Four viable and mature in vitro spores of R. irregulare were deposited on four cardinal directions in Petri dishes filled with water + 3.5% Gelrite^TM (Duchefa Biochemie, The Netherlands), containing or not ABA (1 mM), Ga3 (1 mM), SHAM (1 or 5 mM) or KCN (1 or 5 mM). Six plates were tested for each treatment and incubated in dark at 28°C in upside down position. At 24 days after inoculation (DAI), hyphal germination patterns were observed and the germination rate was calculated. Spore viability was also assessed by iodonitrotetrazolium salt (INT) and spores, which appeared red due to the presence of formazan, were counted. INT is a marker of mitochondrial activity (Walley and Germida, 1995; Mukerji et al., 2002). It is known that INT reduction (the conversion of INT to formazan by two electrons and two protons) is connected to the electron chain transport (Berridge et al., 2005) but the exact process within cell is not yet clearly identified.

The aim of this experiment was to study in a small experimental set the variation of some genes involved in the mitochondrial electron chain and fermentation associated with mycorrhizal performances. In vitro S. tuberosum plantlets (cv K19-99-0012) were grown in growth chambers [20°C/17°C (day/night), 16 h d^-1 photoperiod, 70% relative humidity and 55 μmol m^-2s^-1 photon flux density] on MS medium (Murashige and Skoog, 1962), supplemented with 20 g⋅l^-1 sucrose, 3.5 g l^-1 Gelrite^TM (Duchefa Biochemie, The Netherlands) adjusted to pH 5.6 before autoclaving (121°C for 15 min). Plantlets were pre-treated or not with ABA, a known promoter of the AOX pathway, added in culture medium to 0.1 mM final concentration.

Ten-day-old in vitro plantlets were transplanted in 1 L pots containing 100% sterilized sand (baked twice in a dry oven, at 120°C for 6 h). Plants were inoculated (M) or not (NM) at transplanting time with 100 in vitro spores of R. irregulare INOQ strain QS69 placed at the vicinity of the root and grown in a glasshouse [Loitze, Germany; 32°C/25°C (day/night), natural light and day (July–August)]. In this way, no direct contact between AMF and ABA treatment could occur. Plants were fertilized once a week with 50 ml of a modified Hoagland’s solution without P (1 ppm P was mixed in the sand during pot preparation). The watering was performed when needed with the same volume for all plants. Precautions were taken to avoid watering the shoots. Potato plant and soil harvesting were conducted at 8 weeks after inoculation (WAI). Plant growth parameters such as the shoot, root, and tuber FW were measured. The evaluation of mycorrhizal development was performed according to Trouvelot et al. (1986) method after root staining with China Ink (Vierheilig et al., 1998). The MGD was calculated for several potato parameters (root, shoot, yield and total biomass FW) according to the following formula (Plenchette et al., 1983): [100^∗((M -NM)/M)], expressed as percentage.

A multifactorial experiment was designed to test the effects of different P concentrations and respiratory chain inhibitors on mycorrhizal and plant parameters. Mycorrhizal inoculum was produced in bed cultures with a mix of three plant species (Trifolium pratense, Zea mays, and S. tuberosum) in sterile sand, containing 92,000 propagules/l [determined by the Most Probable Number of mycorrhizal propagules (MPN) test, Gianinazzi-Pearson et al., 1985]. In vitro potato plantlets were grown in growth chambers as described in the Experiment 2 before being transplanted to a glasshouse [Loitze, Germany; 32°C/25°C (day/night), natural light and day (July–August)] and grown for 8 weeks in 1 L pots containing 100% sterilized sand (baked twice in a dry oven, at 120°C for 6 h). Plants were inoculated or not with R. irregulare (INOQ strain QS69) at transplanting time, by mixing 4% of a mycorrhizal inoculum containing spores and mycorrhizal root fragments in the growth substrate.

Five P concentrations were tested: 1, 10, 50, 100, and 300 ppm P (respectively, 0.032, 0.323, 1.614, 3.228, and 9.687 mM as final concentration), corresponding to the concentrations of practical reality in crop field soils (50, 100, and 300 ppm) and in mycorrhizal production under glasshouse conditions (1 and 10 ppm). To achieve those concentrations, KH₂PO₄ was mixed directly into growth substrate. To test the contribution of each respiratory pathway in mycorrhizal and non-mycorrhizal plants at each P concentration, two respiratory chain inhibitors were used: KCN and SHAM (0.1 mM), which inhibit COX and AOX, respectively. The inhibitors were dissolved in sterile water and added to substrate at 08:00 a.m. in one application 7 DAI, which corresponds to the end of the acclimatization period. The experimental design for the inhibitor studies included (1) non-treated plants, non-inoculated (NM) and inoculated plants (M); (2) KCN treated plants that were non-inoculated (KCN) or inoculated (M KCN) plants; (3) SHAM treated plants that were non-inoculated (SHAM) or inoculated (M SHAM) plants. Plants were fertilized once a week with 50 ml of a modified Hoagland’s solution without P, and were watered as needed with the same volume for all plants. Precautions were taken to avoid watering and treating the shoots.

Potato plant and soil harvesting were conducted at 8 WAI. Plant growth parameters including the shoot, root, and tuber FW, were measured. The evaluation of mycorrhizal development was performed according to the Trouvelot et al. (1986) method after root staining with China Ink (Vierheilig et al., 1998). Parameters investigated (by observing with a microscope 30 root fragments slide per sample) included frequency of mycorrhiza (F %), intensity of mycorrhiza, arbuscule, vesicle and intraradical hyphal colonization in whole root system (respectively, M %, A %, V %, and H %) and within mycorrhizal root fragments (respectively, m %, a %, v %, and h %). The spore production was evaluated in M plants for each repetition by isolating and counting spores after performing the wet sieving method (Gerdemann and Nicolson, 1963) using 3 × 10 g of dried substrate. The MGD was calculated for several potato parameters (root, shoot, yield, and total biomass FW) according to the following formula (Plenchette et al., 1983): [100^∗((M – NM)/M)], expressed as percentage. The % FW of root, shoot or tuber was calculated as follows: [Plant organ (root, shoot part or tuber FW)/Total FW] × 100.

Transcript accumulation analyses of plant and fungal genes involved in respiration, nutrient transport and fermentation were performed. Genes of interest are listed in Supplementary Table S1. To obtain the sequences of some targeted genes, a database search was performed using the R. irregulare and S. tuberosum genomes databases^2,3 and the INRA Glomus database⁴. Full-length amino acid sequences of R. irregulare, S. tuberosum and those from fungi and plants were acquired from the JGI database⁵ and GenBank⁶ and were aligned by CLUSTALW and imported into the Molecular Evolutionary Genetics Analysis (MEGA) package version 6 (Tamura et al., 2013). Phylogenetic analyses were conducted using the neighbor-joining (NJ) method implemented in MEGA with the pairwise deletion option for handling alignment gaps and with the Poisson correction model for distance computation. Bootstrap tests were conducted using 1000 replicates.

For both glasshouse trials, 100 mg of root samples were stored in RNAlater (Qiagen) solution at the time of harvest. Total RNA was extracted from 100 mg of roots conserved in RNAlater (Qiagen) using the RNeasy Plant Mini kit (Qiagen). cDNA synthesis was performed with an oligo(dT) primer (Promega) and reverse transcriptase (Masterscript^TM Kit, 5 Prime, Germany) using 250 ng of total RNA. The cDNAs were 1:10 diluted and amplified using the 7500/7500 Fast Real-Time PCR System (Applied Biosystems^®, USA). Amplification reactions were prepared using a SYBR Green PCR Master kit (Maxima) SYBR Green/ROX qPCR Master Mix (Thermo Scientific) using the following concentrations: 6 μl of PCR water, 9 μL of 2x Maxima SYBR Green/ROX qPCR Master Mix (Thermo Scientific) 0.5 μL of each forward and reverse primer (10 mM) and 2 μl of the cDNA template. Three independent biological replicates were analyzed per treatment and each sample was analyzed in duplicate. The specificity of the different amplicons was checked by a melting curve analysis at the end of the amplification protocol. Several candidates were evaluated for further use as reference gene for normalization of the transcript data of target genes, comprising a set of housekeeping genes, rRNA genes and other sequences (data not shown). After evaluation of expression stability using the applications BestKeeper and NormFinder (Andersen et al., 2004; Pfaffl et al., 2004), two genes for potato (StEF1α and StUbc, Gallou, 2011, Supplementary Table S1) and one genes for R. irregulare (GiICL, Lammers et al., 2001, Supplementary Table S1) were chosen as reference genes for our experimental conditions. Expression of target genes was evaluated by efficiency corrected relative quantification for R. irregulare genes and using the geometric normalization factors for potato where two genes are used for normalization (Pfaffl, 2001; Vandesompele et al., 2002). Standard curves of a fourfold dilution series from pooled cDNAs were used for PCR efficiency calculations. All primers used are listed in Supplementary Table S1.

In the Experiment 1–3, differences in plant and fungal growth parameters and gene expression between treatments were examined by a one-way analysis of variance (ANOVA), after log or arcsin transformation of values as indicated in figure legends. Dunnett’s test was conducted to identify significant differences (P < 0.05, symbolized by stars) compared to a specified standard control and Duncan’s multiple range tests were performed to identify significant differences (P < 0.05, symbolized by letters) among P concentrations (after standardization against specified control). Data analysis was performed with the SAS enterprise guide 4.1 (SAS Institute Inc., Cary, NC, USA).

In the Experiment 3, the relationships between P concentration and plant or fungal parameters were investigated by least squares stepwise multiple linear regression with replication using experiment-wise type I error rates of 0.05 for coefficients calculated using the Dunn–Šidák method (Ury, 1976). For each fungal parameter, the complete candidate model included up to the third degree of P concentration or of ln P, three qualitative variables binary coded as 0 or 1 for control, KCN, SHAM plus all first and second level interactions between powers of P and qualitative variables. Whenever considered necessary after graphical exploration of the data, modeling was done with fungal variables or P concentration logarithmically, arcsin transformed or not with the major criteria for model selection being the coefficient of determination (data not shown). Lack of fit was tested for P = 0.05 and coefficients of determination (R²) are presented as proportion of the maximum R² possible (Draper and Smith, 1998). Linear regressions and analyses of variances were done with Statgraphics 4.2 (STSC, Inc., Rockville, MD, USA), all other statistics used in regression were performed in Excel^® (Microsoft Corporation). Least squares stepwise multiple linear regression data are given in Supplementary Table S2 (under each fungal parameter) and Supplementary Table S3 (under each plant parameter).

Compared to non-treated spores, ABA treatment decreased the germination rate (Figure 1A) but increased significantly the viable spore fraction (containing formazan crystals, Figure 1B). Ga3 treatment induced an opposite effect (Figures 1A,B). Both SHAM and KCN, at 1 and 5 mM, inhibited spore germination (Figure 1A), but induced an opposite reaction on INT reduction ability (Figure 1B): red stained spores rate was higher in KCN (for both concentrations) than in SHAM at 1 and 5 mM, and also than in non-treated spores (with 5 mM KCN). Then, ABA induced a significant hyphal branching pattern around the germinated spores (Figures 1C,E) compared to spores treated or not with Ga3 (Figures 1C,D).

The bioinformatic analyses conducted in this study revealed that the AOX family of S. tuberosum is composed of three AOX1 and one AOX2 sequences, called, respectively, StAOX1a, StAOX1b, StAOX1d and StAOX2 according to the AOX classification proposed by Costa et al. (2014). All StAOX isoforms were expressed but harbored tissue specificity in our experimental conditions (Supplementary Figure S3). The R. irregulare genome contained one single AOX sequence (RiAOX).

Two isoforms encoding for cytochrome c (StCytc1 and StCytc2) and COXVb (StCOXIVb1 and StCOXIVb2) were found in S. tuberosum genome, but StCytc2 was not expressed in the roots and therefore it was not considered for further analysis (Supplementary Figure S3). The R. irregulare genome contained one single gene encoding for cytochrome c (RiCytc) and COXVb (RiCOXIVb).

The expression of six (StPT1, 2/6, 3, 4, 5, and 8) out of the ten previously described PT genes belonging to the Pht1 family in S. tuberosum (Leggewie et al., 1997; Rausch et al., 2001; Nagy et al., 2005; Chen et al., 2014) was studied. The expression of StPT2 and StPT6 was not dissociable and was considered as a sum StPT2/6. These two genes shared a very high sequence similarity and are tandemly organized on chromosome 3. On chromosome 6, only two complete sequences were found in potato: StPT7 sequence was incomplete and could not be reconstructed in silico. StPT8, StPT9, and StPT10 were all located in the same region of the chromosome 9. The expression of StPT8, which shares the highest homology with SlPT7 (Chen et al., 2014), was studied. In R. irregulare, the Pht1 family was divided into four different clusters named after the S. cerevisiae sequences therein (Supplementary Figure S4). ScPHO84-like cluster groups had putative high affinity P transporters as ScPHO84 and contained four putative R. irregulare sequences previously described (RiPT1, RiPT2, RiPT3, and RiPT4; Fiorilli et al., 2013; Walder et al., 2016). We found complete and functional sequences only for RiPT1 and RiPT3. Within the putative low affinity P transporters, grouping with the yeast transporters ScPHO87 and ScPHO90 described by Pinson et al. (2004), only one R. irregulare gene was found (RiPT7). The third cluster groups sequences presented homology with the S. cerevisiae high affinity Na⁺/Pi cotransporter ScPHO89 (Sengottaiyan et al., 2013) and two R. irregulare sequences (RiPT5 and RiPT6). RiPT6 presented an incomplete sequence and no expression in our experimental conditions. The last cluster contained ScPHO88-like sequences, which were much shorter (about 189 aa) and did not present a transport activity. One R. irregulare gene matched with this sequence (RiPT8), but as it did not encode a functional transporter, the transcription of this gene was not monitored in this study.

Concerning mitochondrial P transport, four gene sequences encoding a transporter were found in S. tuberosum genome, and one R. irregulare in genome (Supplementary Figure S5). In potato, three of the four genes were expressed in roots (StMPT1a, StMPT1b, and StMPT3). StMPT2 was found to be expressed only in fruits in our experimental conditions (Supplementary Figure S3).

Then, two ldh and three pdc sequences were found in potato genome (Supplementary Figures S6 and S7). It was noted that transcript level was higher for StLDH expression than for StPDC in root, comparing relative values. One ldh sequence was found in the R. irregulare genome (RiLDH).

Following ABA pretreatment, mycorrhizal structures were observed in the primary adventitious root and even in the stem base (Figures 2A–D), which seems to be a rare case in potato roots under normal conditions (Mercy – non-published observation). AMF increased plant total FW biomass in non-pretreated plants by 1.26-fold (MGD = 20.6%) and in ABA pretreated plants by 7.33 (MGD = 86.4%). Pretreatment with ABA during the in vitro phase induced a long-term effect, since a strong plant growth depression was observed (in NM plants, Figure 2E). Mycorrhizal colonization and arbuscule intensity were significantly promoted within ABA pretreated plants (Figure 2F) compared to non-pretreated plants, consistent with observations performed on tomato plants (Herrera-Medina et al., 2007).

The expression analyses of genes involved in the respiratory chain revealed that StAOX1d and StAOX2 were significantly down-regulated while StCytc1 was significantly up-regulated in M versus NM plants in presence of ABA (Figure 2G). The expression of these genes remained unaffected by the presence of AMF and in the absence of treatment. Comparing to NM and M non-treated plants, ABA treatment induced similar trends of up-regulation for StAOX1d and StAOX2 (although changes are not significant), while StCytc1 was down-regulated (significant when comparing NM plants). Contrasting with this observation, the presence of AMF in ABA plants induced significant down-regulation for StAOX1d and StAOX2 and up-regulation for StCytc1 (significant) and CoxIVb2 (tendency) when compared to non-inoculated ABA plants. It was observed a strong up regulation of StLDH1 induced by ABA compared to non-treated plants, inoculated (tendency) or not (significant). This gene was significantly down-regulated in presence of AMF (within ABA plants group). Regarding RiAOX and RiCytc (Figure 2H), no significant effect induced by ABA was observed, although RiCytc tended to be down-regulated.

Dormancy is often related to AOX pathway in plant seeds, but also in fungi. Several examples have been reported, especially within Mucoromycotina (Cano-Canchola et al., 1988; Salcedo-Hernandez et al., 1994) which are phylogenetically closely related to AMF (Hibbett et al., 2007), showing a respiratory shift characterizing spore germination process from AOX (dormancy) to COX (hyphal growth). It is difficult to define AOX/COX interplay during spore germination in R. irregulare as spores can return into dormancy and germinate again several times without obvious phenotypic signs (Giovanetti et al., 2010), and the hyphal tube germination is the physiological consequence of an already implemented respiratory shift, as shown for Mucor rouxii (Cano-Canchola et al., 1988).

Nevertheless, our data (Experiment 1) with KCN suggest that AOX is very likely involved in spore dormancy. Application of SHAM or KCN inhibited the spore germination, supporting the importance of both electron pathways and a possible involvement of a respiratory shift from AOX to COX. However, further work is needed to confirm this statement. AMF spores are sensitive to plant hormones, and their germination responses harbor opposite patterns between the antagonistic hormones ABA and Ga3, as ABA maintained spore dormancy while Ga3 broke it, similarly to plant seeds (White and Rivin, 2000; Linkies and Leubner-Metzger, 2012). Therefore, these observations support the need to apply ABA as pre-treatment (Experiment 2) and the need to apply the respiratory inhibitors some days after plant inoculation (Experiment 3) in order to not disturb the pre-symbiotic developmental phase of R. irregulare. Promotion of hyphal branching pattern in ABA treatment fits with previous observations: Juge et al. (2002, 2009) showed that spores develop g-type pattern germination (fine branching hyphae) around spores when dormancy is incompletely broken or under stress conditions, while G-type pattern (runner hyphal growth pattern) occurs in favorable conditions. Hyphal branching from germ tube generated by ABA seems similar to its effect on the arbuscule formation (Herrera-Medina et al., 2007).

As a remark, INT staining data suggested a low spore viability in Ga3 and SHAM treatment and higher after ABA and KCN (5 mM) treatment. However, no formazan production was observed in germinated spores following Ga3 application, while the germinative hypha was still growing (data not shown). This suggests that INT staining could correspond to a reducing power marker in the cell, and might be associated to AOX metabolism and spore dormancy, rather than solely to a vital staining.

Occurrence of the Arum- and Paris-types are not well understood and corresponds to extremes, which can coexist in a developmental continuum within the same root structure (Smith and Read, 2008). Their formation depends partly on genotypic factors characterizing both partners (Karandashov et al., 2004), partly on environmental factors like phosphate availability (Smith and Read, 2008). In potato roots, both mycorrhizal types and root colonization are commonly influenced by phosphate concentration (McArthur and Knowles, 1992). This is in agreement with our observations since P concentration inhibited mycorrhizal development in a dose-dependent way, and Arum-type structures were occurring at low to medium P concentration (1, 10, and 50 ppm), while Paris-type formation appeared at higher P levels (100 and 300 ppm).

Our data on mycorrhized potato suggest that AOX or AOX-related metabolism is involved in arbuscule/hyphal branching and Arum-type formation, while COX or COX-related metabolism is associated with higher hyphal growth and hyphal-coiled shape (Paris-type) formation. Despite the use of SHAM being controversial because of its possible non-specificity to AOX (Bingham and Stevenson, 1995; Day et al., 1996), our results showed differential and opposite phenotypic patterns in plant and fungal behavior when SHAM and KCN treatments were compared. KCN and ABA are known to stimulate AOX activity (Finkelstein et al., 1998; Choi et al., 2000; Rook et al., 2006; Giraud et al., 2009; Millar et al., 2011; Lynch et al., 2012; Wind et al., 2012). These two molecules promoted both arbuscule intensity and branching, but they caused differential mycorrhizal development (M %), with ABA treated plants yielding a higher M % (with colonization in the primary adventitious root recognized under the experimental conditions even up to the base stem). This could be explained by the fact that COX capacity is not inhibited by ABA unlike KCN. On the other hand, we observed that ABA (Experiment 2) tended to promote the AOX gene transcript levels in non-inoculated potato plantlets, but such response was not clearly identified when using KCN (Experiment 3). It would therefore suggest that the mycorrhizal root colonization, but also transcript regulation of the AOX pathway, are dependent from the functional state on the COX pathway. To summarize, we can deduce that mycorrhizal behavior seems to be linked to the mitochondrial respiratory chain-partitioning environment, probably generated from both partners. As a remark, ethylene is another stress hormone that is able to induce AOX pathway (Simons et al., 1999; Wang et al., 2010; Xu F. et al., 2012), but usually impairs mycorrhizal colonization (using epinastic plant or exogenous application, Zsögön et al., 2008; Fracetto et al., 2013). This phenomenon could be partly explained by the action of cyanide (HCN) produced stoichiometrically (1:1) with ethylene, blocking therefore the COX pathway (Xu F. et al., 2012).

Interpretation of the sporulation rise at 50 ppm P in M plants is challenging, but it seems linked with specific plant metabolism independent of mycorrhizal colonization (M %), as already observed by past studies (Douds and Schenck, 1990), while this last parameter harbored high correlation with P concentration. A particularity is observed at 50 ppm P concentration through the various plant phenotypical parameters studied (plant growth parameters – Supplementary Figure S2; mycorrhizal dependency – Figure 8) and corresponds to the highest value of arbuscule intensity (a %, Supplementary Figure S1) with a predominant Arum-type, concomitant with an up-regulation of RiAOX transcripts (Figure 9) and the highest sporulation (Figure 5).

Taking into account all these observations, we propose a scheme defining the roles of AOX and COX in the different mycorrhizal phenotypical behaviors (Figure 11).

Although many factors can influence the MGD, it is usually recognized that high P concentrations decrease MGD, which may become negative (Smith and Smith, 2011a). In our study, no correlation was found between MGD and P concentration, but respiratory inhibitors induced opposite responses. All possible MGD responses (neutral for non-treated plants, positive under KCN and negative under SHAM) were observed within the same AMF-plant biosystem, illustrating the high plasticity of mycorrhizal behavior linked to the physiological context. We show the possibility to obtain increasing yield with increasing P concentration when respiration in the plant–AMF system is associated with AOX pathway (i.e., in KCN treatments), occurring with a low fungal colonization and associated with relatively stable and almost maximal positive MGD. This goes together with a predominant Arum-type structures and higher arbuscule intensity when compared to treatments where COX pathway would be engaged (i.e., in SHAM treatments), in which Paris-type structure is predominant and associated with a negative MGD. This is consistent with the study of van Aarle et al. (2005), showing that Arum-type structures had higher metabolic activity than Paris-type ones. Moreover, even if SHAM reduced the MGD and the arbuscule intensity, while favoring Paris-type structures, the AMF produced a denser mycelium (h %) and as many vesicles as in non-treated mycorrhizal plants (except at 1 ppm P). AMF responses on plant performance seem therefore to depend mainly on the ability to form hyphal branching (Arum-type), possibly by an increased exchange surface for providing nutrients to plants (Kobae and Hata, 2010; Bapaume and Reinhardt, 2012). However, if MGD responses seem related with arbuscule type in KCN and SHAM treatments, no direct relationship was found between arbuscule intensity and MGD. This leads to the hypothesis that although various membrane transporters (phosphate, amino acid, sugar or nitrogen) were characterized in arbuscules, these fungal structures might not constitute preferential sites for nutrient/element exchanges, and a role should be also attributed to intraradical hypha. This assumption is supported by the fact that glucose transporters were observed not only in arbuscules but also on intraradical hyphae (Helber et al., 2011).

The mycorrhizal dependency variation has been attributed to plant species and cultivars, fungal species and isolates, host and symbiont interplay, mycorrhizal rate, soil phosphorus concentration and environmental conditions (Plenchette et al., 1983; Singh, 2001; Smith and Smith, 2011a). Our data suggest that the electron flow partitioning, which is modulated by environmental stimuli and genetic background, corresponds to a main metabolic component determining the MGD related to the mycorrhizal type. Finally, AOX metabolism, linked usually with higher ABA content, is known to be related to plant growth regulation and has been reported to be connected to growth depression (Sieger et al., 2005). This may partly explain the negative MGD phenomenon that AMF might trigger during early developmental stages in some plant species (Koide, 1985; Graham and Abbott, 2000; Smith et al., 2009; Ronsheim, 2012) during the implementation of the induced systemic response in the plant (Pozo et al., 2002; Hause and Fester, 2005; Hause et al., 2007).

Arbuscular mycorrhizal inducible Pi transporters of the Pht1 family have been described as markers of mycorrhizal development and, in some cases, proposed as markers of mycorrhizal functionality (Javot et al., 2007). Partly in accordance to this statement, our data show a correlation between AM-inducible transporter expression (StPT3, StPT4, and StPT5) and the mycorrhizal development. A gradual repression of the MPU pathway with increasing P availability in non-treated mycorrhizal plants was also observed, which is in accordance with several previous studies on Solanaceae species (Chen et al., 2007; Nagy et al., 2009), suggesting that the MPU is not functional at high P levels. However, the repression of the MPU by P disappeared when plants were treated with KCN, and surprisingly all AM-inducible P transporters were strongly up-regulated at 300 ppm in non-inoculated plants. This last observation deserves further investigations, as no previous works have been done with very high P concentrations to our knowledge.

The mycorrhizal impact on DPU is less clear. A number of studies have shown that AM colonization of plants down-regulates the expression of the DPU Pi transporters (Rausch et al., 2001; Burleigh et al., 2002; Glassop et al., 2005; Requena, 2005; Nagy et al., 2006), while other studies pointed contrasted transcriptional regulations. Chen et al. (2007) showed a mycorrhiza-induced down-regulation of Pht1;1 and Pht1;2 expression under low-P conditions, but up-regulation of Pht1;2 under high-P conditions for pepper, eggplant and tobacco. In tomato, Nagy et al. (2009) did not find any change in the transcript abundance for SlPT1 and SlPT2 upon root colonization. Our data are in accordance with these latter observations, as presence of AMF did not necessarily induced down-regulation of non-AM-inducible transporters, but responses were specific to P concentrations. For example, it was noticed that AMF induced an up-regulation at 100 ppm P but a down regulation at 10 ppm P for StPT1, while an up-regulation of StPT2/6 was observed at 100 ppm P. Only StPT2/6 was regulated by P availability. KCN treatment induced an up-regulation of DPU transporters in mycorrhizal conditions.

Fungal Pht1 genes are known to be expressed in the intraradical phase (Harrison and van Buuren, 1995; Benedetto et al., 2005; Balestrini et al., 2007; Tisserant et al., 2012; Fiorilli et al., 2013; Walder et al., 2016). Fiorilli et al. (2013) showed a slight down-regulation of RiPT1 when exposed to higher P concentrations, and Walder et al. (2016) reported a positive correlation of RiPT5 transcript levels and Pi acquisition in Sorghum. In our study, only RiPT5 was slightly regulated by P levels, with up-regulation in concentrations higher than 50 ppm in non-treated conditions. Few data exist on the translocation of Pi across the inner mitochondrial membrane by Pht3 family in plants and fungi. In our study, no clear role emerged for these transporters within AMF symbiosis. StMPT3, which presented the highest expression levels in the roots, appeared to be constitutive, and only StMPT1a was slightly up-regulated by the presence of R. irregulare in non-treated conditions. RiMPT appeared also to be constitutively expressed in our experimental conditions.

In plants, several transcriptomic analyses revealed that AM establishment can induce the expression of plant N transporters, mainly in arbusculated cells (Gomez et al., 2009; Guether et al., 2009; Kobae et al., 2010; Gaude et al., 2012; Ruzicka et al., 2012). AMTs identified in tomato LeAMT4 and LeAMT5 were reported to be exclusively expressed in mycorrhizal roots and not regulated by NH₄⁺ (Ruzicka et al., 2012). The two potato homologs, StAMT4 and StAMT5, were strongly up-regulated by AMF, and were gradually repressed by P in mycorrhizal plants. This decrease in transcript levels can be attributed to the reduction of fungal structures inside the roots. However, in the case of AM-inducible AMTs, such as AM-inducible Pht1, NM plants presented a strong up-regulation at a high P level (300 ppm). As for P transporters, a suppression of this up-regulation at 300 ppm P in the presence of KCN was observed and could be explained by the need of a functional COX pathway for ammonium and phosphate transport. NH₄⁺ uptake via AMT seems indeed to be accompanied by H⁺ extrusion by the plasma membrane H⁺-ATPases for maintenance of the cytosolic charge balance, and increased fluxes of NH₄⁺ would increase the demand for respiratory ATP (Britto and Kronzucker, 2005). Hachiya et al. (2010) suggested that the ammonium-dependent increase of the O₂ uptake rate can be explained by the up-regulation of the cytochrome pathway, which may be related to the ATP consumption by the plasma-membrane H⁺-ATPases. Similarly, P transport is a proton-dependent process, and the H⁺-ATPase HA1 of Medicago truncatula was shown as essential for P transport in AM symbiosis (Krajinski et al., 2014). StHA1 is also up-regulated at 300 ppm P in non-inoculated plants compared to 1 ppm P and repressed in mycorrhizal plants. It can be therefore hypothesized that an impaired COX pathway would repress ammonium transport via AMT gene family members.

P and N AM-inducible transporters were found to be connected to the mycorrhizal development in non-treated plants. On the other hand, no correlation was observed between the expression of P and N AM-inducible transporters and the presence of either specific fungal structures (arbuscule or intraradical hypha) or any MGD parameters across the treatments (Supplementary Tables S2–S4). In particular, M KCN plants were associated with positive MGD compared to NM KCN plants but the transcript levels of genes encoding AM-inducible transporters were similar to M SHAM treatments. Several studies showed a lack of relationship between Pht1 gene expression and mycorrhizal Pi acquisition (Grace et al., 2009; Grønlund et al., 2013; Walder et al., 2015). It would therefore indicate that MGD and that plant’s Pi acquisition through the MPU is not quantitatively regulated by the expression level of AM-inducible Pht1 genes. These observations, combined with the lack of repression of the MPU by P in presence of KCN, and the strong up-regulation of AM-inducible P transporter genes at 300 ppm in NMs, suggest that these transporters are not suitable markers for a functional symbiosis in field conditions where high P concentrations can occur and natural or anthropogenic source of cyanides can be found in soil or water (with concentrations as high as 100 mg kg DW^-1).

Despite sugar transporters being found and characterized (Helber et al., 2011), AMF are still unable to complete their life cycle even when carbon sources are added in in vitro culture systems without the presence of plant roots. Some publications showed even the opposite: glucose and fructose application under axenic or monoxenic conditions resulted in reduced fungal growth or spore germination rate (Mosse, 1959; Koske, 1981; Hepper, 1982; Siquiera and Hubbell, 1986; Wang et al., 2015), while culture media devoid of sucrose can stimulate spore germination, hyphal growth (D’Souza et al., 2013), as well as, sporulation (St-Arnaud et al., 1996). In many publications, the definition of obligate biotrophy of AMF deals with their dependency on plants for carbohydrate supply (mainly glucose and fructose, Pfeffer et al., 1999). But obviously, this statement seems incomplete since it is not yet strictly proven by the successful implementation of an axenic culture. Genomic and transcriptomic data obtained from R. irregulare were not helpful since this fungal species would possess all the necessary genes to harbor saprobe behavior (Tisserant et al., 2012; Tisserant et al., 2013), although no evidence for gene encoding for fungal multi-domain fatty acid synthase was found (Wewer et al., 2014). Therefore, the fungal needs related to the biotrophy of AMF remains still an open question. In this way, it seems important to emphasize that plants provide, at least, a habitat, i.e., a physical growth support associated with a favorable physiological frame allowing uptake and metabolic assimilation of carbon sources that AMF seems not to encounter without host. This questions the definition of the metabolic frame (induced by stress signals) favorable to AMF colonization, arbuscule development and functions on plant performances. We discuss below the role of COX and AOX, which can shed new light that would allow a better understanding and mastering of mycorrhiza.

This is the first work that extensively studies phenotypic patterns by using respiratory chain inhibitors in mycorrhizal symbiotic associations. Plant growth parameters suggest that “AOX plants” (plants that are able to engage AOX pathway to a higher extent) are more positively responsive to AMF and have higher potentials in carbon allocation to the potato tuber than “COX plants” (plants that show a higher engagement of COX pathway), and plants treated with ABA or with KCN indicated that AMF inoculation is efficient to counteract extreme stresses in terms of growth performance. Under stress-related physiology, AOX might represent the main (or a more engaged) respiratory chain pathway in plant roots, concomitant with higher fermentation that might favor arbuscule development. In analogy to previous observations (Liu et al., 2015), we observed that the presence of AMF reduces plant stress and seems to favor the COX pathway, which is consistent with higher energetic metabolism that associates to positive MGD. Therefore, artificial soft and transitory induction techniques of AOX metabolism in plants, without disturbing COX potential, could represent a physiological target for producers for improving mycorrhizal plant susceptibility and mycorrhizal dependency.

It is very likely that the modulation of respiration partitioning occurs at different stages of AMF life cycle, as shown during presymbiotic phase by the induction of mitochondrial respiration by several plant root exudates (Tamasloukht et al., 2003; Besserer et al., 2008) and AOX expression along the symbiotic establishment (Campos et al., 2015). Optimal mycorrhizal development needs obviously oxygen (KCN data) but seems also related to fermentation (ABA data). This opens the possibility that AMF, at a whole, might belong to obligate aerobic fermentative organisms. Therefore, the role of fermentation in mycorrhizal behavior deserves further investigations providing potentially exciting new avenues for next studies.

While the application of respiratory inhibitors induced significant changes from phenotypical parameters and the expression of some genes involved in transport and fermentation, interpretation of transcriptomic data related to AOX and COX pathways was challenging. However, respiration genes are hierarchically superimposed to specific metabolic pathway genes. Increasing data supports the hypothesis that the most crucial responses of AOX genes to changing environments are induced early during cell reprogramming before specific pathways are initiated and interfere with growth performance (Arnholdt-Schmitt et al., 2006). Recently, this was confirmed for the carrot anti-freezing protein gene under cold treatment (Campos et al., 2016). Furthermore, in A. thaliana, it was shown that limitation of the capacity of the alternative-, cytochrome-, or both termini-oxidases results in unique and overlapping transcriptional responses depending on growth conditions. Nevertheless, this research presented the first clues for the implication of respiratory metabolism in a mycorrhizal symbiosis seen as a holobiont level system. Thus, our research supports further deeper molecular-physiological studies on respiration traits and exploiting genetic approaches through overexpressing and knock-out or respiratory gene editing techniques. However, final progress in order to associate complex respiration metabolism to robust growth performance will absolutely depend also on the development of adequate predictive and phenotyping-screening technologies (Fiorani and Schurr, 2013; Arnholdt-Schmitt et al., 2014, 2016; Nogales et al., 2016).