All used organic solvents [i.e., n-hexane 97%, methanol, trifluoroacetic acid (TFA), and acetonitrile (ACN) (VWR INTERNATIONAL, Leuven, Belgium)] were HPLC/LC-MS grade. Water was purified and demineralized with a Milli-Q system manufactured by Millipore (Bedford, MA, USA). Shikonin (purity >98%) was purchased from Cayman Chemical Company (Biomol GmbH, Hamburg, Germany) and used as internal standard.

For gene expression analysis (i.e., RNA extraction) all non-disposable materials (e.g., glass materials, mortars, and pestles) were first treated with RNase AWAY™ Surface Decontaminant (Thermo Fisher Scientific™, Belgium) and 0.1% diethylpyrocarbonate (DEPC)-treated sterile water. All the reagents were prepared with sterilized (121°C for 15 min) DEPC-treated water. Reagents used in RNA extraction were the following: extraction buffer [100 mM Tris–HCl (pH 8.0), 20 mM EDTA, 1.4 M NaCl, 2% CTAB (w/v) and 2% PVP (w/v)], 2% β-mercaptoethanol (v/v), 5 M NaCl, 1.2 M NaCl, 0.38 M trisodium citrate dihydrate, chloroform, isopropanol, and 70% ethanol (stored at −20°C).

Rhizophagus irregularis (Błaszk, Wubet, Renker and Buscot) C. Walker and A. Schüßler as [“irregulare”] MUCL 41833 and Rhizophagus aggregatus (N.C. Schenck & G.S. Sm.) C. Walker MUCL 49408 were supplied by the Glomeromycota in vitro collection (GINCO).¹ Both AMF were mass-produced on Zea mays L. cv. ES Ballade (Euralis, Lescar, France) in 10 L plastic boxes containing sterilized (121°C for 15 min) lava (DCM, Grobbendonk, Belgium). Two other AMF strains (R. irregularis and Septoglomus viscosum) were isolated from A. tinctoria growing in a suburban pine forest, altitude 50 m of Northern Greece (Evaggelistria, Seih Sou, Thessaloniki, special collection permit obtained by the Institute of Plant Breeding and Genetic Resources, Hellenic Agricultural Organization Demeter – IPBGR, HAO Demeter). They were first trapped on Plantago lanceolata L. (Ecosem, Belgium) and Medicago truncatula Gaertn. (SARDI, Australia) to stimulate the production of numerous spores and then grown as monospores on P. lanceolata at the GINCO premises to obtain single species cultures (see section Isolation and mono-species culture of AMF in Supplementary material, Supplementary Figures 1–4, and Supplementary Table 1). They were finally mass-produced on Z. mays L. in 5 L plastic boxes containing the same sterilized lava substrate as above.

The four AMF strains were grown under greenhouse conditions at 25/18°C (day/night), a relative humidity (RH) of 38%, a photoperiod of 16 h day^–1 and a photosynthetic photon flux (PPF) of 120 μmol m^–2 s^–1.

Alkanna tinctoria unrooted in vitro explants (International Plant Exchange Network – IPEN – accession number GR-1-BBGK-17,5975) were provide by IPBGR, HAO Demeter (Thessaloniki, Greece). The plants were proliferated on Murashige-Skoog (MS) basal medium enriched with plant growth regulators (PGRs) and rooted on Root Culture 1 (RC1) modified medium free of ammonium nitrate (Cartabia et al., 2022). The rooted A. tinctoria plants were further acclimatized following different steps resulting in an optimal plant survival rate (see section “Alkanna tinctoria acclimatization protocol” in Supplementary material and Supplementary Figure 5). The acclimatization protocol was applied for all the experiments in order to have plants of the same age and established under identical growth conditions (first acclimatization from 2nd November 2020 to 7th December 2020 and second from 1st February 2021 to 10th March 2021).

After the acclimatization phase, the plants were carefully removed from the substrate, gently washed under a stream of demineralized water, and transplanted into 1 L pots (11 × 11 × 12 cm) containing a sterilized (121°C for 15 min) substrate mixture (2 peatmoss/2 compost/1 perlite/1 quartz 0.4–0.8 mm/1 quartz 1–2 mm) (see Supplementary Figure 5). Total fresh weight (TFW) of the plants was evaluated before AMF inoculation.

In each experiment, plants were associated [i.e., the mycorrhizal (M) treatments] or not [i.e., the non-mycorrhizal (NM) treatments] with AMF. For the M plants, the substrate was half mixed with the AMF-inoculum substrate above, whereas for the NM treatments, the substrate was half mixed with the same AMF-inoculum substrate above but sterilized (121°C for 15 min). The plants were grown in the substrates for two months under greenhouse conditions set at 24/22°C (day/night), RH of 50%, photoperiod of 16 h day^–1 and LED light (lumigrow) intensity of about 220 μmol m^–2 s^–1.

In Experiment 1, the objective was to evaluate the production of A/Sd and the expression of genes involved in the biosynthetic pathways of A/Sd in A. tinctoria roots colonized (later abbreviated for fluency as M^irr) or not (NM) with R. irregularis MUCL 41833, before (T0) and after 9 (T1) and 37 days (T2) of growth in the S-H cultivation system. These harvesting times were based on the study of Cartabia et al. (2021), conducted with Anchusa officinalis (another Boraginaceae). The authors made a kinetic study (before the transfer of the plants to the S-H cultivation system and after 9 and 30 days of growth in the S-H cultivation system) on the effect of R. irregularis MUCL 41833 on the metabolites profile of A. officinalis. Interesting results were obtained at these harvesting times, and therefore were considered in the present study.

Two-month-old M^irr and NM plants were gently removed from the pots above and their roots cleaned with demineralized water to eliminate substrate debris (see Supplementary Figure 5). Six plants belonging to the M^irr and NM treatments were randomly harvested, and AMF root colonization assessed before (T0) transfer to the S-H cultivation system (see section “2.7. Plants biomass and AMF root colonization”). Fourteen additional plants from the M^irr and NM treatments were transferred to the S-H cultivation system as detailed in Cartabia et al. (2021). The plants were placed in 500 ml plastic bottles (VWR INTERNATIONAL, Leuven, Belgium), cut at the base and covered with a 100 μm size pore nylon mesh (Prosep B.V.B.A., Zaventem, Belgium) glued on the top. The bottles (called containers thereafter) were used bottom-up, filled with 32 g of perlite (KNAUF GMBH, Dortmund, Germany), covered with a superficial layer of black lava rock (1–3 mm) and wrapped in aluminum foil to avoid algae development and inhibition of shikonin production (as shikonin is sensitive to light) (Yazaki, 2017). The containers were transferred randomly in holes made in flex foam supports and were maintained in the greenhouse set at the same conditions as described above. A 90% P-impoverished modified Hoagland solution (see Garcés-Ruiz et al., 2017; and Supplementary Table 2) lacking ammonium nitrate (NH₄NO₃) [as NH₄⁺ is an inhibitor of shikonin production (Yazaki, 2017)] was used at two different concentrations as detailed in Cartabia et al. (2021). After 7 days of acclimatization, the circulatory system was started. Each container was connected to a 1 L glass bottle covered with aluminum foil containing Hoagland^dil100X solution. A 4.6 mm diameter black supply pipe (GARDENA^®, Micro-Drip System, Ulm, Germany) connected a dropper cap fixed on the bottom of the plant container with the glass bottle, and another black pipe of the same diameter connected the glass bottle to the upper part of the plant container via a multichannel peristaltic pump (Gilson’s Minipuls Evolution, France). Once the circulation started, the nutrient solution was pumped from the glass bottle into the plant container and the liquid flowed back by gravity into the bottle. Before starting with the circulation, initial flushing was performed as described in Cartabia et al. (2021). Then, four successive circulations were performed at a velocity of 7.5 ml min^–1 for different durations: 42 h at day 9 (T1), 8 h at day 17, 24, and 31, and finally 42 h at day 37 (T2).

In Experiment 2, the objective was to evaluate the production of A/Sd and the expression of genes involved in the biosynthetic pathways of A/Sd in A. tinctoria roots colonized (later abbreviated for fluency as M^irr, M^aggreg, M^Rhiz, and M^Sept) or not (NM) with R. irregularis MUCL 41833, R. aggregatus MUCL 49408, the native R. irregularis and S. viscosum and grown for 9 days in the S-H cultivation system. This harvesting time was based on the study of Tsiokanos et al. (2022) where different AMF strains were associated to A. officinalis, and plants were harvested after 9 days of growth in the S-H cultivation system. Interesting results were obtained following this harvesting time and therefore considered in the present study. The same procedure as in Experiment 1 was applied. Briefly, after cleaning the roots system, 2-month-old plants (six per treatment) were transferred in the S-H cultivation system, and the containers were randomly placed in the holes made in the flex foam supports. Then, after the acclimatization period of 7 days, a regular circulation was initiated and maintained at 7.5 ml min^–1 for 42 h.

In Experiment 3, the objective was to evaluate the production of A/Sd and the expression of genes involved in the biosynthetic pathways of A/Sd in A. tinctoria roots colonized (M^irr, M^Rhiz, and M^Sept) or not (NM) with R. irregularis MUCL 41833, the native R. irregularis and S. viscosum and growing for 85 days in pots under greenhouse conditions. In this Experiment, R. aggregatus MUCL 49408 was not considered due to inoculum limitation. Six plants per treatment were gently removed from the 1 L pots above and transplanted in 3 L pots (3 × 18.5 × 17 cm) containing a sterilized (2 × 121°C for 15 min) substrate mixture (2 peatmoss/2 compost/1 perlite/1 quartz 0.4–0.8 mm/1 quartz 1–2 mm). They received 200 ml Hoagland^dil200X solution and were randomly moved every week until harvest. The plants were grown under the same conditions as for the S-H cultivation system in Experiments 1 and 2.

For the three experiments, plants were harvested and biomass as well as AMF root colonization evaluated. In Experiment 1, plants were harvested before transfer to the S-H cultivation system (T0) and after 9 (T1) and 37 days (T2) corresponding to 57 and 94 days after AMF inoculation of the plants, respectively. In Experiment 2, they were harvested after 9 days (i.e., 79 days after AMF inoculation of the plants), and in Experiment 3, 85 days after AMF inoculation of the plants.

Whatever the experiment, shoots were separated from roots (cleaned under demineralized water and gently dried). Shoot fresh weigh (SFW) and roots fresh weight (RFW) were then measured. Root colonization was further evaluated by McGonigle et al. (1990). Roots (an approximate of 2 g fresh material) were cut into circa 1 cm length pieces and placed in 50-ml Falcon tubes. Twenty-five milliliter of KOH 10% was added to the roots before incubation at 70°C for 40 min. The KOH solution was then removed and replaced with H₂O₂ 3.5% before incubation at 70°C for 5 min. The roots were then rinsed with demineralized water and HCl 1% was added for 1 min at room temperature. After discarding the solution, the roots were stained with ink 2% (Parker Blue Ink, United States) in HCl 1% (Walker, 2005) by placing the tubes at 70°C for 30 min. The roots were finally rinsed with demineralized water and kept in lactoglycerol (lactic acid/glycerol/distilled water, 1:1:1, v/v) until observation. For colonization assessment, the root fragments were placed on microscope slides and covered with 40 × 22 mm coverslips before observation under a bright field light microscope (Olympus BH2-RFCA, Japan) at ×10 magnification. Around 200 root intersections were observed for each plant in Experiments 1 and 2 and around 300 intersections for the plants in Experiment 3. The total colonization percentage (TC%) of roots (e.g., hyphae, arbuscules, and vesicles/spores), and arbuscules colonization percentage (AC%) was further calculated (McGonigle et al., 1990).

The remaining root system of each replicate was divided in two parts of around 5 g each for metabolites and gene expression analysis. Regarding the analysis of metabolites, roots were freeze-dried during 72 h and kept at −80°C before proceeding with A/Sd extraction (see section “2.8. Quantitative and qualitative analysis of A/Sd in A. tinctoria roots”). For analysis of gene expression, roots were transferred within 5–10 min after sampling into liquid nitrogen and kept at −80°C before proceeding with the RNA extraction (see section “2.9. Analysis of A/Sd target genes expression in A. tinctoria roots”).

For all the experiments, a one-way ANOVA followed by HSD Tukey post-hoc test (p < 0.05) was applied to discriminate between means for plant growth parameters (i.e., SFW, RFW, and TFW), AMF colonization (i.e., TC% and AC%), A/Sd content, and relative genes expression at different time points (Experiment 1) or different AMF treatments (Experiments 2 and 3). Moreover, in Experiment 1, differences between M and NM treatments were highlighted by pairwise comparison with Bonferroni correction (p < 0.05) at each harvesting time (T0, T1, and T2). For all parameters, normal distribution of residuals variance and normality was checked before analyses. Non-normal data were normalized by log10 transformation before analysis. Data analyses were performed by IBM SPSS Statistics for Windows, version 28 (IBM Corp., Armonk, N.Y., USA).

In Experiment 1, no significant differences between harvesting times (before plants transfer to the containers-T0, and after 9-T1 and 37-T2 days of growth in the S-H cultivation system) were noticed for SFW, RFW, and TFW of plants in the M^irr treatment and for SFW of plants in the NM treatment. Conversely, a significantly greater RFW was noticed at T1 and T2 as compared to T0, and greater TFW at T2 compared to T0 for plants in the NM treatment. Irrespective of the harvest time, no significant difference was reported between the SFW of plants in the M^irr and NM treatments, while at T2, the RFW and TFW of the plants in the NM treatment was significantly greater than that in the M^irr treatment. In the M^irr treatment, a significant decrease in TC% was observed at T2 compared to T0 and T1 and in AC% at T1 and T2 compared to T0 (Table 1). No root colonization was observed in plants of the NM treatment.

In Experiment 2, no significant differences were observed in SFW, RFW and TFW and in TC% between the different treatments (M^irr, M^aggreg, M^Rhiz, M^Sept, and NM) after 9 days of growth of the plants in the S-H cultivation system. Conversely, a significant higher AC% was observed for the plants in the M^Rhiz treatment compared to those in the M^aggreg treatment, while intermediate values were noticed in the plants of the M^Sept and M^irr treatments (Table 1). No root colonization was observed in plants of the NM treatment.

In Experiment 3, no significant differences were observed in SFW and RFW of plants between the different treatments (M^irr, M^Rhiz, M^Sept, and NM) after 85 days of growth in pots. Similarly, no significant differences were observed for TC% between the different treatments, while a significantly higher AC% was observed for the plants in the M^Sept treatment compared to those of the M^irr treatment. Plants in the M^Rhiz treatment had intermediate value (Table 2). No root colonization was observed in plants of the NM treatment.

In Experiment 1, whatever the harvesting time (T0, T1, or T2) or treatment (M^irr or NM) no significant differences in content of A/Sd were noticed (Table 1).

In Experiment 2, a significant higher content of total A/Sd expressed as shikonin equivalent was observed in the plants of the M^Rhiz treatment compared to those in the M^irr, M^agreeg, and NM treatments, while an intermediate value was observed for the plants in the M^Sept treatment (Table 1).

In Experiment 3, a significantly higher content of shikonin, whose values were nevertheless below the LOQ (Supplementary Table 2), and in total A/Sd expressed as shikonin equivalent, was reported in the roots of the plants in the M^Rhiz treatment compared to those in the M^irr and NM treatments, while an intermediate value was observed for the plants in the M^Sept treatment (Table 2).

In Experiment 1, whatever the time of evaluation (T0, T1, or T2) or treatment (M^irr or NM) no significant differences were noticed in gene expression of GHQH. Similarly, no significant differences between harvesting times were observed in expression of LePGT1 gene of the plants in the M^irr treatment and expression of LePGT1 and LePGT2 genes of the plants in the NM treatment. Conversely, a significantly higher expression of LePGT2 gene was observed at T2 compared to T1 for the plants in the M^irr treatment, while an intermediate gene expression was noticed at T0. A significant higher relative expression of both LePGT1 and LePGT2 genes was observed in the plants of the M^irr treatment compared to those in the NM treatment at T2 (Figure 1A).

In Experiment 2, whatever the treatment (M^irr, M^aggreg, M^Rhiz, M^Sept, and NM), no significant differences were observed in expression of GHQH and LePGT1 genes. A significant lower relative expression of LePGT2 was observed in the plants of the M^Rhiz treatment compared with the other treatments that did not differ among them (Figure 1B).

In Experiment 3, LePGT2 target gene was not considered due to contamination by the primer dimer. Whatever the treatment (M^irr, M^Rhiz, M^Sept, and NM), no significant differences were observed in expression of GHQH and LePGT1 genes (data not presented).

To better recognize the metabolites produced in A. tinctoria roots, the main chemical compounds were tentatively identified in Experiments 2 and 3 by performing a dereplication strategy based on HPLC-HRMS/MS analysis (Figure 2 and Tables 3, 4) and molecular network representation (Figures 3–5). These analyses were not performed in Experiment 1, since no significant differences were reported in the production of A/Sd.

In Experiment 2, the HPLC-HRMS/MS analysis was performed in positive and negative modes. However, shikonin was not detected in any root sample analyzed. Dereplication analyses of shikonin derivatives were done on positive mode to compare our results with published data available in the literature (Bossard et al., 2022). The major detected compounds were organized in two distinctive clusters: A/Sd naphthoquinones (cluster A, Figures 3A and Table 3) and lipid amides derivatives (cluster B, Figure 3B and Table 3). In cluster A, the first eluted derivative was putatively identified as a methylshikonin isomer (1) which gave a protonated molecular ion [M + H]⁺ at m/z 303 and fragmented to m/z 285, 243, and 233, corresponding to the loss of a H₂O molecule, a C₂H₄O₂ and a C₅H₁₀ fragment from the alkyl chain. Similar fragmentation pattern was observed for compound 4 with a protonated dehydrated molecular ion [M + H-H₂O]⁺, putatively identified as a second methylshikonin derivative. Similar loss was observed for compounds 3 and 6, putatively corresponding to methyl-1′-deoxyshikonin and anhydroalkannin at m/z of 287 and 271, respectively. Data also indicated that the major shikonin derivative produced by the plants under these conditions was anhydroalkannin (6) (Kyogoku et al., 1973; Bai and Jin, 1994). Moreover, a group of lipid amides were also putatively detected from the A. tinctoria roots samples and represented in cluster B. These amides derivatives were tentatively identified as palmitoleamide (7), linoleamide (8), oleamide (9), and stearamide (10) with protonated molecular ions [M + H]⁺ at m/z 254, 280, 282, and 284, presenting a loss of a fragment of m/z 17, corresponding to the loss of an ammoniac molecule (Divito et al., 2012). Other isomeric compounds were also detected by HPLC-HRMS/MS, but in the absence of their fragmentation patterns, putative identities were not proposed. No significant differences were noticed in the metabolites’ production profile of A. tinctoria roots associated or not with different AMF strains (Figure 4).

In Experiment 3, shikonin was detected only in negative mode and this result was corroborated by co-injection with shikonin standard solution (Table 4). Due to the important differences in the fragmentation spectra in positive mode of A/Sd, clusters of these compounds were not formed (Figure 5). In addition, the abovementioned shikonin derivatives isomers (1 and 6) were also observed as deprotonated molecular ion [M − H]^– at m/z 301 with fragmentation signals at m/z 286 and 232 corresponding to the loss of a CH₃ and C₅H₉ groups, for 1, and at m/z 269 with fragment loss at m/z 251 and 241, corresponding to the loss of a H₂O and CO molecules, for 6. The deprotonated molecular ion [M − H]^– of shikonin (11) was observed at m/z 287 with fragmentation signals at m/z 218 and 190, corresponding to the loss of C₅H₉ and the subsequent loss of a CO group (Liao et al., 2015). Moreover, an additional shikonin derivative was observed with a deprotonated molecular ion [M − H]^– at m/z 329, putatively identified as acetylshikonin (13), with fragmentation loss at m/z 269, 251, and 241 corresponding to a McLafferty rearrangement (i.e., neutral loss of acetic acid), and the subsequent loss of a H₂O and CO molecules (Liao et al., 2015). Three others unidentified compounds (12, 14, and 15) were detected under negative ionization mode with deprotonated molecular ion [M-H]^– at m/z 797, 1111, and 1147. However, no clusters were observed between compounds 1, 6, 11, and 13 (shikonin derivatives) and for 12, 14, and 15 (unknown compounds) because they show a low degree of similarities in their fragmentation spectra (in positive ionization mode).

Slight differences in metabolites production were observed between Experiments 2 and 3, when samples were analyzed in the same ionization mode. Nevertheless, the most important difference between the two experiments was the detection of shikonin only in Experiment 3. Analyses performed in negative mode for root samples of Experiment 2 did not show the presence of shikonin. Similarly, in Experiment 3, no significant differences were observed in the metabolites’ production profile of A. tinctoria roots associated or not with different AMF strains, thus producing similar chromatographic profiles.

Whatever the growth system (S-H or pot), an increase in biomass of A. tinctoria was observed in the presence as well as absence of AMF. No significant difference was observed in SFW between the plants in the M and NM treatments after 9 days in the S-H cultivation system (Experiments 1 and 2) or after 85 days in the pots cultivation system (Experiment 3). However, a significantly lower RFW and TFW was reported for the plants in the M treatment compared to those in the NM treatment after 37 days in the S-H cultivation system (Experiment 1), unlike the results obtained with A. officinalis colonized by R. irregularis MUCL 41833 in the same S-H cultivation system (Cartabia et al., 2021). These results are not surprising as it has often been reported that plant growth response to AMF inoculation can vary among AMF species and that the direction (e.g., increase or decrease in plant biomass) and magnitude of the response strongly depends on the combination of plant and AMF taxa (Klironomos, 2003).

Root colonization was observed in each AMF-inoculated plant. However, colonization measured at transfer and after 9 and 37 days in the S-H cultivation system (Experiment 1), decreased steadily, probably because of damages caused to the extraradical mycelium at transfer to the S-H cultivation system with no full recovery and development within the roots throughout the experiment (Cartabia et al., 2021). It is not excluded that this is also a reason for the lower plant biomass reported at the end of Experiment 1 compared to the control, due to an imbalance between the carbon resources transported from plant to the fungus in exchange for nutrients transported from fungus to plant. Interestingly, in both Experiments 2 and 3, the total root colonization was almost similar among the AMF strains, while the percentages of arbuscules were higher in presence of the native R. irregularis and S. viscosum strains, suggesting that they were better adapted to A. tinctoria than the GINCO strains. This has often been reported by comparing commercial strains with local strains; the latter being more prone to rapid and effective colonization (Chenchouni et al., 2020; Wu et al., 2021). Unfortunately, this was not translated into higher plant biomass, probably because of the same justification explained above.

The production of A/Sd was not impacted by the GINCO strain R. irregularis MUCL 41833 (Experiment 1), whereas a significant increase in production was observed with the native R. irregularis strain (Experiments 2 and 3) and an intermediate production between this strain and the GINCO strains with the other native strain S. viscosum (Experiment 2). Similarly, no significant effect was observed with the GINCO strains R. irregularis MUCL 41833 and R. aggregatus MUCL 49408 (Experiments 2 and 3). It cannot be ruled out that these results are related to the fungal genotypes, with the native ones being more prone to stimulate the production of A/Sd than the GINCO ones. The difference between the two native strains further suggests some degree of functional specialization. This is supported by recent studies revealing that various species/strains of AMF can induce different changes in the production of metabolites in the same plant species (Kapoor et al., 2002; Larose et al., 2002; Copetta et al., 2006; Khaosaad et al., 2006; Toussaint et al., 2007; Zubek et al., 2010). For example, Toussaint et al. (2007) showed that Funneliformis caledonium² increased rosmarinic acid and caffeic acid production in O. basilicum, whereas Funneliformis mosseae only increased caffeic acid production. In another study, Jurkiewicz et al. (2010) showed that total phenolic acids accumulation was significantly higher in the roots of Arnica montana L. associated with AMF collected from particular plant’s natural stands (in Kurpie or in Karkonosze), while intermediate increase was observed with AMF isolated from other regions (i.e., R. intraradices UNIJAG PL24-1 and R. intraradices BEG 140 or a mixture composed of the above and of Funneliformis geosporum UNIJAG PL 12-2, Septoglomus constrictum 265-5 Walker and F. mosseae BEG 12). Finally, Frew (2020) reported that the inoculation with a mixture of four commercial species of AMF (Claroideoglomus etunicatum, Funneliformis coronatum, F. mosseae, and R. irregularis) had a greater effect on phenolic concentrations in Hordeum vulgare L. cv. “Hindmarsh” as compared to a single commercial inoculant (R. irregularis). Interestingly, the author also demonstrated that the effects of the commercial mixture was not different from the results obtained with a native multi-species AMF inoculant extracted from field soil. This suggested that a commercial AMF mixture provided little to no additional benefits (Frew, 2020). Conversely, several specialized compounds produced by plants may be synthesized as a chemical defense against the presence of AMF in the roots (Copetta et al., 2006). However, this defense mechanism would be likely to occur with AMF strains that have never been associated to the plant. Our native strains isolated from the wild A. tinctoria in their natural habitat are more adapted to the plant and are likely to have evolved as good associates. Natives R. irregularis and S. viscosum could be better adapted to the presence of A/Sd, which are characterized by different biological activities and by various effects on soil microorganisms (e.g., antifungal activities), and thus better interact with their host plants to regulate and enhance the production of these important therapeutic metabolites (Yan et al., 2019). As reported above, a significant higher AC% was observed with M^Rhiz and M^Sept treatments in Experiments 2 and 3, respectively. An increase in nutrients uptake via the arbuscules can lead to an enhanced production of precursor compounds, such as NADPH, ATP, acetyl-CoA (mevalonic acid pathway), and pyruvate glyceraldehyde and phosphate (methylerythritol-4-phosphate pathway) that are required for the biosynthesis of various SMs (e.g., terpenoids, phenolic, and alkaloids) (Kapoor et al., 2017). In our study, a higher AC% could have led to an increase in the precursor compounds essential to produce A/Sd. However, this was not confirmed in Experiment 3, where the higher AC% in M^Sept treatment was not accompanied by a higher production of A/Sd. If we parallel this to plant biomass, it is well known that the response of plants to inoculation with an AMF may vary depending on the AMF species as previously reported (Klironomos, 2003). It is not excluded that similar effect is observed on metabolites production, although this needs to be demonstrated.

At the end of Experiment 1, a significantly higher relative expression of LePGT1 and LePGT2 genes was observed in A. tinctoria associated with R. irregularis MUCL 41833 compared to the NM treatment. In addition, a significantly higher relative expression of LePGT2 was observed in the AMF-colonized plants at day 37 compared to day 9. However, these increases were not accompanied by an enhanced production of A/Sd. Conversely, in Experiment 2, a significant lower relative expression of LePGT2 was observed in plants associated with the native R. irregularis strain, compared with the other AMF and NM treatments, whereas a significantly higher production of A/Sd was observed with this strain. In both Experiments, no differences were reported for GHQH. In Experiment 3, no significant differences between the treatments were noted for all the genes tested (data not presented).

Although these results may seem surprising, similar outcomes were reported in the literature. For example, Wang et al. (2014) demonstrated that the application of methyl jasmonate (MeJA) on shikonin-deficient A. euchroma cell lines rapidly led to the overexpression of several genes involved in the biosynthesis of A/Sd, including PGT, but was not accompanied by a significant higher production of A/Sd. In another study, Ahmad et al. (2022b) showed a significantly higher expression of LePGT1, LePGT2, and LeGHQH1 in Lithospermum officinale L. roots treated with MeJA, but no higher accumulation of total A/Sd, after eight weeks. Conversely, the same authors (Ahmad et al., 2022a) demonstrated that the bacterium Chitinophaga sp. strain R-73072 significantly upregulated LePGT1 and LePGT2, and a cytochrome P450-LeCYP76B101 genes, resulting in a significant enhanced production of total A/Sd in L. erythrorhizon roots, after 2-3 weeks.

In our study, a few hypotheses can be advanced to explain why upregulation of PGTs was not followed by higher production of A/Sd and, conversely, why downregulation was followed by higher production of A/Sd. Firstly, it is not excluded that the duration of Experiment 1 (37 days) was too short to result in a significant increased production of A/Sd, or that the timing of plant harvesting in our experiments was not optimal (i.e., gene expression behavioral oscillations due to the plant circadian cycle) (Merrow et al., 2005). The PGTs regulate the first biosynthetic steps forming the basic carbon skeleton that leads to A/Sd, and therefore, it is possible that the resulting modulation of A/Sd production took longer after the upregulation of the relative expression of target PGT genes, as supported by Ahmad et al. (2022b), or that the genes oscillations were not timely targeted during our harvest times, as suggested by Wang et al. (2014) (Figure 6). Secondly, in a study by Andrade et al. (2013), a mismatch between alkaloid levels, another important SM group, and gene expression was reported in different Catharanthus roseus (L.) G. Don tissues. In this study, the AMF C. etunicatum had a greater influence (i.e., alkaloids production) in roots than in shoots, and a higher gene expression was reported in the older leaves of M plants as well as in the youngest leaves of NM plants (leaves were harvested at the same time) (Andrade et al., 2013). These results suggest a very precise phenological and spatial regulation process during alkaloid biosynthesis (Mahroug et al., 2007). Moreover, the reported influence of AMF on idioblast and laticifer density in C. roseus plants might have enhanced the expression of enzymes specifically located in these cells (St-Pierre et al., 1999). An increase in glandular trichome density upon mycorrhization was also linked with an enhanced concentration of another group of SMs, the terpenoids (Zhao et al., 2022). This is an interesting aspect that needs to be further verified. In fact, A/Sd compounds are sequestered as granules in the phospholipid layer and are accumulated in the apoplastic spaces, and they can be found in the cork layer of mature roots (Brigham et al., 1999; Singh et al., 2010; Tatsumi et al., 2016). A difference in A/Sd accumulation might have occurred (i.e., higher content in the primary roots as compared to the secondary roots) upon AMF colonization and might have not been fully displayed across A. tinctoria root samples used for our RT-qPCR analysis (i.e., no homogenous samples analyzed with mostly secondary roots utilized). Thirdly, the regulation and accumulation of SMs in plants is usually controlled by a complex network characterized by transcription factors (TFs), which promote or inhibit the expression of multiple genes involved in one or more biosynthetic pathways (Yang et al., 2012; Wu et al., 2021). Transcription factors can act alone or in combination with other TFs to modulate the expression of target genes, and also one TF can regulate the expression of multiple genes participating in one or more biosynthetic pathways (Pinson et al., 2009; Goossens et al., 2017; Hassani et al., 2020). In our study, different TFs might have regulated the target genes (PGTs and GHQH), yet the full landscape of the A/Sd biosynthesis pathway is not entirely understood, with co-expressed genes still to be identified (Suttiyut et al., 2022). Finally, even after a gene has been transcribed, its expression could still be regulated at various stages. Post transcriptional modifications changes might occur to a newly transcribed primary RNA transcript after transcription has occurred and prior to its translation into a protein.

A similar chemical composition was reported in A. tinctoria roots associated with the different AMF and non-colonized control plants in Experiments 2 (S-H cultivation system) and 3 (pot cultivation system) (Figures 4, 5) except for the detection of shikonin in Experiment 3 (Tables 3, 4).

The major compound detected in both experiments was putatively identified as anhydroalkannin (6). This compound is also a main product of shikonin biotransformation by several human intestinal bacteria in aerobic conditions and it was found to be less cytotoxic against a series of human tumor cell lines in vitro, in comparison with shikonin (Meselhy et al., 1994; Min et al., 2000). Methyl A/Sd isomers, as detected in our samples, were also identified from Lithospermum, Alkanna, and Onosma species. 1′-methyl shikonin was isolated and characterized from roots of L. erythrorhizon, exerting moderate antioxidant activity (Han et al., 2008). To the best of our knowledge, 5- or 8- O-methoxyshikonin derivatives have not yet been identified from natural sources. To better clarify the chemical composition of our samples, putative methylshikonin derivatives (1 and 4) must be purified and the methoxyl position elucidated in the structure of these isomers. In addition, in both experiments, four main lipid amides were tentatively identified as palmitoleamide (7), linoleamide (8), oleamide (9), and stearamide (10). Fatty acids amides are a group of nitrogen-containing, lipid-soluble fatty acid derivatives, which act against a variety of diseases such as cancer, bacterial infections, parasitic infection, inflammations, diabetes, and obesity (Kim et al., 2010; Tanvir et al., 2018). Further, it has been reported that fatty acid amides as oleamide from plant root exudates can participate in strong plant-microbe interactions, stimulating nitrogen metabolism in rhizospheric bacteria (Sun et al., 2016).

The data of chemical composition obtained from A. tinctoria growing in our cultivation systems cannot be accurately compared with data from literature (i.e., from nature, commercial samples, and/or from cell plant suspension). Indeed, the proportion and quantity of A/Sd varies depending on the level of stress and microorganisms present in the rhizosphere (Brigham et al., 1999). The influence of the cultivation systems on the production of these methoxy and anhydro A/Sd must be deeply studied to better understand their significance. Secondary metabolites’ production and compositional changes have a strong correlation and association with the environment, and thus synthesized only under specific growth conditions (Peñuelas and Llusià, 1997). Variations in an environmental factor, such as light, temperature, soil water, soil fertility, and salinity, may alter the plant metabolites content (Yang et al., 2018). Indeed, many chemical and physical factors have been found to inhibit shikonin production, such as NH₄⁺, 2,4-dichlorophenoxyacetic acid (a synthetic auxin), low pH, temperature higher than 28°C, and light, especially blue light (Tatsumi et al., 2016). Skoneczny et al. (2017), using a metabolomic approach profiling hydroxynaphthoquinones (HNQs) and pyrrolizidine alkaloids (PAs), demonstrated the influence of high temperatures and water withholding on the accumulation of A/Sd in Echium plantagineum L. Abundance of HNQs, especially deoxyshikonin, shikonin, and dimethylacrylshikonin, rapidly increased in roots exposed to elevated temperatures. Water withholding initially increased NQ abundance, but prolonged drought resulted in reduced total PAs and HNQs (Skoneczny et al., 2017). In our study, A. tinctoria growing in the S-H cultivation system were kept in perlite, while a mixture of peatmoss, perlite, and quartz was used in the pots experiment. Perlite is a growing medium frequently used in hydroponic cultivation system since it has a high-water retention and provide the plants’ roots with strong anchor points for stability and strength. Moreover, in this system, plants received minerals by the circulating nutrient solution flowing directly through the plants’ container, while in the pots it was added at constant intervals and left to be completely absorbed by the plants. Therefore, it is possible that the temperature and the water/nutrient solution retention was higher in the conventional pots than in the plant containers of the S-H cultivation system. Moreover, shikonin and A/Sd might have better accumulated in the pots, while leached out in the S-H system. Finally, plant developmental stage qualitatively and quantitatively influences primary and secondary metabolism. A recent study by Csorba et al. (2022) reported this aspect as the most important driver influencing A. tinctoria metabolites content, revealing a peak content of A/Sd at the fruiting stage. In our study, plants were harvested during the vegetative growth (around five months old), and thus, might have played a role in the chemical profile reported.