Introduction
Polyphenols are the most widespread secondary metabolites in the plant kingdom (Hättenschwiler & Vitousek Reference Hättenschwiler and Vitousek2000) and are involved in nutrient cycling, water retention, soil temperature and P availability, and exert a dominant role in controlling plant-litter-soil interactions in a wide range of ecosystems (Hättenschwiler et al. Reference Hättenschwiler, Tiunov and Scheu2005). They comprise about 40% of the dry weight of leaves and stems and a significant portion of the carbon in terrestrial ecosystems (Herms & Mattson Reference Herms and Mattson1992) and are a key component in many plant ecological processes. Polyphenols can reduce N mineralization through linking of organic N and removal of microorganisms, resulting in lower N leaching (Chapin Reference Chapin1995) and inhibit litter decomposition, increasing persistence of soil organic matter and cationic capacity exchange (Northup et al. Reference Northup, Dahlgren and Yu1995a). Soil nutrients affect polyphenol production (Hättenschwiler et al. Reference Hättenschwiler, Hagerman and Vitousek2003, Haukioja et al. Reference Haukioja, Ossipov, Koricheva, Honkanen, Larsson and Lempka1998, Wright et al. Reference Wright, Jordan, Lee, Duncan, Forsyth and Coomes2010), and higher polyphenol levels are often associated with acid soils with low nutrient content (Haukioja et al. Reference Haukioja, Ossipov, Koricheva, Honkanen, Larsson and Lempka1998, Northup et al. Reference Northup, Dahlgren and Yu1995a, Reference Northup, Yu, Dahlgren and Vogt1995b, Reference Northup, Dahlgren and McColl1998; Kraus et al. Reference Kraus, Zasoski and Dahlgren2004).
The savanna biome extends across Africa, Australia and South America and comprises a mosaic of plant physiognomies, from grasslands to forests (Lehmann et al. Reference Lehmann, Archibald, Hoffmann and Bond2011). The savanna is the second largest biome of Brazil and the world’s largest savanna (Sarmiento Reference Sarmiento1984). It is a global biodiversity hotspot due to high levels of endemism and rapid conversion of natural habitats into agriculture (Myers et al. Reference Myers, Mittermeier, Mittermeier, Da Fonseca and Kent2000). Besides local and regional drivers of change, global changes also affect ecosystem functioning in this biome. Savanna woodland plants have morphological (Bieras & Sajo Reference Bieras and Sajo2009, Sternberg et al. Reference Sternberg, Bucci, Franco, Goldstein, Hoffman, Meinzer, Moreira and Scholz2004) and physiological traits (Franco et al. Reference Franco, Bustamante, Caldas, Goldstein, Meinzer, Kozovits, Rundel and Coradin2005, Goldstein et al. Reference Goldstein, Meinzer, Bucci, Scholz, Franco and Hoffmann2008, Scholz et al. Reference Scholz, Bucci, Goldstein, Moreira, Meinzer, Domec, Villalobos-Vega, Franco and Miralles-Wilhelm2008) associated with growth under a markedly seasonal precipitation regime and low nutrient availability (Haridasan Reference Haridasan, Proctor, Ratter and Furley1992). Responses to these changes may differ among functional and phenological groups. The phenological groups of woody species vary in leaf longevity, internal water balance (Lenza & Klink Reference Lenza and Klink2006), foliar nutrient concentrations (Franco et al. Reference Franco, Bustamante, Caldas, Goldstein, Meinzer, Kozovits, Rundel and Coradin2005), root architecture and temporal patterns of sap flow (Scholz et al. Reference Scholz, Bucci, Goldstein, Moreira, Meinzer, Domec, Villalobos-Vega, Franco and Miralles-Wilhelm2008). Some savanna-woodland tree species have high polyphenol leaf concentrations (Gonçalves-Alvim et al. Reference Gonçalves-Alvim, Korndorf and Fernandes2006, Reference Gonçalves-Alvim, Lana, Ranieri and Fernandes2011; Santos et al. Reference Santos, Costa, Ribeiro, Guimarães, Ferri, Ferreira and Seraphin2002, Skidmore et al. Reference Skidmore, Ferwerda, Mutanga, Van Wieren, Peel, Grant, Prins, Balcik and Venus2010). These are positively correlated with Al soil concentration and negatively correlated with soil pH and concentrations of Ca, Mg, Mn, P and K (Jacobson et al. Reference Jacobson, Santos, Duarte, Farias and Kliemann2005).
Brazilian savanna is limited by N and P (Kozovits et al. Reference Kozovits, Bustamante, Garofalo, Bucci, Franco, Goldstein and Meinzer2007) and nutrient cycling (especially N and P) is very conservative (Bustamante et al. Reference Bustamante, Medina, Asner, Nardoto and Garcia-Montiel2006). Thus, changes in nutrient dynamics can lead to an improved environment for some species, increasing their competitiveness (Bobbink et al. Reference Bobbink, Hicks, Galloway, Spranger, Alkemade, Ashmore, Bustamante, Cinderby, Davidson, Dentener, Emmett, Erisman, Fenn, Gilliam, Nordin, Pardo and De Vries2010). A long-term study of nutrient addition in a savanna woodland demonstrated that increased nutrient availability led to changes in density, dominance and richness of herbaceous and shrub-tree layer, invasion of exotic grasses and increased rates of leaf litter decomposition, altering species composition and nutrient cycling (Bustamante et al. Reference Bustamante, Brito, Kozovits, Luedemann, Mello, Siqueira Pinto, Munhoz and Takahashi2012, Jacobson et al. Reference Jacobson, Bustamante and Kozovits2011, Kozovits et al. Reference Kozovits, Bustamante, Garofalo, Bucci, Franco, Goldstein and Meinzer2007).
Nutrient-rich ecosystems and fertilized environments tend to have lower polyphenolic production (Haukioja et al. Reference Haukioja, Ossipov, Koricheva, Honkanen, Larsson and Lempka1998, Northup et al. Reference Northup, Dahlgren and Yu1995a, Reference Northup, Yu, Dahlgren and Vogt1995b, Reference Northup, Dahlgren and McColl1998). Nonetheless, responses to changes in nutrient cycling rates due to fertilization may vary among phenological groups. Considering that the savanna woodland is a dystrophic environment, strategies to prevent nutrient leaching are important. Thus, because leaf turnover is higher in brevideciduous species than in evergreens, an increase in nutrient concentration in brevideciduous ephemeral leaves may require greater investment in polyphenol production to slow down decomposition and prevent rapid nutrient loss after senescence (Villar & Merino Reference Villar and Merino2001). Therefore, greater nutrient availability may favour species with higher leaf turnover and shift the composition of phenological groups in ecosystems. Based on these considerations and considering the relevance of polyphenols for plant and ecosystem functions, we evaluated the responses of two savanna-woodland plant functional types (based on foliar phenology) to the changes in soil nutrient availability. We hypothesized that (1) N, P and S (present in the formulation of N and P fertilizers – ammonium sulphate and superphosphate) foliar concentrations will increase in response to N, P and N plus P fertilization but responses will differ according to phenology and leaf age; (2) foliar concentrations of total phenols and tannins will differ between phenological group (being higher in brevideciduous species than in evergreen species) and will decrease in response fertilization; and (3) responses will vary according to leaf age (i.e. polyphenol concentrations decreasing with leaf age).
Methods
Study area and fertilization treatments
This study was performed in the Roncador Ecological Reserve, which belongs to the Brazilian Institute of Geography and Statistics (RECOR/IBGE), near Brasília – Federal District, Brazil (15°56'S, 47°53'N, average altitude = 1100 m asl). The soil type is Oxisol (Haplustox), an acidic soil with high Al levels and low cation-exchange capacity (Haridasan Reference Haridasan and Novais-Pinto1994). Annual precipitation during the 2-y study period was 1667 and 1184 mm. Air temperature ranged from 10.1°C (July) to 31.9°C (October). The vegetation is classified as cerrado sensu stricto (savanna woodland) and is the most common savanna vegetation type in Brazil. It consists of a continuous grassy layer and a tree and shrub layer with 20% to 60% canopy cover (Eiten Reference Eiten1983).
The fertilization experiment began in 1998 with a completely randomized experimental design, with four nutrient-addition treatments and four replicates randomly divided into 16 plots of 225 m2 (15 × 15 m) at least 10 m apart from each other. The following treatments were applied to the litter layer without incorporation into the soil: control (C; without fertilization), +N (addition of ammonium sulphate, (NH4)2 SO4), +P (addition of 20% superphosphate, Ca(H2PO4)2 + CaSO4.2H2O), and +NP (simultaneous addition of ammonium sulphate plus 20% superphosphate). Between 1998 and 2006, 100 kg ha−1 of N, P and +NP were applied twice a year (at the beginning and end of the rainy season). The study was conducted after 8 y of fertilization, and in 2005, all the plots in the study area were accidentally burned but the plots were reinstalled immediately after the fire event (Bustamante et al. Reference Bustamante, Brito, Kozovits, Luedemann, Mello, Siqueira Pinto, Munhoz and Takahashi2012, Jacobson et al. Reference Jacobson, Bustamante and Kozovits2011, Kozovits et al. Reference Kozovits, Bustamante, Garofalo, Bucci, Franco, Goldstein and Meinzer2007). The burning duration was short due to the rapid consumption of fine fuel (mainly dry mass of grasses) thus reducing the impacts on the woody layer.
Soil sampling and analysis
Composite soil samples were collected at the beginning of the rainy season (October), consisting of two samples from each plot at five depths (0–10, 10–20, 20–30, 30–40 and 40–50 cm). We determined soil pH in water, total N (micro-Kjeldahl method), available P, K, and exchangeable Ca, Mg and Al. Soil P and K were extracted with Mehlich 1 solution, and Al was extracted with 1 M KCl solution. The available P concentration was estimated by colorimetry with ammonium molybdate. Exchangeable Al was quantified by titration with NaOH. Exchangeable Ca, Mg, and Fe concentrations were determined by atomic absorption spectrophotometry, and K concentration by flame emission (EMBRAPA 1999). Soil pH values were lower in the +N, +P and +NP plots than in the control plots at 0–30 cm depth. Available P concentration was higher in the +P and +NP plots at 0–40 cm depth. Soil Ca concentrations were higher at 0–30 cm depth in +P plots than in control plots. Soil Al concentration was higher at all depths in +NP plots and at 0–30 cm in +N plots. Soil pH values increased with depth in all plots, whereas total N, P, K, Mg and Al concentrations showed the opposite pattern. There were no differences in soil N, K and Mg concentrations between control plots and fertilizing treatments. For more details about soil results, see Jacobson et al. (Reference Jacobson, Bustamante and Kozovits2011).
Woody species
Brazilian savanna plant biodiversity comprises a large number of rare species and few oligarchic species (Ratter et al. Reference Ratter, Bridgewater and Ribeiro2012). In this study, we selected six woody plant species in the experimental plots from two phenological groups (three brevideciduous and three evergreen). Caryocar brasiliense A. St.-Hil. (Caryocariaceae), Blepharocalyx salicifolius (HBK) O.Berg. (Myrtaceae) and Dalbergia miscolobium Benth. (Leguminosae-Papilionoideae) are brevideciduous (with complete canopy replacement during the dry season, and a short period with few or no senescent leaves). The mean leaf lifespan of deciduous and brevideciduous species is 279 days (Souza Reference Souza, Bezerra and Ferreira2012). Ouratea hexasperma (A. St.-Hil.) Baill. (Ochnaceae), Roupala montana Aubl. (Proteaceae) and Styrax ferrugineus Nees & Mart. (Styracaceae) are evergreens. They produce a cohort of new leaves that precedes or is concomitant with the loss of the older leaf cohort (Lenza & Klink Reference Lenza and Klink2006). The mean leaf lifespan of evergreens is 400 days (Souza Reference Souza, Bezerra and Ferreira2012). These species, representing ~38% of community tree density, are among the 10 with the largest importance value index (IVI) (Jacobson et al. Reference Jacobson, Bustamante and Kozovits2011). They are common savanna species in Brazil, as demonstrated by the analysis of 625 Brazilian savanna areas, occurring in 10.7% (B. salicifolius) to 62.2% (R. montana) of the sampled areas (Françoso Reference Françoso2014, Françoso et al. Reference Françoso, Haidar and Machado2016).
Leaf sampling
Twenty individuals of each species were selected and marked (five individuals per treatment, totalling 120 individuals), considering health status and accessibility to foliage. Samples were collected between October and January of 2007 (i.e. beginning to middle of the rainy season). Leaf samples for each individual were collected, at the same time, in three distinct growth stages: newly sprouted leaves (that had emerged less than 2 wk before), young leaves (younger than 2 mo, fully expanded, relatively thin with lighter pigmentation), and mature leaves (older than 4 mo, fully expanded, thicker with darker pigmentation). The leaves were dried at 60°C for 48 h and ground in a Wiley mill (40-mesh). For nutrient analysis, only young and mature leaves were used.
Total phenol extraction and analysis
Leaf extraction was performed at room temperature using 1 g of ground leaf for every 15 ml methanol, 50% (v/v). The samples were extracted over a 24-h period in a dark room and mixed with solvent renewal (+15 ml) (twice for 30 min). The extracts were filtered through 25-μm filter paper (Whatman International, England) and stored at −2°C (Hagerman Reference Hagerman2002). The Folin–Ciocalteu method was used to determine total phenol concentration (Forrest & Bendall Reference Forrest and Bendall1969), by adding 2.5 ml of 10% (v/v) Folin-Ciocalteu reagent (Sigma-Aldrich) to 0.5 ml extract and 2 ml sodium carbonate solution (0.75 mg g−1). The solution was left at room temperature for 1 h and analysed by spectrophotometry at 760 nm (Shimadzu UV-1203; Shimadzu Corporation, Japan). A standard curve was created using tannic acid (Sigma-Aldrich), and concentrations were expressed as percentage (dry mass) of tannic acid equivalents. All measurements were performed in triplicate.
Tannin extraction and analysis
Leaf tannin was analysed using the precipitation method (Hagerman & Butler Reference Hagerman and Butler1978). Briefly, 1 ml extract was added to 2 ml bovine serum albumin fraction V (1 mg ml−1; Sigma-Aldrich). The solution was left at room temperature for 15 min and centrifuged (3000 rpm) for 15 min. After removing the supernatant, the precipitate was dissolved in 4 ml solution of sodium dodecyl sulphate (Sigma-Aldrich) and triethanolamine (Sigma-Aldrich), and 1 ml ferric chloride (0.01 M) was added. After 30 min, absorption was measured at 510 nm using a Shimadzu UV-1203 spectrophotometer (Shimadzu Corporation). A standard curve was created using tannic acid (Sigma-Aldrich), and concentrations expressed as percentage (dry mass) of tannic acid equivalents. All measurements were performed in triplicate.
Foliar nutrients
Leaf samples were digested in a solution of nitric, perchloric and sulphuric acid (10:2:1). We determined the concentrations of sulphur with turbidimetry. Phosphorus foliar concentrations were determined with colorimetric analysis with ammonium molybdate and ascorbic acid, and nitrogen foliar concentrations were determined with the micro-Kjeldahl distillation method (EMBRAPA 1999).
Statistical analysis
Normal distribution of the variables was analysed with Kolmogorov–Smirnov test (α < 0.05). Since data were not normally distributed, even after transformation, they were compared using the Mann–Whitney non-parametric test (α < 0.05). Nutrient concentrations at each soil depth were compared with the F and Student’s t-test (α < 0.05). Comparisons with the control were made for each depth (n = 4). Arcsine square-root transformation was used for total phenols (dry mass %), tannins (dry mass %), and N (mg g−1). Logarithm transformation was used for concentrations of P (mg g−1) and S (mg g−1) (Zar Reference Zar1999). Differences in polyphenol and nutrient concentration between treatments (control, +N, +P, +NP), phenological groups (brevideciduous, evergreens), species and growth stages (newly sprouted leaf, young leaf and mature leaf) were tested by repeated-measures ANOVA using analysis of mixed linear models, followed by the Bonferroni adjustment (α < 0.05) and Dunnett test (α < 0.05) for multiple comparisons. In the mixed linear model, errors and random effects allow the display of correlations between non-constant variables. Factors are assumed to have a linear relationship with the dependent variable. This model can be used to conduct repeated-measures tests, identifying factors that contribute significantly to the model (McCulloch et al. Reference McCulloch, Searle and Neuhaus2008). These analyses were performed using SPSS 15.0 for Windows (SPSS Inc., USA).
Results
Nutrient, total phenols and tannin foliar concentration in control plots
N mean foliar concentration was 14.1 ± 4.1 mg g−1, and ranged from 10.0 to 22.3 mg g−1. Mean N concentration was 16.5 ± 4.4 mg g−1 in brevideciduous and 11.6 ± 1.7 mg g−1 in evergreens. Foliar P concentration was 0.67 ± 0.2 mg g−1, and ranged from 0.5 to 1.3 mg g−1 (0.77 ± 0.3 mg g−1 in brevideciduous and 0.57 ± 0.1 mg g−1 in evergreens). S foliar concentration was 0.41 ± 0.1 mg g−1 (0.38 ± 0.1 mg g−1 in brevideciduous and 0.43 ± 0.1 mg g−1 in evergreens). Total phenol concentration (% dry leaf mass) was 13.5 ±6.3%, and ranged from 3.5 to 26.6% (18.9 ± 3.3% in brevideciduous and 8.1 ± 2.9% in evergreens). Tannin concentration was 5.2 ±3.2%, and ranged from 1.8–10.5% (6.1 ± 3.3% in brevideciduous and 4.4 ± 3.0% in evergreens) (Supplementary material table 1).
Nutrient, total phenol and tannin foliar concentrations in response to fertilization treatments
N, P, S, total phenols and tannin foliar concentration and differences between leaf growth stages in brevideciduous and evergreen species, in control and fertilized plots are presented in supplementary materials tables 1 and 2. Significant differences between N, P, S, total phenols and tannin foliar concentrations between control and fertilized plots, in six savanna woodland species are presented in Table 1.
Table 1. Significant differences (Dunnett, P < 0.05) in N, P, S (mg g−1), total phenol and tannin foliar concentrations (% dry leaf) between treatments (C – control, N – nitrogen addition, P – phosphorus addition, NP – nitrogen plus phosphorus addition) in brevideciduous (Caryocar brasiliense (CB), Blepharocalix salicifolius (BS), Dalbergia miscolobium (DM)) and evergreen species (Ouratea hexasperma (OH), Roupala montana (RM), Styrax ferrugineus (SF)) in a savanna woodland in Central Brazil.
Foliar N (F = 108, n = 120, P < 0.001) and P (F = 72.8, n = 120, P < 0.001) concentrations were higher in brevideciduous species whereas foliar S concentrations were higher in evergreens (F = 20.2, n = 120, P < 0.001).
Only the brevideciduous species Blepharocalix salicifolius (+NP plots) and Styrax ferrugineus (P plots) presented higher P concentrations in P fertilized plots compared with control plots, both in young (F = 3.8, P = 0.02) and mature leaves (F = 4.3, P = 0.01) (Figure 1).
Figure 1. Phosphorus concentration (mg g−1) in young leaves (YL) and mature leaves (ML) in the brevideciduous species Blepharocalyx salicifolius (a), and in the evergreen species Styrax ferrugineus (b) in control plots (C) and fertilized plots (+N, nitrogen, +P, phosphorus; and +NP, nitrogen plus phosphorus) in a savanna woodland in Central Brazil. * Indicates significant differences with control plots (Dunnett, P < 0.05). Error bars represent the standard deviation.
Foliar N:P ratios varied in response to fertilization (F = 6.6, P < 0.001) (Figure 2). Among brevideciduous species, the N:P ratio of B. salicifolius was lower in NP plots than in control plots (12.6 ± 4.4) (P < 0.001) (Figure 2a). The N:P ratio of Dalbergia miscolobium was lower in +P plots (27.9 ± 2.7) than in control plots (32.5 ± 2.0) (F = 7.9, P < 0.001) (Figure 2b). Among the evergreen species, the N:P ratio in leaves of Ouratea hexasperma was higher in +N plots than in control plots (21.4 ± 7.6) (F = 3.3, P = 0.03) (Figure 2c) while it was lower in Roupala montana (17.2 ± 6.6) (F = 4.0, P = 0.01) (Figure 2d) and Styrax ferrugineus (16.3 ± 3.1) (F = 2.2, P = 0.02) (Figure 2e) in +P plots than in control plots (24.6 ± 6.3 in Roupala montana and 21.5 ± 6.0 in S. ferrugineus).
Figure 2. N:P ratio in young leaves (YL) and mature leaves (ML) in brevideciduous species Caryocar brasiliense (a), Blepharocalyx salicifolius (b), Ouratea hexasperma (c), Roupala montana (d) and Styrax ferrugineus (e) in control plots (C) and fertilized plots (+N, nitrogen; +P, phosphorus; and +NP, nitrogen plus phosphorus) in a savanna woodland in Central Brazil. * Indicates significant differences with control plots (Dunnett, P < 0.05). Error bars represent the standard deviation.
Both young and mature leaves of Caryocar brasiliense (F = 22.2, P < 0.001, Figure 3a), Blepharocalix salicifolius (F = 8.3, P < 0.001, Figure 3b), Roupala montana (F = 11.7, P < 0.001, Figure 3e) and Styrax ferrugineus (F = 32.3, P < 0.001) (Figure 3f), showed higher S concentrations in all fertilized plots compared with control plots. The smallest increase in S concentration (28%) was observed in young leaves of Dalbergia miscolobium in +NP plots. The highest increase (6.5 times) was observed in both young and mature leaves of Styrax ferrugineus in +P plots.
Figure 3. Sulphur concentration (mg g−1) in young leaves (YL) and mature leaves (ML) in brevideciduous species Caryocar brasiliense (a), Blepharocalyx salicifolius (b), Dalbergia miscolobium (c) and in evergreen species Ouratea hexasperma (d), Roupala montana (e) and Styrax ferrugineus (f) in control plots (C) and fertilized plots (+N, nitrogen; +P, phosphorus; and +NP, nitrogen plus phosphorus) in a savanna woodland in Central Brazil. * Indicates significant differences with control plots (Dunnett, P < 0.05). Error bars represent the standard deviation.
In all treatments, mean foliar total phenol concentration was 55.5% higher in brevideciduous species, whereas the difference in tannin concentration between phenological groups was half that amount. Brevideciduous species had higher foliar total phenol (F = 578, n = 154 and 170, P < 0.001) and tannin (F = 58.6, n = 154 and 170, P > 0.001) concentrations than evergreen species.
Only the brevideciduous species Blepharocalix salicifolius (total phenol, F = 17.2, P < 0.001 (Figure 4a) and tannin, F = 178.7, P < 0.001 (Figure 4c)) and Dalbergia miscolobium (total phenol F = 25.6, P < 0.001) (Figure 4b) had significant differences in concentrations between fertilized and control plots but responses differed between species and primarily in immature leaves. In leaves of Blepharocalix salicifolius, total phenol (Figure 4a) and tannin (Figure 4c) concentrations increased in +P and +NP plots, whereas in leaves of Dalbergia miscolobium, total phenol concentrations decreased in +P and +NP plots (Figure 4b).
Figure 4. Total phenol concentration (% dry leaf mass) in newly sprouted leaves (SL), young leaves (YL) and mature leaves (ML) in brevideciduous species Caryocar brasiliense (a) and Blepharocalyx salicifolius (b) and tannin concentration (% dry leaf mass) in Blepharocalyx salicifolius (c) in control plots (C) and fertilized plots (+N, nitrogen; +P, phosphorus; and +NP, nitrogen plus phosphorus) in a savanna woodland in Central Brazil. * Indicates significant differences with control plots (Dunnett, P < 0.05). Error bars represent the standard deviation.
Changes in total phenol and tannin concentrations across leaf growth stages were also species-specific. In the evergreen group, total phenol concentration increased in mature leaves of Roupala montana (+P plots) (F = 6.8, P = 0.002) and Styrax ferrugineus (+N plots) (F = 16.3, P < 0.001) (Supplementary material table 2), while tannin concentration decreased with leaf maturity in Ouratea hexasperma (+N and +P plots, F = 15.2, P < 0.001) and Styrax ferrugineus (all plots, F = 28.0, P < 0.001) (Figure 3f). Conversely, the brevideciduous species, Blepharocalix salicifolius showed lower total phenol concentration with leaf maturity (+NP plots, F = 5.5, P = 0.002). Tannin concentration also decreased in mature leaves of B. salicifolius (+N, +P, and +NP plots) (F = 24.9, P < 0.001) and decreased in young leaves relative to newly sprouted leaves in C. brasiliense (+P plots) (F = 212, P < 0.05) (Supplementary material table 2).
Discussion
In dystrophic environments such as the Brazilian savanna woodland, slow-growing species with low foliar nutrient levels might be replaced by species with higher growth rates and greater ability to increase foliar nutrient levels under higher nutrient availability (Aerts & Chapin Reference Aerts and Chapin2000, Bustamante et al. Reference Bustamante, Brito, Kozovits, Luedemann, Mello, Siqueira Pinto, Munhoz and Takahashi2012).
Previous studies in the experimental site indicated more pronounced responses at community level, such as reduced woody plant diversity and increased leaf litter decomposition rates in +NP plots than in +N and +P plots (Jacobson et al. Reference Jacobson, Bustamante and Kozovits2011). Two years after the first fertilization in these plots, N and P concentrations were not significantly altered in leaves of C. brasiliense and O. hexasperma, only B. salicifolius leaves had increased P concentrations (in +P plots) (Kozovits et al. Reference Kozovits, Bustamante, Garofalo, Bucci, Franco, Goldstein and Meinzer2007). Seven years after the first fertilization, N (in +N plots) and P (in +P and +NP plots) concentrations were higher in B. salicifolius and in C. brasiliense (P in +NP plots) (Saraceno Reference Saraceno2006), demonstrating the variation of leaf nutrient concentration in savanna woodland species over time in response to fertilization.
Changes in foliar nutrient concentrations in response to fertilization seems to be more related to differences in the nutritional requirements of the species. The brevideciduous B. salicifolius responded more strongly to fertilization than the other studied species. Thus, changes in soil nutrient availability affect species composition (Jacobson et al. Reference Jacobson, Bustamante and Kozovits2011). In our study, the lack of significant differences in foliar N concentrations between control and fertilized plots may have been due to increased nutrient demand for generation of new tissues after the accidental burning, causing decreased N leaf storage. Changes in N plant cycling due to fire were reported by Nardoto et al. (Reference Nardoto, Bustamante, Pinto and Klink2006) that showed higher N resorption efficiency for Brazilian savanna tree species in burned plots compared to unburned plots. The decreased N:P ratio after P addition suggests plants absorb P at higher rates than N. Only one species (O. hexasperma) increased its N:P ratio after N addition, also suggesting greater P translocation efficiency (Kozovits et al. Reference Kozovits, Bustamante, Garofalo, Bucci, Franco, Goldstein and Meinzer2007). The higher P limitation was also reflected by differences in N and P concentrations between young and mature leaves in this study suggesting that this nutrient is retranslocated even before senescence begins.
The increased foliar S concentration was evident in all studied species in +NP plots, and in four of the six species in +P plots. This result suggests that savanna woody plants may also be limited by S, which has not been reported in the literature before. Jacobson (Reference Jacobson2009) also reported a general pattern of increased S concentration in leaf litter in fertilized plots and a significant decrease in S use efficiency compared with control plots. Sulphur might play a role in strategies to avoid excessive tissue dehydration of savanna woody plants during the dry season as primary S metabolism and associated metabolites are linked to responses to drought stress (Chan et al. Reference Chan, Wirtz, Phua, Estavillo and Pogson2013, Sieh et al. Reference Sieh, Watanabe, Devers, Brueckner, Hoefgen and Krajinky2013).
Although comparison of polyphenolic concentrations across studies is hampered by differences in methods (Escarpa & González Reference Escarpa and González2001, Yu & Dahlgren Reference Yu and Dahlgren2000), the foliar concentrations of total phenols and tannins in this study (3–26.5% for total phenols and 1.8–11% for tannins) are within the ranges found in literature (Hättenschwiler & Vitousek Reference Hättenschwiler and Vitousek2000, Hättenschwiler et al. Reference Hättenschwiler, Hagerman and Vitousek2003, Kraus et al. Reference Kraus, Yu, Preston, Dahlgren and Zasoski2003). Regarding the second hypothesis, only two (both brevideciduous) out of six species showed significant changes in foliar polyphenol concentration in response to fertilization. P and NP addition increased total phenol and tannin concentrations in young B. salicifolius leaves but decreased total phenol concentration in young D. miscolobium leaves. Total phenol concentration increased in mature leaves with P addition only in B. salicifolius. In the brevideciduous species C. brasiliense and B. salicifolius, total phenol concentration reached a quarter of dry leaf weight, indicating that these species allocate much of their resources toward total phenol production. Production of polyphenolic compounds is thought to be an evolutionary advantage for species growing in acidic and highly weathered soils (Chapin Reference Chapin1995, Northup et al. Reference Northup, Yu, Dahlgren and Vogt1995b, Reference Northup, Dahlgren and McColl1998). Wright et al. (Reference Wright, Jordan, Lee, Duncan, Forsyth and Coomes2010) hypothesized that N availability has more influence on polyphenol production than P availability because N limitation decreases protein production and thus competition for phenylalanine, a precursor of many phenolic compounds. P acts as a recyclable co-factor in these reactions, allowing protein and phenolic production to continue under low P conditions. However, we found that P addition had greater effect on foliar polyphenol concentrations than N addition. The production and allocation of polyphenols in leaves in response to increased nutrient availability differed among species, even within the same phenological group. Increased nutrient availability will likely have little effect on this genotypic trait in such a short period, as observed with four species in this study. The species with altered polyphenol concentrations in response to nutrient addition likely have high nutritional requirements, such as B. salicifolius. This finding reinforces the idea that the conceptual models (e.g. Carbon Nutrient Balance Hypothesis (Bryant et al. Reference Bryant, Chapin and Klein1983) or Resource Availability Hypothesis (Coley et al. Reference Coley, Bryant and Chapin1985)) are insufficient to explain variability in plant polyphenol concentrations (Nitao et al. Reference Nitao, Zangerl and Berenbaum2002), especially in ecosystems with extremely high biodiversity, such as the Brazilian savanna.
Contrary to our prediction, the results showed no common pattern for total phenol and tannin concentrations across leaf growth stages. Decreases in total phenol and tannin concentrations during leaf maturation are common in other ecosystems (Covelo & Gallardo Reference Covelo and Gallardo2004), consistent with the results for brevideciduous species C. brasiliense and B. salicifolius in this study.
The observed decrease of T:TP ratio (the ratio between tannin and total phenols) over time in evergreen species may be the result of lower phenol polymerization during leaf maturation. Leaves of evergreen species have longer lifespans than those of deciduous and brevideciduous species (Franco et al. Reference Franco, Bustamante, Caldas, Goldstein, Meinzer, Kozovits, Rundel and Coradin2005, Lenza & Klink Reference Lenza and Klink2006). A lower polymerization rate can maintain polyphenolic compounds longer to protect against oxidation during senescence (McKee et al. Reference McKee, Feller, Popp and Wanek2002). The low specific leaf area of evergreen species may also mitigate the leaching of low-molecular-weight phenolic compounds, since leaves with high leaf mass per area typically have a thicker cuticle to slow water and solute loss (Villar & Merino Reference Villar and Merino2001).
The cost of leaf construction is a key factor in determining the specific competitive characteristics of a phenological group. Evergreens have the advantage of a longer photosynthetic season, whereas deciduous species normally have the advantage of leaves with a high specific area, offsetting the cost of foliar N and P with a higher rate of photosynthesis per unit leaf area (Franco et al. Reference Franco, Bustamante, Caldas, Goldstein, Meinzer, Kozovits, Rundel and Coradin2005, Montes & Medina Reference Montes and Medina1977). Lower investment in lignin and cuticle and greater investment in cheaper structural compounds such as cellulose and hemicellulose allows brevideciduous species to invest more energy in polyphenol production. Our study showed consistent patterns in terms of leaf chemical composition when phenological groups of savanna woodland woody species are compared. However, their responses to nutrient additions showed considerable variations, showing that variability within functional groups should be considered in modelling efforts to understand how tropical and species-rich ecosystems will respond to environmental changes, with consequences for ecosystem function and resilience. In addition, this study suggests that besides the co-limitation by N and P, savanna woodland tree species are also limited by S, indicating that the multi-element limitation deserves further investigation in these ecosystems.
Acknowledgements
We would like to thank Mardocheu Rocha, Maria Regina Sartori da Silva, Gilberto B. Cosac, Lucas E. Gonçalves and the UnB Ecology Lab staff for valuable help. We thank Prof. Gabriela Bielefeld Nardoto (University of Brasília) and Dr Iris Roitman for revising this manuscript. We are also grateful to the administration and staff of RECOR-IBGE for the support for the fieldwork.
Financial support
This study was funded by the Graduate Program in Ecology of UnB, Inter-American Institute for Global Change Research (CRN3 project ‘Nitrogen cycling in Latin America: drivers, impacts and vulnerabilities’ (Nnet)), CNPq (grant 474071/2006-5) and LBA-NASA (ND-07).
Supplementary materials
For supplementary material for this article, please visit https://doi.org/10.1017/S0266467419000257
