molecular data to characterize gene expression profiles in larger basal salt gland cells of S. anglica compared to the parental species, especially concerning genes involved in the xenome.

To test the possible impact of trichomes in organic xenobiotic detoxification, we selected two independent homozygous Arabidopsis glabrous mutant lines SALK_039478.51.10.x and SAIL_1149_D03, which presented T-DNA insertions in the GLABRA1 gene (AT3G27920: MYB protein domain involved in trichome development, Figure 4A ). Glabrous mutant plants are characterized by the absence of trichomes. Plantlets were cultivated under control and phe-induced stress (0 and 25 µm) and compared to their background ecotypes. After 21 days, plantlets were sampled, weighed, and photographed. Representative plantlets ( Figure 4B ) and fresh weights plots are presented ( Figures 4C, D ).

Concerning SALK_039478.51.10.x mutants, plantlet weight was 13.7 ± 0.1 mg under control conditions, and was dramatically reduced (1.9 ± 0.1 mg) under phe-induced stress, while plantlet weight of the corresponding background ecotype Col-0 was 14.1 ± 0.2 mg under control conditions and 4.5 ± 0.1 mg under 25 µm of phe ( Figure 4C ). The same pattern was observed in plantlet weight of SAIL_1149_D03 (14.4 ± 0.2 mg under control conditions and 4.3 ± 0.1 mg under phe-induced stress), and in the corresponding background ecotype Col-3 (14.7 ± 0.2 mg under control conditions against 8.1 ± 0.1 mg under 25 µm of phe; Figure 4D ). However, the biomass of both mutants was significantly reduced under phe-induced stress as compared to background ecotypes ( Figures 4C, D ).

In Arabidopsis, a moderate phe-induced stress (25 µm) induces a significant reduction in plant development. Here, the development of glabrous mutants was further reduced as compared to their background ecotypes under phe-induced stress, which further confirms that trichomes are most likely specialized cells involved in organic xenobiotic detoxification as hypothesized above.

Altogether, these data suggest that in higher plants, cells of the whole organism do not respond equally to phe-induced stress, and that some specialized cells such as trichomes, pavement cells, and basal salt gland cells appear as key components in response to organic xenobiotic-induced stress. Plant specialized cells in detoxification should have increased transcriptomic, metabolic, and morphological abilities to detoxify organic xenobiotics. Interestingly, both cell lineages, trichomes and pavement cells, are known to undergo several rounds of genome doubling without the initiation of mitosis (Melaragno et al., 1993; Schnittger and Hülskamp, 2002; Katagiri et al., 2016). This phenomenon is known as somatic polyploidy, endoreduplication or endopolyploidy (Bateman et al., 2018), and may be of interest for such features. In fact, Barkla et al. (2018) pointed out that “highly metabolically active cells appear to have the highest levels of endopolyploidy, as it has been demonstrated in the endosperm and suspensor cells of the seeds of Arum maculatum and Phaseolus coccineus (24,567 and 8,192C, respectively; Leitch and Dodsworth, 2017)”. Despite a majority of tissues within Arabidopsis endoreduplicate at a basal level (Buchanan et al., 2000), it is important to keep in mind that endopolyploidy-associated growth is sometimes confined to specialized cell types or tissues that perform specific biological functions (e.g. the endosperm in sorghum, see Kladnik et al., 2006; or the pericarp in tomato, see Renaudin et al., 2017; for a review see Barkla et al., 2018). This mostly concerns localized endopolyploids, for which endopolyploidy is punctual in opposition with systemic endopolyploid species like Arabidopsis. In the beginning of their formation, Arabidopsis trichomes cease mitotic divisions and start a multitude of endoreduplication cycles. This multiplies their DNA content several times over, increases their size, and modifies their growth direction (Melaragno et al., 1993; Hülskamp et al., 1994; Kasili et al., 2011). The macroscopic plant trichomes in Arabidopsis are unicellular hairs displaying a high number of genomic copies ranging from 8 to 64C, and pavement cells may reach 16C (Melaragno et al., 1993; Schnittger and Hülskamp, 2002; Hülskamp, 2004).

In contrast to polyploidy, where the duplicated DNA is inherited through the germline and perpetuated over subsequent generations in all cells (Ainouche and Jenczewski, 2010), endopolyploidy arises from variations in the cell cycle where the genome is replicated without cell division, resulting in cells with a single polyploid nucleus and increased DNA content. The increase in DNA content often correlates with increased cell size (Robinson et al., 2018). According to Kondorosi et al. (2000), the increase in cell sizes is widespread in plants and represents a characteristic of endopolyploid cells; the higher the ploidy level, the larger the polyploid cell.

Interestingly, polyploid plants (inherited ploidy) can undergo endoreduplication (somatic polyploidy), as reported for example in trichomes from tetraploid Arabidopsis (del Pozo and Ramirez-Parra, 2015). In Spartina, Levering and Thomson (1971) described the ultra-structural salt gland cells of Spartina, and pointed out that basal salt gland cells contain a very large nucleus (also observed in salt glands from other Poaceae by Somaru et al., 2002). Hence, we hypothesized that basal salt gland cells from polyploid Spartina species still likely undergo endoreduplication, in association with the enlargement of the nucleus and cell size. Beside we reported basal salt gland cells enlargement in the dodecaploid compared to the hexaploid Spartina species, we assumed that both processes may overlap (inherited polyploidy and endoreduplication). This would imply higher ploidy levels in basal salt gland cells compared to other cells for which hexaploid cytotypes in S. alterniflora and S. maritima and dodecaploid cytotypes in S. anglica are expected. In this perspective, endoreduplication in basal salt gland cells could be related with enhanced xenobiotic storage/metabolization potentials (in addition to evolutionary benefits potentially raised by polyploidy), as proposed concerning Arabidopsis trichomes or pavement cells. To date, the relationship between inherited polyploidy and endopolyploidy is still unclear, but Pacey et al. (2020) provided recent evidences that polyploidy reduces endopolyploidy (by comparing Arabidopsis diploid vs. colchicine-treated tetraploid accessions). However, such trade-off was not currently reported between all Arabidopsis genotypes considered by Scholes (2020), depicting that endopolyploidy affected by inherited polyploidy is genotype dependent, and may also increase specifically in tetraploid plants in response to stress damage. Thus, further studies in distant plant lineages such as Spartina would be of interest to decipher how inherited polyploidy combined with endopolyploidy may be beneficial in response to xenobiotics, with a special focus on salt glands. Hence, measurements in ploidy levels of basal salt gland cells between Spartina species are needed to confirm such hypothesis.

Here, we reviewed the putative advantages given by endopolyploidy in specialized cells (largely reported in Arabidopsis trichomes and pavement cells, and hypothesized in Spartina basal salt gland cells) involved in xenobiotic detoxification. We think that endoreduplication represents an interesting scope for future analyses studying the role of specialized cells in xenobiotic detoxification. Notably, it has been suggested that the impact of endopolyploidy is most likely related with an increase in gene expression to sustain high protein production and metabolic activity necessary for development, cell enlargement, or stress response (Scholes and Paige, 2015; Schoenfelder and Fox, 2015). In Arabidopsis trichomes, endoreduplication may explain the enrichment in detoxifying molecules previously described in these structures (Wienkoop et al., 2004; Aziz et al., 2005).

To detect if specialized are involved in organic xenobiotic detoxification in higher plants, we focused our analyses in two contrasting plant lineages, using the diploid plant model Arabidopsis thaliana and species from the polyploid Sporobolus genus subsect. Spartina (Spartina Schreb. Clade in Poaceae). In Spartina, we selected the parental hexaploid species S. alterniflora Loisel (2n = 6x = 62) and S. maritima Curtis (2n = 6x = 60), and their allopolyploid derivative S. anglica C.E. Hubbard (2n = 12x = 120–124) which resulted from genome doubling of their interspecific hybrid. Phe is a common PAH we used throughout the study. Arabidopsis represents a sensitive species to xenobiotics, with a sub-lethal phe concentration of 200 µm (Dumas et al., 2016). In comparison, Spartina species are much more tolerant (S. anglica can tolerate even up to 800 μm phe). The three species were selected because enhanced tolerance abilities to phe were previously reported following allopolyploidization in S. anglica (Cavé-Radet et al., 2019).

Seeds of Arabidopsis thaliana (Col-0) were germinated in Petri dishes on a solid growth medium (half-strength Murashige and Skoog: 0.5 MS; 0.8% agar; pH = 5.6). To ensure that phe is translocated upwards from roots to leaves via the xylem stream, plantlets were grown vertically for 6 days, and then transferred on an identical medium supplemented with 0 and 25 µm phe solubilized in absolute ethanol, with only roots being in contact with the medium. A sterile transparent plastic film was applied between plantlets and the medium to avoid any contact of the rosette-leaves with phe. In total, Arabidopsis plantlets were cultivated in vitro for 21 days before experiments, in phytotronic chambers under fluorescent light/dark regime of 16/8 h, an average ambient temperature of 22/20°C, a light intensity of 8500 Lux and a relative humidity of 60%.

In complement, fragments of Spartina leaves were cultivated using a method previously described in Cavé-Radet et al. (2019). Here, we cultivated leaves from the three species for 2 days in petri dishes on the solid growth medium as previously described for Arabidopsis (0 and 400 µm phe, three biological and technical replicates).

In Arabidopsis, at least 6 plantlets per treatment were filtered under vacuum with 100 µm SOSG reagent in 50 µm phosphate potassium buffer (pH = 7.5). Observations were performed using a spectrofluorometer to reveal singlet oxygen radicals in leaf tissues by green fluorescence, and coupled with epifluorescence microscopy detection (EFMD) to reveal phe.

In Spartina, at least 6 leaves per treatment and species were filtered under vacuum with 35 mg.ml^-1 NBT in phosphate potassium buffer (10 mm, pH = 7.5), followed by three hours of reaction. Then, leaves were immersed for 15 min in acetic-glycerol-ethanol solution (v. 1:1:3) at 100°C, before being bleached in 80% acetone until complete pigment discoloration. Observations were performed using a UV light microscopy to reveal superoxide radicals by blue fluorescent patches in leaf tissues (along the adaxial surface or in semi-thin sections), and coupled with confocal microscopy (CFM) to reveal phe.

Here, we used both singlet oxygen and superoxide radicals as oxidative stress markers (respectively in Arabidopsis and Spartina). As well, phe is detected either using EFMD or CFM by its blue fluorescence emitting from 420–480 nm with a weak but specific signal around 430 nm on 405 nm excitation. As a free molecule, phe is easily visualized under fluorescence, and we took advantage of this to screen specific phe accumulation in a myriad of cells in the two contrasting plant lineages.

To provide additional data supporting that salt glands in Spartina may represent key structures involved in organic xenobiotic detoxification, we aimed to measure salt gland densities and cell sizes between the three species from divergent tolerance abilities to phe (Cavé-Radet et al., 2019). In Spartina leaves cultivated under 400 µm phe, NBT staining allowed us to identify highly localized coloration showing ROS production, specifically in salt glands along the adaxial leaf surface. Thus, at least 6 replicates of phe-treated leaves filtered with NBT were photographed, and blue-stained salt glands were counted inside areas from 0.8 to 2.6 mm² to calculate densities. Salt glands were similarly observed in semi-thin sections on the same samples by UV light microscopy, and basal cell sizes were measured. In addition, cap salt gland cells were observed by scanning electron microscopy (SEM, platform CMEBA UR1). Here, at least 6 replicates of fresh leaves were dehydrated in successive diluted ethanol solutions (70, 80, 90, and 100%) for 10 min. Because salt glands are mainly encountered in the leaf ridges along the adaxial surface, leaves were fractured in liquid nitrogen and fixed prior to SEM observation. Then, cap cells were photographed and measured. Image processing was performed using ImageJ software (v 1.51j8). Statistical analyses were performed between species by pairwise comparisons using Kruskal-Wallis non-parametric test in R 3.5 (R Core Team, 2015).

To validate the impact of trichomes in organic xenobiotic tolerance mechanisms, a functional validation was performed using T-DNA mutant lines affecting GLABRA1 expression (AT3G27920) which disrupt trichome development. Homozygous glabrous mutant (trichomeless) lines SALK_039478.51.10.x (Col-0 background ecotype) and SAIL_1149_D03 (Col-3 background ecotype) were selected (SIGnAL database, http://signal.salk.edu/). Mutants were grown in vitro in Petri dishes (0 and 25 µm phe), stocked for 48h at 4°C before being transferred into phytotronic chambers (same culture protocol and conditions as described for Arabidopsis plantlets in section 4.1). The experiment was performed in biological triplicates. After 21 days, 10 plantlets per treatment were sampled and weighed. Pairwise comparisons were performed between treatments, and mutants were compared to their background ecotypes using Kruskal-Wallis non-parametric tests in R 3.5 (R Core Team, 2015).