This study was conducted in the Breuil-Chenue experimental forest site located in the Morvan (47˚18'N, 4˚5'E, France). The forest is situated on a plateau at an altitude of 638 m . The native forest was clear-felled and replaced in 1976 by monospecific plantations distributed in plots of 0.1 ha of various species such as oak, beech (Fagus sylvatica L.), Norway spruce, and Douglas-fir (Pseudotsuga menziesii [Mirb.] Franco). The soil derives from the “Pierre qui Vire” granite that contains quartz (34.0%), albite (31.1%), K-feldspar (24.2%), muscovite (8.9%), biotite (1.2%), and chlorite (0.5%). The soil has been classified as a Typic Dystrochrept, according to [34] . The soil characteristics of the native forest are described in detail by [35] . The bulk soil has a sandy-loam texture (55% sand and less than 20% clay). The soil is well-aerated and acid (pH_KCl 3.1 - 4.3). The soil cation exchange capacity ranges from 9.2 to 2.7 cmolc∙kg⁻¹ with a base saturation lower than 10%. Humus is present as a moder in the native forest and carbon concentration is 7.3% in the A1 horizon.

Soil samples were collected under beech and Norway spruce stand in November 2007 (autumn), February 2008 (winter) and May 2008 (spring). Soil samples were also collected during summer (July) but the quantity of water extracted by centrifugation from the soils at this period was insufficient for analyses. For each season and both stands, soil samplings were carried out in four replicates in independent plots (about 10 m distance between each soil sample) from pits of 120 × 80 cm . After removing the forest floor, soil samplings were carried out systematically at three depths, i.e., 0 - 3 cm (layer I), 3 - 10 cm (layer II), and 10 - 23 cm (layer III). The choice of 0 - 3, 3 - 10 and 10 - 23 cm was motivated by the fact that these depths correspond approximately to the limits of two distinct and homogeneous soil horizons (from organo-mineral to mineral soil horizons) in both Norway spruce and beech stands [4] . The separation of soil samples into bulk soil and rhizosphere fractions was conducted in the field. In each horizon, soil material was cut and extracted from the profile. At the site, living roots with diameters < 2 mm were carefully removed by hand from each soil layer. The soil without roots was collected to give the bulk soil. Soil aggregates > 1 cm adherent to the roots were removed. The rhizosphere fraction was obtained by gently shaking fresh roots. Bulk soil and rhizosphere were sieved at 200 μm to eliminate the roots and to obtain a comparable particle size distribution in both compartments, and homogenized. These soil samples were placed in air-tight bags and stored at 4˚C for 24 to 48 h. The solutions of the bulk soil and rhizosphere were extracted by centrifugation (15˚C, 20 min, 3000 rpm; JOUAN KR422) for the 3 months (November, February and May), the three soil horizons (0 - 3, 3 - 10, and 10 - 23 cm ) and the two tree species (beech, Norway spruce). Each extract was filtered with a pre-rinsed, 0.45 µm pore diameter, cellulose nitrate filter, and stored at 4˚C. The solutions collected through this extraction protocol correspond to capillary plus gravitational solutions. This method was used because the volume of solution extracted from the rhizosphere would have been insufficient for analysis if we had separated the two types of solution.

The pH of the soil solutions was determined (pHmeter SENTRON, Argus X). Total carbon and nitrogen in soil solutions were estimated using a TOC analyzer (TOC-5050, Shimadzu). The concentration of major cations (K, Ca, Mg, Na, Fe and Mn) and trace elements (Ba, Cd, Ce, Co, Cr, Cs, Cu, Ga, La, Nd, Ni, Pb, Rb, Sr, Th, U, V, Y, and Zn only in soil samples collected in layers I and III) in the solutions was determined inductively coupled plasma atomic emission spectrometer (ICP-AES; Plasma torch JY180 ULTRACE) and by ICP-MS spectrometer (ICP-MS, VG PlasmaQuad PQ2+), respectively.

For each element, rhizosphere effects were calculated as the percentage difference between paired rhizosphere and bulk soil samples for each depth, each season and each tree species. A positive rhizosphere effect indicates a greater flux in the rhizosphere, while a negative rhizosphere effect indicates a greater flux in the bulk soil.

The mean values were calculated from our replicates (n = 4) and are given with standard errors. For each tree species, each depth and each soil compartment, a one-factor variance analysis (ANOVA) was used to assess significant differences between the different seasons, at the threshold of p < 0.05. The normality of distribution and the homoscedasticity of variances were tested. Average comparisons were made using the Student-Newman- Keuls test. Before analysis, all percentages were arcsine transformed. The paired t-Student test was performed to establish significant differences for pH, carbon, nitrogen, major and trace elements between the two soil compartments (not independent samples) for each tree species, each depth and each season at the threshold level of p < 0.05. Statistical analyses were completed with the UNISTAT software (Unistat version 5.0, 2002, England).

The climatic conditions (air temperature and rainfall) measured during the period of the study in the region of the Breuil-Chenue site were presented in [26] . Briefly, the mean temperatures were 4.4˚C, 2.5˚C, and 14.9˚C in November, February, and May 2007, respectively. The mean precipitations were 58, 63, and 122 mm in November, February, and May 2007, respectively. The mean annual temperature and precipitations were about 9˚C and 1300 mm, respectively.

The pH of the rhizosphere solution was significantly inferior to that of the bulk soil solution in winter whatever the soil horizon under beech and only in layer II under Norway spruce (Table 1). This difference is mainly due to a strong increase of the pH in the bulk soil between the autumn and the winter, 4.35 vs. 3.79, 4.36 vs. 4.05, and 5.37 vs. 4.52 for layer I, layer II, and layer III, respectively. The pH of the rhizosphere solution was significantly superior to that of the bulk soil solution in spring in layers II and III under beech, only. No significant difference between the pH of the bulk soil and rhizosphere solutions was observed in autumn.

Figure 1 represents percentages of difference for C and N between the bulk soil and rhizosphere solutions for each depth, each season, and both species. Our results showed an enrichment of the rhizosphere in C and N in layer III whatever the season and the species, with exception of the N in spring under Norway spruce (Figure 1).

In layer I, the rhizosphere is enriched in C in autumn and in N in winter and spring for the beech as well as in C in spring for the Norway spruce. In layer II, the rhizosphere is enriched in C in spring for both species (Figure 1). We did not observe seasonal variations of the rhizosphere effect, except for an increase of this effect on N in spring in layer I under the beech.

Figure 2 represents percentages of difference for major elements between the bulk soil and rhizosphere solutions for each depth, season, and tree species. The rhizosphere solutions were significantly enriched in K compared to those in the bulk soil, independently of the tree species, the depth and the season. No seasonal variation of the rhizosphere effect on K was observed, except for an enrichment of the rhizosphere solution in spring in layer III under beech. For Ca, the rhizosphere solutions were enriched under beech in layers II and III in spring, and under Norway spruce in layer I in spring, and in layer III in autumn and spring. Significant seasonal variations of the rhizosphere effect on Ca were observed under beech in layers II and III, and under Norway spruce in layer I, i.e. higher enrichment of the rhizosphere solution in spring. For Mg, the rhizosphere solutions were enriched under beech in layer II in spring, and under Norway spruce in layers I, II, and III in spring. Significant seasonal variations of the rhizosphere effect on Mg were observed under beech in layer II and under Norway spruce in layers I and II, i.e. higher enrichment of the rhizosphere solution in spring. For Na, the rhizosphere solutions were enriched under beech in layer I in autumn, winter and spring, in layer II in spring, and in layer III in autumn and spring, and under Norway spruce in layers I and II in spring and in layer III in autumn, winter, and spring. Significant seasonal variation of the rhizosphere effect on Na was observed under Norway spruce in

layer II, i.e. higher enrichment of the rhizosphere solution in spring. For Fe, the rhizosphere solutions were enriched under beech in layer I in autumn and in layer II in winter, and under Norway spruce in layer III in autumn, winter and spring. Significant seasonal variations of the rhizosphere effect on Fe were observed under beech in layer I, i.e. higher enrichment of the rhizosphere solution in autumn, and under Norway spruce in layer III, i.e. higher enrichment of the rhizosphere solution in spring. For Mn, the rhizosphere solutions were enriched under beech in layer III in winter, and under Norway spruce in layers I and II in spring and in layer III in winter and spring. Significant seasonal variation of the rhizosphere effect on Mn was observed under Norway spruce in layer I, i.e. higher enrichment of the rhizosphere solution in spring.

Figure 3 represents percentages of difference for trace elements between the bulk soil and rhizosphere solutions for each depth, each season, and both species. In layer I, the general tendency was a depletion of trace elements (except for Rb) in the rhizosphere solution during the winter under both species and an accumulation of trace elements in the rhizosphere solution during the autumn under beech as compared to the bulk soil solution. In layer III, we observed an accumulation of most of the trace elements in the rhizosphere solution whatever the season and the species. Interestingly, the rhizosphere solutions were enriched in Rb whatever the depth, the species and the season.

Our study demonstrated that, whatever the depth (except in layer III under beech) and the species, the rhizo-

sphere solution was enriched in carbon in spring as compared to that of the bulk soil. The rhizosphere solutions were also generally enriched in N. The accumulation of C in the rhizosphere solution can result from the release by roots during the growing period of an abundant amount of carbon into the surrounding soil, called rhizodeposition [36] , which includes the C derived from photosynthates.

Our study also showed that pH in bulk soil solution did not significantly differ from that of rhizosphere solution in autumn and spring for both species. In contrast, the rhizosphere solution is more acidic than that of the bulk soil under beech (layers I, II, and III) and Norway spruce (layer II, only) in winter. The acidification of the rhizosphere solution can be attributed to carbonic and organic acids produced by roots and rhizosphere microflora through respiration and exudation as well as to proton release by roots to compensate for an unbalanced cation-anion uptake at the soil-root interface [37] . Because nutrient uptake is very low or inexistent in winter, it is likely that the decrease of pH in the rhizosphere mainly results from the activity of soil microorganisms. Our results also revealed an enrichment of the rhizosphere in carbon and nitrogen in winter for both species in the 10 - 23 cm horizon. The decrease of pH and the high production of acid compounds, carbon, and nitrogen in the rhizosphere can be explained by the fine-root decomposition but also by root and root-associated microorganism activity that can occur even in winter [38] -[40] . While roots tend to freeze and die at soil temperatures below −6˚C, minimum temperatures for root growth are thought to be between 0˚C and 5˚C. Thus, if soil temperatures warm to or stay above this minimum, as was the case during our study period with a relatively warm winter (mean air temperature > 4˚C), winter roots can break dormancy and become active thus stimulating rhizosphere microflora. In addition, [41] demonstrated that soil microorganisms can maintain both catabolic (CO₂ production) and anabolic (biomass synthesis) processes under frozen conditions and that no significant differences in carbon allocation from [¹³C] glucose into [¹³C] CO₂ and cell organic ¹³C-compounds occurred between +9˚C and −4˚C.

The large quantities of carbon in the rhizosphere could enhance the nutrient availability through an increase of CEC (increase of exchange sites) and a stimulation of microbial growth and activities which favours organic matter mineralization and mineral weathering [19] .

For Norway spruce and beech, the rhizosphere solutions were enriched in K whatever the depth or the season as compared to those of the bulk soil. In addition, the rhizosphere solutions were generally enriched in Ca, Mg, and Na in spring as compared to the bulk soil ones. No significant depletion of the rhizosphere solution in major elements was observed during the year of the study. Our results are in agreement with previous studies led in the Breuil-Chenue site that showed an enrichment of the rhizosphere of Norway spruce, beech, and oak in exchangeable nutrients such as K, Ca and Mg in spring [4] [27] , as well as the enrichment of the rhizosphere clay-sized fraction of Norway spruce and beech in elements such as K, Mg, and Fe [10] . These latter suggest that the formation of mica-like minerals in the rhizosphere in spring results from the presence in the rhizosphere solution of large amounts of K possessing a strong affinity for the high charge expandable phyllosilicate like vermiculite, which become fixed in the interlayer space [24] [42] [43] . Many authors have already observed an enrichment of exchangeable K in the rhizosphere soil and solution of different mature tree species such as Norway spruce, trembling aspen (Populus tremuloides Michx), and Douglas fir [8] [21] [24] [29] [44] . In addition, the accumulation of nutrients in the rhizosphere solution has already been observed in other forest sites under Douglas-fir [24] [45] and Norway spruce [20] . However, a depletion of cations (such as K, Mg and Ca) in the rhizosphere solutions was also observed during tree growth [21] [22] [46] . These authors suggest that nutrient uptake by plants can lead to a decrease of nutrients in the rhizosphere. In our study, the accumulation of nutrients in the rhizosphere solution can thus be attributed to positive differences between the nutrient inputs, principally by mineral weathering, organic matter mineralization and mass flow as well as nutrient output principally by root uptake. In the same site of Breuil-Chenue, a rate increase of organic matter decomposition [47] and of mineral weathering [10] was observed in the rhizosphere, revealing that the element mobilisation is exacerbated in the vicinity of roots due to root and root-associated microorganism activity. Furthermore, previous studies led in the Breuil-Chenue site showed that bacterial strains with a high mineral weathering efficiency were enriched in the ectomycorrhizosphere of beech and oak as compared to the bulk soil [48] [49] . Similarly, from soil samples collected on the same dates as those of our study, [50] demonstrated that the culturable bacterial communities of spruce and beech rhizospheres were characterized by a higher density and a higher potential of mineral weathering compared to the bulk soil ones. Ectomycorrhizal fungi are also known to significantly affect nutrient cycling in soils [51] . For example, [52] demonstrated that ectomycorrhizal fungi enhanced the mineral weathering and contributed thus to the increase of nutrient concentration (such as Ca, Mg, Fe and Mn) in the soil solutions of ectomycorrhizosphere of subalpine fine. These results suggest that high biological activities within the rhizosphere significantly improve soil nutrient availability. This hypothesis is in accordance with the conceptual model of nutrient availability in the soil-root system proposed by [5] and [53] .

In addition, our study reveals that the accumulation of nutrients in the rhizosphere solution was generally higher in spring compared to that in autumn and in winter. This result is in disagreement with the study of [54] who observed a depletion of Ca and Mg in the rhizosphere solution of Douglas-fir in June compared to August, November and February. They suggested that the strong tree uptake during the spring season could cause a decrease of Ca and Mg in the vicinity of roots. Our results, however, corroborate the increase of exchangeable nutrients (especially K) observed in the rhizosphere of Norway spruce and beech in the same soil samples collected in spring compared to the other seasons [26] . These results suggest that even if the root nutrient uptake is high in spring, the intensity of biological activities favouring organic matter degradation and mineral weathering is such as the input-output nutrient budget is higher in spring than in autumn and winter when the root nutrient uptake is low or inexistent. In the beech rhizosphere, [55] showed that enzymatic activities involved in the organic matter degradation (such as N-acetylglucosaminidase, cellobiohydrolase and β-glucosidase) were higher at the beginning of the vegetation period (in June) than in August and September. These authors suggested an effect of a high carbon input due to photosynthetic activity in June, when leaves have been fully developed. In addition, [56] observed an increase of mineral weathering in the rhizosphere of Douglas-fir between March and June. As demonstrated by [26] and [35] , nutrient stocks are very low in the soil of the Breuil-Chenue forest. According to [57] , Ca and Mg pools at the Breuil-Chenue site were amongst the lowest as compared to other monitored sites in France [58] -[60] . Our results thus suggest that in such nutrient-poor soils, trees can enhance nutrient availability in spring in the rhizosphere by a positive feedback loop between roots, microbial communities and soils. This complex process may thus allow maintaining a favorable environment for tree nutrition during the tree growth period.

Interestingly, the dynamics of K in soil solutions differ from those of Ca, Mg, and Na during the year. In fact, the rhizosphere solutions were enriched in K in autumn, winter and spring as compared to the bulk soil ones while the rhizosphere solutions were enriched in Ca, Mg, and Na in spring, only. These results suggest that a weak rhizosphere activity is sufficient to mobilize K to cover tree nutrient needs all throughout the year. In contrast, as soon as the rhizosphere activity slows down, the availability of Ca, Mg, and Na decreases. This highlights the main role of biological rhizosphere processes in the bioavailability of these nutrients. As demonstrated by [56] , soil exchangeable Mg pools are very small and decreased between 1974 and 2001 in the Breuil-Chenue forest, thus the biological rhizosphere processes are going to play a growing role in tree nutrition in the future.

In the present study, the trace element concentrations in soil solutions were significantly different between the rhizosphere and the bulk soil for both species, especially in layer III. These results show for the first time that the tree rhizosphere influences the availability of trace elements in uncontaminated forest soils. According to [61] , the impact of roots on trace element availability can result from several processes, i.e., accumulation/de- pletion of ionic species in the rhizosphere, acidification/alkalinization of the rhizosphere, oxidation/reduction in the rhizosphere, and complexation/chelation in the rhizosphere. As observed in this study for major elements, the rhizosphere solution was generally enriched in trace elements as compared to that of the bulk soil. Because the mineral soil represents the main source of trace elements in terrestrial environments, this enrichment could result from the increase of mineral weathering and organic matter mineralization in the rhizosphere. For example, [29] demonstrated in forest soils that microorganisms contributed to Cu increase in the tree rhizosphere as compared to the bulk soil and suggested that microbial mineralization could partly supply Cu to the solution fraction of the rhizosphere.

A significant depletion of most trace elements was observed in the rhizosphere solution of both tree species in winter and in layer I. This depletion may result from the adsorption of trace elements on oxides and hydroxides of Al, Fe or Mn and on organic matter as demonstrated by [62] -[64] . Furthermore, [65] showed, in contaminated forest soils, that the tree rhizosphere was an accumulation zone of Zn and Cu associated with inorganic amorphous solid phase due to the increase of Fe and Mn oxides in the rhizosphere as compared to the bulk soil. During the winter, the amount of dead roots can equal that of living roots [66] . This organic matter may also have adsorbed trace elements thus decreasing the amount of trace elements in the rhizosphere solution. In addition, it is known that mycorrhizal fungi play a key role in the heavy metal and radionuclide circulation in the soils [67] [68] . For example, [30] demonstrated in a forest soil that Cs was largely accumulated in fungal mycelium compared to bulk soil and rhizosphere. Consequently, the trace elements uptake by mycorrhizal fungi could contribute to the decrease of these elements in the vicinity of roots during the winter. This hypothesis is in agreement with [38] [39] that showed that ectomycorrhizal fungi can be metabolically active during winter period in temperate forest soils. According to [69] , the changes of pH in soils and soil solutions may play an especially significant role in the rate of availability of certain trace elements. Between autumn and winter, we observed a high increase in the pH of the bulk soil solution. These conditions may have decreased the adsorption of trace elements in the bulk soil thus favouring their solubility.

It is interesting to note that Rb was the only trace element enriched in the rhizosphere solution independently of the tree species, the depth, and the season. This result confirms the enrichment of the rhizosphere solution in K we observed. Indeed, both K and Rb show the same uptake kinetic by plants and ⁸⁶Rb⁺ has often been used in studies of K uptake [70] as it seems to emulate K to a high degree due to rather similar physicochemical properties (i.e., valence and ion diameter).