Introduction

Root exudates play a critical role in nutrient acquisition, either by directly mobilising nutrients from the growth medium or indirectly acquiring nutrients through microbial actions1,2. However, root exudates are underexploited in the quest to breed nutrient-use-efficient crops, not least due to a limited understanding of the exudates of wild crop relatives and landraces that are a large potential source of information and genetic material for modern breeding programmes3,4. In the context of phosphorus use efficiency (PUE), low-molecular-weight exudates such as organic acids and phytosiderophores or enzymes (e.g., phosphatases) have been widely studied5. However, many exudates remain unidentified, and the emerging field of exudomics is addressing this gap through advanced technologies, including non-target Fourier Transform Ion Cyclotron Resonance Mass Spectrometry (FT-ICR-MS) to connect exudomic data with genomics, offering valuable insights for crop breeders6,7. Root exudates are relevant for the emerging circular nutrient economy that repurposes fertiliser nutrients from waste streams, with efficiency gains expected by aligning genotypes with next-generation fertilisers for target pedoclimatic conditions and agronomies. Thus, root exudome analysis can support the transition towards nutrient use efficiency, expanding ‘omics tools.

Next-generation fertilisers are gaining momentum as they reduce a region’s or nation’s dependence on conventional P fertilisers. Most currently used mineral P fertiliser are mined from non-renewable mineral deposits at great environmental cost, are subject to price volatility and supply chain uncertainties with mineral P deposits restricted to a few regions globally, and are often inefficient, with only half the applied P fertiliser incorporated by global crops8. Leveraging knowledge of root exudates has the potential to advance use of a broad range of P sources, including sewage ash, struvite, hazenite, and organic materials—such as manures and composts from food wastes—as viable alternatives to conventional P fertilisers9,10,11. Recycled P sources such as sewage ash and struvite can supply crops when paired with suitable crops and growth environments10,12, with the circular P economy enabled by consortia of engineers, fertiliser designers, crop breeders and agronomists.

Numerous crop traits determine P efficiency. Internal PUE refers to the plant’s ability to use acquired P efficiently through metabolic adjustments in response to P limitation, including remobilising P from source to sink tissues13,14. External PUE is the efficient acquisition of P from the growth medium, which is facilitated by root architecture, mycorrhizal associations, microbial interactions and root exudates, for example, carboxylates and enzymes that mobilise P from inorganic and organic sources, respectively15,16. Phosphorus sources such as rock phosphate and some recyclates have low solubility, which can limit their effectiveness as fertilisers. For example, rock phosphate17, sewage ash (AshDec®, Raniro et al.10) and struvite18 can exhibit low relative fertiliser efficiency compared to more readily water-soluble commercial fertilisers such as triple superphosphate. However, when paired appropriately with target crops and environments, low-solubility P sources can perform as well as, or even outperform, conventional fertilisers10,18,19. Since fertiliser effectiveness depends on complex interactions between crop genotype, P source and environment, it is essential to understand the causalities.

We used FT-ICR-MS non-target analysis to develop exudomics as a tool to improve the PUE of sorghum. Sorghum was chosen as a versatile, climate-resilient crop used for (gluten-free) food, livestock feed, as biofuel and biomaterial production20. Genetic improvement of sorghum is well advanced, as sequencing of numerous genotypes has revealed a large pool of genetic diversity and domestication history linked to functional traits21.

Root exudates, including phenolics, flavonoids and others, can promote microbial activity, including P-solubilizing bacteria, and can have allelopathic effects, as seen with ferulic acid22 and sorgoleone23,24. Sorghum-specific sorgoleone promoted root colonisation by arbuscular mycorrhizal fungi25,26, and the increased phenolic exudation of sorghum mutant BTx623 enriched beneficial rhizosphere bacterial communities23. To isolate the direct effects of exudates on P mobilisation, our study minimised microbial interactions in controlled experimentation.

Here, we explored root exudation as a tool for breeding PUE sorghum. We compared two landraces and one improved inbred line, hypothesising that the landraces, having been selected in lower P environments than the inbred line, outperform the inbred line under conditions of low P availability. Analysis of growth, P acquisition and root exudates using P sources of different solubility and lines with contrasting genetic backgrounds has the potential to advance understanding of how sorghum modulates exudate profiles.

Methods

Seed preparation and germination

We selected three Sorghum bicolor (L.) Moench lines from a previously characterised diversity panel, predominantly consisting of genotypes from the Sorghum Conversion Program27. The genotypes included (i) SC283-14E a converted landrace from Tanzania (mixed race) characterised by high root surface area that was positively correlated with PUE under low P conditions28, (ii) SC648-14E a converted landrace from South Africa (kafir) characterised by its relationship with arbuscular mycorrhizal fungi and high P uptake capacity29, and (iii) BTx623 (PI 659985 MAP) a widely used inbred line (mixed race) developed under high-input agricultural conditions, serving as the reference genome for sorghum and exhibiting extensive variations30,31. Seeds were sterilised with a solution containing 100 ml bleach, 100 ml autoclaved milli-Q water, and 50 µL Tween 20 detergent32 and suspended in the sterilisation solution for 1.5 h using a rotary rube mixer (Ratek, RSM7DC) at 1.5 rotations per minute. Seeds were taken to a sterilised laminar flow cabinet where the sterilisation solution was removed via pipette. The seeds were rinsed with autoclaved Milli-Q water and lightly suspended before being removed via pipette. The rinsing step was repeated 6 more times. Seeds were then placed into Petri dishes with filter paper (Whatman®, 1001-085) moistened with autoclaved milli-Q water and positioned in a sterilised growth cabinet and semi-uncovered. Germinates were transferred into the microcosms (see below) upon radicle emergence.

Experimental details: growth medium, microcosms and fertiliser

The growth medium comprised of a 1:1 (v/v) mix of fine-grain silicate sand (crystalline silica, Richgro washed play sand) and coarse river sand (washed Brisbane River sand provided by the University of Queensland glasshouse services) to ensure a fine but sufficiently porous substrate. The sand was washed with Decon 90, a P-free surfactant, followed by repeated rinsing with autoclaved milli-Q water. Potassium chloride was used to neutralise the growth medium to a final pH of 6.85 and placed in an oven at 110 °C to dry for 48 h33. The microcosm design was a modified 50 mL Falcon Tube34 with an open top and base of 27 and 21 mm diameter, respectively. The base was inset with two layers, which included a 0.2 mm thick stainless steel mesh underlaying glass microfibre filter paper (Whatman®, GF/A). Microcosms were filled with 40 g of sand, followed by fertiliser deposited 2.5 cm below the surface, and filled with an additional 35 g of sand and placed in an oven at 62.5 °C for 72 h.

Phosphorus sources included granules with 1.5 ± 0.1 mm diameter (rock phosphate (RP), hazenite (Haz), and single superphosphate (SSP)). Each microcosm received 5 mg P at the start of the experiment (except the mineral solution control, see below). Selected fertiliser pellets were placed in the oven at 62.5 °C for 48 h followed by UV exposure for 30 min for sterilisation. The control plants received mineral solution (MS) with an initial 0.1 mL 91.4 mM KH2PO4 and 8.5 mM K2HPO4 in the first week, which incrementally increased to 0.25, 0.5 and 1.0 mL after each consecutive week (totalling 20 mg P at the end of the 4-week experiment). All treatments received a complete (no added P) nutrient solution every second day, with 1 mL nitrogen solution containing NH4NO3 initially at 2.5 mM and incrementally increasing to 5, 10 and 20 mM each consecutive week. All other nutrients were provided as solution with 1 mM K2SO4, 1 mM MgSO4, 1 mM CaCl2, 200 µM FeEDTA, 10 µm MnSO4, 10 µm H3BO3, 1 µm CuSO4, 3 µm ZnSO4 and 0.35 µm Na2MoO4.

Treatments were arranged in a randomised block design. Plants were grown in a growth cabinet (NVIRON, A1000) for 28 days post-planting and received full-spectrum artificial light for 18 h at 30 °C, and no light for 6 h at 28°C. Photosynthetic photon flux density in the growth cabinet equalled 200 for the first 14 days, and 480 for the remaining 14 days (PPF-UV = 2.28, PPF-B = 98.71, PPF-G = 189.97, PPF-R = 192.42, PPF-NIR = 20.33). Correlated colour temperature, lux and absorption peak equalled 4689 K, 29871 lm m2 and 660 nm, respectively. Water holding capacity (WHC) of the microcosms was calculated daily by weighing each tube and maintained at 90%. The microcosm surface was lightly sprayed with autoclaved milli-Q at mid-day and late afternoon to avoid drying out. Microcosms were flushed at the end of weeks 1, 2 and 3 at 200% WHC to mitigate potential salinity.

Exudate collection and FT-ICR-MS analysis

On day 28, the leachate was collected within the same light and temperature conditions between 10:00 and 16:00 to minimise diurnal effects35. Autoclaved milli-Q water was used to flush 150% WHC, and 2 × 1.5 mL of leachate was collected in 2 mL Thermo ScientificTM Screw Cap Micro Tubes with an altered pierced cap for air-transfer sublimation during the freeze-drying step. Samples caps were covered with parafilm and immediately transferred to a −80 °C freezer for 72 h. The parafilm was then removed from the sample caps and placed in a freeze dryer (VirTis®, BenchTop Pro with Omnitronics™) at 71 mT, −48.5 °C, for 48 h.

Freeze-dried exudate samples were rehydrated using 400 µL milli-Q water and vortexed until solubilised. Sample pH was adjusted to 2 using 0.5 µL HCl (32%). A further 50 µL of methanol was leached through an OMIX C18 100 µL tip, followed by 100 µL of acidified water (pH 1.5). The exudate solution was leached through a C18 tip, followed by 100 µL of acidified water. Lastly, 50 µL of methanol was flushed through the C18 for analysis. Samples were loaded and analysed by the FT-ICR-MS at a pump injection rate of 2 µL min−1 (methanol).

We chose non-target analysis to detect biochemicals in the range of molecular masses of 125–2000 amu. Samples were analysed on a high-field FT-ICR-MS from Bruker Daltonics with a 12-T magnet from Magnex. A time domain transient with four million data points (4 MWords), representing the recorded signal from ion oscillations in the magnetic field, was obtained and Fourier-transformed into a frequency domain spectrum. The frequency domain was afterward converted to a mass spectrum by the SolariX Control program of Bruker Daltonics. The ion excitations were generated in broadband mode (frequency sweep radial ion excitation), and 500 scans were accumulated for each mass spectrum in a mass range of 120–1000 amu. Ions were accumulated for 300 ms before ICR ion detection. The pressure in the quadrupole/hexapole and ICR vacuum chamber was 3 × 10−6 mbar and 6 × 10−10 mbar, respectively.

Data were analysed using the software Bruker Compass DataAnalysis 5.0 SR1 (x64) version 203.2.3586, with a smooth width (m/z) of 0.0002 (Da) with 10 cycles. Internal calibration search ranges from 0.005 (m/z) ±0.2 ppm. FT-ICR-MS with m/z from 120 to 1000 amu were calibrated externally and internally to preclude alignment errors. Subsequently, the mass spectra were exported to peak lists at a signal-to-noise ratio ≥3. Elemental formulas were calculated combinatorially within a mass accuracy window of ±0.2 ppm for each peak in batch mode by an in-house software tool and validated via the senior-rule approach/cyclomatic number, assuming valence 2 for sulphur (S).

FT-ICR-MS analysis enables highly resolved (R > 106 at m/z 200) and accurate chemical mass analysis of electrospray-generated ions within a 200 ppb error window over a wide mass range from m/z 100 to 1000. The weight of the ions is measured with a precision lower than the mass of an electron (Δm/z = 0.0003 amu), and the specific signals can be differentiated with the same mass precision due to the ultrahigh resolution. These exact masses of the ions can routinely be converted into unique compositional formula bearing the light elements carbon (C), hydrogen (H), nitrogen (N), oxygen (O), and S (or any other element in the target), also taking account of their possible chloride and sodium adducts and of their natural isotopic abundance. The solvent extracts generated over 20,000 thousands of individual signals that were converted into elementary compositions (elemental family) and represented in van Krevelen type of diagrams (H/C vs. O/C) or related (H/C vs m/z) in which each formula is represented by a dot (the size of the dot is proportional to its abundance) as a projection of the relative oxygenation degree (O/C) and saturation degree (H/C) for various classes of compound types (CHO, CHNO, CHOS, CHNOS). Analysed compounds were limited to a maximum of four N and P metabolites were excluded from the analysis.

Biomass and root length analysis

After leachate collection, above-ground biomass was harvested and placed into an oven at 39 °C for 72 h. The growth medium was washed, and roots removed by rinsing with deionised water. Roots were placed on a transparent plastic dish and scanned to quantify root characteristics (Epson Expression 10000XL, B11B168031) at 16-bit Greyscale with 800 dpi resolution. Scanned root images were prepared using Adobe Photoshop (version 24.7.4) to remove residual growth medium and background incidentals. Scanned root images were analysed using WinRHIZOTM software (version 2019a) with 15 root classes with 0.1 mm intervals (<0.1 to >1.4 mm). After scanning was completed, roots were dried at 39 °C for 72 h in an oven. Shoot and root dry weight was recorded before ICPOES analysis (5% nitric acid digestion, Thermo iCAP PRO XP) to quantify total P.

Statistical analyses and graphical representations

To analyse the measured plant traits across genotypes and P treatments, we performed a two-way analysis of variance (ANOVA) for total biomass dry weight, shoot and root biomass, total root length, and total plant P content, followed by a Tukey test for post hoc multiple comparisons at a significance level of p ≤ 0.05. Principal component analysis (PCA) and partial least squares-discriminant analysis (PLS-DA) were conducted using R statistical software to investigate the clustering and separation of samples based on their chemical compositions. Confidence ellipses were calculated based on a 95% confidence interval to illustrate the distribution of the samples within each group. Both analyses were performed with ‘mixOmics’ and ‘ggplot2’ packages in R. A hierarchical cluster analysis (Hierarchical Clustering Explorer Version 3.5) allowed group samples based on their similarities across the dataset, where the dendrogram illustrates the relationships between treatments and respective genotypes by successively merging clusters according to their proximity in feature space.

Results

Plant growth and phosphorus acquisition

The effect of sorghum genotypes and P treatment (No P, rock phosphate RP, single superphosphate SSP, hazenite Haz, mineral solution MS) were analysed with the response variables of P acquisition, total biomass, shoot and root allocation, and root length (Fig. 1, Supplementary Table 1). The two-way ANOVA revealed a significant interaction (F8, 45, 2.60, p ≤ 0.05) between genotype (F2, 45, 28.3, p ≤ 0.05) and P treatment (F4, 45, 251.9, p ≤ 0.05). Similarly, comparing shoot and root weight, a significant interaction (F8, 45, 4.22, p ≤ 0.05 and F8, 45, 3.28, p ≤ 0.05, respectively) occurred between genotypes (F2, 45, 14.5, p ≤ 0.05 and F2, 45, 29.0 p ≤ 0.05, respectively) and P treatment (F4, 45, 432, p ≤ 0.05 and F4, 45, 101.7, p ≤ 0.05, respectively). Two-way ANOVA determined no significant interaction (F8, 45, 0.879, p = 0.541) between genotype and P treatment on total root length, with differences between genotype (F2, 45, 12.1, p ≤ 0.05) and P treatment (F4, 45, 12.4, p ≤ 0.05). Total root length was largely consistent across all genotypes and treatments, except for SC283-14E and SSP, Haz, and MS treatments, which exceeded the SC648-14E and BTx623 No P and RP treatments.

Fig. 1: The effect of sorghum genotypes and P treatment (No P, rock phosphate RP, single superphosphate SSP, hazenite Haz, mineral solution MS).
Fig. 1: The effect of sorghum genotypes and P treatment (No P, rock phosphate RP, single superphosphate SSP, hazenite Haz, mineral solution MS).The alternative text for this image may have been generated using AI.
Full size image

Biomass and phosphorus content of sorghum genotypes SC283-14E, SC648-14E and BTx623 (n = 4): a dry weight biomass (g plant−1), b shoot and root dry weight (g plant-1), c total root length (cm plant−1), and d total P accumulated biomass (mg P plant−1). Data was analysed with two-way ANOVA and post-hoc Tukey test. Superscript letters represent significant (p ≤ 0.05) differences between treatments. Error bars represent standard deviation. See Supplementary Table 1 for a detailed statistical analysis.

Plant P content showed significant interactions (F8, 45, 3.07, p ≤ 0.05) between sorghum genotypes (F2, 45, 4.61, p ≤ 0.05) and treatments (F4, 45, 1430, p ≤ 0.05). The total P content was approximately four-times greater in the MS control and Haz treatments, consistent with the four-times higher P supply with MS compared to the other P treatments (20 and 5 mg P microcosm−1, respectively). No P control, RP, and SSP treatments had mostly similar P contents.

MS treatments had significantly more shoot biomass than the other treatments, except BTx623, which matched Haz. Root and shoot biomass of No P and RP treatments were consistently lower than all other treatments, and root biomass of SC283-14E grown with RP, SSP and MS was significantly (p ≤ 0.05) higher than BTx623 biomass.

Exudome—similarities and clustering analysis

Leachates were analysed with FT-ICR-MS to examine root exudate profiles (Fig. 2). A principal component analysis (PCA, Fig. 2a) and partial least squares discriminant analysis (PLS-DA, Fig. 2b) were conducted to identify P treatment clusters against plant responses denoted by the detected compounds. PC 1 and PC 2 (respectively) scores for the PCA and PLS-DA account for 11.2% and 7.1%, and 10.4% and 4.4%, respectively, of the total variance. The PCA illustrates that RP and MS treatments cluster separately from the other groups, while the remaining treatments have less distinct clusters. The clustering patterns in the PLS-DA reflect greater distinctions between treatments, and similar to the PCA, the MS treatment separates from the other treatments.

Fig. 2: Leachate analysis of sorghum root exudation profiles using FT-ICR-MS.
Fig. 2: Leachate analysis of sorghum root exudation profiles using FT-ICR-MS.The alternative text for this image may have been generated using AI.
Full size image

Exudome analysis of sorghum genotypes SC283-14E, SC648-14E and BTx623 with different P treatments: principle component analysis (a) and partial least squares discriminant analysis (b) with score plots illustrating the distribution of compounds between P treatments (dot = sample), c van Krevelen diagram depicting the distribution of all detected compounds (H/C versus O/C) from an FT-ICR-MS analysis; the size of the bubbles indicates the signal intensity, the colours represent different molecular compositions (blue = CHO, green = CHOS, orange = CHNO, and red CHNOS), and the donut chart represents the total and distribution of the compounds, and d hierarchal cluster analysis: the red and blue bars represent grouping and the similarities between exudate clusters; the top coloured rectangle overlaying the red and blue bars match the same colour scheme and treatment profile of the panels ‘a’ and ‘b’ (see colour key for treatment details); the bottom rectangles represent the different genotypes and include light grey = SC283-14E, grey = SC648-14E, and dark grey = BTx623.

The van Krevelen diagram (Fig. 2c) shows 5240 detected compounds with elemental composition, summarised in the donut chart, with the distribution centred on CHO and CHNO compounds. The cluster of compounds and compound families with high signal intensity between 1.25–2.0 H/C and 0.2–0.5 O/C indicate a diverse presence of compounds, including, but not limited to, low molecular weight organic acids, polyketides, polyphenols, flavonoids, terpenoids, lipids, peptides, amino acids, and their respective derivatives and additional functional groups (see Discussion).

A hierarchical cluster analysis (Fig. 1d) correlates and profiles sorghum genotypes and their respective treatments by their detected compound similarities. The MS treatment is organised under one tree cluster (blue, right), individualised by the respective genotypes SC283-14E (light grey), SC648-14E (grey) and BTx623 (dark grey). Similarly, RP falls within its own tree cluster (red, right), although the SC283-14E and BTx623 branches slightly delineate and include one repetition anomaly found in a different tree. SC648-14E and BTx623 genotypes in the Haz treatment fall into the same tree cluster (red, left), and SC283-14E into another tree (3 of 4 replicates, blue, left). The control nutrient (Control N, left, red) and water (Control W, left, blue) treatments were found between tree clusters.

Genotype-specific exudomes

The van Krevelen diagram for genotype SC283-14E (Fig. 3a) shows the sporadic presence of CHOS compounds with ratios between 1.5 and 1.6 H/C (likely sulphonated polyketides), CHNOS compounds in the ~1.5 H/C ratio region (terpenoids) and a dominant CHO compound could be annotated as an abscisic acid catabolite (i.e., phaseic acid). The signal intensity for phaseic acid was most prominent for the RP treatment, although still relatively high in the other treatments, comparatively. Notably, the SC648-14E and BTx623 genotypes also produced phaseic acid with the RP treatment (Fig. 3b, c), although only a fraction of the signal strength compared to the SC283-14E genotype.

Fig. 3: Van Krevelen diagrams depicting the distribution of defining detected compounds (H/C versus m/z, and H/C versus O/C).
Fig. 3: Van Krevelen diagrams depicting the distribution of defining detected compounds (H/C versus m/z, and H/C versus O/C).The alternative text for this image may have been generated using AI.
Full size image

The size of the bubbles indicates the signal intensity, the colours represent different molecular compositions (blue = CHO, green = CHOS, orange = CHNO, and red CHNOS), and the bar graphs illustrate the signal intensity of the defining detected compounds respective to other treatments. a SC283-14E, bar graph = phaseic acid only, b SC648-14E, bar graph = all detected compounds, and c BTx623, bar graph = all detected compounds.

For genotype SC648-14E (Fig. 3b), two primary compounds were annotated from a cluster of exclusively CHO compounds (likely polyphenols) in the H/C ratio range between 0.6 and 1.0 and on an O/C ratio of 0.3–0.5; these were catechin and 4-O-caffeoylshikimic acid. 4-O-caffeoylshikimic acid is an ester of caffeic acid, belonging to the hydroxycinnamic acid family, and shikimic acid, a cyclohexene-derived acid. The signal intensity of the total compounds detected suggests SC648-14E has a greater exudation profile of CHO compounds compared to SC283-14E and BTx623.

The BTx623 genotype is characterised by its greater presence of CHNO compounds (Fig. 3c). The CHNO compounds are likely represent a diverse mixture of amino acids and/or amino sugars (1.0–2.0 H/C and 0.4–0.8 O/C). The affiliate bar graph shows the signal strength was higher in the No P, RP and SSP treatments, and less so in the Haz and MS treatments.

Exudome responses to hazenite

The van Krevelen diagrams illustrate the differentiating compounds in exudomes of SC283-14E, SC648-14E and BTx623 (Fig. 4a–c) growth with Haz.

Fig. 4: Van Krevelen diagrams depicting the distribution of defining detected compounds (H/C versus m/z, and H/C versus O/C) grown with Haz.
Fig. 4: Van Krevelen diagrams depicting the distribution of defining detected compounds (H/C versus m/z, and H/C versus O/C) grown with Haz.The alternative text for this image may have been generated using AI.
Full size image

Three sorghum genotypes are shown: a SC283-14E, b SC648-14E, and c BTx623. Size of the bubbles indicates the signali ntensity, colours represent different molecular compositions (blue = CHO, green = CHOS, orange = CHNO, and red CHNOS). Dataextraction profile search at 0.85 threshold.

SC283-14E showed a relatively uniform but wide distribution of CHO and CHON compounds (Fig. 4a). For the CHO family, the H/C ratios ranging from 0.8 to 1.8, strict O/C ratio ending at 0.7, and m/z primarily ranging from 350 to 550, are likely a mixture of flavonoid dimers, tannins, phenolic acid dimers and/or their respective derivatives.

In contrast, SC648-14E shows a distinct shift towards lower m/z values (200–300, Fig. 4b). The broader O/C range suggests enhanced oxidation or oxygen-rich metabolite profiles, possibly including carboxylate or esterified species compared to SC283-14E. In accordance with the H/C and O/C ratios (0.5–1.5 and 0.4–0.6, respectively), the CHO cluster is likely a mixture of phenolic dimers and/or flavonoids. The CHNO compounds are likely N-containing species such as amide and/or peptide derivatives (some of which may represent conjugates or dimers).

BTx623 has a metabolic profile that is distinguished by a broader range of compound families, including CHOS and CHONS compounds. The CHONS family suggests a robust presence of sulphur and nitrogen conjugates, potentially indicating the presence of sulfonated peptides or amino acids, including a more complex metabolite profile involving multi-functionalised compounds and conjugates as discussed below.

Discussion

Total biomass, biomass allocation to roots and shoots, P accumulation and root exudomes confirm that the sorghum genotypes respond differently to P sources and provide evidence for our hypothesis that landraces are more P efficient than an inbred line. SC283-14E had growth and P accumulation traits that distinguished it from BTx623, and SC648-14E had an intermediate position. We could tentatively confirm the previous assessment of SC283-14E with a large root surface area conveying PUE, as this genotype exceeded the root weight of the other genotypes in some treatments. Perhaps unsurprising, SC648-14E had an intermediate performance because mycorrhizas conveying PUE were excluded from the experiment. Promisingly, root exudomes differed between all genotypes, including between the two landraces that have the Kafir genetic background.

Plants sense P availability and adjust root architecture36. In crops, including sorghum, root length and P uptake can be correlated37. A quantitative trait loci study analysing root architecture alleles found 58% of the enhanced PUE for genotype SC283-14E was due to its enlarged root surface area28. In the studied conditions, all sorghum genotypes had similar root length and root P concentrations. However, SC283-14E had higher root biomass in the RP, SSP and MS treatments than BTx623, which can be caused by greater root diameter, more lateral root branching, and more or longer root hairs to enhance uptake of P and other nutrients38. Higher root biomass without an increase in root length might indicate that the plant is allocating more resources to lateral root growth and/or increasing its root branching rather than expanding it further, which is a strategy to optimise P acquisition36,39. Interestingly, significant differences in root biomass and P concentrations manifested in the least and most bioavailable P treatments (RP and MS, respectively) when comparing SC283-14E and BTx623, supporting the notion that traits of commercial crops differ from landraces, having been selected in higher P environments typical of high-input agriculture.

SC283-14E exhibited significantly higher total P concentrations in the MS treatment, indicating a superior capacity for P accumulation compared to SC648-14E and BTx623 when provided with soluble P (but did not result in greater biomass production, except when compared to BTx623 in the SSP treatment). On the other hand, SC648-14E produced significantly more biomass in the MS treatment than BTx623, suggesting that SC648-14E may possess enhanced internal PUE. The significantly higher shoot biomass of this genotype, coupled with similar total P content in the no P treatment, further supports the notion of greater internal PUE. Together, these results underscore the different PUE traits of the studied genotypes, including external and internal PUE.

The analysis of compounds in background controls (i.e., microcosms without plants and water only, without plant with nutrient and water) annotated 2100 compounds (Supplementary Fig. 1) that the PCA nestled with the No P, SSP and Haz treatments, and centred within the hierarchical tree clusters. A greater number of compounds was detected in all other treatments (No P: 3400 compounds; RP: 3800 compounds; SSP: 3850 compounds; Haz: 2400 compounds), while the MS treatments had 1600 compounds. Thus, up to 1262 ± 582 detected compounds differentiate sorghum genotypes and their respective responses to P treatments. The number of detected compounds detected in our study matches previous research analysing exudates profiles of alfalfa, bean, barley, cucumber, ryegrass, maize, pumpkin and wheat using similar FT-ICR-MS parameters6,40. Lv et al.40 used FT-ICR-MS to compare rhizosphere and bulk soil samples and found a similar number of unique detected compounds but almost double the quantity of shared compounds (i.e., detected in both rhizosphere and bulk soil), suggesting microbes produce a similar quantity and compound profiles to maize, cucumber and ryegrass.

The PCA showed RP treatments clustering largely independent from the other treatments, with greater variation of plant responses denoted by the large confidence ellipses, and relatively heterogenous clustering. This suggests that RP activates a distinct and diverse response between sorghum genotypes. The overlap between No P and RP treatments may suggest similar plant responses with low bioavailability of RP, consistent with the significantly lower biomass and P content compared to all other treatments. This notion is supported by the low solubility and poor bioavailability of RP17,41. Both SSP and Haz treatments had less variability, which indicates closer similarities in detected compounds, possibly due to similar bioavailability with water, citrate and other carboxylates12.

The MS treatment had the lowest number of detected compounds and differed from other treatments in our analyses (PCA, PLS-DA, hierarchal tree clustering), indicating that plants had a unique response to the non-limiting supply of soluble P. This difference between MS and all other treatments may be explained by (i) plants exuding fewer compounds and/or at lower rates when adequately supplied with P as opposed to exuding compounds that may mobilise P and/or as a nutrient stress response, (ii) MS-supplied plants producing P-containing compounds which were not detected in our analysis, or (iii) RP, SSP, or Haz having chemical constituents that were detected in the analysis.

PLS-DA detected similarities between compounds specific to RP and SSP treatments, which may be explained by both substrates having different bioavailability, but also similarities with constituent elements such as calcium present in considerable quantities in both treatments (~37% and 20% Ca in RP and SSP, respectively)42. Other elements are specific to particular P sources, e.g., sulphur with 10% of SSP. Additional sulphur in the SSP and RP treatments may have increased S compound families detected within the SSP treatment and/or altered the plant’s exudation profiles in response to these secondary compounds.

Although previous research has characterised sorghum exudates and exudation profiles (Table 1), these studies did not cover the large range of biogeochemical compounds detected in our analysis. Therefore, we have broadened the search to exudates detected in other crops to identify key compounds and their potential function for P mobilisation. The compounds and their families annotated highlight the primary differences between genotypes and their responses to P treatments. These elemental families are a mixture of metabolites denoted by their elemental structures, some of which likely represent P acquisition mechanisms either directly liberation P from soil or indirectly through microbial relationships.

Table 1 Compilation of known sorghum exudates and their putative effects
Full size table

SC283-14E had phaseic acid (i.e., oxidised abscisic acid) as a dominant compound (Fig. 3). Abscisic acid in its carboxylate form increased P uptake in rice43, wheat and barley44. Phaseic acid also has been correlated with mitigating abiotic stressors, including drought in sunflower45 and heat in mandarin plants46. In the Haz treatment, SC283-14E (Fig. 4) showed conjugates of aromatic carbons and/or organic acids with benzoic acid and tartaric acid as possible compounds. Benzoic acid altered the community composition of soil bacteria47 and fungi48,49. Conifers upregulated exudation of tartaric acid under low P supply50, and tartaric acid correlates with increased microbial stimulation and root colonisation51. Thus, these compounds are candidates for P mobilisation, either directly or by stimulating soil microbes possibly adapted for acidic soils with low native soil P seen across Tanzania52,53. With SC283-14E categorised as PUE and showing the highest PUE of the tested genotypes, the next step of research could involve genome and transcriptome analysis to explore to what extent the acids annotated here contribute to PUE.

PUE of SC648-14E has been defined by its relationship with arbuscular mycorrhizal fungi, which correlates with enhanced P uptake under low P conditions29. This genotype had a uniform profile composed of flavonoids and/or polyphenols. Catechin, a flavonoid and polyphenol, can act as an allelochemical antimicrobial agent, a chelating compound supporting P liberation, and has autoinhibitory properties54. Inderjit et al.55 showed that catechin lowered the presence of microbial communities and reduced plant growth, illustrating its conditional effects. Possible polyphenols include 4-O-caffeoylshikimate (i.e., esterified chlorogenic acid), which increased in tomato exudates of plants inoculated with arbuscular mycorrhizal fungi56. Similarly, polyphenol coumaric acid was upregulated in the presence of arbuscular mycorrhizal fungi, which improved P uptake of maize57, and boosted some fungi and bacteria communities while inhibiting others58. Another detected putative polyphenol, caffeic acid, can serve as an inhibitor of bacteria59 and also mitigate abiotic stresses60.

A primary candidate for the largest CHO compound detected in the SC648-14E and Haz treatment may be a ferulic acid derivative. When ferulic acid undergoes conjugation or esterification, it can form various derivatives depending on the nature of the conjugation. Ferulic acid is a hydroxycinnamic acid exuded by roots with a wide range of allelopathic functions, including affecting fungi concentrations24,61, lowering bacterial diversity in cucumber62, and mitigating fungi disease in tobacco plants63. Thus, the detected flavonoids and polyphenols, including ferulic acid derivatives, may contribute to PUE via various means.

There was notable clustering of the CHON family with high saturation and low m/z ratios. These compounds likely reflect peptides or conjugated amino acids with hydroxyl or carboxyl groups and/or moderately saturated alkaloids with possible esterification or functional group attachments. These compounds represent a large option of potential candidates (Table 2), especially considering the extensive range of peptides, amino acids and their respective conjugates and derivatives. Potential compounds include leucylproline (i.e., leucyl, the free radical of leucine) and glutamine. Leucine occurs in higher concentrations in plants under P stress43 and can increase bacteria root colonisation64. Leucine and glutamine were produced in greater amounts by arbuscular mycorrhizal fungi in low P conditions65, and glutamine has been correlated with improved P and nutrient uptake66,67,68. Thus, there is evidence that the compounds annotated as characterising SC648-14E are involved in P mobilisation, including interactions with mycorrhizal fungi (which were not studied here).

Table 2 Analysis of different elemental families and specific compounds not previously characterised in sorghum
Full size table

The BTx623 genotype was largely differentiated from the other genotypes by a mixture of CHO (likely polyketides) and CHON (likely amino acids) elemental families across all P treatments, and notable CHOS and CHONS families in the Haz treatment. Within the Haz treatment, the high H/C ratios suggest the CHOS compounds are sulphur-containing fatty acid derivatives or thioesters with long alkyl chains and an attached functional group. Similarly, the detected CHONS compounds may be complex conjugates involving a sulphur-containing amino acid (such as cysteine or methionine) with an added functional group or conjugate. Cysteine and methionine are exuded by roots to provide sulphur to microbes69, and are compounds for unidirectional root-microbial nutrient cycling of nitrogen and sulphur70,71. Under non-limiting hydroponic conditions, BTx623 exuded higher concentrations of mono- and di-saccharides when compared to other sorghum varieties, which promoted a distinct microbial community composition72. While these sugars were not prominent in our analysis, Miller et al.73 demonstrated that soil type influences exudation profiles annotated as rhizosphere-conditioners and microbial recruitment compounds. This may suggest BTx623 exudation profiles are oriented toward rhizosphere conditioning and microbial interactions over direct P mobilisation.

Taken together, our study presents evidence that the sorghum genotypes have different exudate profiles. Many of these compounds have previously been annotated and are involved in plant-microbe interactions, nutrient mobilisation including P mobilisation, and exuded in response to stress. Subsequent research can build on these findings, connecting to crop breeders’ tools and evaluate contrasting sorghum lines in controlled and real-world environments to deepen causal understanding and harness root exudates for direct and indirect P mobilisation.

This study demonstrated that non-target root exudome analysis can identify distinct compounds produced by sorghum genotypes, illustrating that specific compounds may reflect each genotype’s unique adaptations to low P environments and contribute to P mobilisation, beneficial microbial relationships and stress mitigation. Greater biomass and/or P uptake of landraces compared to the inbred line studied here confirms that landraces have superior PUE, which is likely the result of having been selected in low P environments. Thus, there is an opportunity to enhance PUE through the selection of genotypes paired with P sources, including less soluble recyclates that are becoming available as nations transition from the current linear to a circular phosphorus economy. Hazenite emerges as an effective P source and candidate to substitute superphosphate fertiliser. Integrating powerful ‘omics approaches has much potential to advance PUE of crops and enable P-efficient cropping and the next-generation fertilisers generated in the circular P economy.