Poor plant performance under simulated climate change is linked to mycorrhizal responses in a semiarid shrubland

Summary

1. Warmer and drier conditions associated with ongoing climate change will increase abiotic stress for plants and mycorrhizal fungi in drylands worldwide, thereby potentially reducing vegetation cover and productivity and increasing the risk of land degradation and desertification. Rhizosphere microbial interactions and feedbacks are critical processes that could either mitigate or aggravate the vulnerability of dryland vegetation to forecasted climate change.

2. We conducted a four-year manipulative study in a semiarid shrubland in the Iberian Peninsula to assess the effects of warming (~2.5ºC; W), rainfall reduction (~30%; RR) and their combination (W+RR) on the performance of native shrubs (Helianthemum squamatum) and their associated mycorrhizal fungi.

3. Warming (W and W+RR) decreased the net photosynthetic rates of H. squamatum shrubs by ~31% despite concurrent increases in stomatal conductance (~33%), leading to sharp decreases (~50%) in water use efficiency. Warming also advanced growth phenology, decreased leaf nitrogen and phosphorus contents per unit area, reduced shoot biomass production by ~36% and decreased survival during a dry year in both W and W+RR plants. Plants under RR showed more moderate decreases (~10-20%) in photosynthesis, stomatal conductance and shoot growth.

4. Warming, RR and W+RR altered ectomycorrhizal fungal (EMF) community structure and drastically reduced the relative abundance of EMF sequences obtained by high-throughput sequencing, a response associated with decreases in the leaf nitrogen, phosphorus and dry matter contents of their host plants. In contrast to EMF, the community structure and relative sequence abundances of other non-mycorrhizal fungal guilds were not significantly affected by the climate manipulation treatments.

5. Synthesis: Our findings highlight the vulnerability of both native plants and their symbiotic mycorrhizal fungi to climate warming and drying in semiarid shrublands, and point to the importance of a deeper understanding of plant-soil feedbacks to predict dryland vegetation responses to forecasted aridification. The interdependent responses of plants and ectomycorrhizal fungi to warming and rainfall reduction may lead to a detrimental feedback loop on vegetation productivity and nutrient pool size, which could amplify the adverse impacts of forecasted climate change on ecosystem functioning in EMF-dominated drylands.

Introduction

Current climatic models predict drastic changes in the climate of the Mediterranean region over the next decades as a consequence of anthropogenic greenhouse gas emissions (Collins et al., 2013). Forecasted changes for this period include temperature increases of 2-5º C, reduced rainfall amounts, increased vapor pressure deficit and evapotranspiration, and more frequent occurrence of extreme climatic events (Meehl & Tebaldi 2004; Giorgi & Lionello, 2008; NOAA 2015). These projected changes will increase aridity conditions and reduce soil moisture availability in the Mediterranean region and other dryland areas (Dai 2013, Huang et al., 2016), thus reducing plant primary productivity and vegetation cover (Delgado-Baquerizo et al., 2013; Williams et al., 2013; Ahlström et al., 2015; Guiot & Cramer, 2016) and increasing the risk of land degradation and desertification (Le Houérou, 1996; Safriel & Adeel, 2005).

Most manipulative field studies assessing climate change impacts on Mediterranean-type vegetation have evaluated the effects of warming (Peñuelas et al., 2004; Prieto et al., 2009; León-Sánchez et al., 2016) and rainfall reduction (Peñuelas et al., 2001; Miranda et al., 2011; Peñuelas et al., 2013; Tielbörger et al., 2014) in isolation from each other. Furthermore, few of them have investigated the response of Mediterranean semiarid ecosystems to climate change, as most research conducted so far in this region has focused on areas with dry or sub-humid climate (e.g., Peñuelas et al., 2004; Sardans et al., 2008; Prieto et al., 2009; but see Tielbörger et al., 2014). Experimental warming generally enhances plant nutrient status, photosynthesis and growth in Mediterranean plant communities during winter and early spring, but has neutral or negative effects during late spring and summer (Llorens et al., 2003; Prieto et al., 2009; Wu et al., 2011). Experimental rainfall reduction has predominantly negative effects on plant nutrition, photosynthesis, growing season length, and biomass accumulation (Sardans et al., 2008; Prieto et al., 2009; Wu et al., 2011), although coexisting plant species may differ widely in their sensitivity to this factor (Llorens et al., 2003). On the other hand, Mediterranean dryland plants have evolved a wide array of adaptive phenotypic plasticity mechanisms to cope with heat and drought stress, which could buffer any negative impacts of climate warming and drying on plant performance (Nicotra et al., 2010; Bussotti et al., 2014; Nardini et al., 2014).

Rhizosphere microbial interactions and feedbacks are critical processes that could either buffer or aggravate the vulnerability of dryland plant communities to forecasted climate warming and drying (Sardans & Peñuelas, 2013). In particular, symbiotic mycorrhizal fungi largely mediate plant nutrient uptake, carbon dynamics and vegetation responses to environmental change (Allen et al., 2003; Leake et al., 2004; van der Heijden et al., 2008; Vargas et al., 2010). Mycorrhizal fungi can promote host plant tolerance of multiple abiotic stresses (including heat and drought stress; Augé, 2001; Bunn et al., 2009; Lehto & Zwiazek, 2011), but could also be themselves highly vulnerable to the detrimental impacts of forecasted climate change (Querejeta et al., 2009; Millar & Bennett, 2016). However, relatively few studies have examined in detail the coupled responses of dryland plants and their associated mycorrhizal fungi to climatic manipulations (Compant et al., 2010; Mohan et al., 2014), so further experimental research in this field is needed.

Woodland and shrubland communities composed of ectomycorrhizal sclerophyllous woody plants from the genera Cistus, Halimium and Helianthemum, (Cistaceae), Arbutus (Ericaceae), Quercus (Fagaceae) and Pinus (Pinaceae) are widespread throughout the Mediterranean Region and other dryland areas. Whereas some of these genera can also form other types of mycorrhizal associations (arbuscular, arbutoid or ectendomycorrhizal), all of them are thought to be primarily ectomycorrhizal (Thanos et al., 1992; Brundrett, 2009; Zambonelli et al., 2014). Many ectomycorrhizal fungi (EMF) show rather low tolerance to soil warming and/or drying (Cline et al., 1987; Querejeta et al., 2009), as even moderate increases in heat or drought stress can trigger large decreases in EMF root colonization, extraradical mycelial production and physiological activity (Runion et al., 1997; Jany et al., 2003; Kennedy & Peay, 2007; Navarro-Ródenas et al., 2011). Field studies have shown that severe heat and drought stress can negatively affect the abundance and diversity of EMF communities in Mediterranean-type ecosystems (Querejeta et al., 2009; Büntgen et al., 2012; Allen & Kitajima, 2013; de la Varga et al., 2013; Salerni et al., 2014; Ágreda et al., 2015), and elsewhere (Swaty et al., 1998; Gehring et al., 2006; Valdés et al., 2006). Increased heat and drought stress severity could negatively affect EMF fungal communities both through the direct detrimental impacts of soil warming and drying on the fungi themselves and through the indirect effects of reduced carbon assimilation and availability for allocation to mycorrhizae in stressed host plants.

Semiarid plant communities growing on gypsum soils are rich in rare and endemic plant species of high conservation value (Escudero et al., 2015). Gypsum soils are particularly stressful for plants and soil microbiota due to their low fertility and unfavourable physical and chemical properties, and very little is known about the potential impacts of forecasted climate change on native plant and soil microbial communities from gypsum habitats (Maestre et al., 2013, 2015; León-Sánchez et al., 2016). To contribute to fill this knowledge gap, we carried out a four year manipulative field experiment in a semiarid ecosystem in Central Spain dominated by the tussock grass Stipa tenacissima L. and the gypsophilous shrub Helianthemum squamatum (L.) Dum (Fig. S1). This shrub forms symbiotic mycorrhizal associations with both EMF and arbuscular mycorrhizal fungi (AMF) simultaneously (Gutiérrez et al., 2003; Brundrett, 2009), making this species a suitable model system to evaluate the impacts of climate warming and drying on a rich and complex mycorrhizal fungal community. We simulated the climate conditions projected for the second half of the 21st century (Collins et al., 2013) by using open top chambers (~2.5ºC temperature increase), rainout shelters (~30% rainfall reduction), and their combination, and assessed the effects of climate manipulation on plant gas exchange, nutrient status, growth and survival, as well as on the community structure of the rhizosphere fungal community.

Our overarching hypothesis is that climate change will have multiplicative detrimental effects on nutrient cycling and vegetation productivity through strong plant-soil feedback mechanisms involving mycorrhizae (Ehrenfeld et al., 2005). We hypothesize that a warmer and drier Mediterranean climate will trigger a detrimental feedback loop on the primary productivity of EM and mixed EM/AM plants (Fig. 1). In this hypothesized conceptual model, increased heat and drought stress for plants during the growing season strongly decrease their net photosynthetic rates, leading to large reductions in photosynthate availability for allocation to mycorrhizae, which in turn results in altered EMF community structure and decreased relative abundance of EMF sequences in the rhizosphere fungal community (Kiers et al., 2011), as well as decreased EMF carbon-sink strength (Wright et al., 2000) and nutrient-uptake capacity (Leake et al., 2004). Eventually, decreased host plant nutrient status combined with decreased carbon-sink stimulation of photosynthesis by EMF further decrease carbon assimilation rates, water use efficiency and biomass production (Guehl & Garbaye, 1990), thereby closing the hypothesized detrimental feedback loop on plant productivity.

Figure 1.

Hypothesized conceptual model of a mycorrhizal role in modulating host plant responses to climate change, which may lead to a detrimental feedback loop on the productivity of semiarid shrublands under a warmer and drier climate. EMF: ectomycorrhizal fungi.

Materials & Methods

Study site and experimental design

The study was carried out near Aranjuez, in central Spain (40°02′N–3°32′W, 495 m altitude). The study area has a continental Mediterranean climate, with a mean annual temperature of 15°C and average rainfall of 349 mm. Soils are derived from gypsum, have pH values ca. 7 and are classified as Gypsiric Leptosols (IUSS Working Group WRB 2006). Soils are shallow (4-10 cm deep overlying weathered gypsum bedrock) and show a thin organic horizon (1-2 cm thick). Vegetation is a native grassland and shrubland community dominated by S. tenacissima, and contains gypsophilous shrubs such as H. squamatum and Gypsophila struthium L. Perennial vegetation cover is lower than 40%.

To evaluate the effects of projected climate change conditions on the performance of native plants and mycorrhizal fungi, we established an experiment in February 2011 to examine the effects of three different climate manipulation treatments (warming [W], rainfall reduction [RR] and the combination of both [W+RR]) on pre-existing H. squamatum shrubs and their associated mycorrhizal fungi. The experiment includes 10 replicates per each climate manipulation treatment plus 30 control plots, making a total of 60 experimental plots distributed across a 100 x 50 m area. These plots were randomly assigned to the different treatments, and were at least 2 m distant from each other. Plot size is ~1 m². The target shrub H. squamatum is the dominant (often the only) plant species present in the experimental plots (Fig. S1). Each plot contained 1-4 adult H. squamatum individuals at the time of experimental set-up. New H. squamatum individuals recruited in the experimental plots during spring in every year of the study, which compensated for the drought-induced mortality events that occurred during the dry summers, particularly in 2012.

The warming treatment simulates the predictions derived from six atmosphere general circulation models for the second half of the twenty-first century in the Western Mediterranean region (De Castro et al., 2005), and was achieved by installing ventilated open-top chambers (OTCs), which are hexagonal chambers with sloping slides of 40 cm x 50 cm x 32 cm (Fig. S2). These chambers were made of transparent methacrylate, which transmits about 92% of visible light, has a reflection of incoming radiation of 4%, and passes on ca. 85% of incoming energy (information provided by the manufacturer; Decorplax S.L., Humanes, Spain), and have been used in previous field warming experiments (Maestre et al., 2013, 2015). Upon installation in the field, the OTCs were suspended ~3 cm above the ground by a metal frame to allow free air circulation and exchange with the surrounding environment, which minimizes undesirable experimental effects (Hollister & Webber, 2000). The mean air and soil temperature increases achieved within the OTCs range from 1-2º C (winter) to 4-7º C (summer), which is in good agreement with climate change projections for the second half of the XXIst century in the study area (Collins et al., 2013).

To simulate projected reductions in precipitation (De Castro et al., 2005), we used passive rainout shelters that intercept and exclude ~30% of the incoming rainfall from the plots. The permanent (non-moveable) rain exclusion shelters are made of transparent methacrylate troughs covering ~30% of the area of the experimental plots. Rainfall reduction is achieved by suspending the methacrylate troughs over an aluminum frame above the experimental plots (height 130 cm, width 100 x 100 cm, Fig. S2). Intercepted rainwater is diverted through collection pipes, stored in tanks placed next to the experimental plots and removed after each rainfall event. Finally, the combined W+RR treatment is achieved by installing both OTCs and rainfall exclusion shelters over the same experimental plot (Fig. S1).

Air temperature and relative humidity and soil moisture and temperature were continuously monitored using replicated automated sensors (HOBO^® U23 Pro v2 Temp/RH and TMC20-HD sensors, Onset Corp., Pocasset, MA, USA, and EC-5 soil moisture sensors, Decagon Devices Inc., Pullman, WA, USA, respectively).

Plant measurements

Net photosynthetic rate (A), stomatal conductance (g_s), transpiration rate (E), maximum efficiency of photosystem II under light conditions (Fv’/Fm’) and the quantum efficiency of photosystem II (∅PSII, which measures the proportion of light absorbed by chlorophyll associated with photosystem II that is used for photosynthesis; Baker & Rosenqvist, 2004) were measured in 2012 (February, May, June and July), 2013 (February, April, June and October), 2014 (May) and 2015 (May) with a LI-6400-40 Leaf Chamber Fluorometer and a LICOR 6400-01 CO₂ injector (Li-Cor, Lincoln, NE, USA). Leaf gas exchange was measured on fully expanded leaves that were placed in a 2 cm² leaf cuvette. During these measurements, air CO₂ concentration was controlled using the injection system and compressed CO₂-cylinders with a CO₂ concentration of 390 µmol mol^-1 CO₂. Measurements were conducted at a saturating light of 1500 µmol m^-2 s^-1, and at ambient air temperature and relative humidity. The air flow was set to 250 µmol s^-1. All leaf-gas exchange measurements were conducted between 8:00 and 11:00 am (GMT), when photosynthetic rates are highest. For warmed plants, all leaf gas exchange measurements were conducted under the prevailing microclimatic conditions within the OTCs (i.e., elevated temperature and vapor pressure deficit relative to ambient). All the leaves used for gas exchange measurements were collected thereafter to measure their area using an image scanner program (Image Pro Plus, Media Cybernetics, Inc. Rockville, USA). On each date, leaf gas exchange measurements (A, g_s, E, Fv’/Fm’, ∅PSII) were conducted on 8-10 H. squamatum individuals per climate manipulation treatment (generally on the same individuals in successive dates). Intrinsic water use efficiency (WUE_i) was calculated as the ratio between net photosynthetic rate and stomatal conductance (A/g_s).

To obtain a time-integrated measure of water use efficiency (Cernusak et al., 2013), fully sun-exposed adult leaves were collected in April-May of each study year to determine their carbon isotope ratios (δ¹³C). Leaf samples were oven dried at 60ºC and finely ground with a ball mill before being weighted and encapsulated into tin capsules for analysis. The δ¹³C and C and N concentrations of leaf material were measured using elemental analyzer/continuous flow isotope ratio mass spectrometry (ANCA/SL elemental analyzer coupled with a Finnigan MAT Delta PlusXL IRMS). Isotope analyses were conducted at the Centre for Stable Isotope Biogeochemistry, University of California, Berkeley (USA). Leaf δ¹³C values are expressed in delta notation (‰) relative to the reference standard V-PDB. Long-term external precision for δ¹³C analyses is 0.14‰. Foliar P concentrations were measured by inductively coupled plasma optical emission spectrometry (ICP-OES, Thermo Elemental Iris Intrepid II XDL, Franklin, USA) after a microwave-assisted digestion with HNO₃:H₂O₂ (4:1, v:v) at the Ionomics facility at CEBAS-CSIC.

In April 2013, 2014 and 2015, four leaves per target shrub were collected to measure their area (cm²), dry mass (mg) and leaf mass area (LMA, mg cm^-2). A picture of the fresh leaves was taken using a Canon 3000D (Canon Inc, Tokyo, Japan) and the resulting image was processed using Image Pro Plus (Media Cybernetics, Silver Spring, USA) software to measure leaf area. Leaves were thereafter oven dried at 60ºC for 24h to determine their dry weight. LMA was calculated as the ratio between leaf dry weight and leaf area (Poorter et al., 2009). Nitrogen and phosphorus contents per unit leaf area (N_area, P_area) were calculated using leaf N and P concentrations and LMA values.

In late winter 2012 and 2013, three-four terminal shoots of each target shrub were labeled with red tape to measure their elongations during the growing season. We measured shoot elongation in mid-Spring (April) and again at the end of the growing season (June). The late/early spring growth ratio was then calculated as the quotient between the shoot elongations recorded during the latter and earlier parts of the growing season (May-June and March-April, respectively), which provides an indication of changes in shoot growth phenology with climate manipulation. In October 2015, one representative terminal shoot of approx. 10 cm length per target H. squamatum shrub was sampled to evaluate the effects of the climate manipulation treatments on shoot biomass production. We measured the total dry mass (48 hours at 60ºC), total number of leaves and total leaf area of these shoots, and standardized the values per 10 cm shoot length. We did not carry out an extensive destructive sampling of aboveground plant biomass to prevent severe disruption of the experimental plots in this long-term field experiment.

Finally, we measured post-summer plant survival rates after the first autumn rainfalls in each study year. Plant survival rate was estimated as the percentage of H. squamatum individuals present in spring in each experimental plot that were still alive at the end of the summer drought period.

Characterization of the rhizosphere fungal community

Helianthemum squamatum is the only known ectomycorrhizal host species present in the local plant community, as all the other coexisting plant species are arbuscular mycorrhizal (Brundrett, 2009). We focused on the mycorrhizal fungal community present in the rhizosphere of H. squamatum shrubs. Rhizosphere soil and root samples were collected on 30 June 2014 by inserting a hand auger corer (5 cm diameter, 0-5 cm deep) into the soil under the canopy of focal H. squamatum shrubs (8 replicated plots per treatment). This sampling depth was considered suitable because the fertile organic horizon is only 1-2 cm deep, and also because sampling deeper rhizosphere soil and roots would have been technically challenging and would have caused unacceptable disruption of the plant-soil system in this long-term field experiment. However, we acknowledge that H. squamatum roots can penetrate deeper than 5 cm into the soil profile (maximum rooting depth ~ 1 m; Guerrero-Campo et al., 2006), and that mycorrhizal fungi are often distributed more deeply than our sampling depth (Bornyasz et al., 2005; Querejeta et al., 2009; Pickles & Pither, 2014), which represents a limitation of our study. The 32 sampled plots had similar density of H. squamatum plants (1-2 adult individuals) and similar focal plant size across treatments. The rhizosphere samples were thereafter transported to the laboratory on ice and stored at 4 ºC, and rhizosphere DNA was extracted within the following two weeks from 1 g subsamples containing a mixture of H. squamatum fine roots and rhizospheric soil adhering to them, in order to assess the impact of the climate manipulation treatments on both the radical and extraradical phases of the ectomycorrhizal mycelium. DNA extractions were performed using the UltraClean™ DNA isolation kit (MO BIO Laboratories, CA, USA). Extracted DNA was checked in 1 % agarose gels run in 0.5 X TAE buffer (Tris-acetate-EDTA; 100 V, 15 min). To analyze the rhizosphere fungal communities we used a tag-encoded FLX-titanium amplicon pyrosequencing (TEFAP) approach. The internal transcribed spacer (ITS) region was PCR amplified by using the fungal universal primers ITS1F (5’-CTTGGTCATTTAGAGGAAGTAA-3’) and ITS4R (5’-TCCTCCGCTTATTGATATGC-3’) and a HotStarTaq Plus Master Mix Kit (Qiagen, CA, USA) under the following conditions: 94°C for 3 minutes, followed by 28 cycles of 94°C for 30 seconds, 53°C for 40 seconds and 72°C for 1 minute, after which a final elongation step at 72°C for 5 minutes was performed. Samples were randomized prior to PCR amplification. Amplicons from all samples were mixed in equal concentrations, purified using Agencourt Ampure beads (Agencourt Bioscience Corporation, MA, USA) and sequenced with Roche 454 FLX titanium instruments and reagents.

Sequences were sorted out according to their barcodes. Sequences with barcodes showing one or more mismatches were discarded. Barcodes and primers were trimmed, and a total 176,702 raw fungal sequences were deposited in EMBL with the study accession number PRJEB12618 (http://www.ebi.ac.uk/ena/data/view/PRJEB12618). Low-quality sequences (shorter than 200 bp, including ambiguous base calls, with homopolymer runs exceeding 9 bp, or with an average quality score below 20) were removed. Sequences were then de-noised and chimeras removed. De-noising was aimed at removing a variety of errors producing non-target sequences including sequencing errors, PCR errors and chimeras, although this process may reduce alfa and beta diversity by decreasing the number of usable sequences and OTUs detected. According to some authors, de-noising can alter the results of fungal community analyses by biasing against sequences that are not in the reference database and/or changing genuine sequences into sequences that the algorithm prefers (Hart et al., 2015). Operational taxonomic units (OTUs) were defined by clustering at 3% sequence divergence using USEARCH v5.1 (Edgar, 2010), and singleton OTUs removed. Sequencing and basic bioinformatics was performed by MR DNA (Shallowater, TX, USA). Sequences corresponding to 1738 fungal OTUs, which had a mean sequence length of 445±55 bp (mean±SD), were taxonomically classified using BLAST and the UNITE database v.7.1 (Kõljalg et al., 2013) in QIIME (Caporaso et al., 2010). Fungal OTUs were assigned to ecological guilds using FUNGuild v1.0 (Nguyen et al., 2016a). We separately analyzed the fungal OTUs identified as ectomycorrhizal (EMF), arbuscular mycorrhizal (AMF), dark septate endophytes (DSE), saprotrophs (soil, wood, dung and undefined saprotrophs), plant pathogens, other fungi (mycoparasites, animal pathogens, lichenized and foliar epiphytes) and the remaining unmatched fungal OTUs. FUNGuild assignments had mostly ‘highly probable’ and ‘probable’ likelihoods for all guilds: EMF (80%), AMF and DSE (100%), saprotrophs (89%), pathogens (96%) and other guilds (74%). To account for uncertainties associated with functional assignment, we repeated all analyses after excluding all OTUs assigned as ‘possible’ and found that the main results remained unchanged. For each rhizosphere sample, we obtained the total number of sequences corresponding to each fungal OTU, and calculated their relative sequence abundances based on the total number of sequences in the sample so as to account for the differential sampling depths. It is important to note that the relative sequence abundance of a fungal species or guild (e.g. EMF) generated from high-throughput sequencing does not necessarily equate the actual relative abundance of a fungal species or guild as the term is traditionally used in ecology (e.g. number of individuals, hyphal biomass, number or proportion of colonized root tips, etc.), and therefore the relative sequence abundance data should be interpreted with some caution. Finally, OTU richness was estimated by an individual-based multinomial model, which samples without replacement at a given sampling depth, using QIIME based on Colwell et al. (2012).

Statistical analyses

Repeated measures ANOVAs (RM-ANOVAs) were used to evaluate the effects of the two experimental factors (warming and rainfall reduction), time, and their interactions, on leaf gas exchange parameters (A, gs, WUE_i, ØPSII, Fv’/Fm’, E), leaf nutrients (%N, %P, N_area, P_area), leaf dry mass, LMA, δ¹³C, shoot elongation during late spring, late/early spring growth ratio, and survival rate. Temperature (warming vs. ambient) and rainfall (reduced vs. ambient) were used as between-subject factors and time was the within-subject factor in the RM-ANOVAs, followed by post-hoc tests (LSD pairwise means comparisons) to assess significant differences between treatments (Control, W, RR and W+RR). For each plant response variable analyzed by RM-ANOVA, the effect size of the climate manipulation factors was estimated using a partial eta squared (η_p²) if a significant effect was found. Analyses of variance (ANOVA) were used to test the effects of the different climate manipulation treatments on shoot biomass production per unit length and its components, followed by LSD post-hoc tests. Linear regression analysis was used to examine the relationship between A and g_s in each climate treatment, and the intercepts and slopes of the fitted regression lines were compared among treatments with analysis of covariance (ANCOVA) using Statgraphics Plus 5.1 (Manugistics, Rockville, MD, USA). Non-normal variables were log-, square root or arcsin transformed prior to analysis. We used SPSS 22.0 software (SPSS Inc., Chicago, IL, USA) to conduct regression, ANOVA and RM-ANOVA analyses. Changes in the community structure of EMF and saprotrophic fungal communities were tested by permutational multivariate analysis of variance (PERMANOVA) using Bray Curtis dissimilarity matrices with the adonis function in the vegan package for R (Oksanen et al., 2015; R Core Team, 2016). PERMANOVA was carried out using pairwise orthogonal contrasts comparing the OTU x plot relative sequence abundance matrices of each climate manipulation treatment with that of the control treatment. We used the betadisper function in vegan to determine whether significant differences were based on the centroid location or dispersion of each treatment. Non-metric multidimensional scaling (NMDS) of the Bray Curtis dissimilarity matrix was used to visualize differences in fungal community structure among treatments in vegan with the isoMDS function of the MASS package for R (Venables & Ripley, 2002). Climate manipulation treatment effects on the relative sequence abundance of the different fungal ecological guilds in the rhizosphere of H. squamatum shrubs were tested using ANOVAs followed by LSD post-hoc tests.

Results

Treatment effects on microclimatic variables

Throughout the study period, the warming treatment (W and W+RR) increased mean air temperature by ~2.5ºC (Fig. S3). Warming treatment effects were greatest during summer, when midday temperatures inside the OTCs were increased up to 6-7 °C on some days. The warming treatment also increased surface soil temperature by ~2.5ºC on average. Vapor pressure deficit was also higher in plots exposed to warming (W and W+RR) than in those exposed to current ambient temperature (Control and RR) throughout the study (1311 vs 1042 Pa, respectively); this difference was greatest during spring (1488 vs 1138 Pa, respectively) and summer (2924 vs 2330 Pa, respectively). Rainout shelters did not substantially alter air/soil temperature, as average differences between RR and control treatments were consistently below 0.4 °C throughout the study period (Fig. S3). Topsoil water content (0-5 cm depth) closely followed the rainfall events and was decreased to the same extent by the OTCs and the rainout shelters across the three climate manipulation treatments (by 2-3% on average, Fig. S4), relative to the control.

Experimental climate change impacts on plant gas exchange

Across the study period, warming reduced the mean net photosynthetic rate (A) of H. squamatum shrubs by 29% and 32.5% in the W and W+RR treatments, respectively (Fig. 2a; η_p²=0.587; P<0.001, Table S1). Net photosynthetic rates were consistently lower in shrubs exposed to warming (W and W+RR) than in those exposed to current ambient temperature throughout the study, except during a cold period in winter when A was higher under warming (February 2012; Fig. 2a; significant Warming x Time interaction in Table S1). Differences in net photosynthetic rates were greatest at the peak of the growing season (April-May), when A was, on average, 30.5% and 42% lower in the W and W+R treatments (respectively) than in the controls. Across dates, the warming treatments (W and W+RR) also reduced ØPSII by 8.6% relative to the controls (Fig. S5b; η_p²=0.304; P<0.001, Table S1). However, Fv’:Fm’ was not significantly affected by warming across dates (Fig.S5a; P=0.836, Table S1). Stomatal conductance and transpiration rate increased with warming (Figs 2b and S5c; η_p²=0.485; P<0.001 for g_s; Table S1), with mean increases of 33% in W and W+RR plants relative to the controls. The combination of decreased A with increased g_s under warming led to a drastic reduction in WUE_i, which was on average 55.2% and 46.6% lower in the W and W+RR plants, respectively, than in the controls throughout the study (Fig. 2c; Table S1; η_p²=0.953; P<0.001). The linear regression between mean A and g_s values across dates had significantly lower intercept and slope in W and W+RR plants than in Control or RR plants (ANCOVA, P<0.001 and P=0.008, respectively; R² values shown in Fig. 3).

Figure 2.

Mean net photosynthesis rate, stomatal conductance and intrinsic water use efficiency of Helianthemum squamatum throughout the four years of the study in each experimental treatment. Treatments are control, warming (W), rainfall reduction (RR) and warming + rainfall reduction (W+RR). Data represent means ± SE (n= 8-10 for W, RR and W+RR; n=15-20 for C). For each variable, the last cluster of bars represents the long-term average values across dates in each treatment, and different letters indicate significant differences (P<0.05) according to ANOVAs and LSD post-hoc tests. Data on g_s and WUEi not available in February 2012.

Figure 3.

Linear regressions between the mean net photosynthesis rate (A) and mean stomatal conductance (g_s) of Helianthemum squamatum in all the treatments across measurements dates. Error bars represent standard errors. Treatments are control, warming (W), rainfall reduction (RR) and warming + rainfall reduction (W+RR).

Across dates, rainfall reduction moderately decreased net photosynthetic rates (Fig. 2a; η_p²=0.107; P=0.042) and stomatal conductance (Fig. 2b; η_p²=0.101; P=0.049), and enhanced intrinsic water use efficiency (Fig. 2c; η_p²=0.204; P= 0.004). This treatment did not have, however, any effects on the ØPSII or the Fv’:Fm’ of H. squamatum (Fig. S5a, b; Table S1). On average, RR plants had 9.6% lower A, 12.1% lower g_s, 9.2% lower E and 4% higher WUE_i than control plants across dates. The detrimental effects of the RR treatment were greatest at the peak of the growing season (April-May), when RR plants had on average 17.2% lower A and 19.3% lower g_s than the controls across years.

Changes in leaf nutrient status, δ¹³C, LMA, shoot growth and post-summer survival with experimental climate change

Leaf N_area and P_area were consistently decreased by warming across years (W and W+RR treatments; η_p²=0.516, P<0.001 for N_area; η_p²=0.496, P<0.001 for P_area; Tables 1, S2). Leaf N_area was also significantly decreased by the RR treatment in 2014 (R²=0.418, P<0.001). Despite large variability in environmental conditions within and among measurement dates, net photosynthetic rate and intrinsic water use efficiency correlated positively with both N_area and P_area across treatments (Fig. S6; R² and P values shown in figure). Plants subjected to W and W+RR treatments also had lower average leaf N and P concentrations than plants under current ambient temperatures across the study period (Tables S2, S3; η_p²=0.194, P=0.004 for N; η_p²=0.188, P=0.008 for P). In the last growing season of the study (May 2015), leaf N and P concentrations were 11-25% lower in the W, RR and W+RR treatments than in control plants (R²=0.492, P<0.001 for N; R²=0.213, P=0.021 for P; Table S3).

Table 1.

Leaf nitrogen and phosphorus contents per unit area (N_area, P_area), leaf dry mass and leaf mass area, and shoot elongation during late spring (May-June) of Helianthemum squamatum shrubs in the control (C), warming (W), rainfall reduction (RR) and warming + rainfall reduction (W+RR) treatments. Data represents means ± SE (n= 8-10 for W, RR and W+RR; n=15-20 for C).

	Leaf Narea (g m-2)
YEAR	C	W	RR	W + RR
2013	4.03 ± 0.08	3.31 ± 0.11	3.83 ± 0.10	3.21 ± 0.10
2014	3.72 ± 0.09	2.90 ± 0.14	3.16 ± 0.15	3.08 ± 0.15
2015	3.36 ± 0.11	2.78 ± 0.14	3.41 ± 0.16	3.05 ± 0.14
	Leaf Parea (g m-2)
2013	0.183 ± 0.025	0.143 ± 0.029	0.176 ± 0.026	0.139 ± 0.022
2014	0.125 ± 0.016	0.103 ± 0.025	0.113 ± 0.015	0.110 ± 0.008
2015	0.118 ± 0.005	0.088 ± 0.007	0.123 ± 0.008	0.106 ± 0.007
	Leaf Dry Mass (mg)
2013	12.3 ± 0.8	9.4 ± 1.0	9.4 ± 1.0	10.7 ± 1.0
2014	18.4 ± 0.9	11.3 ± 1.4	15.4 ± 1.4	15.0 ± 1.5
2015	12.0 ± 0.7	10.7 ± 0.9	13.4 ± 1.0	10.6 ± 0.9
	Leaf Mass Area (mg cm-2)
2013	18.1 ± 0.5	15.0 ± 0.7	18.0 ± 0.7	15.4 ± 0.7
2014	18.3 ± 0.5	14.4 ± 0.8	15.6 ± 0.8	15.7 ± 0.9
2015	17.0 ± 0.7	17.0 ± 0.9	21.7 ± 1.0	17.1 ± 0.9
	Shoot Elongation in late spring (cm)
2012	4.8 ± 0.3	3.7 ± 0.6	4.5 ± 0.6	3.7 ± 0.6
2013	8.4 ± 0.6	4.7 ± 1.0	7.5 ± 0.9	7.2 ± 1.0

Decreases in leaf nutrient concentrations translated into even larger decreases in total foliage N and P pools due to large concurrent decreases in plant biomass production under the W and W+RR treatments (R²= 0.299, P<0.001 for foliage N pool; R²=0.275, P<0.001 for foliage P pool; Table 2). Four years after the start of the experiment, total dry biomass per unit shoot length was 33% and 38% lower in the W and W+RR treatments, respectively, than in the control (R²=0.227, P=0.002; Table 2). Decreased biomass production under warming was primarily due to large reductions in foliage (R²=0.254, P=0.001) without changes in stem biomass. Plants in the W and W+RR treatments experienced reductions in both the number of leaves (28%; R²=0.211, P=0.011) and the total leaf area per unit shoot length (36%; R²=0.157, P=0.002) compared to control plants. Plants in the RR treatment also had a lower number of leaves per unit shoot length than control plants (16% decrease; R²=0.211, P= 0.036).

Table 2.

Mean dry biomass production in Helianthemum squamatum shrubs at the end of the four-year study period (October 2015). Data on total shoot dry mass (leaves plus stems), leaf dry mass, stem dry mass, leaf number, leaf area and foliage N and P pools per 10 cm shoot length are shown. Data represents means ± SE in the control, warming (W), rainfall reduction (RR) and warming + rainfall reduction (W+RR) treatments (n= 10 for W, RR and W+RR; n=30 for C). Values in the same column sharing letter are not significantly different according to post-hoc LSD tests.

	Shoot Dry Mass (mg)	Leaf Dry Mass (mg)	Stem Dry Mass (mg)	Leaf Number	Leaf Area (cm2)	Foliage N pool (mg)	Foliage P pool (mg)
CONTROL	452.9 ± 26.5 a	324.4 ± 20.3 a	128.5 ± 8.5 a	126.5 ± 7.7 a	18.5 ± 1.4 a	5.71 ± 0.35 a	0.21 ± 0.01 a
W	299.5 ± 38.9 b	196.9 ± 29.8 b	102.7 ± 12.5 a	94.8 ± 11.3 ab	12.0 ± 2.1 b	3.30 ± 0.52 b	0.14 ± 0.02 b
RR	364.2 ± 44.4 ab	259.6 ± 33.9 ab	104.6 ± 14.2 a	100.0 ± 12.8 ab	17.5 ± 2.3 a	4.83 ± 0.56 a	0.20 ± 0.02 a
W+RR	283.1 ± 44.4 b	184.2 ± 33.9 b	98.9 ± 14.2 a	72.5 ± 12.9 b	11.7 ± 2.4 b	2.84 ± 0.59 b	0.11 ± 0.02 b

The leaf δ¹³C values of H. squamatum were consistently lower in the W and W+RR treatments than in the control and RR treatments throughout the study (Fig. 4; Table S2; η_p²=0.360, P<0.001), indicating decreased time-integrated water use efficiency under warming. Leaf δ¹³C data associated positively with WUEi data (obtained by leaf gas exchange measurements) across individual plants and treatments (e.g. R²=0.224, P=0.001 in May 2013; Fig. S7), thus confirming that W and W+RR plants operated at a lower leaf-level water use efficiency than control and RR plants. Warming (W and W+RR) reduced LMA across years (Tables 1, S2; η_p²=0.428, P<0.001) relative to the control treatment, a response largely due to a 21% average reduction in mean leaf dry mass (Tables 1, S2; η_p²=0.166, P=0.007). Mean shoot elongation during May-June was also reduced by warming (24% mean decrease; Tables 1, S2; η_p²=0.112, P=0.019), and thereby both W and W+RR plants had lower ratios of late spring/early spring shoot elongation than plants exposed to ambient temperatures (Table S2; η_p²=0.201, P= 0.001), indicating an advanced shoot growth phenology and an earlier cessation of the growing season. Shoot growth phenology was not significantly affected by the RR treatment (Table S2).

Figure 4.

Mean leaf carbon isotope ratios (δ¹³C) of Helianthemum squamatum throughout the four years of the study. Treatments are control, warming (W), rainfall reduction (RR) and warming + rainfall reduction (W+RR). Data represent means ± SE (n= 8-10 for W, RR and W+RR; n=15-20 for C). Consistently lower δ¹³C values in warmed plants (W and W+RR treatments) indicate decreased intrinsic water use efficiency.

Post-summer plant survival rate across treatments was lower in 2012 (60%) than in the other years (89% in 2013; 90% in 2014; 85% in 2015), probably because 2012 was the driest year of the study (262 mm annual rainfall, vs. 326-458 mm in the other years). Across years, mean post-summer survival was not significantly affected by the climate manipulation treatments according to RM-ANOVA (Table S2). However, whereas post-summer survival was largely unaffected by climate manipulation in wetter years (2013, 2014 and 2015), it was significantly reduced by the W and W+RR treatments in the driest year of the study (2012; R²=0.107, P=0.002, Fig. 5). Post-summer plant survival rate was very similar between RR and control plants across years.

Figure 5.

Mean post-summer plant survival rate of Helianthemum squamatum during the driest year of the study (2012), in which survival was significantly reduced by warming (P=0.001). Treatments are control, warming (W), rainfall reduction (RR) and warming + rainfall reduction (W+RR). Data represent means ± SE (n= 10 for W, RR and W+RR; n=30 for C). Different letters indicate significant differences (P<0.05) among treatments according to ANOVAs and LSD post-hoc tests.

Changes in rhizosphere fungal communities with experimental climate manipulation

Pyrosequencing analyses detected a total of 1738 fungal OTUs in the rhizosphere of H. squamatum shrubs, including 30 EMF, 17 AMF, 18 DSE, 453 saprotrophs, 202 plant pathogens and 54 OTUs belonging to other ecological guilds (lichenized, mycoparasites, epifoliar, animal pathogens), as well as 964 unmatched OTUs. Rarefied EMF OTU richness per sample was 9.8±1.3 (mean±SD), with no significant differences among treatments. The most frequent and abundant EMF OTUs were identified as members of the genera Geopora, Picoa, Cortinarius and Ceratobasidium (Fig. 6).

Figure 6.

Rank abundance plot showing the impact of the climate manipulation treatments on the relative abundance of sequences of different ectomycorrhizal fungal (EMF) genera in the rhizosphere of Helianthemum squamatum shrubs. Means and standard errors per treatment (n=8) are shown.

The analysis of the EMF community through OTU-based distance metrics revealed significant effects of all climate manipulation treatments on EMF community structure relative to the control. Effects were significant for W (PERMANOVA: F_1,28=2.35, P<0.03, R²=0.08), RR (F_1,28=2.08, P≤0.05, R²=0.07) and W+RR (F_1,28=1.84, P≤0.05, R²=0.07), as can be visualized in the NMDS plot shown in Fig. 7. Differences in EMF community structure between the control and the climate manipulation treatments were not due to differences in dispersion among treatments, as suggested by the non-significant result of the follow-up betadisper analysis (ANOVA: F_3,26=0.96, P=0.42). In contrast to EMF, the analysis of the community structure of other fungal guilds through OTU-based distance metrics did not detect any significant effects of climate manipulation (all PERMANOVAs: F_1,28<1.5, P>0.08, R²≤0.05).

Figure 7.

Nonmetric multidimensional scaling (NMDS) plot of ectomycorrhizal fungal community structure in the control and climate manipulation treatments. Standard deviation ellipses allow visualizing multivariate dispersion in ectomycorrhizal community structure. Treatments (n= 8) are control, rainfall reduction (RR), warming (W) and warming + rainfall reduction (W+RR). The 2D stress value shown in the figure provides a global measure of the quality of fit of the NMDS plot to the data, with a low value indicating a good fit.

Across treatments, the relative abundance of EMF sequences in the rhizosphere fungal community was unrelated to focal H. squamatum plant size or total plant cover in the experimental plots (P>0.7 in both cases). The relative abundance of EMF sequences in the rhizosphere was reduced by more than 66% by all the climate manipulation treatments compared to the control (R²=0.391, P=0.003; Fig. 8a), and this reduction was largely driven by declines in the most dominant OTUs of the local EMF community, particularly those belonging to the genera Geopora sp. and Cortinarius sp. (R²=0.314, P=0.013 and R²= 0.338, P=0.008, respectively; Fig. 6). Interestingly, the relative abundance of EMF sequences in the rhizosphere of H. squamatum shrubs was positively associated with mean leaf N content in host plants across treatments (Fig. 9). The relative abundance of EMF sequences in the rhizosphere was also positively associated with mean leaf P content (R²=0.172, P=0.018) and mean leaf dry mass (R²=0.195, P=0.011) in host plants across treatments. In contrast to EMF, the relative sequence abundances of all other fungal guilds in the rhizosphere (saprotrophs, plant pathogens, AMF, DSE, others) were not significantly affected by the climate manipulation treatments (Fig. 8b, c; Fig. S8).

Figure 8.

Box-and-whisker diagrams showing the relative abundance of fungal sequences ascribed to different fungal ecological guilds in the rhizosphere of Helianthemum squamatum shrubs in the climate treatments: a) ectomycorrhizal (EMF); b) saprotrophs; c) plant pathogens. For each climate treatment (n= 8), boxes indicate the upper (75%) and lower (25%) quartiles of the data, and the thick black line in the middle of box represents the median value. Lines extending vertically from the boxes (whiskers) indicate the maximum and minimum values, and atypical values (outliers) are plotted as individual points. For each fungal guild, different letters indicate significant differences (P<0.05) among treatments according to ANOVAs followed by LSD post-hoc tests. Treatments are control, warming (W), rainfall reduction (RR) and warming + rainfall reduction (W+RR).

Figure 9.

Relationship between the relative abundance of ectomycorrhizal fungal (EMF) sequences in the rhizosphere of Helianthemum squamatum shrubs and leaf N content across climate treatments (N=32). All data were obtained during spring 2014. Treatments are control, rainfall reduction (RR), warming (W) and warming + rainfall reduction (W+RR).

Discussion

Warming (W and W+RR) strongly decreased the net photosynthetic rate of H. squamatum, with no indication of progressive photosynthetic acclimation to warming with time (Gunderson et al., 2010). Warming (W and W+RR) reduced the carbon assimilation rate of H. squamatum shrubs at any given stomatal aperture (Fig. 3), revealing the existence of strong non-stomatal limitations on photosynthesis (nutritional, biochemical, metabolic and/or diffusional; Feller et al., 1998; Flexas & Medrano, 2002; Salvucci & Crafts-Brandner, 2004; Sage & Kubien, 2007). Net photosynthetic rate was positively associated with both N_area and P_area across treatments (Fig. S6), indicating a key role of poor leaf nutrient status in decreasing photosynthesis in the climate manipulation treatments relative to the controls (Galmés et al., 2013; León-Sánchez et al., 2016). The simultaneous decreases of leaf N_area and P_area likely caused interactive and mutually reinforcing detrimental effects on the photosynthetic biochemistry of W and W+RR plants, given the key importance of N and P for Rubisco carboxylation capacity and for ribulose-1.5-bisphosphate regeneration, respectively (Reich et al., 2009; Warren, 2011). Leaf photosynthetic capacity in plants correlates with leaf N and P contents because of the crucial role that N-rich enzymes, particularly Rubisco, play in the biochemical fixation of CO₂ (Wright et al., 2004). Moreover, biochemical reactions involved in photosynthesis require orthophosphate as a substrate, and N allocation to Rubisco decreases with declining P supply (Warren & Adams, 2002).

Stomatal conductance and transpiration rates were strongly enhanced by warming, especially during wet periods when soil moisture was plentiful, as found in other studies (White et al., 1999; Misson et al., 2004). Increased g_s and E with warming may help prevent or minimize leaf overheating and heat-induced damage of the photosynthetic machinery through increased evaporative leaf cooling (Crawford et al., 2012). Increased g_s and E with warming may also represent an adaptive mechanism for enhancing transpiration driven mass flow of soil nutrients to roots in the nutrient-stressed W and W+RR plants (Cramer et al., 2009; Matimati et al., 2014). Additionally, we speculate that warming could have favored the release of crystallization water contained in gypsum rocks and soil, which can account for up to 21% of gypsum’s weight (Palacio et al., 2014), thus rendering it available for plant water uptake. This could have further contributed to increased g_s and E in W and W+RR plants during spring and summer. However, higher g_s and E rates in warmed plants likely accelerated soil moisture depletion, thereby shortening the growing season through an earlier onset of drought stress (as indicated by reduced shoot elongation during the latter half of spring in W and W+RR plants). Moreover, leaf-level intrinsic water use efficiency was halved by warming as a result of the large decreases in A and increases in g_s in the W and W+RR treatments, which may represent a key detrimental feedback whereby leaves need to transpire more water to assimilate less carbon in a climate change scenario.

Foliage biomass production per unit shoot length was strongly decreased in the W and W+RR treatments, a result in agreement with findings from climate manipulation studies conducted in semiarid ectomycorrhizal ecosystems from the USA (Adams et al., 2015; García-Forner et al., 2016). However, despite the large decreases in leaf nutrient status, photosynthesis and shoot growth observed in W and W+RR plants, they achieved similarly high post-summer survival rates as control plants in years with above- or near-average rainfall. Several phenotypic plasticity mechanisms many have helped ensure plant survival under warming. These include advancing shoot growth phenology to escape the intensification of late season drought and heat stress (Badeck et al., 2004; Menzel et al., 2006; Parmesan 2007), building fewer leaves with lower dry mass and LMA to decrease the carbon cost of growth (Poorter et al., 2009) and increasing g_s and E per unit leaf area to reduce leaf overheating (Crawford et al., 2012) while at the same time reducing foliage area to decrease whole plant transpiration (Limousin et al., 2009), as found in our study. However, the buffering capacity of these presumably adaptive responses was clearly overwhelmed during a dry year (2012), as indicated by large increases in summer mortality in the W and W+RR treatments.

Despite moderate decreases in photosynthetic rates and shoot growth, RR plants achieved similar post-summer survival rates as the controls throughout the study (even in a dry year), which adds to the mounting evidence that Mediterranean semiarid shrubs are remarkably tolerant and resistant to rainfall reduction because they are well preadapted to high (intra- and interannual) precipitation variability and recurrent drought stress (Miranda et al., 2011; Tielbörger et al., 2014). In addition to an enhanced WUE_i, other adaptive mechanisms may have contributed to the resilience of H. squamatum against rainfall reduction, such as decreasing leaf number per unit shoot length to reduce whole canopy transpiration (Limousin et al., 2009 and this study), increasing rooting depth or volume, and/or increasing root/shoot biomass ratios (Nardini et al., 2014). Interestingly, very few significant interactions between warming and rainfall reduction were found in this study (Tables S1 and S2). The combined detrimental effects of warming and rainfall reduction on plant performance were not greatly exacerbated when applied simultaneously and were not even additive, which is similar to the results of other studies conducted in semiarid EMF ecosystems of the US Southwest (Adams et al., 2015; García-Forner et al., 2016).

The findings of this study are in agreement with those of previous reports showing links between host plant performance and ectomycorrhizal fungal community responses to climate manipulation. Several studies have shown that EMF abundance and richness generally increase with warming in high latitude tundra and boreal ecosystems where photosynthesis is limited by low temperature (Clemmensen et al., 2006; Deslippe et al., 2011; Treseder et al., 2016). In contrast, the opposite response (i.e. decreased plant photosynthesis along with decreased EMF relative abundance and richness under warming) has been recently reported in a boreal/temperate ecotone ecosystem located near the lower latitudinal limit of the host plant range where productivity is limited by water availability (Fernández et al., 2017). We found that warming, rainfall reduction and their combination altered ectomycorrhizal community structure and led to drastic decreases in the relative abundance of EMF sequences in the rhizosphere of H. squamatum, which is the only EMF host plant species at our study site. In contrast to EMF, the community structure and relative sequence abundances of other non-mycorrhizal fungal guilds (e.g. saprotrophs) did not change significantly in the climate manipulation treatments. This strongly suggests that altered EMF community structure and decreased relative abundance of EMF sequences with climate manipulation was linked to plant responses and largely driven by decreased carbon supply to mycorrhizae from their stressed host plants under hotter and/or drier conditions (Fernández et al., 2017; Castaño et al., 2017).

The dominant fungi in the studied EMF community have medium distance (Cortinarius), short distance (Geopora, Picoa) or contact (Ceratobasidium; Veldre et al., 2013) hyphal exploration types (Agerer, 2001; Tedersoo & Smith, 2013). Cortinarius showed the largest proportional reduction in relative sequence abundance with the W+RR treatment compared to the control (over tenfold), suggesting that EMF with more abundant extraradical mycelium production (and thus higher carbon demand and sink strength) could be particularly vulnerable to decreases in host plant photosynthesis under increasing heat and drought stress (Fernández et al., 2017). Interestingly, mean leaf dry mass and N and P contents in H. squamatum shrubs decreased with decreasing relative abundance of EMF sequences in the rhizosphere across treatments, and this was associated with decreased host plant photosynthesis and productivity, thereby supporting the conceptual model outlined in Fig. 1. Additional studies assessing the impact of climate manipulation on percentage colonization of roots and extraradical hyphal production by mycorrhizal fungi would be useful to further corroborate the validity of this conceptual model. Tracer (¹³C, ¹⁵N, ³²P) experiments aimed at measuring the impact of climate manipulation on the exchange of carbon and nutrients between the plant and mycorrhizal fungal partners would also be useful to further test the model.

In our study, altered EMF community structure and decreased relative abundance of EMF sequences in the rhizosphere fungal community under climate change conditions suggests a shift in the competitive balance and partitioning of soil nutrients between the saprotrophic and EMF functional guilds to the detriment of the latter, as indicated by large decreases in foliage nutrient pool sizes in EMF host plants. Reduced leaf N and P contents may decrease plant nutritional quality for herbivores and slow down litter decomposition and nutrient recycling under a warmer and drier climate (Allison et al., 2013; I. Prieto & J.I. Querejeta, unpublished data). Moreover, declining relative abundance of EMF sequences and plant biomass growth with climate aridification might also stimulate soil organic matter decomposition through an increased soil nutrient availability for competing fungal and bacterial decomposers (Fernandez & Kennedy, 2016), which would be expected to further accelerate land degradation and desertification in drylands.

In contrast to EMF, the relative abundance of AMF sequences under H. squamatum was not significantly affected by climate manipulation, possibly due to their lower carbon requirements and cost to the host plant compared to EMF (Jones et al., 1998; Leake et al., 2004; Gehring et al., 2006; Soudzilovskaia et al., 2015b) and to their better adaptation and tolerance to drought and heat stress in soil (Allen et al., 1995; Querejeta et al., 2009; Soudzilovskaia et al., 2015a). The low relative abundance of AMF sequences found by not using AMF-specific primers (Schlaeppi et al., 2016) is an important caveat of our study, although it should be noted that H. squamatum typically shows considerably lower AMF colonization than other coexisting plant species in semiarid shrubland communities such as our study area (Alguacil et al., 2009).

In conclusion, the interdependent responses of plants and their associated EMF to warming and rainfall reduction appear to have mutually amplifying effects that strongly decrease plant nutrient status, photosynthesis, WUEi, biomass production, and drought survival. The linked responses of EMF and their host plants to simulated climate change lead to large declines in vegetation nutrient pool size, with potentially cascading detrimental effects on ecosystem nutrient retention and cycling (van der Heijden et al., 2010). The observed adverse impacts of forecasted climate change on plant performance and rhizosphere-microbial interactions could thereby send drylands into a degradation trajectory in which carbon and nutrient flows through fungal-plant mycorrhizal associations spiral downward, leading to an alternative state of greatly decreased vegetation productivity. Our findings highlight the vulnerability of native plants and their symbiotic mycorrhizal fungi to forecasted climate change in semiarid shrubland ecosystems, and point to the importance of a deeper understanding of plant-soil feedbacks to accurately predict dryland responses to forecasted aridification.