|Main authors:||Giulia Bongiorno, Barbara Thürig, Else K. Bünemann, Paul Mäder, Lijbert Brussaard, Ron de Goede, Lucius Tamm, Joeke Postma|
|iSQAPERiS editor:||Jane Brandt|
|Source document:||Bongiorno G., Bünemann E., de Goede R., Brussaard L., Mäder P. (2018) Screening of novel soil quality indicators. iSQAPER Project Deliverable 3.4, 66 pp|
Note: The following publication is based on the material contained in this section of iSQAPERiS
- Giulia Bongiorno, Joeke Postma, Else K. Bünemann, Lijbert Brussaard, Ron G.M. de Goede, Paul Mäder, Lucius Tamm, Barbara Thuerig. 2019. Soil suppressiveness to Pythium ultimum in ten European long-term field experiments and its relation with soil parameters. Soil Biology and Biochemistry 133, 174-187. https://www.sciencedirect.com/science/article/pii/S0038071719300732
The capacity of soils to regulate soil borne plant pathogens is an essential element of soil quality. In this article we assess the level of natural disease presence and general soil suppressiveness in 10 European Long term field experiments (LTEs), and understand if the disease suppressiveness is due to biotic origins; test if general soil suppressiveness is affected by long term tillage and fertilization in these LTEs; assess the relationship between soil suppressiveness and chemical, physical and biological soil quality parameters and determine which are the most important parameters that explain soil suppressiveness, and understand the direct and indirect relation between soil suppressiveness and labile organic carbon fractions.
|2. Materials and methods|
The capacity of soils to regulate soil borne plant pathogens is an essential element of soil quality (Bünemann et al., 2018). This capacity of the soil is called soil suppressiveness and it has been related to chemical, physical and biological soil parameters (Janvier et al., 2007). Previous investigations have showed strong evidences that biological, and in particular microbiological, properties have a crucial role in determining disease suppressiveness (Bonanomi et al., 2010; Bonilla et al., 2012). Soil suppressiveness as a whole is probably a combined effect of general and specific soil suppressiveness (Postma et al., 2008; Yadav et al., 2015). General disease suppressiveness relates to activity, biomass and diversity of soil organisms and is based on the collective capacity of soil and plant microbiome to compete and be antagonistic to pathogens. General disease suppressiveness is often effective against a broad range of pathogens, it is not transferable, is eliminated upon soil sterilization and can be enhanced by agricultural practices that have a positive effect on soil microbial communities, activities and abundance (Schlatter et al., 2017). Specific disease suppressiveness is the results of the presence of specific microorganisms antagonist to pathogens and, contrary to general disease suppressiveness, it can be transferred to a disease conducive soil (Schlatter et al., 2017). General disease suppressiveness is considered to be more persistent than specific suppressiveness, and this is due to the diversity of the pathogen-antagonistic population that can act complementarily against pathogens. Soil suppressiveness is difficult to measure due to the specificity of pathogens present in different areas and the incomplete understanding of the mechanisms behind this phenomenon.
Several studies have shown that agricultural management can affect soil suppressiveness because it affects soil physical, chemical and biological properties (Sánchez-Moreno and Ferris, 2007). Examples are compost addition (Alfano et al., 2011; Boehm et al., 1993; Chen and Nelson, 2008; Lumsden et al., 1987; Pane et al., 2011; Pascual et al., 2002; van Os and van Ginkel, 2001), manure addition or organic amendments (Darby et al., 2006; Stone et al., 2003), tillage (Pankhurst et al., 1995), residue retention (Medvecky et al., 2007), and crop rotation (Manici et al., 2005). Intensive agricultural management has the potential to decrease soil biodiversity, including natural enemies and competitors of pathogens, and subsequently decrease soil suppressiveness (van Elsas et al., 2002). However, many studies focused on the effect of short term compost additions (after days, weeks or months) and few investigated how tillage and organic matter addition practices shape general soil suppressiveness in the long term (after years) (She et al., 2017; Tamm et al., 2010). This knowledge could contribute to creating a more sustainable disease control in agricultural settings and it should be a research priority.
One of the biggest challenges of soil quality assessments in agricultural lands is the assessment of soil parameters which can be used as indicators to measure soil suppressiveness (Bonilla et al., 2012). Many studies tried to find relationships between soil suppressiveness and soil parameters (Darby et al., 2006; Höper and Alabouvette, 1996; Postma et al., 2008). However, contrasting correlations between soil suppressiveness and various soil chemical, physical and biological parameters have been observed depending on the pathogen and the system studied (Höper and Alabouvette, 1996; Janvier et al., 2007). For example, the quality of the organic matter seems to play an important role in disease suppression (Stone et al., 2001). Recently, disease suppression has been associated to labile carbon, and this is particularly interesting taking into account the value of this soil parameter in soil quality assessments (Darby et al., 2006).
The objectives of the current study were to
- assess the level of natural disease presence and general soil suppressiveness in 10 European Long term field experiments (LTEs), and understand if the disease suppressiveness is due to biotic origins;
- test if general soil suppressiveness is affected by long term tillage and fertilization in these LTEs;
- assess the relationship between soil suppressiveness and chemical, physical and biological soil quality parameters and determine which are the most important parameters that explain soil suppressiveness, and
- understand the direct and indirect relation between soil suppressiveness and labile organic carbon fractions.
We hypothesised that disease suppressiveness in the ten LTEs was due to biotic origin, that the more sustainable management (reduced tillage and organic fertilization) would increase soil suppressiveness and that the most important parameters in explaining soil suppressiveness will be biological parameters related to microbial biomass and activity. We used the pathosystem Pythium ultimum-Lepidium sativum as a model for general disease suppressiveness. Pythium ultimum is a necrotrophic-saprophytic oomycete and several studies found that general disease suppressiveness is an important mechanism active against this pathogen (Cook and Baker, 1983; Pane et al., 2011; Postma et al., 2000). In fact, Pythium spp. are r-strategist organisms but they are poor competitors relative to other root-colonizing and saprophytic organisms, and as such they are very sensitive to microbial activity (Chen et al., 1988; Postma et al., 2000). Moreover, Pythium ultimum has been used as a model pathogen (or indicator) for general disease suppressiveness in previous studies (Craft and Nelson, 1996; Löbmann et al., 2016; Manici et al., 2005; Tamm et al., 2010; Thuerig et al., 2009). Lepidium sativum is recognized to be a sensitive test plant, and it grows very fast, permitting a high throughput screening of soil suppressiveness (Bonanomi et al., 2006; Erhart et al., 1999; Pane et al., 2011).
Soil suppressiveness bioassay with pooled LTEs samples
For this part of the project, we used the samples coming from the 0-10 cm layer for trials which had as a factor tillage (66 samples), and the 0-20 cm layer for the trials which had as a factor only fertilization (35 samples) (total samples n=101). Equal parts of soil (approximately 100 ml) were collected from each sample. The samples belonging to the same LTE were pooled and mixed in order to obtain 1 L of soil for each LTEs (ten pooled samples in total). These composite samples are referred to as pooled LTEs samples and they were used to assess the natural disease pressure and disease suppressiveness in the different LTEs. Moreover, we wanted to prove the biological nature of general disease suppressiveness in our LTEs. For this purpose, half of each sample was autoclaved at 121º C for 20 minutes to kill the major part of soil microorganisms, including soil pathogens. The other half was kept fresh and both 0.5 L samples were stored for one or two days at 3ºC. The soil suppressiveness bioassay was performed using the Pythium ultimum - Lepidium sativum pathosystem model based on the protocol of Tamm et al. (2010) and modified as follows. Ten days before the sowing of the cress (L. sativum), inoculum of Pythium ultimum isolated from tomato (provided by Biointeraction and Plant Health, Wageningen Plant Research, The Netherlands) was produced on millet (24 g of sterile millet used as a substrate plus 20 mL of demineralized water) and incubated in a dark room at 20º C. Nine days before the sowing of the cress, the autoclaved and non-autoclaved soil was taken out of the cold room and incubated at 20º C for one week to acclimatize and permit the reactivation of microorganisms. After eight days of mycelium growth, and two days before the sowing of the cress, the mycelium/millet culture was chopped and homogenized with a sterilized metal spatula. The homogenized mycelium/millet culture was then mixed with sand (1:80 mycelium/sand dilution factor) to obtain a final density of 0.125 g of P. ultimum /millet seeds per litre soil. Subsequently, this mixture was used to inoculate the test soils. The test soil did not receive any additional fertilization. The experimental setup included 10 autoclaved and 10 non-autoclaved pooled LTEs samples, two dosages of P. ultimum inoculum (0 and 0.125 g L-1) and 4 replicates per P. ultimum inoculum concentration. The soils were placed in plastic polypropylene containers (Ø 133 cm, 500 ml ) perforated at the top and were pre-incubated in the dark at 20º C for two days. After this pre-incubation time, 4 pots (Ø 6 cm, 95 mL) for each perforated plastic container (the replicates) were filled with 0.3 l of inoculated or non-inoculated soil and were sown with 0.5 g untreated biological seeds of L. sativum (De Bolster, Epe, The Netherlands) each. Each pot was placed on an individual plant saucer to avoid cross-contamination between different soils and treatments. Pots were completely randomized and incubated in a growth chamber at 23ºC (day) and 18ºC (night) with a day-length of 16 hours and 80% relative humidity (Unifarm, Wageningen University, The Netherlands). For the first two days after sowing a plastic sheet covered the pots to prevent evaporation and ensure 100% relative humidity for germination. After two days, the plastic sheet was removed and the pots were irrigated from below when needed. Seven days after sowing, shoot fresh weight in each pot was assessed by cutting the shoots with scissors directly above the ground. Weight reduction (WR) was calculated according to Tamm et al. (2010), as:
where n=natural, a = shoot weight in pots with natural soil not inoculated with P. ultimum, and b = mean shoot weight in autoclaved soil not inoculated with P. ultimum, and
where i=induced, c = shoot weight in pots with natural soil inoculated with P. ultimum, and d = mean shoot weight in natural soil not inoculated with P. ultimum.
The WR-n refers to the natural disease pressure present in the soil, and the WR-i to the disease suppressiveness induced by an experimentally added pathogen. Weight reduction of cress plants has been used throughout the manuscript as a measure for soil suppressiveness (i.e. more weight reduction means less disease suppressiveness).
Soil suppressiveness bioassay for management treatments within the LTEs
The soil suppressiveness bioassay described in section 2.4 was also used to determine WR-i for the agricultural treatments separately within each LTE. For these bioassays the weight reduction (%) was calculated as in equation 2) only with natural soil with Pythium relative to natural soil without Pythium, because soils were not autoclaved. Bioassays were run in 10 separate batches, one for each LTE. The experimental setup included only non-autoclaved soil samples, two dosages of P. ultimum inoculum (0 and 0.125 g l-1) and 4 replicates per P. ultimum inoculum concentration for each sample. During all the performed bioassays (pooled and individual LTEs) a standard potting soil was included and used as a control for the reproducibility of the experiments (Potting soil Nº 4-102060, Unifarm, Wageningen University, The Netherlands; for details on the potting soil Table S12; results of the potting soil Table S13).
The effect of the four different treatments (natural soil, natural soil with Pythium, sterilized soil, and sterilized soil plus Pythium) on the fresh weight of plants growing in soil with pooled samples per LTE was analysed with one-way ANOVA.
The effects of the agricultural treatments on the soil suppressiveness (expressed in % weight reduction) were assessed by linear mixed effect models (LMEs). The LMEs were run independently for each LTE. Tillage and/or organic matter addition were included as fixed factors while, depending on the trial, block, main plot and subplot were introduced as random factors to take into account the nested design of the experiments. The results were considered statistically significant at P < 0.05. The effects of tillage and organic matter addition and their interaction on % of weight reduction were addressed by performing an analysis of variance (function anova) on the linear mixed effect model. For all the models, normality and homogeneity of variances of the residuals were checked both visually and with the Shapiro-Wilk and Levene’s tests. For these tests, results were considered statistically significant at p<0.05. When the ANOVA indicated a statistically significant effect at p<0.05, Tukey HSD post-hoc test was used to assess significant differences between treatments. The function emmeans was used to compute the estimated least squares means from the model.
Spearman’s rank order correlation was used to examine relationships between weight reduction and biological, physical, and chemical soil quality parameters, including labile carbon fractions. For the correlation analyses, data from the individual treatments from all ten LTEs were used (n=101). The relationship between labile carbon fractions and soil parameters was validated using partial correlations, correcting for variation caused by the intrinsic differences of the LTEs (pedoclimatic zones). The soil parameters for which we had measurements from all plots of the LTEs were used to construct a multiple linear regression model to select the most important variables explaining weight reduction. Variables with a VIF ≥ 5 were removed one by one from the model to control for strong collinearity. Based on the selected variables, a stepwise selection procedure was used to obtain a final regression model (Min and Toyota, 2013). The model was afterwards checked for normality and homogeneity of the residuals and these assumptions were met allowing for different variance structures for the residuals of the different LTEs and fitting a generalized least squares model (gls), due to the variation present between the LTEs (Zuur, 2009). The Akaike Information Criterion (AIC) was used to select the final model (Field et al., 2012). T-values are reported to quantify the contribution of each predictor to the model (Field et al., 2012). In parallel to multiple linear regression, a regression random forest analysis (Breiman, 2001) was executed to further assess the importance of the soil parameters in explaining variation in the weight reduction (WR) (Supplementary material). The random forest is a technique quite robust to collinearity between explanatory variables and it determines the relative importance and the statistical significance of each predictor variable by measuring the decrease in prediction accuracy (i.e. the increase in the mean square error between the observation and the out-of-bag predictions) when the data for that predictor are randomly permuted. The parameters specified in the random forest model were mytry=8 (p/3 where p=number of explanatory variables and n trees=1000). The methodology used to find relationships between WR-i and soil parameters followed a descriptive approach, without a pre-distinguished hypothesis (Postma et al., 2008).
Structural equation modelling (SEM) was used to evaluate the direct and the indirect effects of the labile carbon fractions on the %WR-i. We established an a priori model including the main physical, chemical and biological variables and labile carbon fractions that resulted to be important for weight reduction from the correlations and the multiple regression model (Figure S1). The data matrix was fitted using the maximum-likelihood estimation method. The Chi-square test (χ2; the model has a good fit when 0 ≤ χ2/degrees of freedom. ≤ 2 and p≥0.05) and the root mean square error of approximation (RMSEA; the model has a good fit when RMSEA 0.05 and p≥0.10) were used to test the overall goodness of fit for the SEM (Schermelleh-Engel et al., 2003). We calculated the total standardized effects of the predictors on WR-i. In addition, we confirmed the fit of the SEM performing Bollen-Stine bootstrap test (the model has a good fit when the P value of the test is high (traditionally > 0.1) (Schermelleh-Engel et al., 2003).
All statistical calculations were carried out using R Development Core Team version 3.3.2. For the linear mixed effects model and the generalized least square model, the packages nlme, and emmeans were used, for the multiple linear regression and the correlation analysis the packages car and stats were used (Pinheiro et al., 2018). For the random forest regression analysis the randomForest package (Liaw and Wiener, 2002) was used, and for the structural equation model the lavaan package was used (Rosseel, 2012).
Soil suppressiveness bioassay with samples pooled per LTE
In CH3, NL2, SL1, PT1, and ES4, the shoot weight of the cress had increased (p<0.05) due to autoclaving compared to not inoculated natural soil, while this had no effect in CH2, NL1, HU1 and HU4 (Figure 3, green; pink and blue boxes). In CH1, the autoclaving process decreased (-79%, p<0.05) the weight of the cress plants compared to the natural soil. Soils with the highest natural disease are those of PT1, ES4 and NL2, followed by SL1, CH2 and CH3 (Figure 4). In these LTEs, shoot fresh weight in non-inoculated natural soil was significantly (P<0.05) lower than shoot fresh weight in non-inoculated sterilized soil. In NL1, HU1 and HU4, this difference was not significant.
Inoculation of natural soil with P. ultimum resulted in a decrease in cress fresh weight compared to plants growing in natural soil without P. ultimum. The biomass reduction was statistically significant (P<0.05) in all the trials except for ES4 and HU4 (Figure 3, pink and green boxes). The soils from LTE SL1, NL2 and CH3 were least capable to counteract the growth reduction induced by the inoculation of the pathogen (growth reduction 53, 42, and 40%, respectively; Figure 4, Table S14). Plant grown on soil from ES4 were least sensitive to inoculation with P. ultimum (10% growth reduction), however, the natural disease pressure in the natural soil of ES4 was already very high (56% growth reduction; Figure 3 and 4).
For all LTEs, cress plants grew better (p<0.05) in autoclaved soil that were not inoculated with P. ultimum than in autoclaved soils with the disease (93% growth increase averaged over all LTEs, Figure 1, blue and violet boxes). In addition, in all the LTEs shoot fresh weight in sterilized soil inoculated with Pythium was statistically lower than in natural soil with Pythium (Pp<0.05) (Figure 3, green and violet boxes).
There was no statistically significant relationship between cress shoot weight reduction (WR-i) upon pathogen addition and cress shoot biomass when grown in natural soil without P. ultimum inoculation (data not shown).
Soil suppressiveness bioassay for management treatments in the LTEs
Tillage did not affect the cress shoot biomass, nor the weight reduction (WR-i) (Table 7, 9). However, for HU4 the significance level was close to 0.05 (p=0.08), with less weight reduction (WR-i) in reduced tillage plots compared to conventional tillage plots.
Table 7. Effect of tillage on cress shoot biomass fresh weight (g) in natural soil and weight reduction (%, WR-i) in natural soil with inoculation of P. ultimum compared to natural soil without P. ultimum inoculation in LTEs CH1, CH2, and HU4. Least squares means, standard errors (in parenthesis) and F and p values for mixed linear effect models are reported for conventional tillage (CT) and reduced tillage (RT). Differences are considered significant at p≤0.05.
In contrast, fertilization affected cress shoot biomass and P. ultimum induced cress shoot growth reduction, but only in a selection of LTEs (Table 8, 9).
ES4 shoot growth reduction due to P. ultimum inoculation was lower in plots that were managed organically compared to plots that were managed conventionally (p=0.04). Also, the shoot biomass of cress plants in natural ES4 organic soil was higher than shoot biomass of cress plants in soil from the conventional treatments. The latter was true also for cress grown in soil from the biodynamic treatments in LTE CH3 compared to the conventional treatments (Table 8). For the organic matter addition trials (PT1, HU1, NL1, and SL1) the results were more complex. In NL1 and SL1, the cut and carry fertilizer and the bio-waste application, respectively, did not affect the P. ultimum induced weight reduction (WR-i) of cress plants (Table 9). In LTE PT1 we found a marginally significant (p=0.06) higher WR-i when biochar (either with or without addition of compost) was added to the soil as compared to non-fertilized control soil (Table 8). In LTE HU1, P. ultimum cress shoot WR-i was significantly lower (p=0.006) in plots that received mineral N fertilization as compared to plots that did not receive an additional mineral N fertilization (36.6% and 54.7%, respectively; Table 8). However, in this trial the organic fertilization did not affected P. ultimum cress shoot WR-i mineral N fertilization. Using the data from all the trials we found a negative relationship between cress shoot WR-i and shoot biomass in natural soil to which no pathogen was added (rs=-0.33; p=0.0007) (Figure S2).
Table 8. Effect of different organic matter additions (OM) on cress shoot biomass fresh weight (g) in natural soil and weight reduction (%, WR-i) in natural soil with inoculation of P. ultimum compared to natural soil without P. ultimum inoculation in LTEs CH3, PT1 and ES4. Least squares means, standard errors (in parenthesis) and F and p values mixed linear effect models are reported for the different type of fertilization. Differences are considered significant at p≤0.05.
Table 9. Effect of tillage and organic matter addition (OM) on cress shoot biomass fresh weight (g) in natural soil and weight reduction (%, WR-i) in natural soil with inoculation of P. ultimum compared to natural soil without P. ultimum inoculation in the LTEs NL1, NL2, SL1, and HU1. Least square means, standard errors (in parenthesis) and F and p values for mixed linear effect models are reported for the tillage and organic matter treatments and their interactions. Differences are considered significant at p≤0.05.
Correlation with soil parameters
WR-i (derived from the separate experiments for each of the LTEs) was negatively correlated with chemical parameters: pH, total N, CEC, Ca and K, and positively with the soil C to N ratio (Table S15). A negative correlation was also found for the soil physical parameters WHC, silt, clay, and penetration resistance, while a positive correlation was found for bulk density and sand. WR-i was negatively correlated with biological parameters: microbial biomass (MBC, MBN) and activity (soil respiration), with the microbial quotient (qMic), and decomposition measured as tea bag index. A weak positive correlation was found for the metabolic quotient (qCO2). Correlations with labile organic carbon fractions were negative and strongest for HWEC, followed by Hy-DOC and POXC, while DOC and POMC were not correlated with WR-i. In addition, WR-i was weakly positively correlated with DOC SUVA and the hydrophilic-DOC fraction SUVA (Hy-DOC SUVA) (Table S15). The partial correlation showed that after normalization for structural differences between the LTEs (the pedoclimatic characteristics), total N, MBC, soil respiration, qMic, POXC, HWEC and POMC were negatively correlated with WR-i, while the C to N ratio, tea bag decomposition and DOC SUVA were positively correlated with WR-i (Table 10).
Table 10. Partial correlation coefficients (ρ) between the weight reduction (WRi) and other chemical, physical and biological parameters used as dependent variables, corrected for the long term field experiments (LTEs). The number of samples used in the analyses was 101.
TOC total organic carbon, TON total nitrogen, C/N carbon to nitrogen ratio, CEC cation exchange capacity, WSA water stable aggregates, WHC water holding capacity, BD bulk density, MBC microbial biomass carbon, MBN microbial biomass nitrogen, qCO2 metabolic quotient, qMic microbial quotient, Hy hydrophilic carbon, Hy SUVA specific ultraviolet absorbance of hydrophylic carbon, DOC dissolved organic carbon, DOC SUVA specific ultraviolet absorbance of dissolved organic carbon, POXC permanganate oxidizable carbon, HWEC hot water extractable carbon, POM-C particulate organic matter carbon.* p ≤ 0.01, **p ≤ 0.001, ***p ≤ 0.0001
Multiple regression and structural equation model (SEM)
A multiple regression model to explain pathogen induced cress shoot growth reduction only selected microbial biomass N (MBN) and hot water extractable carbon (HWEC) as statistically significant explanatory variables (Table 11). The final model had a lower AIC compared to previous models taken into account (Table S16). Moreover, the comparison of the models with anova analysis revealed that the inclusion of the LTEs in the fixed part of the model significantly increased the fit of the model (Table S16). The random forest analysis corroborated the results of the multiple regression models by selecting MBN and HWEC as being important predictor variables. However, in addition to the multiple regression model, the random forest analysis also identified MBC, qMIC, POM-C and sand as important predictor variables (Figure S3).
Table 11. Multivariate generalized least squares model determined from soil parameters measured in the 101 soil samples. The dependent variable was the % weight reduction of cress plants upon addition of Pythium ultimum (WR-i) compared to natural soil without the addition of the pathogen. Significant model parameters (p<0.05) explain variation in weight reduction (WR-i), and are derived from a reduced regression model after a stepwise selection procedure.
The structural equation model (SEM) fitted to investigate the direct and indirect effects of the labile carbon fractions, nutrients and texture on the cress shoot WR-i (%) indicated that the labile carbon fractions HWEC and POXC have an indirect negative effect on cress shoot WR-i through their positive effects on microbial biomass carbon (Figure 5). Microbial biomass carbon (MBC) has a positive effect on soil respiration (RS) which has a direct negative effect on cress shoot WR-i. Also, we found indirect and direct effects of texture (sand fraction) and an indirect effect of CEC on the cress shoot WR-i. The SEM explained 38% of variation in the WR-i.
Soil suppressiveness in pooled LTEs samples
Plants growing in natural autoclaved soil grew better than in natural soil, which was most likely due to the elimination of plant pathogens with the autoclaving (Thuerig et al., 2009). Comparing the growth of plants in these two soils gave us an indication of the disease present in the field. In this respect, we found different levels of natural disease in the LTEs, pointing out the heterogeneity of pathogen communities assemblages in the different sites (Figure 4). Since autoclaving can eliminate also organisms that can act as natural enemy of soil pathogens, the addition of Pythium to autoclaved soil seriously diminished plant growth (Figure 3). In these soils Pythium had the opportunity, in terms of space and nutrient utilization, to grow undisturbed. This observation confirmed the results of previous studies which indicated the biological nature of soil suppressiveness against Pythium (Gravel et al., 2014; Knudsen et al., 2002; Löbmann et al., 2016; Lumsden et al., 1987; Thuerig et al., 2009; van Os and van Ginkel, 2001).
Effect of soil management practices on soil suppressiveness
In general, we did not find strong effects of soil management on disease caused by Pythium. This result is in accordance with previous studies (Knudsen et al., 2002; Löbmann et al., 2016; Vestberg and Kukkonen, 2014). Organic matter additions can have positive, negative or neutral effects on soil suppressiveness (Bonanomi et al., 2007). These effects can vary with time, the suppressive capacity of organic material added to the soil can disappear some months after its application and between different batches of the same applied material (Litterick et al., 2004). For example, Darby et al. (2006) found that the disease severity of root rot of sweet corn increased with time in soil which received organic amendment and lightly decomposed fPOM. This suggests that changes in the nature of the organic matter (nutrients, quality and stage of decomposition, time of application) and in the soil environment are important for disease suppression (Boehm et al., 1993; Charest et al., 2005; De Ceuster and Hoitink, 1999). Therefore, it is possible that our soil management (in particular organic matter addition) had a short term effect which was lost in the long term. Organic matter should be decomposed, but not excessively, in order to support soil suppressiveness (Litterick et al., 2004; Tuitert et al., 1998). For example, the biochar in PT1 could lack readily-available substrates which reinforce antagonistic microbial activity. Biowaste, which was used in SL1, has been shown to be variable in its general disease suppressive capacity (Veeken et al., 2005). In addition, the pathogen might have used its saprophytic capacities to grow on the organic material added and increase its inoculum density if other soil organisms were not competitive.
Remarkably, in ES4 the organic system had less weight reduction than the conventional system, which could be due to the retention of more complex and readily available organic substrates. Complex substrates, for example lignocellulosic substrates, can increase the presence of natural enemies like other Pythium species and Trichoderma, and more readily available substrates could increase general microbial activity (Medvecky et al., 2007). He et al. (2010) found that in organic managed soil the disease suppressive effect of compost was higher and more stable compared to conventionally managed soil. However, in CH3 and NL2 we did not find more disease suppressiveness in the biodynamic and integrated system, respectively, compared to conventional ones. At an earlier sampling of the CH3 trial, Tamm et al. (2010) found that at high fertility input the conventional system was more suppressive than the biodynamic.
Regarding the tillage management, SL1 and HU4 tended to have more disease suppression in reduced than in conventional tillage. This could be due to the higher microbial biomass and activity in reduced tillage compared to conventional tillage plots, and to the positive effect of soil properties created by the reduced tillage on plant growth. In contrast, often studies found that reduced tillage had less disease suppression than conventional tillage (Garbeva et al., 2004). Pankhurst et al. (1995) ascribed this observation to the positive effect of soil properties created by the reduced tillage on Pythium growth. In general, it is possible that the applied management did not change the microbial community composition, i.e. did not favour microorganisms which were antagonistic to Pythium, or did not stimulate the indigenous antagonistic microbial population.
The negative relationship between growth reduction and cress fresh weight in the natural soil without Pythium addition can be explained by the fact that plants already damaged by other pests or disease present in the soil, could have predisposed plants to Pythium infection (Medvecky et al., 2007). A similar effect could also be due to soil physical and chemical properties not favourable for seedling growth, such as low soil pH. Broders et al. (2009) suggested that the prediction of disease incidence across a broad area may not be possible due to inherent variability ad that it would be better to assess it based on a field by field basis. Another element that could have affected the results is the crop that was growing in the fields, which has not been taken into account in this study. Plants can exert a very strong influence on soil microbial communities and therefore on soil suppressiveness (Mazzola, 2004; van Elsas et al., 2002).
Correlations between growth reduction and soil parameters
We correlated the cress weight reduction (WRi) with chemical, physical and biological parameters taking into account the influence of the LTE on the relationship between the parameters with partial correlations. Contrary to the results of the bivariate correlations, with the partial correlations only few parameters resulted to be correlated with weight reduction, e.g. total organic nitrogen, C to N ratio, microbial biomass C, soil respiration, qMic, tea bag decomposition, DOC SUVA, POXC, HWEC and POMC. This result underline the strong site effect on the disease suppressiveness in our study. The negative correlations (bivariate and partial) found between cress weight reduction (WRi) and the biological parameters, i.e. microbial biomass C, soil respiration and qMic suggest that reduced microbial populations and activity are associated with an increased severity of the disease. This is in support of the hypothesis that soil biota are involved in the suppression of soilborne-borne diseases. Suppression of Pythium has often been associated with the biomass and activity of the entire microbial community (Craft and Nelson, 1996; Darby et al., 2006; Gravel et al., 2014; Inbar et al., 1991; Lumsden et al., 1987; Scheuerell et al., 2005; van Os and van Ginkel, 2001) but less often, it has been associated with labile organic carbon fractions. In our study we found correlations with various labile carbon fractions (negative correlations for POMC, HWEC and POXC, and positive correlations for C/N, and DOC-SUVA), but we did not find a correlation between cress weight reduction upon pathogen addition (WR-i) and TOC. However, in previous studies, both organic matter and labile carbon fractions were found to be positively correlated to general disease suppressiveness, an effect that is ascribed to their positive impact on the competitive potential of soil microbial communities against pathogens (Broders et al., 2009; Mazzola, 2004; Schlatter et al., 2017). The mechanism through which labile organic carbon can favour soil suppressiveness is mainly related to the support of an active soil microbial community, which will compete for nutrients and space and can thrive on nutrients released by the pathogen during plant’s attack (Pascual et al., 2002).
Multiple regression between soil suppressiveness and soil parameters
Adding the LTE as a fixed factor permitted to find important parameters for explaining the variation in weight reduction independent from the site. Microbial biomass N and HWEC were the most important parameters, pointing once more at the essential role played by microbial communities in soil suppressiveness against Pythium. An active soil microbial community is able to compete for nutrients and space and can thrive on nutrients released by the pathogen during plant’s attack (Pascual et al., 2002). On the other hand, when the soil microbial community is not well established, nutrient concentration is high, and soil conditions are not favourable for their growth, Pythium can grow saprophytically on dead soil organic matter , overcoming in such a way the suppressive effect of microbes (Hendrix and Campbell, 1973). Other mechanisms can be involved in disease suppression beside antagonism for nutrients such as parasitism, predation, production of specific compounds and volatiles, fungistats, host mediated resistance, siderophore and enzyme production (Charest et al., 2005; Mazzola, 2002; Pane et al., 2011; Van Agtmaal et al., 2017). Also the composition of the microbial community and the presence and the activity of less abundant and more specific microbial groups or taxa which have antagonistic characteristics affect soil suppressiveness (Mazzola, 2002). Soil substrate composition is very important for the expression of disease suppressivenes, being a key determinant of microbial community composition (Garbeva et al., 2004). Copiotrophic bacteria can compete with Pythium for soil resources, decreasing the potential of the pathogen to cause plant infection (He et al., 2012). Saprophytic Fusarium, Pseudomonas, Actinomycetes and fungi (including Trichoderma) are active against Pythium (Boehm et al., 1993; Lumsden et al., 1987; Postma et al., 2000). Also Manici et al. (2005) and Van Agtmaal et al. (2017) found that Pythium was negatively correlated with (saprophytic) fungi present in the soil. She et al. (2017) showed that, under long term continuous cropping, soil properties have an impact on the microbial communities and that bacterial diversity and community composition correlate with tobacco bacterial wilt disease rate.
Relation between soil suppressiveness and labile carbon with structural equation model (SEM)
HWEC and POXC resulted to have an indirect effect on weight reduction through a direct positive effect on microbial biomass and, subsequently, microbial activity. Labile carbon is considered the primary energy source for microorganisms, and probably it contain part of the microbial biomass and microbial by-product. Unfortunately, we cannot assign a cause-effect relation between labile carbon and soil suppressiveness because they were measured at the same time, and the relationships assessed by the structural equation model do not per se imply causality. Nevertheless, this result confirms that the quality of the soil carbon and its effect on soil microorganisms is essential for soil suppressiveness.
We found only a weak effect of long term agricultural practices on soil suppressiveness compared to a very strong effect of LTE site. This observation points to the necessity to study and select more carefully site-specific agricultural practices, or combination of practices, which can favour soil suppressiveness . The selection of agricultural management in our sites was, in fact, not driven by the scope of alleviating diseases. As we explained above, the choice of the type of organic material added to the soil could have a big influence on disease development. Also, the variability between sites makes the comparison between them very difficult, and this is also the reason why we decided not to merge them in an overall analysis. In addition, it might be that short-term effect of agricultural practices could be more important in disease suppression than the long-term effect.
Nevertheless, we found that growth reduction across LTEs was linked mainly to microbial biomass and activity and to labile carbon (in particular HWEC) which is a biochemical parameter strictly linked with microbial parameters. Moreover, most likely, also the abiotic parameters (i.e. Sand, pH, soil nutrients, bulk density etc.) which we measured were linked to weight reduction through their influence on microbial parameters and plant growth.
Weight reduction could not be entirely explained by the soil parameters measured, suggesting that other more specific mechanisms could be behind soil suppressiveness. This mechanisms might be related to specific microbial assemblages, which could be more easily addressed with novel molecular methods. Future studies should focus on more specific hypotheses on the relation between soil suppressiveness and microbial characteristics, which could help to distinguish direct and indirect influences, and primary and secondary factors in disease suppressiveness. Moreover, they should make use of other pathosystems (both plant and pathogen) and agricultural practices in order to be able to generalize the results on disease suppressiveness.
Note: For full references to papers quoted in this article see