Tracing phosphorus from soil through mycorrhizal fungi to plants

A recent paper about carbon and phosphorus (P) transfer between arbuscular mycorrhizal (AM) fungi and plants (Lekberg et al., 2024) was cited by Spohn & Wanek (2025; pp. 443–445, in this issue) to highlight potential pitfalls of using P radioisotopes (in this case ³²P, although equally relevant for ³³P as they behave similarly; Frossard et al., 2011). Specifically, the paper by Spohn & Wanek (2025) states that without knowing the specific activity in soil solution (i.e. the ratio of ³²P : ³¹P), the conclusion in Lekberg et al. (2024) that more P was delivered by AM fungi in high-P than low-P soils may not be valid if the three low-P soils sorbed more ³²P. We agree with Spohn & Wanek (2025) that tracer experiments should be interpreted with caution, and we appreciate the opportunity to explore some of the pitfalls highlighted in that paper in more detail.

Sorption of inorganic orthophosphate (H₂PO₄⁻ and HPO₄²⁻) – the main P form taken up by plants, fungi, and prokaryotes from soil solution (Bucher, 2007) – is determined by pH, organic matter, hydrous oxides of aluminum and iron, and calcium carbonate (Frossard et al., 1995; Daly et al., 2001; Barrow, 2017). In Lekberg et al. (2024), soils were collected from two regions, one with high-P availability and one with low-P availability. In both regions, soils were classified as fine to gravely, loamy Mollisols, and neither soil pH (6.80 ± 0.38 vs 6.53 ± 0.38, means ± SE) nor organic matter concentrations (4.50 ± 0.69% and 5.03 ± 2.00%) differed (P > 0.5) between the high-P and low-P soils, respectively. Total organic P stocks also did not differ significantly between the high-P and low-P soils (1178 ± 302 and 684 ± 79 mg kg⁻¹) and the strong regional difference in P availability based on Bray 1 extractions were likely due to soil mineralogy.

There is another reason why differences in sorption likely did not influence results in Lekberg et al. (2024). One of the main assumptions of plant–soil–isotope experiments is that the concentration of the isotope-tracer must be negligible compared to that of the nonlabelled nutrient in solution. If the amount of ³²P added is similar to ³¹P, the solution and solid-phase P equilibrium is disrupted and net sorption of total and ³²P will occur. The solution concentration where no net sorption or desorption occurs in soil is known as the ‘equilibrium P concentration at net zero sorption’ (EPC₀). Specific to each soil, EPC₀ depends on the same variables that determine sorption, as well as P desorption kinetics, and microbial P immobilization and release. A recent meta-analysis by Simpson et al. (2021) illustrates that many soils have an EPC₀ of 0.1 mg l⁻¹, and the lowest value reported was 0.5 μg l⁻¹. The ³²P concentration used in Lekberg et al. (2024) was 3.2 pmol l⁻¹ or 0.1 ng l⁻¹; at least three orders of magnitude lower than the lowest EPC₀ ever reported in the literature. Net sorption of radioisotopes due to disrupted soil solution and solid-phase equilibrium is, therefore, unlikely to occur when using ‘carrier-free’ tracers as was done in Lekberg et al. (2024).

Even without net sorption of P, isotopic dilution will occur through time after tracer addition due to isotopic exchange of ³²P between the liquid and solid-phase with more ³²P being sorbed over time, accompanied by desorption of ³¹P, radioactive decay, and microbial immobilization and release (Helfenstein et al., 2018). Desorption of ³¹P was not specifically highlighted in Spohn & Wanek (2025), although ‘P release’ was mentioned, which also includes processes like P mineralization and solubilization. However, the implication of variation in P desorption is apparent in their fig. 1, which shows higher ³¹P concentration in the P-rich soil. Given the 10-fold greater background concentration of bioavailable ³¹P in the high-P soils in Lekberg et al. (2024), and because the same amount of ³²P was added to all pots, for every ³²P disintegration measured in plants growing in the low- and high-P soils, 10 times more ³¹P would have been taken up and delivered by AM fungi in the high-P than in the low-P soils. Thus, as discussed in Lekberg et al. (2024), the estimated higher AM fungal P delivery in the high-P soils based solely on the measured ³²P is, if anything, likely a conservative estimate.

While patterns observed in Lekberg et al. (2024) are substantiated due to the reasons outlined above, we agree with Spohn & Wanek (2025) that estimates of resource uptake would be strengthened by measurements of both ³¹P and ³²P in the bioavailable orthophosphate fraction through time, especially if different soils are used. However, proper execution of such measurements through time is challenging for several reasons. First, estimating specific activity of ³²P assumes even spatial distribution of ³²P, which is almost always violated given the low mobility of P and because radioisotopes are often added as a point source (including in Lekberg et al., 2024). Second, repeated sampling of soil after tracer addition involves some degree of disturbance that could affect soil structure, possibly promote mineralization, and disrupt fungal hyphae and roots. This would inevitably change conditions following the first sampling as a function of the amount of soil sampled and degree of disturbance. Also, there is some debate about how to best assess and quantify the plant- and/or mycorrhiza-available P pool (Bühler et al., 2003; Mason et al., 2013). Using extractants (e.g. Bray 1) as suggested by Spohn & Wanek (2025) and used in Lekberg et al. (2024), can generate artifacts. This last issue could be avoided by directly quantifying P in soil solution, but collecting and reliably analyzing inorganic orthophosphate in microliters of soil solution is extremely difficult given its low concentration and possible interferences due to soil colloids and dissolved organic P. To acquire specific ³²P activity in soil solutions of the different soils over time, an alternative option is to conduct parallel laboratory batch experiments and use isotope exchange kinetics (Frossard & Sinaj, 1997) although such an approach may not be fully representative of patterns in soils with plants and mycorrhizal fungi. Related to this, whether ³²P specific activity that changes over time in the soil solution reflects P acquisition by AM fungi at all time points (e.g. 1 min, 1 h, and 1 d after tracer addition, or during daytime vs nighttime) is uncertain given that so little is known about temporal and spatial acquisition patterns and mechanisms (but see attempts by Dessureault-Rompré et al., 2007), interactions with other soil biota that immobilize, mineralize, and solubilize P, and effects of water and other nutrient availabilities. Add to this the difficulty of measuring declining levels of ³²P radioactivity across longer timescales with confidence given radioactivity from naturally occurring isotopes such as ⁴⁰K, ¹⁴C and ³H. Given these challenges, perhaps it is not surprising that most models fail to accurately predict P acquisition in plants (Hinsinger et al., 2011), although different estimates of soil available P often correlate reasonably well with P uptake and each other (Steinfurth et al., 2021). Ultimately, we need to refine the assumptions and better understand the physical, chemical, and biological properties of soils that govern P availability to make radioisotope tracing of P acquisition from soil to AM fungi and plants truly quantitative. This is a difficult, albeit not impossible, task.

Alternative methodologies to radio-phosphorus tracers exist, and hydroxyapatite conjugated to quantum dots is increasingly being used to assess P transfer in AM (e.g. Whiteside et al., 2019; van't Padje et al., 2021). Quantum dots are appealing in many ways due to their greater longevity than radioactive P isotopes and the possibility of in vivo tracking on cellular levels (Färkkilä et al., 2021). Previous research has shown acquisition and delivery of quantum dots to plants by AM fungi and have elucidated fascinating patterns involved in plant–fungal interactions. However, relying on these approaches is not without challenges and uncertainties because the uptake and delivery mechanisms may differ from those that govern the uptake and delivery of orthophosphate. It is assumed that most P is acquired by AM fungi and delivered to plants via substrate-specific membrane transporters (Ezawa & Saito, 2018). Transported molecules are ionic and have a hydrated diameter of < 1 nm, whereas quantum dots coated with apatite are 8 nm diameter, which sometimes are combined into complexes reaching 200 nm in diameter (van't Padje et al., 2021). In this size range, quantum dots are likely taken up by endocytosis (Raven, 2022), which is a completely different mechanism compared to inorganic orthophosphate transporters. We consequently echo the suggestion in Raven (2022) that isotopes could be an excellent way to assess if and how quantum dots accurately reflect phosphate uptake and delivery by AM fungi.

With increasing realism in experiments comes increasing difficulty. As we continue to try to understand the functioning of mycorrhizal symbioses in nature, it is essential that we use appropriate methods and are aware of potential limitations. As such, the discussion generated by Spohn & Wanek (2025) is very timely.

None declared.

YL conceived of the idea and wrote the first draft, YL and BPC conducted the sorption tests, and BPC calculated sorption. YL, BPC, JJ, DJ, PM and CP edited the manuscript draft and provided critical input.