Biomass partitioning of plants under soil pollution stress

Pollution gradients, soil series, and glasshouse conditions for the empirical study

Soils used for this experiment were collected from a wood preservation site (6 ha). In this site, the use of creosote, and various Cu-based salts has resulted in soil Cu-contamination over the whole site and large patches of PAH in smaller areas⁵⁹. Former studies have shown the ecotoxic impact of this contamination on vegetation biomass, cover, richness and diversity⁶⁰ and the reduction of soil enzymatic activity⁶¹. In February 2016, 65 kg of soil were collected from two areas of the site, one known for its contamination in Cu, and the other previously identified for its contamination in PAH⁵⁹. An additional control soil was collected from the grassland next to the site. The control corresponds to an alluvial sandy soil (Fluviosol – Eutric Gleysols, World Reference Base for soil resources) and the contaminated soils were developed on this Fluviosol. Soils were transferred to a glasshouse nearby and spread out thinly on a tarpaulin for 15 days to ensure complete air drying. Ten samples of each soil were analysed for their PAH, Cu, C, N, and P concentrations. N concentration was higher in the control soils, while polluted soils showed higher P-availability (Supplementary Table 1). Regarding PAH, the 16 regulatory PAH were quantified. The range of soil properties and contamination values (889 ± 10 mg Cu.kg_soil⁻¹ and 657 ± 331 mg PAH.kg_soil⁻¹ for the first contaminated soil, 4276 ± 209 mg Cu.kg⁻¹ and 3142 ± 419 mg PAH.kg_soil⁻¹ for the second contaminated soil) showed Cu and PAH contamination in both cases, with higher contamination of the second soil. In this study, these soils are referred to as Cu-PAH soil and HIGH-Cu-PAH soil, respectively.

To create the soil series, both soils were mixed by combining one third and two-thirds of air-dried contaminated soils with the control soil (March 2016) giving seven soil treatments: Control, 1/3 Cu-PAH, 2/3 Cu-PAH, Cu-PAH, 1/3 HIGH-Cu-PAH, 2/3 HIGH-Cu-PAH, and HIGH-Cu-PAH. Each of these seven soils was divided into 25 pots (10 × 10 × 15 cm) containing 800 mg of soil, giving a total of 175 pots. In order to inoculate all potted soils with similar micro-organism populations (especially Rhizobium populations), 1 g of control soil was added to the pots with undiluted polluted soils and vice versa. All pots were watered and weighed to determine their water holding capacity and left for 2 weeks to enable micro-organisms populations to react.

To ensure that the environment was as homogeneous as possible during the whole experiment, a whitened glasshouse was used to favour diffuse and homogeneous solar radiation, and to limit differences in temperature and Vapour Pressure Deficit (VPD). In the case of a temperature increase above 25 °C, the glasshouse was also cooled by automatic ventilation and misting was used to avoid an increase of VPD above 1 kPa. An air temperature and humidity probe (U23 Prov V2 ®Hobo) was used to monitor VPD variations (kPa) during the whole course of the experiment.

Plant cultivation and monitoring of plant development

The dwarf bean (Phaseolus vulgaris, cv. Oxinel, ® Vilmorin) was chosen as a model plant species because of its known plasticity of biomass allocation, both for wild and selected genotypes²⁹. This plasticity has been detected in response to soil resources²⁹ and also light regimes⁶². In addition, it is a species commonly used as bio-assays in ecotoxicology due to its sensitivity to soil pollution (see ref. ²⁶ using soils from the same site as this study). Seeds of similar weight [0.22; 0.30 g] were selected to avoid large differences in seed reserves. After soaking for 4 h in tap water, three seeds were sown in each of the pots on March 21. Germination took 11–16.2 days depending on the soil treatment and this time increased with soil contamination. As a large majority of the seeds germinated, 1 seedling per pot was selected randomly and kept for the experiment. For each soil, we planned to harvest five plants at five different development stages (stage 1: end of cotyledon opening, stage 2: first trifoliate leaf, 3: second trifoliate leaf, 4: 3–4 trifoliate leaves, 5: 5–6 trifoliate leaves) giving 25 plants for each soil. Pots were watered every 2 or 3 days and weighed to maintain the water holding capacity (WHC) of soil at 60%⁶³. Plants were harvested for analysis when they reached the desired development stage. In most phytotoxic soils, plants did not reach the fourth or fifth stage by the end of the experiment, thus they were harvested and classified into their real development stage at harvest (see for instance Table 1 shows that most plants of the High-Cu-PAH soil did not grow and were classified in development stage 1).

Biomass partitioning, root and shoot (specific) areas

On the day of harvest, plant parts were separated (stem, leaves, and roots). Roots were washed gently with water and nodule numbers were counted. All organs were scanned and analysed to determine their area (software Winfolia for leaves and stems, WinRhizo for roots, Regents Instruments, Quebec, Canada). Then all plant samples were dried and weighed. The whole process determined the dry biomass of plant parts, their area, as well as Specific Leaf Area (SLA, cm².g⁻¹) and Specific Root Area (SRA, cm².g⁻¹). Analysis of SLA and SRA is important because: (i) they may also be involved in plant response along resource gradients to maintain a functional equilibrium. For instance, SLA can increase strength in the shade to maintain light capture area¹³, and SRA can increase to maintain water uptake during water stress⁶⁴; and (ii) they may be impacted by soil pollution. A decrease in SRA is part of the root syndrome in phytotoxic soils because of decreasing root elongation and root thickening⁴³.

Indicators of resource acquisition

To estimate light capture and potential acquisition of photo-chemical energy, we assessed chlorophyll a, b and another carotenoid synthesis by determining their leaf concentrations (See Supplementary Information for more details regarding corresponding methods).

Water uptake and transpiration: To limit water evaporation, the soil in each pot was covered with a small plastic sheet (10 × 10 cm). At each watering, the mass of water added to maintain the pot at 60% of SWHC was recorded as the amount of water taken up and transpired since the last watering. The last 10 days before harvest were considered for analysis of plant transpiration. Independently of soil treatments, the amount of water transpired could be impacted by the leaf area (and the number of stomata), and by the variation of VPD occurring in the glasshouse despite cooling and water misting. Therefore, the weight of water transpired per leaf area, per day and per kPa of VPD (({{{{{{rm{mg}}}}}}}_{{{{{{{rm{H}}}}}}}_{2}0}).cm⁻².day⁻¹. kPa_VPD⁻¹) was calculated.

Nitrogen acquisition and Symbiotic Nitrogen Fixation (SNF): To estimate N acquisition by plants, their leaf N concentration and an indicator of their SNF were determined. After drying and grinding (Retsch PM4 planetary grinder, Retsch, Haan, Germany), leaf N concentration was measured by an elemental analyser (NA 1500 NCS, Carlo Erba, Milan, Italy) for a subset of 112 samples encompassing all soil treatments and a wide range of plant size. Regarding SNF, most plants in this experiment did not initiate nodulation (because of toxicity and their small size). It was planned to use ¹⁵N soil labelling and the isotopic dilution method to estimate the efficiency of the SNF⁶⁵, but it was not applicable in our experiment because of the low nodulation and the small amount of N derived from the atmosphere. Instead, at harvest, roots were cleaned gently and the number of nodules was determined.

Statistics and reproducibility

All pots were placed randomly in the glasshouse at the beginning of the experiment. They were moved randomly every 15 days to avoid any spatial dependency between sample units. All statistical analyses were performed with R software (R Core Team, 2016). Regarding the biomass of plant parts, stems and leaves were considered together in a single shoot compartment when analysing the results. Bean stems are also photosynthetically active, and separate analyses for leaves gave consistent results.

All measured plant traits (SRA, SLA, water transpiration, nodule number, leaf N and chlorophyll concentration) can vary with plant ontogeny and plant size. Thus, variation of these traits (dependent variables) was analysed considering both plant size and soil treatments (explanatory variables). Shoot biomass was used as a surrogate for size for aboveground traits (SLA, water transpiration, leaf N, and chlorophyll concentration). Root biomass was used for belowground traits (root nodule number). Water transpiration was analysed by ANCOVA (shoot biomass as a regressor, soil treatment as a covariate). Leaf nitrogen and chlorophyll concentrations were first analysed by segmented linear modelling (segmented package) because of a radical change in the relationship between shoot biomass and these leaf traits at some size threshold. Then soil treatment effect on these traits was analysed on the residuals of the segmented relationships by ANOVA. Note that similar responses were observed for the different kinds of pigments, and only the results for chlorophyll_a+b concentrations are reported in this study. Similarly, we used a segmented linear model (segmented package) for the relationship between root nodule number and root biomass, and the soil treatment effect was analysed on this first model residuals. As to SLA and SRA, plants were grouped according to their shoot and root biomass tertile respectively. Then for each tertile, ANOVA was used to test the difference between SLA and SRA with soil treatment. When performing ANCOVAs, in the case of significant effects of soil treatment and interaction with plant size, post-hoc pairwise comparisons were used to test the difference of intercept or slopes between soil treatments (emmeans package). When performing ANOVAs, Post-hoc Tukey HSD pairwise comparisons were performed in case of significant effect of soil treatment. All variables were log-transformed when necessary to respect the condition of application of linear modelling.

When investigating allometric relationships between root and shoot biomass, or root and shoot areas, the interest is related to the analysis of how root biomass (or area) scales against shoot biomass (or area), rather than predicting the value of one variable from another. Standard Major Axis (SMA) regression (smatr package) on log-transformed variables was used accordingly to study this allometric scaling and its changes with soil treatments⁶⁶. When changes in α scaling exponent with soil treatment are significant, estimation of differences in β (proportionality coefficient) between treatments is not enabled by SMA regression⁶⁶. In that case, after estimating α with SMA regression, we estimated β value for each soil treatment using non-linear least square modelling (see Supplementary Table 2) because changes in β values have a biological meaning in our context (delay in early root development).

Meta-analysis of literature dealing with plant biomass partitioning and modelling of changes in root: shoot ratio

Collection of published studies and case studies: we used the ISI Web of Science database to locate published studies on the effect of soil pollution on plant biomass partitioning. We entered a general query made using the combination of two phrases, one regarding biomass partitioning, and the other regarding soil pollution. We used several equivalent phrases regarding both terms, leading to the following query: (“biomass partitioning” OR “biomass partition” OR “biomass allocation” OR “root: shoot”) AND (“pollution” OR “contamination” OR “heavy metals” OR “PAH” OR “phytoremediation” OR “phytomanagement”). Some additional studies were picked out from the reference list found in the studies collected from our query. From the first selection of 53 potential studies (from their title and summary), the final collection made after careful reading of the entire studies comprised only 15 references (Table 2). From these studies, we identified 25 case studies suitable for the meta-analysis that was conducted in a large variety of geographical locations and climates. Studies and case studies were excluded from our database when M_R: M_S could not be calculated, when they dealt with air pollution (not our subject), when no phytotoxic effects were shown (no decrease in plant growth), when no statistical analyses or tests had been done for the reported results regarding M_R: M_S and root and shoot parts. When plant growth was reported both in hydroponic and for growth in soil substrates, we assumed that results from soil substrates where more suitable for analysing the biomass partitioning response. When other treatments were used (for instance mycorrhizae inoculation), we averaged the response to these treatments at each level of soil pollution. Finally, we considered one case study as being the unique combination of one team of researchers, one studied plant species, and one contaminant at stake. In one study, we made an exception and considered two case studies for two populations of the same plant species being exposed to the same contaminant, the two populations being reported as being metallicolous and non-metallicolous and which showed contrasting responses.

The statistics and information recorded: we aimed to answer three questions: Is there any general pattern (increase or decrease) of the M_R: M_S ratio reported in the literature? Do changes in M_R: M_S depend on some explicative factors (for instance the contaminant type)? and Do M_R: M_S variations depend on pollution effect on plant size? This last question is important in our study which aimed to distinguish changes due to simple allometric effects rather than plant response. For each case study, the main indicator of biomass partitioning available was the M_R: M_S ratio, either provided directly, or calculated from root and shoot biomasses. Total plant dry biomasses were also recorded. Then, we calculated two statistics to enable the comparison of studies that were not originally designed to be compared. Firstly, we calculated the effect size metric (referred to as relative response in this study) to estimate the effect of pollution on the M_R: M_S ratio as follows:

Values close to 0 are associated with a negligible effect of the treatment, while negative and positive values indicate negative and positive effects of the treatment, respectively. The relative response is a reliable approach to quantify the effects of treatment compared to control and is regularly used in plant science (e.g. ref. ⁶⁷). Secondly, we calculated a phytotoxic effect by normalisation of the effect of pollution on plant growth as follows:

Values close to 0 are associated with a negligible effect on plant growth, while values close to 1 indicate a strong decrease in plant growth. Additionally, we report relevant information to analyse its potential influence on M_R: M_S results. We reported the plant species involved, its functional group (monocotyledonous grass, dicotyledonous forb, and woody species), its life cycle (annual, perennial), the experiment duration (and if several measurements were made at different times) and the kind of contaminant at stake (see Table 2).

The meta-analysis: regarding the general pattern of M_R: M_S changes_, they were classified on the basis of the statistical results reported in the different studies as follows: (i) stable: no change of the M_R: M_S value was reported; (ii) variable: increase or decrease of the M_R: M_S ratio was reported for a pollution treatment compared to the control, and other treatments with a higher level of pollution showed no or opposite effects; (iii) increase: increase of the M_R: M_S ratio was reported for pollution treatment, and other treatments with higher levels of pollution also showed an increase compared to the control; (iv): decrease: an opposite situation to increase described above. Additionally, we tested the effect of the contaminant type, plant functional type, and plant life cycle on the relative M_R: M_S by ANOVA. Finally, we tested the dependence of the relative M_R: M_S with biomass loss for case studies showing an increase or decrease in this ratio by using linear modelling. This was done to compare results among case studies (one averaged value per case study). For results within a case study (when several soil pollution levels were available per case study), relative M_R: M_S and biomass loss compared to the control was calculated for each pollution level. Rate of relative M_R: M_S changes (Δ M_R: M_S /Δ biomass loss) was calculated and compared to 0.

Modelling of changes in root: shoot ratio with exposure to pollution stress: we modelled the changes in the root: shoot ratio compared to a control situation without exposure to excess contaminants. We considered the change of allometric relationships (Eq. 1.) between root and shoot parts by the three potential drivers related to soil effect and plant response, and we followed the following steps. First, we calculated the growth of shoot parts as follows:

gr represents plant growth rate (it can concern shoot biomass or area) per day; d is the duration of the growing period (in days); gr_decrease is the phytotoxic effect on plant growth (interval [0,0.8] is considered here); S is the number of shoot parts produced after the corresponding duration d.

Second, we calculated corresponding root parts as follows

With β and α the parameters of the allometric relationship of a given plant species in a control soil; β_decrease (the interval [0;0.5] is considered) is the effect of pollution stress on the early root development; α_increase (the interval [0;0.5] is considered) corresponds to plant response with increasing biomass partitioning in favour of roots; and R is the number of root parts produced.

Finally, changes in root: shoot ratios were calculated by dividing the root: shoot ratio obtained on polluted soils by the root: shoot ratio obtained in a control situation (gr_decrease; β_decrease; and α_increase set to 0).

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Source: Ecology - nature.com