Introduction

Soils are not only providing the largest terrestrial carbon (C) pool (Lal 2008), soil CO2 fluxes and belowground C allocation are also major pathways in the global C cycle (Ryan and Law 2005). Soil respiration (microbial & root) is responsible for most soil C losses and largely controls ecosystem C sequestration (Schlesinger 1997). In turn, it strongly depends on energy inputs from aboveground C sources. The major interface of C input and turnover is the rhizosphere, where soil microorganisms rapidly decompose rhizodeposits (Hütsch et al. 2002; Gunina and Kuzyakov 2015). In contrast to the bulk soil, the rhizosphere is characterized by intensified C and nutrient turnover with high spatial and temporal heterogeneity along and across the root (Oburger and Schmidt 2015; Kuzyakov and Razavi 2019). Recent application of imaging approaches have greatly improved our understanding of this small-scale heterogeneity, and established the term ‘rhizosphere hotspots’ for particularly active areas (Kuzyakov and Blagodatskaya 2015; Kuzyakov and Razavi 2019). Despite often assumed, the direct relationship between these small-scale hotspots and soil C balances on the next higher scale level (i.e. plant individual) is still unknown.

Rhizosphere properties are controlled by root activity and by rhizodeposition (Jones et al. 2009). The size and extension of the rhizosphere have been well described (Kuzyakov and Razavi 2019). Low molecular root exudates are mainly released from the root tip and their concentration is highest within around 2 mm distance from the root surface (Holz et al. 2017; Kuzyakov and Razavi 2019). As they serve as easily available energy sources for microorganisms (Gunina and Kuzyakov 2015) and are linked to root growth (Holz et al. 2017, 2018), the occurrence of low molecular root exudates is likely associated with high CO2 production. Usually, glucose is used as a model substance to simulate root exudation, comonly increasing CO2 efflux rates after application (Blagodatskaya et al. 2011; Mason-Jones et al. 2018). This has established the assumption, that increased rhizodeposition generally results in higher microbial activity and soil respiration (Phillips et al. 2011). However, exudates not only consist of monosaccharides but include a broad variety of monomeric and polymeric substances. Chemically more complex root exudates (e.g. mucilage) and root necromass, are made bioavailable by microbially produced exoenzymes which break down polymeric components, such as cellulose, hemicellulose and other polysaccharides (Allison et al. 2011). Previous studies have reported positive relationships between enzyme activities and bulk soil CO2 efflux (Phillips et al. 2011; Mancinelli et al. 2013). Therefore, it is assumed that this translates to spatially resolved patterns in the rhizosphere. Those areas with high activity of enzymes mediating SOC decomposting (i.e., enzyme hotspots) might also indicate areas with high CO2 production.

Accordingly, cocentrations of rhizodeposits and microbial biomass decrease from the root towards the bulk soil (Schenck zu Schweinsberg-Mickan et al. 2010; Holz et al. 2019; Kuzyakov and Razavi 2019). Major C and N turnover processes were found to be significantly higher in rhizosphere soil compared to the bulk soil (Finzi et al. 2015). However, such findings do not consider the spatial complexity within the rhizospere, and do not allow for conclusion on the effect of rhizosphere processes on CO2 efflux because both measurements are commonly not coupled. Quantifying CO2 efflux alone does not account for process dynamics whithin the rhizosphere hotspots, while the quantification of rhizosphere hotspots alone might not translate to CO2 balances on larger scale levels.

Assessing the importance of rhizosphere-imaging derived information for the C balances in plant-soil interactions, requires the application of flux measurements and imaging methods in unison. To bridge this gap between small-scale rhizosphere processes and individual plant-soil interaction, we applied different imaging methods on the root area of maize (Zea mays) plants grown in sandy arable soil. Root exudation was visualized by 14C-imaging (Pausch and Kuzyakov 2011; Holz et al. 2017), and β-glucosidase activity was imaged by soil zymography (Spohn et al. 2013; Razavi et al. 2016). The cumulative 14CO2 and total CO2 efflux rates were quantified over 1 day before imaging.

Our objectives were to assess (1) whether visibile hotspots of root exudates and enzyme activity are a dominant scource and thus suitable to predict total soil C respiration, and (2) whether it is possible to estimate the contribution of different rhizophsere processes to C respiration by coupling imaging and CO2 flux measurents. We hypothesize that the intensity of root exudate and enzyme hotspots coincide with high CO2 fluxes and that hotspots of rhizosphere activities are better predictors of CO2 efflux than bulk soil or plant related variables alone.

Material and methods

Experimental setup

Rhizoboxes with an inner size of 20x40x1 cm were filled with an arable soil (each about 1 kg soil with a bulk density of 1.41) collected from a field close to Göttingen, Germany. All boxes were treated the same to avoid clustering or covariate effects of confounding factors. Soil organic carbon content was 20.0 g kg−1, total nitrogen was 1.7 g kg−1 and the pHH2O was 4.9. Soil particle size was distributed as follows: Clay: 8.6%, silt: 18.5%, sand: 73% (Holz et al. 2017). Maize seeds (Zea mays - KWS 2376) were immersed in a 10% H2O2 solution for 10 min to avoid seed-borne diseases before germination. The seedlings were placed on a filter paper and were germinated in a petri dish 3 days in the dark. After 3 days, one seedling was planted per rhizobox (n = 17). During plant growth, the boxes were inclined by 55° to make the roots grow along the transparent front cover (Fig. 1a). This is relevant for measuring the distribution of enzymes around the roots during zymography and for 14C imaging of roots and root exudates. The plants were kept in a climate chamber during the total growth period and watered daily to maintain a volumetric water content of 20–23%. The temperature was 25 °C during the day, 22 °C during the night, the photoperiod was 14 h, and the light intensity was 300 μmol m−2 s−1. These growth conditions were used as they represent optimal growth conditions for Zea mays.

Fig. 1
figure 1

Exemplary images of (a) maize roots in an open rhizobox, (b) distribution of 14C within the maize root system and in (c) root exudates imaged by phosphor imaging and (d) distribution of β-glucosidase activity in the same sample

Full size image

14C labelling and CO2 measurements

After four weeks of plant growth each plant was labelled with 0.5 MBq 14CO2 (specific activity of 59.6 mCi mmol C−1) for 5 h. Before starting the labelling, rhizoboxes were opened and a moist filter paper (Whatman, 1001–917, 11 μm) was attached in order to capture exudates from the exposed root surface. After closing the rhizoboxes again, each rhizobox was packed in a plastic bag, which was closed with modelling clay at the lower part of the stem of the plants. Inside the bag, a 20 mL 1 M NaOH trap was placed to trap the 14CO2 released from soil. For 14C labelling, plants were placed in an acrylic glass chamber with a size of 0.6 × 0.6 × 0.8 m. A ventilating fan was used to distribute the 14CO2 homogenously within the chamber. The label (Na214CO3 dissolved in 1 M NaOH) was placed into a glass vial, which was connected with the chamber by plastic tubes. After adding phosphoric acid (50%) to the label, 14CO2 was released and pumped into the chamber where it was circulated for 5 h. 24 h after beginning of labelling the plastic bags were opened and the NaOH traps were removed for determination of total CO2 as well as 14CO2. The rhizoboxes were opened and the moist filter paper was carefully removed from the rhizobox surface and oven dried at 60 °C to avoid microbial decomposition of root exudates collected in the filter paper. Imaging plates (Storage phosphor screen, BAS-IP MS 2040 E, VWR, 20 × 40 cm) were placed on the filter paper, which was covered by a thin plastic film for 14 h. After this, the screens were scanned (FLA 5100 scanner, Fujifilm) with a spatial resolution of 50 μm to visualize the distribution of 14C in root exudates (hereafter referred to as “Rhizosphere 14C hotspots” – Fig. 1b). After removal of the filter paper, the rhizobox surface was covered with a thin plastic film and a storage phosphor screen was placed on each rhizobox surface for 14 h. After this, the samples were scanned as described above to visualize 14C allocated to roots and root exudates (hereafter referred to as “Total 14C hotspots” – Fig. 1c).

To determine 14C-CO2 in the NaOH traps a scintillation cocktail (Eco Plus) was added to NaOH with the ratio of 4:1. 14C activity in NaOH was determined using a liquid scintillation counter (Hidex, 300 SL). Total CO2 respiration from soil was measured from a subsample (2 ml) of the NaOH trap by a TOC analyzer (TOC 5050, Shimadzu Corporation, Kyoto, Japan).

Soil Zymography

The distribution of β-glucosidase activity was determined by soil zymography. The measurements were conducted according to Spohn and Kuzyakov (2013) but without a agarose gel as in Razavi et al. (2016). Briefly, polyamide membrane filters (Tao Yuan, China) with a pore size of 0.45 μm were cut into 20 × 40 cm pieces. Each membrane filter was placed in a solution of 20 mL of water containing 6 mg 4-Methylumbelliferyl-b-D-glucoside (MUF-G) which is a substrate for β-glucosidase. After hydrolysis of the substrate, fluorescence of MUF can be measured. The membrane was soaked with the solution and was then attached to the rhizobox surface for 60 min. Subsequently, the membrane was removed from the soil surface and the remaining soil was carefully removed with a small brush. The membranes were placed under UV light, and pictures were taken at 360 nm wavelength (Fig. 1d) with a digital camera. In addition to the actual membrane, a small piece of membrane (1.5 × 6 cm) soaked with water instead of substrate solution was also imaged to control for the background signal. The final images had a spatial resolution of 40 μm per pixel.

For calibration, solutions with increasing 4-methylumbelliferone (MUF) concentrations were prepared: 0, 386, 967, 1934, 2901 and 3868 pmol cm−2 h−1. Pieces of membranes (4 × 4 cm) were cut, soaked with the solution, and imaged as described for the soil zymography. From the amount of solution taken up per cm of membrane and the MUF concentration of the solution, the concentration of MUF per area was calculated. For the calculation of the β-glucosidase activity, the amount of MUF per area was divided by the incubation time. The equation obtained from the calibration was applied to all images to convert the grey value to β-glucosidase activity. The signal of the control membrane was subtracted from the image to remove the background signal.

Plant harvest

After zymography measurements, plant shoots were cut and dried at 40 °C. The roots were taken out of the soil and the soil attached to the roots after being gently shaken (rhizosheath) was collected by carefully washing it from the roots. Roots were dried at 40 °C. Rhizosphere soil and bulk soil (the soil not adhering to the roots) were freeze dried to avoid microbial degradation of labile carbon compounds. To measure 14C activity, ground shoots, roots, rhizosheath and bulk soil samples were combusted in an Oxymat OX500. The released 14CO2 was captured in a scintillation cocktail (C400, Zinsser) and quantified using a liquid scintillation analyzer (Tricarb, 3180, PerkinElmer). Shoot and root biomass was determined gravimetrically.

Image analysis

For quantification of 14C in images, the images were converted from a log into a linear system by applying the following equation:

$$ PSL={\left(\frac{Res}{100}\right)}^2\ast \frac{4000}{S}\ast {10}^{\left(\frac{QL}{G}-\frac{1}{2}\right)} $$
(1)

Where PSL (photo stimulated luminescence) is the quantified value of the image in linear scale, Res is the resolution of the image in μm (Res = 50 μm), S is the sensitivity (S = 5000), L is the latitude (L = 5) and G is the gradation (G = 65,535). After conversion of the images, the background noise was removed: The part of the image where the screen was not in contact with the sample was selected and subtracted from the part of the image where the root system was visible to remove the background noise. Rhizosphere 14C hotspots as well as total 14C hotspots were identified based on the contrast between regions of high 14C activity and regions with low 14C activity and based on the longitudinal shape of the roots using the program Roottracker2D. A detailed description of the approach can be found in Menon et al. (2007). The hotspot area and intensity were calculated using MATLAB (The MathWorks). Enzyme hotspots were also identified using the program Roottracker2D (Menon et al. 2007). For an objective selection of rhizosphere hotspots, rhizosphere extension was specified to the location where phosphatase activity had decreased to 5% of its maximum activity at the root center (Holz et al. 2020). As for 14C imaging, hotspot area and intensity were calculated using MATLAB (The MathWorks).

Statistics

All data were checked for normal distribution and consistency, and measures of central tendency and dispersion were calculated (Supplementary Table 1). Variable interrelations were assessed by Pearson correlation and visualized by multidimensional scaling of the correlationmatrix (network plot), where stronger correlated variables are ordinated closer to each other. Single effects of predictor variables (biomass and imaging parameters) on soil respiration (CO2 and 14CO2 efflux) were univariately assessed by simple linear regression. Residual diagnostics were conducted visually, as well as by using Shapiro-Wilk test and Cook’s distance = 0.5. For all analyses the significance level was defined as p < 0.05.

We used random forest models (RF) to preselect explanatory variables for multiple linear regression. Variables with no, or negative % increase in mean square error were excluded. To avoid excess degrees of freedom due to missing values, 14C activity in the rhizosheath was tested and compared iteratively. Variables in each multiple linear regression model were checked for variance inflation (VIC) and stepwise removed if VIC > 10. The akaike information criterion (AIC) was used to select the best performing multiple linear model. Models with better AIC values were only considered when significant (p < 0.05). The final model was visualized by partial regression plots, each showing the effect of one predictor variable on the response variable while other variable effects were kept constant.

All statistical analyses were conducted in R v3.6.1 (R Core Team 2017) using ‘randomForest’ package for building random forest model (Liaw & Wiener 2002), ‘corrr’ package for correlation and network analyses (Kuhn et al. 2020), and ‘ggplot2’ for data visualization (Wickham 2016).

Results

Data verification and range

Exept for shoot biomass, all predictor and explanatory variables followed a normal distribution among the boxes (Supplementary Fig. 1), and no data annomalies, such as outliers or clustering, occured. The maize plants reached an average shoot drymass of 1.6 g (Supplementary Table 1). Their average root biomass was 0.67 g, ranging between 0.2 and 1.4 g, with a visible surface area between 0.8 and 8.2% of the whole rhizobox surface. From these boxes, we measured total CO2-efflux rates between 2 and 5 mg kg−1 soil d−1 (Supplementary Table 1). The cummulative 14CO2 activity after 24 h ranged between 15 and 53 kBq. The 14C-imaging provided a data range of 34,817 to 3,238,628 PSL for the total hotspots, and 25,537 to 145,708 PSL for the rhizosphere hotspots alone. The β-glucosidase activities in bulk soil ranged between 556 and 2541 μmol h−1, while rhizosphere specific activities ranged between 5.16 and 114.93 μmol h−1. The average areas of rhizosphere and bulk soil were 4.29% and 95.71% respectively.

The response variables (CO2 and 14CO2 efflux) were positively correlated to each other (r = 0.41, p = 0.02). Between the explanatory variables, significant interactions were detected (Fig. 2): The shoot and root biomass were directly related (r = 0.80, p = 0.001) and root biomass weakly correlated with total 14C hotspots (r = 0.46, p = 0.06). The total number of 14C hotspots strongly correlated with rhizosphere 14C hotspots (r = 0.72, p = 0.002). No interactions were found between biomass variables and rhizosphere 14C hotspots as well as enzyme activities. Correlations between 14C-imaging variables and zymography variables were weak or completely absent.

Fig. 2
figure 2

Network plot of Pearsson correlations. Stronger correlations between variables are indicated by stronger paths (red = negative, blue = positive), and by appearing closer to each other (distances are determined by multidimensional scaling). Only correlations with |r| > 0.4 are shown and significance levels are indicated by * = 0.05, ** =0.01 and *** = 0.001

Full size image

Biomass related variables (i.e. root surface area, shoot and root biomass) showed strong correlation between each other, but where weakly related to CO2 flux and imaging variables (Fig. 2). Zymography derived variables clustered with total CO2 efflux, while 14C imaging variables clustered with 14CO2 efflux.

Variable effects on total CO2 and 14CO2 efflux

Zymography showed a clear negative effect of β-glucosidase activitiy on total CO2 efflux (Fig. 3a, b). Bulk and rhizosphere enzyme activity each explained more than 50% of the total CO2 efflux variance. The 14C-imaging hotspots were not related to the total soil respiration, neither in the whole soil, nor in the rhizosphere specific area (Fig. 3c, d). The total CO2 efflux was also independent from root surface area and biomass (Fig. 3e, f), although the latter showed a positive tendency (p = 0.08). In contrast to total CO2 efflux, the measured 14CO2 efflux was completely independent from zymography variables (Fig. 4a, b). The number of 14C activities in the total root zone and in the rhizosphere, however, were positively related to 14CO2 efflux and explained 33% and 26% of the response variance, respectively (Fig. 4c, d). Yet again, the root biomass and root surface area had no effect on the efflux rate (Fig. 4e, f). The 14C activity of root-attached soil was unrelated to CO2 or 14CO2 efflux (Supplementary Fig. 2).

Fig. 3
figure 3

Effect of β-glucosidase activities in bulk soil (a) and rhizosphere (b), 14C-hotspots (c, d), and belowground-biomass variables (e, f) on total CO2 efflux over 24 h. Significance level (p), coefficient of determination (R2) and area of 95% confidence (grey) for simple linear regression are annotated

Full size image
Fig. 4
figure 4

Effect of β-glucosidase activities in bulk soil (a) and rhizosphere (b), 14C-hotspots (c, d), and belowground-biomass variables (e, f) on 14CO2 efflux. Significance level (p), coefficient of determination (R2) and area of 95% confidence (grey) for simple linear regression are annotated

Full size image

Variable importance

Root surface area, 14C activity of root-attached soil and shoot biomass showed either negative or no effects on RF predictive accuracy for total CO2 as well as 14CO2 fluxes and were excluded from further analyes. This equally applied for root biomass in 14CO2 predictions.

The best performing models for predicting total CO2 and 14CO2 efflux from imaging methods alone, included enzyme acitivtiy and total 14C hotspots. This variable combination slightly increased the explained variance of total CO2 efflux from R2adj = 0.50 (AIC = 25.9) to R2adj = 0.56 (AIC = 25.1) compared to the best performing single explanatory variable (i.e. total enzyme activity). Adding root biomass as a 3rd predictor for total CO2 efflux, strongly increased model performance (R2adj = 0.75, AIC: 18.8) (Fig. 5). The 14CO2 efflux explained by rhizosphere-enzyme acitivtiy and total 14C hotspots was R2adj = 0.65 (AIC: 82.0), compared to R2adj = 0.29 by total 14C hotspots alone (AIC = 124.0). Combinig enzyme variables to explain total CO2, and 14C hotspots to explain 14CO2 efflux, lead to lower R2adj values or non-significance due to multicolliniarity effects, respectively.

Fig. 5
figure 5

Partial regression plot for total CO2 efflux explained by bulk β-glucosidase activity, rhizosphere 14C activity and root biomass. Each trenline shows an explanatory variable effect when other variables are kept constant

Full size image

Discussion

There is an emerging view that plants, and in particular the allocation of C to roots and rhizosphere, strongly control soil C cycling (Kuzyakov and Cheng 2001; Hütsch et al. 2002; Jones and Hinsinger 2008). Rhizosphere-imaging approaches are powerful tools for spatially allocating hotspots of the major controlling processes (Oburger and Schmidt 2015; Roose et al. 2016). Yet, the overall importance of these hotspots for larger-scale C balances remains unclear. We used a combination of imaging approaches as proxies for root exudation and exo-enzymatic C turnover, to bridge the gap between rhizosphere-scale hotspots and soil C respiration on an individual plant scale.

Bulk soil β-glucosidase activities in combination with rhizosphere 14C activity and root biomass, explained up to 75% of total CO2 efflux variance. This indicates that combining imaging hotspots of root exudation with hotspots of β-glucosidase activity allows to roughly predict soil respiration rates. Complementing these variables with root biomass measurements accounts for three-dimensional variability in the rhizosphere which may not be well depicted by 2D images. However, other variables – such as rhizosheath 14C-activity – did not seem to fulfil the same purpose.

In our study, root exudates were visualized through increased 14C-activity outside the root specific area. These exudates serve as easily available energy sources for microorganisms in the rhizosphere and are therefore expected to enhance SOM turnover and CO2 efflux (Kuzyakov and Cheng 2001; Hütsch et al. 2002; Gunina and Kuzyakov 2015). This effect has been frequently reported, and is directly related to microbial activity and thus microbial respiration. Furthermore, root exudation can be linked to root growth, i.e. is a proxy for fine root activity and root respiration (Sun et al. 2017). Both of these CO2 sources can explain why increasing rhizosphere 14C activity (i.e., increased root exudation) directly led to higher total soil respiration rates, when other variables were kept constant (Fig. 5).

In contrast to rhizosphere 14C activity, β-glucosidase activities were negatively related to total CO2 efflux and were the only variables to show significant univariate relationships. Given the equally negative correlations between β-glucosidase and 14C activities, this suggests that high root exudation leads to reduced enzyme activity, counterbalancing effects on total soil respiration. This contradicts our initial expectation, considering that β-glucosidase activity promotes SOM break down in bulk soil (Wick et al. 2002), and also contradicts the observation in previous studies where β-glucosidase activity was positively related to CO2 efflux (Gispert et al. 2013; Liang et al. 2015). However, Gispert et al. (2013) also reported a negative relationship between β-glucosidase activity and CO2/SOC ratio, indicating that these positive correlations may actually be artifacts of SOC to CO2 relationships. Since our study design (mostly) eliminates the impact of different bulk soil SOC contents, the negative β-glucosidase – CO2 relationships might instead be a result of soil microbial reaction to substrate availability. Microbes predominantly invest in producing exo-enzymes for OM decomposition when labile C sources (such as root exudates) are scarce, leading to higher β-glucosidase activity and lower bacterial growth rate (del Giorgio and Cole 1998). Consequently, we found lower β-glucosidase activity in the presence of root exuates (i.e. rhizosphere 14C activity) and a stronger relation between bulk β-glucosidase activity and CO2 than between rhizosphere β-glucosidase activity and CO2. In turn, this suggests that zymography β-glucosidase hotspots might be associated with turnover of native SOC rather than plant derived C. This is supported by the strong relationships between β-glucosidase hotspots and total CO2 efflux. In contrast, 14C imaging variables were closely related to 14CO2 efflux, emphasizing their importance for root, as well as rhizo-microbial respiration (de Vries et al. 2019).

Those variables that are directly dependent on the presence of plant roots (i.e. root surface area and root biomass) where weakly correlated to total CO2 and 14CO2 efflux, as well as to all imaging variables (Fig. 2). Because root exudation strongly varies along the root axis, with maximum intensities directly at the tip and shortly behind the tip (Dennis et al. 2010; Holz et al. 2017), exudation effect on CO2 fluxes and related imaging variables are presumably decoupled from root biomass and root surface area. However, using solely imaging-derived variables to explain total soil respiration provided medium correlation strength and predictive power of simple, as well as multiple regression models, with just about half the variance explained. The inclusion of root biomass significantly increased the predictive power of the model. This might also indicate that a mediating factor is required to translate results from 2D-imaging methods to processes on a 3D-scale.

Conclusions

In this study, we used imaging methods to relate hotspots of root exudation and enzyme activity to total soil respiration on an individual plant scale. We conclude that these hotspots are reasonable predictors for soil CO2 efflux and perform better than static variables, such as root biomass alone. Soil respiration was mainly related to bulk enzyme activity (i.e., SOC turnover), while 14C-hotspots mainly accounted for plant derived and rhizosphere 14CO2 efflux. However, the relationships between hotspots and total soil respiration contain large uncertainties, and a combination of methods is required to achieve a meaningful representation of related processes. Further research is needed to evaluate these results concerning type, function and distribution of soil and rhizosphere hotspots as well as relationships to nutrient fluxes and root morphology. Increasing the range of values by including treatments, such as plant varieties differing in root exudation (Tawaraya et al. 2013; Holz et al. 2017) or fertilization levels, may also lead to stronger statistical relationships.