Growing Medium Type A ﬀ ects Organic Fertilizer Mineralization and CNPS Microbial Enzyme Activities

: Managing plant fertilization is a major concern of greenhouse growers to achieve sustainable production with growing media (GM). Organic fertilization is popular but is more di ﬃ cult to control, since organic compounds need ﬁrst to be mineralized by microbes. After 7, 14, 28, and 56 days of incubation, we investigated the response of microbial activities and nutrient releases from three frequently used organic fertilizers (horn and two plant-based fertilizers) in three frequently employed GM types (peat, coir, and bark). We measured pH, electrical conductivity, nutrient contents (NH 4 + -N, NO 3 − -N, PO 43 − -P, SO 42 − -S), and enzyme activities ( β -1.4-glucosidase, urease, acid phosphatase, arylsulfatase). After fertilization, microbes in coir expressed all the C, N, P, and S functions studied, making related nutrients available. In peat and bark, some C, N, P, and S-related pathways were locked. Peat presented high NH 4 + -N and PO 43 − -P releases linked to high acid phosphatase and β -glucosidase activities, while bark showed high nitriﬁcation rates but weak enzyme activities. Fertilizer types modulated these responses with lower activities and nutrient releases with horn. Our results contributed to better understanding mineralization processes in GM, showing di ﬀ erent microbial responses to fertilization. This study pointed out the necessity to look deeper into microbial functions in GM optimizing biological and physicochemical properties.


Introduction
Horticultural crops frequently use organic materials in growing media (GM) formulation to physically support plant growth while ensuring appropriate solid/air/water balance and nutrient supplies for healthy roots [1,2]. The range of growing media constituents and stand-alone substrates includes peat, coir pith, wood fibers, bark, composted materials (e.g., green waste, bark). Among these organic materials, peat has been widely used in growing media during the last decades due to its reliable properties such as low bulk density, high biochemical stability, high porosity, and high air and water-holding capacity, making this substrate particularly suitable for growing a large number of vegetables and ornamentals [1,[3][4][5]. However, the availability of this natural resource is nowadays under pressure due to increasing demand for GM, increasing regulation policies about the preservation of peatland as carbon sinks, and transportation costs [2,5,6]. Researchers and the GM industry are working to use peat more wisely and sparingly [7] by partially or totally substituting it with other renewable or sustainable organic materials (e.g., coir, bark, wood fibers, green composts) [4,5,8,9]. Another step toward sustainable soilless systems involves the use of organic fertilizers as a substitute to synthetic fertilizers [9]. Dealing with nutrient excess or deficiency is extremely challenging and has been so far an obstacle to organic fertilization in horticulture [9,10].
To survive and reproduce in the environment, microbes degrade complex organic molecules as electron and energy sources for ATP production, which is needed for cellular reactions and new synthesis using carbon and nutrients from mineralization [30]. Microbes mediate their resource allocation toward targeted substrates through C-, N-, P-, or S-acquiring enzyme production in order to meet their stoichiometric needs [31,32] by adopting various strategies to detect and efficiently use these substrates [33]. The constrained and stable elemental composition of microbes makes fertilizer's quality an important factor that is supposed to drive biochemical cycles and thus nutrient availability [32,34]. Usually, the organic fertilization of soil stimulates microbial growth and enhances enzyme activities [35]. Increasing N availability can promote C-or P-acquiring enzyme production, but higher P availability does not necessarily increase N-acquiring enzymes [32]. Recent works already showed differences in the ability of different growing media (e.g., green waste compost, coir, and peats) used alone to ensure microbial N mineralization (i.e., ammonification) and nitrification [36]. Moreover, adding organic fertilizer (vegetable and animal-based materials) to GM was found to increase the number of amoA copies, indicating an increase of nitrifier abundance [34,36], but subsequent nitrification rate was not determined.
Managing nutrient availability in GM constituents with organic fertilizers is a question of microbial ecology (i.e., plant-microbes-fertilizer interactions) and thus is difficult to predict compared to mineral fertilization. Past studies dealing with the biological properties of GM lack comprehensive insight on microbial ecology, and the relationships between microbes and organic fertilizer in soilless cropping systems received only recent attention [34,36]. In such organic fertilized systems, plant nutrition will depend on the resulted amount and form of available nutrients from fertilizer mineralization and nitrification mediated by microbes. As a matter of fact, available C is often the most limiting factor for microbial growth in soil, and in some cases, N and P can also be limiting [37]. The availability of C to microbes was also suggested as the main driver of microbial decomposition rates and thus N immobilization in GM [38]. Indeed, the addition of glucose to GM was found to increase microbial activity [39], but the immobilization of N or P can also rapidly occur in GM [27][28][29]. Thus, as microbes in GM seem to have multiple limitations, it is essential to assess microbial activities involved in mineralization processes driving nutrient availability in growing media depending on fertilizer type through an overview of C, N, P, and S cycles.
The aim of this paper was to assess microbial functions in three contrasted materials frequently used as stand-alone growing media (peat, coir, and bark) combined with three different organic fertilizers (horn meal and two different plant-based) that would lead to distinct nutrient availability dynamics. We suspected that GM type with strongly unbalanced stoichiometry ratios would affect the microbial functional response to fertilization. More specifically, we hypothesized that higher C:N, C:P, and C:S (i.e., low N, P, and S content) ratios of a GM type (Table 1) would increase microbial activities through enzyme production to get access to nutrients. In addition, we expected that high fertilizer elemental ratios and recalcitrance (Table 1 and Table S1) would slow down microbial activities and organic matter mineralization rates (N, P, and S releases), limiting thereby nitrification process.

Experimental Design and Treatments
A microcosm-scale experiment was designed in order to test interactions between 3 contrasted GM types among the most used in soilless horticulture (peat, coir, and composted bark, called "bark" thereafter) and 3 fertilizer types varying by its origin and chemistry (F1, F2, and Horn meal) ( Table 1, Tables S1 and S2). Nutrient availability (ammonium (NH 4 and sulfate (SO 4 2− -S)), pH, electrical conductivity (EC), and enzyme activities (β-1.4-glucosidase, urease, acid phosphatase, and arylsulfatase) were monitored over time (after 7, 14, 28, and 56 days). Physicochemical properties of the GM and fertilizers are indicated in Table 1 and supplemented in Tables S1 and S2. Horn meal is an animal-based and high N fertilizer (13.6% N w:w). F1 and F2 (commercial names are confidential) are two plant-based (including mainly cacao shells, meal of oil-press cake from grape seeds and soy, vinasse) granular fertilizers that are well NPK-balanced (Table 1) but the latter being poorer in nutrient content (NPK 6-1.3-3.3 and 2-0.2-2.1, respectively following commercial values).
Synthetically, a GM sample was weighted to obtain 200 mL using bulk density, which was disposed in a container (12 × 8.5 × 6.5 cm) that was covered with plastic film. Holes in the plastic film allowed air circulation in and out of the container. Samples were pre-incubated during 7 days to allow equilibration at 25 • C and at a water suction of −5 kPa (pF 1.7, ≈60% water-holding capacity (WHC) and 45-50% water-filled pore space (WFPS)), corresponding to optimal conditions for microbial activity [25,40,41]. After this pre-incubation period, GM and fertilizers (applied at 300 mg N L −1 of GM, corresponding to 4.3, 4.8, and 1.6 mg N g −1 dry mass (DM) in peat, coir, and bark, respectively) were mixed together. Treatments consisted of 3 GM types × 4 fertilizer types (including no fertilizer application) × 4 times × 3 replicates for a total of 144 destructive samples. Then, the samples were incubated in the dark during a maximum of 56 days at 25 • C and maintained at 60% of WHC with deionized water. After 7, 14, 28, and 56 days of incubation, 36 samples were retrieved from incubation, homogenized, and sampled for chemical and enzymatic assays. Samples of each GM were also retrieved after pre-incubation (day 0) for initial chemical and enzymatic assays (n = 9). Table 1. Physicochemical properties of organic growing media and organic fertilizers (n = 3, mean ± sd). A property with no sd value was measured on a composite sample (i.e., with no replicate).

Chemical Analyses
At each sampling date, a 20 mL subsample measured on a weight basis according to bulk density was retrieved from samples (200 mL). The subsample was immediately extracted with 30 mL of deionized water by shaking during 1 h on an orbital shaker (350 rpm); then, it was centrifuged at 4000 rpm during 2 min 30 s, filtered (LAB-ONLINE ® ashless filter no. 13), and stored at −20 • C before analyses. Extracts were analyzed for ammonium (NH 4 + -N) and nitrate (NO 3 − -N) contents by the nitroprusside-salicylate method and nitrosalicylic acid method, respectively [53,54]. Available phosphorus (PO 4 3− -P) was measured with molybdenum blue method [55] and available sulfur (SO 4 2− -S) was measured with turbidimetric analysis (i.e., precipitation of sulfate with barium ions under acid condition) [56,57]. Nutrient contents were expressed as µg of element released per gram dry mass for each treatment. Nutrient availabilities on a volume basis were also added in Supplementary Materials ( Figure S1). The pH and EC were analyzed on deionized water extract (GM:water, 1:5 v:v) following AFNOR Standards, NF EN 13,037 [47] and NF EN 13,038 [48], respectively.

Enzyme Activity Assays
Four enzyme activities-β-1.4-glucosidase (β-Glu) hydrolyzing glycoside linkages of cellulose, acid phosphatase (acid-P) releasing phosphoryl groups from various molecules, arylsulfatase (aryl-S) hydrolyzing sulfate esters formed in various molecules, and urease releasing NH 4 + -N from urea hydrolysis-were performed by colorimetric assays as described by Eivazi and Tabatabai (1988) [58], Tabatabai and Bremner (1969) [59], Tabatabai and Bremner (1970) [60], and Tabatabai and Bremner (1972) [61], respectively. β-1.4-glucosidase [58], acid phosphatase [59], and arylsulfatase [60] activities were measured by quantifying the amount of p-nytrophenol (pNP) released from enzyme substrates: p-nitrophenyl β-D-glucopyranoside (pNPG; 5 mM), p-nitrophenyl phosphate (pNPP; 5 mM) and p-nitrophenyl sulfate (pNPS; 5 mM) respectively. One mL of enzyme substrate solution (pNPG, pNPP, or pNPS respectively) was added to 0.5 g of sample (fresh mass) and incubated in a water bath at 37 • C in sodium acetate buffer (0.2 M, pH 5) during 60, 30, or 120 min, respectively, depending on the enzyme type. The reaction was stopped by adding 1 mL of 0.5 M CaCl 2 and 4 mL of 0.5 M NaOH. The absorbance of the supernatant was measured at 412 nm after filtration (Rotilabo ® filter no.601). β-1.4-glucosidase, acid phosphatase, and arylsulfatase activities were expressed as µmol of pNP released per minute and per gram of dry mass. Urease activity [61] was measured by adding urea (20 mM) in acetate sodium solution (pH 6) to a subsample (0.5 g sample fresh mass). The suspension was incubated for 2 h in a water bath at 37 • C and then filtered (Rotilabo ® filter no.601). The amount of NH 4 + -N was quantified in the filtrate following the nitroprusside-salicylate method [54]. Initial amounts of NH 4 + -N were assayed following the same procedure without urea addition. Urease activity was calculated by subtracting initial amount of NH 4 + -N from samples incubated with urea and expressed as µmol of NH 4 + -N released per minute and per gram of dry mass. Enzyme activities on a volume basis were also added in Supplementary Materials ( Figure S3).

Statistical Analyses
We analyzed the effects of growing media (GM), fertilizer type (Fert.) over time (Time) with a 3 way-ANOVA (p < 0.05) on pH, electrical conductivity, ammonium, nitrate, phosphate, and sulfate contents, and the four enzyme activities using Statistica 13.0 [62]. Among GM types, mean comparisons between fertilizer types at each time point were tested with Tukey's HSD (honestly significant difference) test at p < 0.05. Illustrations (Figures 1-3) were drawn using the ggplot2 package [63] in R free software (version 3.5.1) [64].
Effects of GM, Fert., and Time and their interactions on nutrient content and enzyme activities were tested by main and pairwise tests in permutational multivariate analysis of variance (PERMANOVA) using Primer e software v6 (Primer-E Ltd., Plymouth, UK). PERMANOVA is a routine non-parametric analysis for testing the response of many variables to one or more factors, in an ANOVA-like experimental design, on the basis of any resemblance measure, using permutation methods [65]. PERMANOVA eliminates the requirement of normal distribution that must be satisfied for the analysis of variance, but this is not the case for many kinds of multivariate ecological data [66].
Correlation iconography analysis was performed using the software CORICO©. It calculates partial and total correlations between variables as followed [67][68][69][70]: as a total correlation between the variable A and the variable B, and as a partial correlation between A and B with regard to a variable C (i.e., after deduction of the contribution of C).
With this method, we selected "remarkable" Pearson correlations between dependent (GM, Fert., and Time) and independent variables (pH, EC, inorganic N, P, and S content and C, N, P, and S enzyme activities). Correlations are "remarkable" when the correlation lasts even if any of the other available variables are kept constant. A link between two variables (A and B) is remarkable if the correlation between them does not result in the dependency to a third variable C. In other words, to avoid redundancy, the link between A and B is "remarkable" if and only if the total correlation between them is higher than a given threshold and if the partial correlation between A and B in respect to any other variable C is also higher in absolute values than this threshold and with the same sign as the total correlation [68]. The threshold is fixed by the user between 0 and 1 as a compromise between clarity of the information needed and the number of links depicted depending on the dataset. Increasing the threshold decreases the number of links but can also decrease the information, especially if the variables of interest depend on the others. A significant correlation between two variables is not necessarily "remarkable" if it is caused by a third one. Considering a single coefficient of Pearson's correlation does not take into account the context of the link between two variables depending on all other variables. Thus, selecting remarkable correlations is stricter and better represents correlations in the dataset. This method is more stable and robust than factorial analysis (Principal component analysis, Partial least squares regression, etc.), which is weakly affected by the introduction of a new variable in the dataset [68] and does not require a choice of projection [60][61][62][63]. The risk of promoting wrong similarities between observations is very low considering "remarkable" correlation [69]. We conducted this analysis twice. The first one included the independent variables (GM, Fert., and Time) in the correlation calculations to draw links with dependent (pH, EC, nutrients, and enzyme activities) variables at two different thresholds. A threshold of 0.3 was used to underline the most conservative links (i.e., less sensitive to the variation of other variables), and 0.1 was used to depict supplementary links to complete the information given by the first one. Straight lines depict positive links and dotted lines depict the negative ones. We used a second analysis to test if the repartition of individuals was related to our factors (independent variables) by using all dependent variables as representative of individuals. In this analysis, a cluster indicates a greater number of remarkable correlations.

Growing Media and Fertilizer Effects on pH and Electrical Conductivity
Fertilizer type significantly affected the electrical conductivity (EC) depending on the GM types and whatever the time (GM × Fert. interaction F = 35.1, p < 0.001; Table 2). Indeed, the EC was the highest with F2 (500-600 µS cm −1 ), intermediate with F1 (200-550 µS cm −1 ), and the lowest with the horn (100-300 µS cm −1 ) addition. Moreover, both horn and F1 additions greater increased EC in bark samples compared to coir and peat samples, and F2 similarly increased EC in the three GM types ( Figure 1A-C).  Water pH was significantly influenced by the GM, fertilizer types, and time (F = 17.9, p < 0.001, GM × Fert. × Time interaction; Table 2). More specifically, each GM type showed specific dynamics. In peat samples, fertilization increased the pH until 5.5 with F1 and F2, and horn had a weaker effect, whereas the unfertilized sampled lost 0.5 units of pH in 56 days ( Figure 1D). In coir samples, the pH increased after fertilization (from 7.0 until 8.0) after 7 days, but then, the pH strongly and continuously decreased below the unfertilized treatment, reaching a pH of 5.3 with F1, 6.0 with horn (25 and 15% lower than control, respectively). F2 was not different from unfertilized samples after 56 days ( Figure 1E). In bark samples ( Figure 1F), pH did not vary with time and was significantly lower in F1 and horn compared to the unfertilized and F2 treatments.  Table 2).
More specifically, NH 4 + -N content in peat samples significantly and strongly increased over PO 4 3− -P content increased strongly and significantly in peat samples over time, reaching 0.9 mg PO 4 3− -P g −1 DM for both F1 and F2 reaching 75% P added ( Figure S2), being 80% higher than PO 4 3− -P content (0.5 mg PO 4 3− -P g −1 DM) with horn but also reaching nearly 75% P added (being lower than the supply with granular), and it was maintained close to 0 in unfertilized samples ( Figure 2G). Coir samples also showed a significant increase in PO 4 3--P content, less sharply, with a fertilization intensity effect in the following order: F2 was about two times higher than F1 being about three times higher than horn, the latter being similar to the unfertilized PO 4 3− -P content ( Figure 2H), reaching respectively a bit less than 50, 25, and nearly 0% P added ( Figure S2). Bark samples showed almost no contrast of PO 4 3− -P content between treatments ( Figure 2I).   . Columns refer to GM types (peat, coir and bark, respectively) and rows refer to nutrient content (ammonium, nitrate, phosphate, and sulfate content; mg N, P, or S per g dry mass (DM), respectively). Each graph depicts differences between fertilizer types: Unfertilized (control without fertilizer depicted by dotted lines); F1, F2: plant-based fertilizers (dashed and dotted-dashed lines, respectively) and horn (gray lines).  Figure S2) and around 0.4 mg SO 4 2− -S g −1 DM with horn (50% S added, Figure S2). SO 4 2− -S content slightly increased with F1 and F2 fertilizers (0.3-0.6 mg SO 4 2− -S g −1 DM) in bark compared to horn and unfertilized treatments ( Figure 2L). SO 4 2− -S content was near 0 in the 3 GM types with no fertilization ( Figure 2J,K,L). SO 4 2− -S content reached 75% S added with F1 and around 30% S added with F2 and horn ( Figure S2).

Effect of Growing Media and Fertilizer Types on Enzyme Activities
Acid phosphatase (acid-P) and arylsulfatase (aryl-S) activities were affected by GM, fertilizer type, and time (p < 0.05, GM × Fert. × Time interaction, Table 3), while 1,4-β-glucosidase (β-Glu) activity was affected by GM and fertilizer type whatever the time (p < 0.001, GM × Fert. Interaction, Table 3) and urease weekly responded to the experimental design (p < 0.05, Fert. × Time interaction, Table 3). Table 3. Effects of growing media (GM), fertilizer (Fert.), Time, and their interactions on enzyme activities (β-1.4-glucosidase (β-Glu), urease, acid phosphatase (Acid-P), and arylsulfatase (Aryl-S); µmol min −1 g −1 dry mass). Except for bark samples that showed no change in the four tested enzyme activities ( Figure 3C,F,I,L), we observed specific patterns in peat and coir depending on the enzyme type, the fertilizer, and time. More specifically, acid-P in peat decreased over time even with fertilizer addition except with horn remaining relatively steady ( Figure 3G). In coir samples, we observed an increase in acid-P in the three fertilizer types (highest increase with F2) until 28 days of incubation, and then, it stabilized for F1 and horn. Acid-P decreased with F2 until 56 days ( Figure 3H). Aryl-S in peat did not change over time nor with fertilization ( Figure 3J) but significantly increased after 7 days in coir samples with F1 and F2 and followed by a decrease in activity until the end of the experiment. On the contrary, horn addition increased this activity over time with a least intensity ( Figure 3K). The β-Glu in peat strongly increased with F1 and F2 fertilizers and then was maintained steady over time, whereas β-Glu only showed a moderate increase with horn addition ( Figure 3A). In coir samples, this activity increased consistently and steadily whatever the fertilizer type ( Figure 3B). Values are means and error bars standard deviations (n = 3). Columns refer to GM types (peat, coir and bark, respectively), and rows refer to each enzyme (µmol min −1 g −1 dry mass (DM)). Each picture depicts differences in fertilizer type: Unfertilized (control without fertilizer depicted by dotted lines); F1, F2: plant-based fertilizers (dashed and dotted-dashed lines, respectively) and horn (gray lines).

Multivariate and Correlation Analyses between Nutrient Releases and Enzyme Activities
We ran a permutational analysis of variance (PERMANOVA) to statistically test and synthesize the behavior of the GM types combined with different fertilizers (Fert.) over time (Time) on pH, electrical conductivity (EC), nutrient contents, and enzymes activities. This analysis indicated a significant (p < 0·001) main effect of the three factors (Pseudo-F = 51.7, 12.2, and 4.3 for GM, Fert., and Time, respectively) and its interactions at level 2. Indeed, GM × Fert. was the most significant and interesting interaction (Pseudo-F = 2.84, p < 0.001), whereas Time was not significant when considered . Columns refer to GM types (peat, coir and bark, respectively), and rows refer to each enzyme (µmol min −1 g −1 dry mass (DM)). Each picture depicts differences in fertilizer type: Unfertilized (control without fertilizer depicted by dotted lines); F1, F2: plant-based fertilizers (dashed and dotted-dashed lines, respectively) and horn (gray lines).

Multivariate and Correlation Analyses between Nutrient Releases and Enzyme Activities
We ran a permutational analysis of variance (PERMANOVA) to statistically test and synthesize the behavior of the GM types combined with different fertilizers (Fert.) over time (Time) on pH, electrical conductivity (EC), nutrient contents, and enzymes activities. This analysis indicated a significant (p < 0·001) main effect of the three factors (Pseudo-F = 51.7, 12.2, and 4.3 for GM, Fert., and Time, respectively) and its interactions at level 2. Indeed, GM × Fert. was the most significant and interesting interaction (Pseudo-F = 2.84, p < 0.001), whereas Time was not significant when considered with other factors (Pseudo-F = 0.51, p > 0.05). Pairwise tests indicated for each GM type that horn was significantly different to F1 and F2, with F1 inducing the highest changes (p < 0.001), but it was not significantly different to F2.
Iconography of correlations on individuals depicted the contrasts between GM types ( Figure 4A): Indeed, the strongest differences were found between peat (right side) being far from coir and bark (left side). We also observed a significant effect of fertilizer type depending on the GM type ( Figure 4A,B). In bark, fertilized treatments were separated from unfertilized treatment with a gradient: Horn is the closest to unfertilized treatment and F2 is the farthest. Peat weakly responded to F1 and F2 fertilizer, and except for a few individuals, horn was hardly different from unfertilized treatment. In coir, the effect of fertilizers seemed dependent on time (even if we did not find this interaction with PERMANOVA, see above). Indeed, horn depicted several groups according to the time effect (magenta arrow) differentiating from F1 and F2 (blue and green arrows), both starting from a first group (blue circle, Figure 4B) including unfertilized and horn after 7 days. with other factors (Pseudo-F = 0.51, p > 0.05). Pairwise tests indicated for each GM type that horn was significantly different to F1 and F2, with F1 inducing the highest changes (p < 0.001), but it was not significantly different to F2. Iconography of correlations on individuals depicted the contrasts between GM types ( Figure  4A): Indeed, the strongest differences were found between peat (right side) being far from coir and bark (left side). We also observed a significant effect of fertilizer type depending on the GM type ( Figure 4A,B). In bark, fertilized treatments were separated from unfertilized treatment with a gradient: Horn is the closest to unfertilized treatment and F2 is the farthest. Peat weakly responded to F1 and F2 fertilizer, and except for a few individuals, horn was hardly different from unfertilized treatment. In coir, the effect of fertilizers seemed dependent on time (even if we did not find this interaction with PERMANOVA, see above). Indeed, horn depicted several groups according to the time effect (magenta arrow) differentiating from F1 and F2 (blue and green arrows), both starting from a first group (blue circle, Figure 4B) including unfertilized and horn after 7 days. A second iconography of correlations focusing on both independent and dependent variables showed which dependent variables were mostly influenced by GM types (threshold of 0.3, Figure  5A). Indeed, peat differentiated to other GM types by the highest acid phosphatase levels and pH (i.e., lowest levels), bark discriminated to the others by the β-Glu and acid-P as the lowest levels, and coir revealed the highest levels of aryl-S and pH. For fertilizer types, F2 showed the strongest links with SO4 2− -S content and EC (i.e., highest levels). We also detected relationships between dependent variables such as β-Glu strongly and positively related to SO4 2− -S and NH4 + -N contents, the latest being correlated to PO4 3− -P content. Aryl-S positively correlated with SO4 2− -S and pH. NO3 − -N content was also strongly related to EC and time. At a lower threshold (i.e., 0.1 depicting links with more dependency on other variables but still remarkable), more relationships appeared ( Figure 5B). Peat correlated with NH4 + -N content, while bark correlated with EC and NO3 − -N content. F2 correlated with PO4 3− -P, SO4 − -S contents, and EC. F1 was related to NH4 + -N and NO3 − -N content. Unsurprisingly, unfertilized treatment correlated with the lowest values of nutrient contents and EC. Concerning dependent variables, an iconography of correlations indicated positive relationships between Once the reader focuses on one color corresponding to a GM type ( Figure 4A), the reader can then identify fertilizer types in Figure 4B.
A second iconography of correlations focusing on both independent and dependent variables showed which dependent variables were mostly influenced by GM types (threshold of 0.3, Figure 5A). Indeed, peat differentiated to other GM types by the highest acid phosphatase levels and pH (i.e., lowest levels), bark discriminated to the others by the β-Glu and acid-P as the lowest levels, and coir revealed the highest levels of aryl-S and pH. For fertilizer types, F2 showed the strongest links with SO 4 2− -S content and EC (i.e., highest levels). We also detected relationships between dependent variables such as β-Glu strongly and positively related to SO 4 2− -S and NH 4 + -N contents, the latest being correlated to PO 4 3− -P content. Aryl-S positively correlated with SO 4 2− -S and pH. NO 3 − -N content was also strongly related to EC and time. At a lower threshold (i.e., 0.1 depicting links with more dependency on other variables but still remarkable), more relationships appeared ( Figure 5B). Peat correlated with NH 4 + -N content, while bark correlated with EC and NO 3 − -N content. nutrient contents from organic matter mineralization (i.e., NH4 + -N, PO4 3− -P, and SO4 2− -S linked all together) as well as β-Glu to other enzyme activities (urease and acid-P).

Nutrient Availability from Fertilizer Mineralization Is Driven by the GM Type
Organic fertilizer mineralization is recognized as primarily controlled by humidity and temperature that were kept constant and supposedly optimal [40,41] in our experiment. Indeed, water content between 50 and 70% of the water-holding capacity (WHC) is known to be an optimal air-water equilibrium for aerobic microbes [71] and supporting enzymatic functions providing energy and nutrients for microbial respiration and growth. In the current study, we maintained the three GM at pF 1.7 (≈60% of WHC and 45-50% WFPS) corresponding to a closely air content of 51 ± 1% of total pores spaces (v:v). Thus, water and dioxygen were assumed to be optimal, and the significance of GM impacts was mainly attributed to the biological and/or chemical properties of the GM types.
Microbes convert organic N compounds (e.g., -NH2 or -NH) into inorganic forms by ammonification releasing mostly nitrogen as NH4 + -N in solution (very little is in NH3-N form at our pH conditions). Thus, ammonification uses H + and releases OHfrom water, increasing the pH of the solution (Figure 1). Especially, peat strongly accumulated ammonium content due to weak nitrification ( Figure 2). This special link between peat and pH was indeed confirmed by the iconography of correlations ( Figure 5A,B). Then, ammonium oxidation by AOB and/or AOA and nitrite oxidation by NOB through nitrification produce two times more protons (acidifying effect) than ammonification. It has to be noticed that nitrification is thus limited by NH4 + -N content and can slow down if ammonification intensity decreases or nitrification is faster. Thus, the balance between ammonification and nitrification processes would be a decrease in pH where nitrification was strongly active such as in coir and bark mostly with F1 fertilizer and also horn at a lower intensity. On the contrary, F2 fertilizer did not sustain an increase in nitrification after 28 days in coir

Nutrient Availability from Fertilizer Mineralization Is Driven by the GM Type
Organic fertilizer mineralization is recognized as primarily controlled by humidity and temperature that were kept constant and supposedly optimal [40,41] in our experiment. Indeed, water content between 50 and 70% of the water-holding capacity (WHC) is known to be an optimal air-water equilibrium for aerobic microbes [71] and supporting enzymatic functions providing energy and nutrients for microbial respiration and growth. In the current study, we maintained the three GM at pF 1.7 (≈60% of WHC and 45-50% WFPS) corresponding to a closely air content of 51 ± 1% of total pores spaces (v:v). Thus, water and dioxygen were assumed to be optimal, and the significance of GM impacts was mainly attributed to the biological and/or chemical properties of the GM types.
Microbes convert organic N compounds (e.g., -NH 2 or -NH) into inorganic forms by ammonification releasing mostly nitrogen as NH 4 + -N in solution (very little is in NH 3 -N form at our pH conditions). Thus, ammonification uses H + and releases OHfrom water, increasing the pH of the solution (Figure 1). Especially, peat strongly accumulated ammonium content due to weak nitrification ( Figure 2). This special link between peat and pH was indeed confirmed by the iconography of correlations ( Figure 5A,B). Then, ammonium oxidation by AOB and/or AOA and nitrite oxidation by NOB through nitrification produce two times more protons (acidifying effect) than ammonification. It has to be noticed that nitrification is thus limited by NH 4 + -N content and can slow down if ammonification intensity decreases or nitrification is faster. Thus, the balance between ammonification and nitrification processes would be a decrease in pH where nitrification was strongly active such as in coir and bark mostly with F1 fertilizer and also horn at a lower intensity. On the contrary, F2 fertilizer did not sustain an increase in nitrification after 28 days in coir (ammonium availability limiting the process, Figure 2B,E) and then resulted in a neutral effect on pH, confirming a very sharp link between N cycle and pH in GM. Moreover, nitrifier's activities highly depend on pH conditions being optimal at 7.5-8 [14]. Thus, organic fertilization stimulating nitrification could in turn create acidic conditions unfavorable to their survival, if nitrate is not taken up by plant roots [15]. During the plant production phase, root uptake would produce OH − balancing acid/alkali conditions. In production, there is a need to manage N transformations and especially slowing down nitrification at an early stage of the plant growth (low nutrient demand), avoiding losses by lixiviation and/or denitrification. This can be achieved by controlling ammonification with a mixing of GM types or managing nitrifier's activities by decreasing pH (best at pH 7.5-8 ± 2 units) and thus nitrification. The burst in electrical conductivity (EC) after fertilization is classically monitored in horticulture and is considered as a risk inducing plant damages or death and can be controlled with a well-known behavior of organic fertilizer or irrigation in case of critical salinity [72]. Thus, buffering the burst of nutrient availability is very important and can strongly depend on the GM properties. The cation exchange capacity (CEC) indicates how many nutrients, mostly cations (e.g., Ca, Mg, K, Na, NH 4 , etc.), can be stored by adsorption in the GM (i.e., on the negative charges). The pH is indicative of how much the CEC is filled by these nutrients, an acidic pH revealing a non-saturated CEC where most of available exchange sites are filled with protons [73,74]. The CEC of a GM can exceed 96 cmol + kg −1 , but since CEC is expressed on a weight basis and GM have a low bulk density (more than 10 times lower than soil), the GM capacity to retain cation is low [74] (Table 1). In this study, the three GM types were similar in terms of CEC (Table 1), indicating no potential differences for buffering nutrient release, as revealed by similar changes in EC increasing after fertilization (Figure 1) driven by fertilizer type (F2 > F1 > horn). Indeed, coir and bark had saturated CEC with mostly Ca 2+ (Table S1). Only peat, presenting a non-saturated CEC (48 cmol + kg −1 and 65% saturated, Table S1) due to its very acidic pH (pH = 4), slightly attenuated the increase in EC under the influence of fertilization ( Figure 1D), the latter releasing nutrients and replacing protons on the negative charges of the solid phase. Anions such as phosphates or sulfates may rapidly be available for plant and microbes and can increase EC and salinity [74]. Our results were consistent with a high accumulation of these anions ( Figure 2), but other ions not monitored in this study can also contribute to the increase of EC. Moreover, the capacity to retain anions is very low in GM [1] generally containing little native aluminium, iron, or calcium, bridging them to the adsorbent complex [74]. It is also possible that precipitation occurs in GM at low pH (e.g., iron or aluminium with phosphates) or on the contrary under alkaline conditions (e.g., calcium with phosphates). Precipitation was probably weakly involved in nutrient availability with our range of pH ( Figure 1D-F), being optimal for nutrient availability [74].

The Role of Fertilizer's Biochemical Composition on Nutrient Availability
Usually, C:N and C:P ratios are a good predictor of N and P mineralization/immobilization [75,76]. Consequently, a faster release of N and P was expected with F1 compared to F2 (Figure 2) due to higher C:N and C:P ratios of F2 compared to F1 (Table 1). This was confirmed with coir and partially with peat. In peat, ammonification was not affected by F1 and F2 chemistry (Figure 2A), but in coir, F2 seemed to slow down ammonification rates after 14 days (F1 > F2; Figure 2B) with a direct negative feedback on nitrate production. In bark, only the C:N ratio seemed to induce feedback control with faster nitrification with F1 (not visible with ammonification near 0).
We showed in the current study that horn is a very interesting organic fertilizer in terms of nutrient supply with no burst effect and the response intensity depends, as hypothesized, on the GM type. Despite the lowest C:N and C:P ratios, horn induced the weakest releases of nutrient as hypothesized ( Figure 2) mainly due to its physical (high density material and hard fibers) and chemical recalcitrance necessitating the protein depolymerization of keratin (i.e., proteins) (Table S1) by microbes [77][78][79]. Horn does not only contain nitrogen, as frequently stated, but it does contain other nutrients necessary for plants and microbes such as P, among many others (Table 1; Table S1) with typical dynamics depending on the GM type ( Figure 2). Thus, its mineralization needs to be considered into organic fertilizer formulations and should be added as a slow N release compared to granular.
The two granular fertilizers increased EC much higher than horn, but overall, they were only barely reaching the second class of risk in salinity according to Durand's classification [80]. Basil was successfully grown in the similar conditions (unpublished data), indicating that the dose of fertilizer used in this study, on a professional recommendation basis (N = 300 mg L −1 of GM), was well established (first negative effects were observed from 600 mg L −1 and mortality above 900 mg L −1 , unpublished data). Thus, the nutrient content ratios and biochemistry of fertilizers are important to consider as long as the GM type modulates the microbial response to fertilization.

Enzyme Activities Are Driven by the GM Type
When organic resources are added to a growing media usually limited by nutrients, one can expect that microbes would transform it in available nutrients for their metabolism. However, an elemental composition of organic resources rarely meets the elemental demand of microbes. Thus, microbes can produce specific extracellular enzymes to achieve its resource demand and regulate their element use efficiency by releasing nutrients in excess [32] that benefit other organisms such as plants. Thus, the stoichiometry of organic resources should be considered regarding the stoichiometry of microbial biomass (C:N:P around 42:6:1 and 60: 7:1) [81,82] constrained by a highly stressful environment.
C-to-nutrient (N, P or S) ratios of GM types (Table 1 and Table S1) are indeed much higher compared to microbial stoichiometry [36] and in our study particularly regarding P (C:P ratio of 2716, 1512, and 608 for peat, coir, and bark, respectively). According to evolutionary-economic theory, investment (i.e., nutrient allocation and energy) into the enzyme synthesis of microbes is controlled by targeted nutrient availability and should be directed toward mining for scarce elements [83]. In the nutrient-limited conditions of GM, we showed a strong effect of the organic fertilizer directly linked to C:N:P:S ratios and biochemistry (Table 1 and Table S1) on microbial enzyme activities [84] depending on the GM type. Fertilizer addition can increase labile organic carbon stimulating microbial growth and activities [85]. Enzyme activities involved in C, P, or S acquisition increased in coir (only partially in peat) after organic fertilizer addition, following economic theory [86]. This would imply that enzyme production is induced to reach greater resource acquisition. Organic fertilization stimulated the 1.4-β-glucosidase activity, NH 4 + -N, PO 4 3− -P, and SO 4 2− -S releases and were remarkably linked ( Figure 5A) together after F1 and F2 addition, specifically in peat and coir (Figure 2A,B,J,K). Especially, the increase of N availability is known to stimulate C-acquiring enzymes as 1.4-β-glucosidase activity [86,87]. However, enzyme activities did not systematically increase with organic fertilization (no increase at all for bark), but nutrient content increased, thus making it difficult to link enzymes to nutrient availability, as also recently reported by Guerra et al. (2018) [88] in a mix of coir fiber and composts. If enzyme substrates are not available, microbes do not strongly regulate enzyme production, or enzyme costs are low enough to allow continuous production [86]. This would corroborate more complex relationships and that numerous types of enzymes and different regulation pathways are involved [35,86]. These complex relationships could be analyzed by an iconography of correlations that consider the links between all variables without simplifying the information (Figures 4 and 5).

Remarkable Relationships between Enzyme Activities and Nutrient Availabilities in Each Growing Media
In our study, microbial enzyme activities were greater related to GM type and/or nutrient availabilities (especially NH 4 + -N and SO 4 2− -S) than fertilizer type itself, as shown by the iconography of correlation (Figures 4 and 5). One exception concerned urease activity. High available N can repress urease production and thus the activity as reported previously [89,90] and could explain the low sensitivity of this measure in our study for all the GM types. Peat was especially linked to acid phosphatase activity highlighted by the iconography of correlations ( Figure 5A). Margalef et al. (2017) [91] reported in a meta-analysis that phosphatase activities were better predicted by N availability rather than P, reflecting an increase in P needs (i.e., allocation of N toward P-acquiring enzyme). After N fertilization, P can become limiting for microbial growth [91]. The high acid phosphatase activity over the experiment ( Figure 3G) could be inherited from peat bog presenting a huge C:P ratio (Table 1). Indeed, peatland is renowned for high phosphatase activity (0.78 µmol min −1 g −1 on average), indicating P rather than N limitations under these conditions [91]. This high level of activity might have been sufficient in our experiment to ensure complex molecule decomposition and P mineralization in a peat sample. Moreover, a microbial need for C might have exceeded the need for P, as suggested by increasing P availability throughout the experiment, which was previously reported for P-poor soil [92]. Then, the resulting high availability of P after 14 days with F1 and F2 could have repressed acid phosphatase activity, explaining the decreasing trend over time except with horn ( Figure 3G). In this latter case, the slower release of PO 4 3− -P could explain the steady level of this activity until the end of incubation ( Figure 3G).
Bark was negatively correlated with enzyme activities ( Figure 5A,B), being low but detectable and steady throughout the experiment ( Figure 3C,F,I,L). These weak activities and investment in enzyme productions in bark samples might indicate the presence of labile organic compounds, containing the carbon and nutrients need for microbial growth [90]. Indeed, the synthesis of enzymes represents an important investment in energy for microbes and might not be always necessary [86]. Composted barks rich in polysaccharides as well as free mono and disaccharides favored an active, indigenous, and varied population of microorganisms [1,11], which are even active without any fertilization (unpublished data). Carlile and Dickinson (2004) [26] indicated a higher dehydrogenase activity and basal respiration compared to peat. Thus, microbes in bark might be more initially limited by nutrients than C, as revealed by low NH 4 + -N (limiting nitrification) and PO 4 3− -P ( Figure 3C,I,K). Microbes are also prone to store in their biomass limiting elements [93], adjusting their own stoichiometric ratios. For instance, C can be stored as lipids and glucan [32], and P can be immobilized as polyphosphates in bacteria cells [94,95]. The immobilization of S in microbial biomass [96] might also explain lower S availability in bark ( Figure 3K). A shift in the microbial community structure (and thus biomass stoichiometry) may also generate stoichiometric plasticity to microbial communities and adaptivity to resource type [32]. Organic fertilizer mineralization that occurred (i.e., increasing NO 3 − -N and SO 4 2− -S) might also be explained by other pathways involving different enzymes than the ones investigated here [35]. In coir, weak initial available nutrients and enzyme activities could reveal C and nutrient limitation for microbes and a need to invest into enzyme production [86], which benefits the outweighed costs in C and N necessary for protein synthesis [32]. This was consistent with the slower N and P mineralization that we observed in this GM. Furthermore, the iconography of correlations depicted an interaction between fertilizer type and time ( Figure 4A,B), reflecting dynamics of enzyme and nutrient releases as dependent on fertilizer type (i.e., slower with horn vs. F1 and F2). A remarkable relationship with arylsulfatase activity was highlighted by the iconography of correlation ( Figure 5A), its dynamics differing strongly in coir compared to other GM ( Figure 3J,K,L). Arylsulfatase production is affected by microbial composition and is not necessarily dependent on available S. Some microbes can produce arylsulfatase independently of S availability, while other sulfate-starving microbes can successively induce their production and then repress it when S availability is high [97]. In our study, we could suspect the latter case of a repressive effect of sulfate on this activity in coir samples, where we discovered a decrease in this activity after 7 days ( Figure 3K) subsequently due to the high release of sulfate for F1 and F2 ( Figure 2K). The decrease in pH might also be a factor of influence ( Figure 5A) as depicted by a positive correlation between pH and this activity as also previously reported for soil [98,99].

Conclusions
Our study provides evidence that GM chemical and biological properties drive N mineralization and nitrification after organic fertilization and consequently the amount of nutrients and their forms potentially available for plant uptake. Ammonification is not a limited process in our GM types, but nitrification is limited in peat because of the acidic pH. The microbial communities of each GM responded differently to organic fertilization expressed by both different enzymatic strategies and nutrient release patterns. Although we expected a strong relationship between enzyme activity and the related nutrient availability, our results suggested that enzyme activities in GM might not be strictly stimulated or repressed by nutrient availability but can also result from a constitutive production or a decoupled regulation. An iconography of correlations allowed us to highlight the particularity of each GM compared to the others. Peat was specifically related to ammonium accumulation due to weak nitrification and high acid phosphatase activity, which was probably inherited from peat bog and a strong C:P ratio. On the other hand, bark presented weak enzyme activities, but a strong nitrification capacity; however, this was limited by the ammonium content. Coir had an intermediate profile regarding mineralization processes but also showed a repressive arylsulfatase activity due to high sulfate content and a decrease in pH. Mixing different GM types seems a promising way to optimize microbial functions and thus N-P-S availabilities. This hypothesis needs to be tested, as the physicochemical properties will change in mixes, affecting microbial communities and functions.
Furthermore, fertilizer chemistry modulated mineralization rates and influenced enzyme strategy at some points. Iconography highlighted especially the high release of anions and high EC induced by F2 addition. F1 seemed to provide more suitable conditions for N release as ammonium in peat, as nitrate in bark, and both in coir. Overall, N being the most limiting element in GM, its status has to be managed carefully. Indeed, mining for N can induce a risk of salt toxicity for roots due to an excessive release of other nutrients. Further studies on microbial functions related to N and C cycles (and their interactions) with a wider range of N-and C-related enzymes (e.g., proteases) as well as microbial catabolism (i.e., actual activity) would provide additional clues for better understanding nutrient mineralization in GM.

Supplementary Materials:
The following are available online at http://www.mdpi.com/2073-4395/10/12/1955/s1, Table S1: Additional chemical information on growing media and fertilizers. Table S2: Chemical properties on growing media on volume basis. Figure S1: Effect of growing media and fertilizer types on nutrient availability over time as expressed on a volume basis (mg N, P, or S per L DM). Figure S2: Effect of growing media and fertilizer types over time on nitrogen, phosphorus, and sulfur availability as expressed as a percentage of respective supply (% N, P, S added). Figure S3: Effect of growing media and fertilizer types on enzyme 1.4-β-glucosidase (β-Glu), urease, acid phosphatase (Acid-P), and arylsulfatase (Aryl-S) activities, over time, as expressed on volume basis (µmol min −1 L −1 DM).