We used two pyrolysis biochars (hereafter: BGR, USB) to assess differences between biochar types. The BGR biochar was produced from forest harvesting residues (Pinus resinosa and P. banksiana) from the Upper Peninsula of Michigan, USA, and pyrolyzed in a rotary reactor system at 650°C for 30 min. The USB biochar was produced from waste wood pallets (southern yellow pine species), pyrolyzed in a continuous carbonizer at 550°C for 18 min. To eliminate mineral ash effects from fresh biochar, we soaked biochars with 0.1 M HCl for 72 h and thoroughly rinsed with water. After this weathering treatment, biochar pH was 7.26 ± 0.15 and 7.14 g ± 0.01 and dry densities were 0.19 and 0.30 g/cm³ for BGR and USB, respectively.

We implemented a factorial mesocosm experiment in a greenhouse, using three biochar treatments (BGR, USB and a No Biochar control) and four nutrient addition treatments, with five replicates per treatment (60 total mesocosms). We created biochar treatments by mixing each biochar type with a commercially washed and screened sand (3% silt) at 12% volume, which equates to a field application rate of 50 Mg ha⁻¹. We chose this high rate to help determine upper thresholds where carbon sequestration is maximized, and plant health is maintained. Biochar and sand mixtures were combined in a cement mixer to ensure even mixing and then steam-sterilized overnight.

We created mesocosms using 12.7 cm × 12.7 cm × 30.5 cm square pots that we divided into three compartments: two side “plant” compartments and a center “nutrient” compartment. The compartments were separated with plastic sheeting, into which we established a 10 cm × 10 cm window comprised of 50 μm nylon mesh (Elko Filtering Company) secured with silicon caulk. The mesh window allowed fungal hyphae to pass through but excluded plant roots. We filled each mesocosm compartment with the appropriate biochar + sand treatment (all compartments within a given pot received the same biochar + sand treatment). We then placed AMF whole inoculum (1.2 g per mesocosm, International Culture Collection of Vesicular Arbuscular Mycorrhizal Fungi, West Virginia University) in 7.5 cm deep holes in each plant compartment and filled the holes leaving a shallow indentation (~0.75 cm deep) for seeds. The whole inoculum included roots, spores, hyphae, and growth medium of AMF species Gigaspora rosea (120–150 spores per g), Rhizophagus clarus (220–250 spores per g), and R. irregularis (250–300 spores per g). Velvetleaf seeds (collected from MSU's Agriculture Research Farm, 42°42′38.2″N, 84°28′16.6″W) were planted on May 3rd in each shallow hole on top of the inoculum to ensure contact with seedling roots and covered with media. We planted additional seeds in compartments that had not germinated on May 10th and May 16th, with all germination occurring by May 20th. Then, we thinned seedlings to two plants per mesocosm, one in each plant compartment. Germination did not occur in some plant compartments (~11%), so we transplanted the extra seedlings from other mesocosms within the same biochar treatment. Mesocosm treatments were placed randomly on the greenhouse table and maintained under supplemental lighting (16 h:8 h L:D) at ~20°C, with daily watering through an automated sprinkler system. To ensure all plants had access to basic nutrient supply needed to establish initial growth, we fertilized mesocosms with 200 ml of 0.5X Hoagland solution (Hoagland and Arnon, 1950) twice a week from May 3, 2019 to July 23, 2019 (12 weeks). Because carbonates in the sand led to a media pH ~ 9, we acidified the Hoagland solution to pH 5.5 with HCl to create conditions more favorable to plant growth.

On July 23, 2019 we stopped regular Hoagland applications and initiated four nutrient addition treatments to test how N and P uptake by mycorrhizae influences velvetleaf competitiveness. Each nutrient treatment combined (a) a weekly application (100 mL) of a modified Hoagland's solution with soluble N and/or P, accessible to both AMF hyphae and plants, and (b) an insoluble, organic N and/or P substrate tube added to each mesocosm's central compartment, accessible to AMF hyphae but not plant roots. This was designed to ensure that plants could only utilize the organic substrates if AMF proliferated through the mesh windows into the central compartment, where they could then solubilize the organic nutrient sources (e.g., excreting extra-cellular enzymes or stimulating other microbes to do so) and then transport the resulting inorganic ions to the plant (Frey, 2019). By limiting nutrient acquisition to AMF only, we can better quantify symbiosis impact on plants. Treatments were: (1) Soluble NP (Sol NP), (2) Insoluble N and Soluble P (Insol N + Sol P), (3) Insoluble P and Soluble N (Insol P + Sol N), and (4) Insoluble NP (Insol NP) (Figure 1).

To create the organic substrate tubes, we constructed 50 μm fine nylon mesh windows (Elko Filtering Company) on 50 mL centrifuge tubes (polypropylene, Fisher Scientific). We then used a blender to homogenize organic N (gelatin, 3 g per tube, 0.06 g/ml) and/or P (calcium phytate, 0.3 g per tube, 0.006 g/ml) with biochar + sand media and filled the tubes with these mixtures. Organic substrate concentrations were calculated to equal the total amount of soluble N or P applied via modified Hoagland solutions over the course of the experiment. We deployed the substrate tubes (top of tube flush with soil surface) in the center mesocosm compartment. We modified Hoagland solutions by excluding soluble N and/or P, depending on treatment.

After 11 weeks of nutrient treatments, we harvested the plants and separated shoots from roots at the root collar. We collected media samples from each mesocosm by shaking the residual biochar + sand media from the roots of both mesocosm plants into 50 mL tubes. We then clipped five 1.0 cm samples of fine roots from each of the two plants in the mesocosm and stored at 4°C. Plant samples were rinsed with water, dried at 60°C for 48 h and weighed before grinding (Wiley mill, 1.0 mm mesh screen).

Ground plant samples were analyzed for total P concentration by ashing samples at 500°C for 5 h and then digesting with 3N nitric acid. Samples were then diluted (1:9 with 0.3N sodium hydroxide) before analysis via the ascorbic acid method (John, 1970). We pulverized ground plant tissues using a roller-mill and determined total C and N concentrations (Costech ECS 4010 CN analyzer, Valencia, CA USA). We calculated above- and belowground biomass values by weighing both mesocosm plants and dividing by two to get average biomass per mesocosm. We then added above- and belowground biomass values to get total net primary production (NPP).

We measured AMF colonization of plant roots to interpret AMF benefits on nutrient uptake and plant response (Treseder, 2013). Although colonization rates can vary across AMF species (Treseder, 2013), our plants were inoculated with the same three species. We rehydrated roots 24 h before staining and colonization counting procedures (Phillips and Hayman, 1970; McGonigle et al., 1990). Root tissues were cleared with 10% KOH, stained with 0.05% Trypan Blue, de-stained with 5% acetic acid, and mounted on microscope slides. Each slide contained five 1.0 cm roots from each mesocosm. We examined slides under a compound microscope at 400x magnification, with 20 fields-of-view per root. For each field-of-view, we scored presence or absence of mycorrhizal hyphae, vesicles, and/or arbuscules as colonized or not colonized, respectively. We calculated colonization percentage as the total number of colonized views divided by total field-of-views multiplied by 100. We did not differentiate between AMF species and assumed that steam sterilization removed most all other mycorrhizae besides those in our inoculum (Brito et al., 2009).

We also measured extraradical hyphal length density (ERH) in organic substrate tubes as a proxy for AMF activity (Jakobsen et al., 1992; Staddon et al., 1999). We rinsed each organic substrate tube contents with H₂O and decanted through a sieve stack (top: 500 μm, bottom: 212 μm). We then stained residues on the 212 μm sieve with 0.05% Trypan blue stain (lacto-glycerol) and incubated for 30 min. The stained residue was rinsed and mixed with 200 ml of H₂O, and 20 mL was collected and vacuum-filtered (0.45 μm nylon filter). Two filters for each mesocosm were mounted on microscope slides and scored with 25 random fields-of-view at 10x objective magnification, 100x total magnification. For each field-of-view, we counted the number of times an AMF hypha crossed any gridline present in the reticle (1 cm per side, 10 × 10). We then used the average hyphal intersection counts per filter to calculate average hyphal densities (calculations in Supplementary Material).

We extracted inorganic N in media from each mesocosm using 25 mL of 0.5 M K₂SO₄ and 5 g media, and measured nitrate (NO3-) via an enzyme reduction method (Patton and Kryskalla, 2011) and a microplate protocol for ammonium (NH4+) (Sinsabaugh et al., 2000) (Biotek synergy H1, Winooski, VT, USA). We extracted phosphate (PO4-) using 40 mL of 0.5 M NaHCO₃ and 8 g of media, and analyzed concentrations using the molybdenum blue-ascorbic acid protocol (John, 1970) adapted for a microplate assay (Song et al., 2019). We oven-dried media subsamples at 105°C for 48 h to calculate gravimetric soil moisture and reported nutrient concentrations as a dry mass basis (μg g⁻¹). We measured media pH as 1:2 w:v in H₂O.

For all response variables, we performed fully interactive general linear models (multiple two-way analysis of variance) with biochar type and nutrient treatment as explanatory variables. Because of the unbalanced design (due to plant mortality, etc.), we used type III sum of squares. Statistical significance was determined when p < 0.05. We performed pairwise comparisons using Tukey's tests, adjusted for multiple comparisons. For AMF colonization comparisons, we used a beta-distributed (response bounded by 0 and 1) mixed-effects linear model with a random effect of mesocosm ID to account for variation between the root replicates collected per mesocosm. For ERH comparisons, we used a normally-distributed mixed-effects linear model with a random effect of mesocosm ID to account for the variation between the two filters processed per mesocosm. We checked model assumptions visually by plotting residuals and predicted values and log-transformed any response variables that did not conform to normality assumptions. All analyses were performed using R statistical software version 3.5.3.

Belowground biomass was significantly affected by nutrient treatment (ANOVA, p = 1.07 × 10⁻⁴). For treatments without biochar, belowground biomass in Insol NP treatments was 63% higher than Sol NP treatments (p = 4.48 × 10⁻³), 29% higher than Insol N + Sol P treatments (p ≥ 0.05), and 69% higher than Insol P + Sol N treatments (p = 2.35 × 10⁻³) (Figure 2B). However, this relationship was not significant across biochar treatments (all p ≥ 0.05). Additionally, there were no significant effects of nutrient treatment, biochar, or the nutrient × biochar interaction on aboveground biomass or total NPP (Figures 2A,C).

Nutrient treatment also significantly affected root:shoot biomass ratio (ANOVA, p = 7.85 × 10⁻⁷) with similar trends to belowground biomass. For example, without biochar, root:shoot ratios were the highest in Insol NP treatments (Figure 2D). Root:shoot ratios in Insol NP treatments were 60% larger than ratios in Sol NP (p = 1.43 × 10⁻⁴), 38% larger than ratios in Insol N + Sol P (p = 0.04), and 61% larger than ratios in Insol P + Sol N (p = 1.24 × 10⁻⁴). Additionally, compared to No Biochar treatments, BGR and USB decreased the root:shoot ratio in Insol NP treatments by 42 and 27% respectively (p = 1.66 × 10⁻³, 0.05), but had no significant effects in other nutrient treatments (all p ≥ 0.05).

Nutrient treatment significantly affected aboveground plant tissue N (ANOVA, p = 0.04); however, post-hoc testing did not reveal statistically significant differences between treatments (all p ≥ 0.05). Nutrient treatments had significant effects on belowground plant tissue N (ANOVA, p = 1.22 × 10⁻³) that were similar to aboveground N tissue, but did not vary across biochar treatments (interaction, p ≥ 0.05). Therefore, regardless of biochar treatment, tissue N concentrations were the highest in Insol N + Sol P treatments compared to all other nutrient treatments in both above- and belowground tissue (Figures 3A, 3C); although, this relationship was only significant for BGR treatments in belowground tissue. In BGR, belowground N tissue in Insol N + Sol P treatments was 37% higher than Sol NP (p = 0.07) and 52% higher than both Insol P + Sol N (p = 1.99 × 10⁻³) and Insol NP treatments (p = 0.05).

Biochar, nutrient, and biochar × nutrient interaction (ANOVA, p = 4.21 × 10⁻⁸, 6.7 × 10⁻⁴, 9.26 × 10⁻⁷) all significantly affected belowground tissue P concentrations. Overall, BGR biochar decreased belowground tissue P compared to No Biochar treatments (Figure 3D). For example, BGR significantly decreased belowground plant tissue P by 38% for both Sol NP treatments (p = 9.71 × 10⁻⁶) and Insol NP (p = 6.6 × 10⁻⁴). Additionally, plant tissue P was 31% lower in BGR compared to USB Sol NP treatments (p = 5.85 × 10⁻³). For USB, belowground plant tissue P was 33% lower in Insol P + Sol N treatments when compared to No Biochar (p = 1.51 × 10⁻³), although USB Sol NP treatments did not significantly differ from No Biochar Sol NP treatments (11% difference, p ≥ 0.05). Nutrient treatments had different effects on belowground tissue P within biochar types (Figure 3D). Within BGR, belowground plant tissue P was 53% higher in Insol N + Sol P compared to Sol NP treatments (p = 3.43 × 10⁻⁴), however this relationship was reversed in No Biochar treatments (p ≥ 0.05). In USB biochar, plant tissue P was generally lower in Insol P + Sol N treatments with 32% less than Sol NP (p = 2.41 × 10⁻³) and 37% less than Insol NP (p = 6.86 × 10⁻⁵). Aboveground tissue P concentrations were significantly affected by biochar × nutrient interaction (ANOVA, p = 0.03), however post-hoc tests did not reveal differences between treatments (all p ≥ 0.05, Figure 3B).

Nutrient treatment significantly affected AMF root colonization (ANOVA, p = 7.31 × 10⁻⁵) with rates generally highest in Insol NP or Insol N + Sol P treatments and lowest in Insol P + Sol N treatments (Figure 4A). In No Biochar treatments, root colonization was on average 161% higher in Insol NP treatments than Insol P + Sol N (p = 1.51 × 10⁻³). There were no significant effects of any treatment on AMF ERH (Figure 4B).

Inorganic N and P media concentrations were low (NO3- and NH4+ < 4 μg g⁻¹; PO4- < 1 μg g⁻¹) but were generally higher in treatments with soluble N and soluble P, for NO3- and PO4- respectively (Supplementary Figures 1, 2). Despite this, none of our treatments significantly affected media NO3-, NH4+, total inorganic N (NO3- + NH4+) nor PO4- (Supplementary Table 1).