journal homepage: www.elsevier.com/locate/scitotenv 1. Introduction Cerrado is a floristically rich savanna located in the central region of Brazil. Its vegetation is composed of a mosaic of physiognomies, ranging from open grasslands to woodlands, with intermediary forms between them (Ratter et al., 1997). One of the main environmental determinants of the Cerrado vegetation is rainfall seasonality. Rainfall occurs mainly in spring-summer (October–March), when monthly average ranges from 150 to 500 mm. However, in autumn-winter (April–September) monthly rainfall is reduced to between 0 and 50 mm (Silva et al., 2008), which depletes soil-water supplies, mainly for herbaceous spe- cies (Rossatto et al., 2013). Besides climatic seasonality, fire incidence is another key feature of the Cerrado, which are more frequent in open physiognomies than in woodlands, and occur mainly in the dry season (Pereira-Júnior et al., 2014). These restrictions affect herba- ceous-shrubby plants, in which most of the species lose the aerial or- gans in the dry season and resprout during the rainy season (Mantovani and Martins, 1988). Frequent and unpredictable changes in environmental factors re- quire traits that allow plant survival. Carbohydrate storage may buffers plants against temporal changes in external as well as in internal growth conditions (Suzuki and Stuefer, 1999). Variations in environmental fac- tors throughout the year may induce seasonal fluctuations of plant re- serves, including non-structural carbohydrates (NSC). NSC pools vary according to phenological patterns and the source-sink balance (Würth et al., 2005). Therefore, the presence of well-developed peren- nial underground organs, with plenty reserves and viable buds are im- portant adaptive traits of Cerrado-native plants that assure survival after long periods of water shortage and fire occurrence (Appezzato-da-Glória et al., 2008; Mantovani and Martins, 1988; Ratter et al., 1997). Several Cerrado herbaceous species have different types of under- ground organs, with distinctive morphological characteristics, propor- tion of storage tissues, and the type of stored compounds (Moraes et al., 2016). Among them, several Asteraceae species have thickened bud-bearing underground organs, which accumulate fructans as the major reserve (Appezzato-da-Glória et al., 2008; Figueiredo-Ribeiro, 1993). Fructans are fructose polymers, occurring in nearly 15% of the Angiosperms, including the derived families Poaceae and Asteraceae (Hendry, 1993). These carbohydrates are prominent plant NSC, surpassed only by starch and sucrose (Hendry and Wallace, 1993). They can be found at high levels in storage organs, reaching up to 80% of the dry mass of certain species, such as Helianthus tuberosus L. (Edelman and Jefford, 1968), Cynara cardunculus L. (Raccuia and Melilli, 2010) and Chrysolaena obovata (Less.) Dematt. (previously named Vernonia herbacea (Vell.) Rusby) (Carvalho and Dietrich, 1993). These and other Asteraceae species accumulate inulin-type fructans, which have β(2,1)-linked fructosyl units on sucrose (Hendry, 1993). In- ulin synthesis starts with the transfer of one fructosyl unit from a donor to an acceptor sucrose, catalyzed by the enzyme sucrose:sucrose 1- fructosyltransferase (1-SST), producing the trisaccharide 1-kestose. Inu- lin polymerization results from the action of the enzyme fructan:fructan 1-fructosyltransferase (1-FFT), which transfers fructosyl units from a donor to an acceptor fructan molecule with degree of polymerization (DP) ≥ 3 (Edelman and Jefford, 1968; Van den Ende, 2013). The activity of these enzymes results in fructan molecules with unequal length (Vijn and Smeekens, 1999). Inulin catabolism occurs with the removal of ter- minal fructosyl units from fructans, catalyzed by the enzyme fructan 1- exohydrolase (1-FEH) (Van den Ende et al., 2000). As storage compounds, fructans support high energy demanding developmental phases, such as sprouting and reproduction (Mellado-Mojica and López, 2012; Raccuia and Melilli, 2010). Further- more, they are essential for plant recovery after disturbances, such as fire occurrence, that cause damage to aerial parts (Asega and Carvalho, 2004; De Roover et al., 1999; Oliveira et al., 2012). In addition to its re- serve function, fructan metabolism is associated with resistance to cold 405 L.V. de Almeida et al. / Science of the Total Environment 598 (2017) 404–412 and drought, through osmotic regulation, membrane protection and scav- enging of free radicals (Garcia et al., 2011; Hincha et al., 2007; Hisano et al., 2004; Peshev et al., 2013), protecting underground organs during dry seasons. Variations in total fructose contents and in fructan molecular mass were described in distinct phenological phases for Cerrado-native spe- cies of Asteraceae and Amaranthaceae growing in the subtropical region (Carvalho and Dietrich, 1993; Isejima and Figueiredo-Ribeiro, 1993; Silva et al., 2013; Vieira and Figueiredo-Ribeiro, 1993). All these studies showed high oligosaccharide levels during sprouting and an increase of polysaccharides in the dry season, following the senescence of the aerial parts. Recently, the same pattern was shown for the temperate Helianthus tuberosus L. (Krivorotova and Sereikaite, 2014) and for the Mediterranean Cynara cardunculus L. (Raccuia and Melilli, 2010). How- ever studies on seasonal variations of fructans in species growing in the Cerrado core with distinct types of underground organs are still scarce. Ichthyothere terminalis (Spreng.) Blake (Asteraceae–Millerieae) is abundant in cerrado rupestre, a Cerrado phytophysiognomy where plants grow on rocky outcrops, with acidic and nutrient-poor soil, scarce water and frequent fires (Abdalla et al., 2016; Ribeiro and Walter, 2008). Its thickened underground system is dimorphic and constituted of a proximal region located close to the soil surface, with several buds from which aerial stems originate. Below this portion, the main axis (root origin) is orthogravitropic (growing vertically oriented). The me- dian region is soft and has a fleshy texture, with both short thin and long thickened diagravitropic roots, growing alongside the soil surface. All regions of this underground system accumulate inulin-type fructans (Abdalla et al., 2016). In this study, we collected orthogravitropic and diagravitropic roots of field-grown I. terminalis plants for 12 months, which comprised rainy and dry seasons, as well as different phenological phases. Our objectives were to analyze how abiotic environmental factors and plant phenology influence fructan dynamics in field grown plants, and to verify if fructan metabolism differs in the two root types during the year. As dimorphic roots allow plants to optimize water absorption in different substrate regions between seasons (Oliveira et al., 2016), we hypothesize that seasonality of environmental factors would affect fructan concentra- tions in both root types in different ways. 2. Materials and methods 2.1. Study site This study was carried out in a preserved area in Cerrado core at Reserva Biológica “Prof. José Ângelo Rizzo” of the Universidade Federal de Goiás, Mossâmedes, Goiás, Brazil (16°22′–15°48′S and 50°44′– 49°55′W). This area includes a gradient of Cerrado physiognomies, such as rocky outcrops savanna, known as cerrado rupestre. The altitude varies from 700 to 1080 m (Almeida et al., 2015) and the climate is Köppen's Aw (Silva et al., 2008). Underground organs of adult plants of I. terminalis were collected monthly, from February 2012 to January 2013. The voucher was depos- ited in the Herbarium of the Universidade Federal de Goiás (UFG, 47501). The climatic data of this period from the closest meteorological sta- tion (Goiás, GO, A014) were accessed in “Instituto Nacional de Meteorologia” homepage (www.inmet.gov.br) (Fig. 1). Adult plants were marked with wooden sticks to facilitate finding the underground organs after abscission of aerial parts during the dry season. The most recent fire that occurred in the study area was in August 2010, seven- teen months before the beginning of plant collection. 2.2. Plant material Underground organs of four plants of I. terminalis with similar aerial developmental stages were collected each month as mentioned above. The median region of thickened orthogravitropic and diagravitropic roots were chosen for this study because inulin spherocrystals are wide- ly distributed in these tissues (Abdalla et al., 2016). Roots were washed immediately after harvest to remove soil par- ticles. Healthy and viable parts from the median regions of the two root types were pooled independently. Samples (2 g) of each part were cut in small pieces and immediately boiled in aqueous ethanol (80%) for enzyme denaturation and stored (−20 °C) for subsequent soluble carbohydrate extraction. The ethanol with dissolved carbo- hydrates was kept and incorporated to the crude extract. Aliquots (1 g) were oven dried (60 °C) to constant weight for dry mass (DM) determination. Water content (WC) was determined as a per- centage of the fresh mass. 2.3. Soluble carbohydrate extraction and analyses Samples boiled in ethanol were homogenized and the suspension filtered. The filtrate was reserved and the plant residue was re-extracted twice with boiling ethanol (80 °C) for 5 min. The residue was then ex- tracted twice with distilled water at 60 °C for 30 min (modified from Pollock and Jones, 1979). In all steps filtrates were separated from resi- dues by vacuum filtering through filter paper. Crude extracts, which consisted of the pooled soluble phases, were vacuum-concentrated in a rotary evaporator at 37 °C. Aliquots of the crude extracts (1 mL) were fractionated into fructo-oligosaccharides (FOS) and fructo-poly- saccharides (FPS) by cold precipitation by adding to the extracts three volumes of ethanol (100%), stored overnight at −20 °C and centrifuga- tion at 177 ×g for 10 min at 5 °C. The supernatants were composed of mainly hexoses, sucrose and low molecular weight sugars while the precipitates contained the FPS fractions (Abdalla et al., 2016; Carvalho et al., 1998). 406 L.V. de Almeida et al. / Science of the Total Environment 598 (2017) 404–412 Fig. 1. Monthly means of climatic parameters registered from February 2012 to January 2013 in Serra Dourada, Mossâmedes, Goiás, Brazil, according to the meteorological station closest to the study site. Maximum (circles), mean (filled diamonds) and minimum temperatures (squares) (A). Rainfall (bars) and relative air humidity (circle) (B). Water soluble carbohydrates (WSC) in the crude extracts were quantified by a colorimetric assay with phenol-sulfuric acid using glucose as a standard (Dubois et al., 1956). Total fructose (TF) was determined in crude extracts, FOS and in FPS fractions using a kestose-specific assay with anthrone reagent and fructose as stan- dard (Jermyn, 1956). Crude extracts were deionized using ion exchange resins (Amberlite cationic - IRA 120 and anionic - IRA 410, VETEC, Duque de Caxias, Brazil). Following pH neutralization with ammonium hydroxide, the extracts were vacuum-dried at 37 °C and solubilized in ultrapure water (18.2 MΩ) to the final concentration of 400 μg mL−1 of fruc- tose equivalents. Deionized extracts were filtered through 0.45 μm membranes prior the analysis by high performance anion exchange chromatography with integrated pulsed amperometric detection (HPAEC/IPAD) on a Dionex system ICS-5000, using a CarboPac PA100 column (4 × 250 mm). Soluble neutral carbohydrates were separated by a gradient of sodium acetate (500 mM) in sodium hy- droxide (100 mM), as described by Silva et al. (2015). Extracts of C. obovata rhizophores were used as a reference for inulin series (Carvalho and Dietrich, 1993). 2.4. Statistical analysis Average multiple comparisons were performed by Kruskal- Wallis one-way ANOVA on ranks, with Dunn's post hoc test (p b 0·05). To perform multivariate analysis, roots metabolite contents (re- sponse matrix, 96 × 6), as well as the monthly climatic averages (five parameters) of the sampling time and the corresponding plant phenological stages, as categorical variable (explanatory ma- trix, 96 × 6) were submitted to the canonical redundancy analysis (RDA). RDA revealed an ordination of the response data constrained by explanatory variables, which accounts for the patterns of the only explained variation between data sets. Unrestricted Monte Carlo permutation tests (999 permutations) were performed to assess the significance of canonical axes, showing the relationships be- tween metabolite variables and the selected climatic and phenolog- ical predictors (Lepš and Šmilauer, 2003). In all the analyses, the variance inflation factor of the variable (VIF) with Bonferroni's cor- rection was used to determine the selection of the explanatory vari- ables. VIF-values ≥ 10 were considered as multicollinears (Lepš and Šmilauer, 2003). The hierarchical cluster analysis (HCA) was used aiming to detect the natural sample grouping from the RDA analyses and their intra and intergroup relations. Euclidean distance was used as similarity index and the hierarchical grouping was done according to the variance minimization method (Ward, 1963). Total variation partitioning of response data was obtained by partial RDAs (pRDAs), using the explanatory matrix reordered in two sets of se- lected variables: phenological (senescence, sprouting and reproduction) and environmental (average monthly minimum temperature and rainfall precipitation) data. This method resulted in distinct variation fractions with or without overlap between both sets of variables (Borcard et al., 1992; Legendre and Legendre, 2003). In addition to the previous techniques, principal response curve (PRC) was applied to investigate the effects of the response variables and their alterations in time (Ter Braak and Šmilauer, 2012; Van den Brink and Ter Braak, 1999). In PRC, sampling months were used as categorical covariable and the interaction between sampling time and treatment (roots) were used as explanatory variables. The anal- ysis produces a diagram which shows the time gradient in x-axis and the first or the second canonical axes of metabolite differences of the treatment in relation to a control, here attributed to orthogravitropic roots, in y-axis (Moser et al., 2007). Monte Carlo permutation tests were performed to evaluate if the PRC explains a significant part of the variance of the treatment in relation to control in all the time series (999 permutations), as well as to verify if the treatment results in one metabolite significantly altered in each sampling time (499 permutations). Before multivariate analyses, data were pre-processed: the response matrix was centered in the mean and standardized (all variables have 407 L.V. de Almeida et al. / Science of the Total Environment 598 (2017) 404–412 Fig. 2. Box plot diagrams showing water content (A, B), total water soluble carbohydrates (WSC) (C, D), total fructose (E, F), fructo-oligosaccharides (FOS) (G, H), fructo-polysaccharides (FPS) (I, J) in orthogravitropic (A, C, E, G, I) and diagravitropic roots (B, D, F, H, J) of field-grown I. terminalis. Crosses represent medians (n = 4). Phenological phases were: Reproduction (R), senescence of the aerial organs (Se), dormancy (D), sprouting (Sp) and vegetative growth (VG). means equal to zero and variances equal to one), so that they have the same importance in the multivariate analysis. The explanatory matrix was also centered and standardized. The multivariate analyses were conducted in Canoco (Canonical Community Ordination, version 5.0, Biometrics, The Netherlands, 2012). 3. Results and discussion 3.1. Phenology and water content I. terminalis has a developmental cycle with vegetative growth and reproduction in the rainy season and senescence of aerial parts and dor- mancy of the underground organs in the dry season. In the present study, in February 2012, most of I. terminalis population in the Biological Reserve was at reproduction, exhibiting inflorescences and fruits until March 2012 (end of the rainy season). Senescence of aerial parts began in April 2012, and was followed by dormancy of underground or- gans. Buds resprouted and originated new aerial stems at the end of September, restarting the developmental cycle. Flowering began in January 2013. The phenological cycle of Cerrado herbaceous species is highly influ- enced by rainfall, as in other seasonal neotropical environments (Chambers et al., 2013; Mantovani and Martins, 1988). In the Cerrado, rainfall is positively correlated with the number of flowering herba- ceous species (Batalha and Martins, 2004). However, in several Cerrado species some phenological events are synchronized by occurrence of fire (Coutinho, 1990). In I. terminalis flowering was observed 1to 2 months after a fire event (Abdalla et al., 2016 and field observations). During the present study, I. terminalis plants were not subjected to fire, since the previous fire occurred in August 2010. Changes in water content of the underground organs in I. terminalis were influenced by rainfall precipitation. Water content in roots ranged from 61 to 85%, with lower levels in dry season and higher levels at the onset of rainfall (Fig. 2A, B). Higher water contents were found in No- vember and December in orthogravitropic roots, and in March and De- cember (p b 0.05) in diagravitropic roots. Lower levels were detected in May, July and August (p b 0.05) for orthogravitropic roots and in August (p b 0.05) for diagravitropic roots. In September and October, even with low rainfall, water content in roots started to increase. Both root types showed similar oscillations in water content, except in March, when it increased in diagravitropic roots. The difference between seasons is probably due to variations in soil moisture as pointed out by Dimitrakopoulos and Bemmerzouk (2003). 3.2. Water soluble carbohydrate dynamics In orthogravitropic roots, WSC levels were higher in October, reaching 393.38 mg g−1 DM, during vegetative growth, and lower in February, May and June (p b 0.05) (Fig. 2C). In diagravitropic roots, WSC ranged from 123.73 mg g−1 DM in August to 229.92 mg g−1 DM in October (Fig. 2D). Both root types appeared to retain at least 10% of their dry mass in WSC throughout the year. Resprouting species, such as I. terminalis, have increased amounts of NSC stored in underground organs to support resprouting (Clarke et al., 2013). Stored carbohy- drates allow plants to develop, even during phases when photosynthet- ic aerial organs are absent (Landhäusser and Lieffers, 2003). Carbohydrate storage occurs when assimilation exceeds demands for maintenance and growth (Chapin et al., 1990). In I. terminalis, water sol- uble carbohydrates oscillated in the analyzed period, indicating alterna- tion of phases with high carbohydrate production and phases of carbohydrate mobilization and use (Fig. 2). Inulin-type fructans are the main water soluble carbohydrates accu- mulated in underground organs of I. terminalis (Abdalla et al., 2016). In this study, fructans were the major component (35–80%) of the WSC pool (Fig. 2 E, F), present as a linear homologous series that co-eluted with the inulin series from C. obovata rhizophores. Fructan storage in underground organs of Asteraceae from Cerrado is an adaptive feature to environmental constrains (Carvalho et al., 2007). Fructans support the high-energy demanding phenological phases, such as sprouting and flowering (Carvalho et al., 2007; Edelman and Jefford, 1968; Silva et al., 2013). 408 L.V. de Almeida et al. / Science of the Total Environment 598 (2017) 404–412 The highest total fructose concentrations were in the beginning of the rainy season. Interestingly, in April, at the beginning of aerial organs senescence, total fructose levels increased in both root types, especially the FPS fraction (Fig. 2I, J). This could be due to sugar mobilization to the underground organs, since a large fraction of carbon compounds and mineral nutrients are translocated to the underground organs during leaf senescence (Lapointe, 2001). These carbohydrates seem to be the energy source to promote rapid vegetative development and reproduc- tion after constraints, such as prolonged drought periods and fires. Throughout the developmental phases, the fructan profile displayed a balance between synthesis and mobilization in both root types, ev- idenced by the variation in FOS and FPS proportions (Fig. 2G, H, I, J). In high-demanding phenological phases, such as sprouting, vegeta- tive growth and reproduction, FOS contents increased. However, taking into account the ratio FOS:FPS (Fig. 3A, B), FPS prevailed from April to October/November (orthogravitropic/diagravitropic), when FOS:FPS was ≤1. Inversely, in December, just before inflores- cence development, a sharp increase occurred and the ratio remained N1 during reproduction. Chromatograms showed glucose, fructose, sucrose, and inulin-type fructans present throughout all the analyzed months (Fig. 4). Monosac- charide peaks increased at the end of dormancy and remained higher during sprouting and vegetative growth until the onset of reproduction in January 2013. At the end of the vegetative growth (December 2012), fructan peaks were the lowest among all the analyzed phases, and sucrose and 1-kestose peaks predominated. By the end of the reproduc- tion in March, and during senescence and dormancy, these peaks were lower while fructo-oligo- and fructo-polysaccharide peaks were higher, especially in orthogravitropic roots, and DP reached values N 40 at the end of dormancy (August 2012). Intermediary non-identified peaks, probably of the reducing inulo-n-ose series, were observed in both Fig. 3. Box plot showing fructo-oligosaccharides:fructo-polysaccharides (FOS:FPS) ratio in orthogravitropic (A) and diagravitropic (B) roots of field grown I. terminalis. Crosses represent medians (n = 4). Dotted lines indicate FOS:FPS = 1. Phenological phases were: Reproduction (R), senescence of the aerial organs (Se), dormancy (D), sprouting (Sp) and vegetative growth (VG). root types in all phenological phases, except in senescence. This series lacks the terminal glucose and is found in several Asteraceae species, generally in periods of intense inulin hydrolysis (Gupta and Kaur, 2000; Carvalho et al., 1997; Ueno et al., 2011; Van Arkel et al., 2012). 3.3. Relationships between climate, phenology and fructans In phenological studies, it is relatively common to detect correlations between climatic variables and a particular phenological response; however multiple climatic factors are expected to co-vary (Forrest and Miller-Rushing, 2010). Thus multivariate analyses are frequently used to detect interactions among several factors simultaneously. The variability patterns of I. terminalis samples were evaluated by the RDA of the root metabolites levels conditioned to climatic parame- ters, as well as the plant phenological phases. At the end of response data modelling by RDAs, two climatic parameters were obtained (min- imum temperature and rainfall precipitation) in addition to three plant phenological phases (senescence, sprouting and reproduction) as ex- planatory variables (Fig. 5). RDA results indicated that correlations between both data matrices were higher in the first two canonical axes (R1 = 0.703 and R2 = 0.453) and with variance inflation factors considered low (VIF b 2.3), suggesting no multicollinearity among the variables in multivariate re- gression models (Lepš and Šmilauer, 2003). Monte Carlo permutation 409 L.V. de Almeida et al. / Science of the Total Environment 598 (2017) 404–412 Fig. 4. HPAEC/IPAD profiles of soluble carbohydrates extracted from dimorphic roots of field-grown I. terminalis during an annual developmental cycle. Arrows indicate inulo-n-ose peaks. Phenological phases were: Reproduction (R), senescence of the aerial organs (Se), dormancy (D), sprouting (Sp) and vegetative growth (VG). Glucose (G), fructose (F), sucrose (S), 1- kestose (1-K). tests (999 permutations) showed highly significant results for the first two canonical axes (RDA1: 20.9% of the explained variation, F = 23.8, p = 0.001; RDA2: 5.6%; F = 6.8, p = 0.008), indicating that variation patterns of the original matrices did not arise by chance. The sum of the canonical axes was also significant (sum = 0.2860; F = 7.2; p = 0.001), so that 28.6% of the total variance of the metabolites was retained by selected explanatory variables. According to RDA triplot distribution, the increase in RDA1 axis is as- sociated, especially, to a decline in metabolites as rainfall decreases, co- inciding with dry and cold winter, independently of the sampled root type. These conditions are mainly related to samples harvested in the period of the aerial organs senescence and suggest a seasonal influence on RDA1. The results support the recognition of rainfall as one of the main drivers for phenological shifts in tropical South American seasonal environments (Chambers et al., 2013), since these changes may be in- fluenced by the assimilation and allocation of reserves (Rathcke and Lacey, 1985). Differently, an increase on RDA2 is related to higher levels of total water soluble carbohydrates, FPS and total fructose, which occurred mainly in plants in senescence and sprouting phases, independently of the sampled root. Higher FOS levels, water content and the increase in FOS:FPS ratio are related to reproduction, coincidently with months with higher rainfall precipitation. Thus, while RDA2 shows mainly the phenological variations of the samples, especially in senescence, sprouting and reproduction, RDA1 described, mainly, the variation of the metabolites in I. terminalis roots, as a response to water shortage. In this way, three natural groups of samples were suggested by RDA, which were corroborated by the HCA, using the RDA's scores as vari- ables. Fig. 6 shows the similarities between individual roots in terms of Euclidean distances, in which samples collected during the dry period (April–September), in senescence and sprouting phases constituted class 1. The samples collected between October and December, with higher minimum temperatures provided class 2, while samples in re- production, during January and March, composed class 3. Although RDA/HCA contributed significantly to the general compre- hension of data, suggesting a strong influence of the seasonal effect followed by a lower contribution of the phenological phases in I. terminalis, little can be deduced about the importance of each one to Fig. 6. Dendrogram of similarity representing the hierarchical clustering of I. terminalis samples based on RDA scores, using Euclidean distance and variance minimization technique. The samples are represented by the month of collection and orthogravitropic (O) or diagravitropic (D) roots. 410 L.V. de Almeida et al. / Science of the Total Environment 598 (2017) 404–412 Fig. 5. RDA triplot showing the distribution of I. terminalis samples according to root metabolites (small arrows) explained by phenological stage (senescence, sprouting and reproduction) and climate (rainfall and minimum temperature). The centroids of the phenological stages are represented by triangles, while the climatic variables are represented by large arrows. the total explained variance by the canonical axes, that is, those involv- ing response and explanatory matrices. One method developed by Borcard et al. (1992) and improved by Peres-Neto et al. (2006) allows the partitioning of the explained vari- ance, using partial RDAs (pRDAs). The procedures assess the relative im- portance of constraining matrix after adjusting the variability of other data sets which are regarded as covariate. In this study, variation partitioning was performed on two separate sets, one containing the plant phenological phase (senescence, sprouting and reproduction) and the other containing environmental data (selected climatic vari- ables) as explanatory matrices (Table 1). Results indicated that the total explained variation in metabolites re- sponse matrix (28.6%; fraction [a + b + c], Table 1) can be partitioned by the two descriptor sets, resulting in a model whose residue was 71.4% ([d]). Among explained variation's fractions with purer contribu- tion, without overlapping, the environmental set stood out (21.0%; [a]), and phenology represented 7.4% ([b]) of the total variance, all highly significant (p ≤ 0.002). These results are in accordance with RDA analy- sis, whose variances explained by the two first canonical axes (RDA1: 20.9%; RDA2: 5.6%) were attributed to the environmental effects and the phenological influence, respectively. It can also be noticed that the total contribution of the environmental set corresponds to 21.2% ([a + c]; p = 0.001), while the phenological predictor ([b + c]; p = 0.011) explains 7.6% of the total variability. The fraction constituted of the overlapping of descriptors sets effects showed a small contribution (0.2%; [c]). One alternative to better understand the partitioning of the total ex- plained variation can be obtained by the visualization of Venn's diagram (Fig. 7). The results of the total variation partitioning indicated that ap- proximately three-quarters of the explained variation can be modeled by climatic variables, that is, the variation in I. terminalis root metabo- lites are mainly determined by the environment. Aiming to evaluate if the differences between metabolite levels in roots are altered along sampling time, PRC was performed between orthogravitropic (control) and diagravitropic roots (treatment) for the 12 month sampling period (Fig. 8). The results indicated that 95% of the total variation of the data can be explained by treatment effect, 81.7% of which are shown in the first PRC diagram (F = 11.6, p = 0.023; 999 permutations). The differences among the sampling months contributed to the majority of total variance (50.5%), and the remaining part (31.6%) can be attributed to the differences among samples. The variables that mostly influenced the first PRC axis were the FPS and WSC levels, followed by FOS and total fructose levels, with higher weights. The FOS:FPS ratio and water did not show any contribution to the differentiation between samples. The second PRC axis was not significant and is not presented. When the data were restricted to each time period, a discrete difference (0.054 b p b 0.082) was observed in the period between July and September, time that coincides with dor- mancy and sprouting phases. In the beginning of this period, the metab- olites had a sharp decrease in their levels in diagravitropic roots (treatment), followed by an increase at the end relative to the control. Phenology influences spatial differences in fructan metabolism, with in- creased activity of fructan hydrolytic enzymes during sprouting near the proximal and bud-bearing region (Portes and Carvalho, 2006). These differences may indicate that in phases with high energetic demand, both root types, despite having similar fructan dynamics, exhibit a fructan metabolism modulated by local and temporal changes attending to the source-sink relationships. 4. Conclusions During a 12 month period without a drastic disturbance, such as fire, naturally field-grown plants of I. terminalis stored approximately 10– 40% of inulin-type fructans in dimorphic roots. This means that orthogravitropic and diagravitropic roots share the same function as storage organ. Fructan dynamics have similar patterns described for other Asteraceae species, with the proportional increase of polysaccha- rides with the senescence of the aerial organs, as exampled byC. obovata Table 1 Variance partitioning summary of I. terminalis through partial RDAs. Effects and variables F Pa Total effect Environmental and phenological Covariables Fraction of variation Explained variation (%) [a + b + c] 28.6 7.2 0.001 Partial effects Environmental Phenological [a] 21.0 13.3 0.001 Environmental [a + c] 21.2 12.5 0.001 Phenological Environmental [b] 7.4 3.1 0.002 Phenological [b + c] 7.6 2.5 0.011 Joint effects Environmental and phenological [c] 0.2 Residuals [d] 71.4 a Based on the Monte Carlo permutation test (999 permutations). Phenological pre- dictor: senescence, sprouting and reproduction; environmental predictor: average monthly rainfall and minimum temperature. Fig. 7. Venn diagram of total variation partitioning in metabolites from I. terminalis roots which is attributable to two sets of explanatory variables: phenological phases and environmental predictors. Percentages refer to explained variation by each fraction of variation, according to Table 1. Unexplained variance corresponds to residuals. 411 L.V. de Almeida et al. / Science of the Total Environment 598 (2017) 404–412 and Gomphrena species (Carvalho and Dietrich, 1993; Silva et al., 2013; Vieira and Figueiredo-Ribeiro, 1993). However, in this study, multivari- ate analyses showed that environmental factors, such as rainfall de- crease, had a stronger influence on metabolite levels in both root types than phenological shifts. Future experiments with controlled con- ditions and water withholding are necessary to confirm this finding. Only slight differences were found in fructan dynamics between orthogravitropic and diagravitropic roots, thus they may have similar fructan metabolism regulation. These differences may reflect distinct microclimatic conditions in both regions and also represent the influ- ence of sink-strength in carbon mobilization. Thus, both root types can be analyzed in futures studies concerning fructan metabolism in this promising fructan accumulator species. Acknowledgments This study was supported by FAPEG (PRONEX AUX PESQ 007/2009). L.V. Almeida is grateful to CAPES, Brasília, Brasil for a study grant. P.H. Ferri is a CNPq research fellow. The authors are grateful to Dr. Maria Angela M. Carvalho for critical and English review and to Cinara F. Abraão for lab support. References Abdalla, D.F., Moraes, M.G., Rezende, M.H., Hayashi, A.H., Carvalho, M.A.M., 2016. Morpho- anatomy and fructans in the underground system of Apopyros warmingii and Ichthyothere terminalis (Asteraceae) from the cerrado rupestre. J. Torrey Bot. Soc. 143:69–86. http://dx.doi.org/10.3159/TORREY-D-14-00050.1. Almeida, V.O., Carneiro, R.V., Carvalho, M.A.M., Figueiredo-Ribeiro, R.C.L., Moraes, M.G., 2015. Diversity of non-structural carbohydrates in the underground organs of five Iridaceae species from the Cerrado (Brazil). S. Afr. J. Bot. 96:105–111. http://dx.doi. org/10.1016/j.sajb.2014.10.003. Appezzato-da-Glória, B., Cury, G., Soares, M.K.M., Rocha, R., Hayashi, A.H., 2008. Under- ground systems of Asteraceae species from the Brazilian Cerrado. J. Torrey Bot. Soc. 135:103–113. http://dx.doi.org/10.3159/07-RA-043.1. Asega, A.F., Carvalho, M.A.M., 2004. Fructan metabolising enzymes in rhizophores of Vernonia herbacea upon excision of aerial organs. Plant Physiol. Biochem. 42: 313–319. http://dx.doi.org/10.1016/j.plaphy.2004.02.005. Batalha, M.A., Martins, F.R., 2004. Reproductive phenology of the cerrado plant communi- ty in Emas National Park (central Brazil). Aust. J. Bot. 52:149–161. http://dx.doi.org/ 10.1071/BT03098. Borcard, D., Legendre, P., Drapeau, P., 1992. Partialling out the spatial component of eco- logical variation. Ecology 73:1045–1055. http://dx.doi.org/10.2307/1940179. Carvalho, M.A.M., Dietrich, S.M.C., 1993. Variation in fructan content in the underground organs of Vernonia herbacea (Vell.) Rusby at different phenological phases. New Phytol. 123:735–740. http://dx.doi.org/10.1111/j.1469-8137.1993.tb03784.x. Fig. 8. Diagram showing the first component of the PRC of the differences in measured chemical parameters between diagravitropic (square) and orthogravitropic (circle, control) roots along sampling times (months). Chemical weights (b k ) (on the right) can be interpreted as a relative contribution of individual chemicals to the response given in the diagram. For each sampling time the Monte Carlo permutation test (499 permutations) was performed for comparison between roots. Carvalho, M.A.M., Zaidan, L.B.P., Dietrich, S.M.C., 1997. Growth and fructan content of plants of Vernonia herbacea (Asteraceae) regenerated from rhizophores. New Phytol. 136, 153–161. Carvalho, M.A.M., Pinto, M.M., Figueiredo-Ribeiro, R.C.L., 1998. Inulin production by Vernonia herbacea as influenced by mineral fertilization and time of harvest. Rev. Bras. Bot. 21:275–280. http://dx.doi.org/10.1590/S0100-84041998000300006. Carvalho, M.A.M., Asega, A.F., Figueiredo-Ribeiro, R.C.L., 2007. Fructans in Asteraceae from the Brazilian Cerrado. In: Shiomi, N., Benkeblia, N., Onodera, S. (Eds.), Recent ad- vances in fructooligosaccharides research. Research Signpost, Kerala, pp. 69–91. Chambers, L.E., Altwegg, R., Barbraud, C., Barnard, P., Beaumont, L.J., Crawford, R.J.M., Durant, J.M., Hughes, L., Keatley, M.R., Low, M., Morellato, P.C., Poloczanska, E.S., Ruoppolo, V., Vanstreels, R.E.T., Woehler, E.J., Wolfaardt, A.C., 2013. Phenological changes in the southern hemisphere. PLoS One 8. http://dx.doi.org/10.1371/journal. pone.0075514. Chapin, F.S., Schulze, E.D., Mooney, H.A., 1990. The ecology and economics of storage in plants. Annu. Rev. Ecol. Syst. 21:423–447. http://dx.doi.org/10.1146/annurev.es.21. 110190.002231. Clarke, P.J., Lawes, M.J., Midgley, J.J., Lamont, B.B., Ojeda, F., Burrows, G.E., Enright, N.J., Knox, K.J.E., 2013. Resprouting as a key functional trait: how buds, protection and re- sources drive persistence after fire. New Phytol. 197:19–35. http://dx.doi.org/10. 1111/nph.12001. Coutinho, L.M., 1990. Fire in the ecology of the Brazilian cerrado. Fire in the Tropical Biota. Springer, pp. 82–105. De Roover, J., Van Laere, A., Van den Ende, W., 1999. Effect of defoliation on fructan pat- tern and fructan metabolizing enzymes in young chicory plants (Cichorium intybus). Physiol. Plant. 106:158–163. http://dx.doi.org/10.1034/j.1399-3054.1999.106202.x. Dimitrakopoulos, A.P., Bemmerzouk, A.M., 2003. Predicting live herbaceous moisture con- tent from a seasonal drought index. Int. J. Biometeorol. 47:73–79. http://dx.doi.org/ 10.1007/s00484-002-0151-1. Dubois, M., Gilles, K.A., Hamilton, J.K., Rebers, P.A., Smith, F., 1956. Colorimetric method for determination of sugars and related substances. Anal. Chem. 28, 350–356. Edelman, J., Jefford, T.G., 1968. The mechanism of fructosan metabolism in higher plants as exemplified in Helianthus tuberosus. New Phytol. 67:517–531. http://dx.doi.org/ 10.1111/j.1469-8137.1968.tb05480.x. Figueiredo-Ribeiro, R.C.L., 1993. Distribuição, aspectos estruturais e funcionais dos frutanos, com ênfase em plantas herbáceas do cerrado. Rev. Bras. Fisiol. Veg. 5, 203–208. Forrest, J., Miller-Rushing, A.J., 2010. Toward a synthetic understanding of the role of phe- nology in ecology and evolution. Philos. Trans. R. Soc. Lond. Ser. B Biol. Sci. 365: 3101–3112. http://dx.doi.org/10.1098/rstb.2010.0145. Garcia, P.M.A., Asega, A.F., Silva, E.A., Carvalho, M.A.M., 2011. Effect of drought and re- watering on fructan metabolism in Vernonia herbacea (Vell.) Rusby. Plant Physiol. Biochem. 49:664–670. http://dx.doi.org/10.1016/j.plaphy.2011.03.014. Gupta, A.K., Kaur, N., 2000. Fructan metabolism in Jerusalem artichoke and chicory. In: Gupta, A.K., Kaur, N. (Eds.), Carbohydrate Reserves in Plants - Synthesis and Regula- tion. Elsevier, pp. 223–248. Hendry, G.A.F., 1993. Evolutionary origins and natural functions of fructans - a climatolog- ical, biogeographic and mechanistic appraisal. New Phytol. 123:3–14. http://dx.doi. org/10.1111/j.1469-8137.1993.tb04525.x. Hendry, G.A.F., Wallace, R.K., 1993. The origin, distribution, and evolutionary significance of fructans. In: Suzuki, M., Chatterton, N.J. (Eds.), Science and Technology of Fructans. CRC Press, Boca Raton, pp. 119–139. Hincha, D.K., Livingston, D.P., Premakumar, R., Zuther, E., Obel, N., Cacela, C., Heyer, A.G., 2007. Fructans from oat and rye: composition and effects on membrane stability dur- ing drying. Biochim. Biophys. Acta 1768:1611–1619. http://dx.doi.org/10.1016/j. bbamem.2007.03.011. Hisano, H., Kanazawa, A., Kawakami, A., Yoshida, M., Shimamoto, Y., Yamada, T., 2004. Transgenic perennial ryegrass plants expressing wheat fructosyltransferase genes ac- cumulate increased amounts of fructan and acquire increased tolerance on a cellular level to freezing. Plant Sci. 167:861–868. http://dx.doi.org/10.1016/j.plantsci.2004.05. 037. Isejima, E.M., Figueiredo-Ribeiro, R.C.L., 1993. Fructan variations in tuberous roots of Viguiera discolor Baker (Asteraceae): the influence of phenology. Plant Cell Physiol. 34, 723–727. Jermyn, M.A., 1956. A new method for the determination of ketohexoses in the presence of aldohexoses. Nature 177, 39. Krivorotova, T., Sereikaite, J., 2014. Seasonal changes of carbohydrates composition in the tubers of Jerusalem artichoke. Acta Physiol. Plant. 36:79–83. http://dx.doi.org/10. 1007/s11738-013-1388-5. Landhäusser, S.M., Lieffers, V.J., 2003. Seasonal changes in carbohydrate reserves in ma- ture northern Populus tremuloides clones. Trees 17:471–476. http://dx.doi.org/10. 1007/s00468-003-0263-1. Lapointe, L., 2001. How phenology influences physiology in deciduous forest spring ephemerals. Physiol. Plant. 113, 151–157. Legendre, P., Legendre, L., 2003. Numerical Ecology. second ed. Elsevier, Amsterdam. Lepš, J., Šmilauer, P., 2003. Multivariate Analysis of Ecological Data Using CANOCO. Cam- bridge University Press. Mantovani, W., Martins, F.R., 1988. Variações fenológicas das espécies do cerrado da Reserva Biológica de Moji Guaçu, Estado de São Paulo. Acta Bot. Brasilica 11, 101–112. Mellado-Mojica, E., López, M.G., 2012. Fructan metabolism in A. tequilana Weber Blue Va- riety along its developmental cycle in the field. J. Agric. Food Chem. 60:11704–11713. http://dx.doi.org/10.1021/jf303332n. Moraes, M.G., Carvalho, M.A.M., Franco, A.C., Pollock, C.J., Figueiredo-Ribeiro, R. de C.L., 2016. Fire and drought: soluble carbohydrate storage and survival mechanisms in 412 L.V. de Almeida et al. / Science of the Total Environment 598 (2017) 404–412 herbaceous plants from the Cerrado. Bioscience 662:107–117. http://dx.doi.org/10. 1093/biosci/biv178. Moser, T., Römbke, J., Schallnass, H.-J., Van Gestel, C.A.M., 2007. The use of the multivariate principal response curve (PRC) for community level analysis: a case study on the ef- fects of carbendazim on enchytraeids in terrestrial model ecosystems (TME). Ecotox- icology 16, 573–583. Oliveira, V.F., Silva, E.A., Zaidan, L.B.P., Carvalho, M.A.M., 2012. Effects of elevated CO 2 con- centration and water deficit on fructan metabolism in Viguiera discolor Baker. Plant Biol. 15:471–482. http://dx.doi.org/10.1111/j.1438-8677.2012.00654.x. Oliveira, R.S., Abrahão, A., Pereira, C., Teodoro, G.S., Brum, M., Alcantara, S., Lambers, H., 2016. Ecophysiology of campos rupestres plants. In: Fernandes, G.W. (Ed.), Ecology and Conservation of Mountaintop Grasslands in Brazil. Springer, Switzerland: pp. 1–567 http://dx.doi.org/10.1007/978-3-319-29808-5. Pereira-Júnior, A.C., Oliveira, S.L.J., Pereira, J.M.C., Turkman, M.A.A., 2014. Modelling fire frequency in a Cerrado savanna protected area. PLoS One 9, e102380. http://dx.doi. org/10.1371/journal.pone.0102380. Peres-Neto, P.R., Legendre, P., Dray, S., Borcard, D., 2006. Variation partitioning of species data matrices: estimation and comparison of fractions. Ecology 87, 2614–2625. Peshev, D., Vergauwen, R., Moglia, A., Hideg, E., Van den Ende, W., 2013. Towards under- standing vacuolar antioxidant mechanisms: a role for fructans? J. Exp. Bot. 64: 1025–1038. http://dx.doi.org/10.1093/jxb/ers377. Pollock, C.J., Jones, T., 1979. Seasonal patterns of fructan metabolism in forage grasses. New Phytol. 83:9–15. http://dx.doi.org/10.1111/j.1469-8137.1979.tb00720.x. Portes, M.T., Carvalho, M.A.M., 2006. Spatial distribution of fructans and fructan metabo- lizing enzymes in rhizophores of Vernonia herbacea (Vell.) Rusby (Asteraceae) in dif- ferent developmental phases. Plant Sci. 170:624–633. http://dx.doi.org/10.1016/j. plantsci.2005.10.017. Raccuia, S.A., Melilli, M.G., 2010. Seasonal dynamics of biomass, inulin, and water-soluble sugars in roots of Cynara cardunculus L. F. Crop. Res. 116:147–153. http://dx.doi.org/ 10.1016/j.fcr.2009.12.005. Rathcke, B., Lacey, E.P., 1985. Phenological patterns of terrestrial plants. Annu. Rev. Ecol. Syst. 16:179–214. http://dx.doi.org/10.1146/annurev.es.16.110185.001143. Ratter, J.A., Ribeiro, J.F., Bridgewatter, S., 1997. The Brazilian cerrado vegetation and threats to its biodiversity. Ann. Bot. 80, 223–230. Ribeiro, J.F., Walter, B.M.T., 2008. As principais fitofisionomias do bioma Cerrado. In: Sano, S.M., Almeida, S.P. de, Ribeiro, J.F. (Eds.), Cerrado Ecologia e Flora. Embrapa Cerrados, Planaltina, pp. 151–212. Rossatto, D.R., Sternberg, L.S.L., Franco, A.C., 2013. The partitioning of water uptake be- tween growth forms in a Neotropical savanna: do herbs exploit a third water source niche? Plant Biol. 15:84–92. http://dx.doi.org/10.1111/j.1438-8677.2012.00618.x. Silva, F.A.M. da, Assad, E.D., Evangelista, B.A., 2008. Caracterização climática do Bioma Cerrado. In: Sano, S.M., Almeida, S.P. de, Ribeiro, J.F. (Eds.), Cerrado Ecologia e Flora. Embrapa Cerrados, Planaltina, pp. 69–88. Silva, F.G., Cangussu, L.M.B., Paula, S.L.A., Melo, G.A., Silva, E.A., 2013. Seasonal changes in fructan accumulation in the underground organs of Gomphrena marginata Seub. (Amaranthaceae) under rock-field conditions. Theor. Exp. Plant Physiol. 25, 46–55. Silva, T.M., Vilhalva, D.A.A., Moraes, M.G., Figueiredo-Ribeiro, R.C.L., 2015. Anatomy and fructan distribution in vegetative organs of Dimerostemma vestitum (Asteraceae) from the campos rupestres. An. Acad. Bras. Cienc. 87:797–812. http://dx.doi.org/10. 1590/0001-3765201520140214. Suzuki, J.-I., Stuefer, J., 1999. On the ecological and evolutionary significance of storage in clonal plants. Plant Species Biol. 14:11–17. http://dx.doi.org/10.1046/j.1442-1984. 1999.00002.x. Ter Braak, C.J.F., Šmilauer, P., 2012. Canoco reference manual and user's guide: software for ordination, version 5.0. Microcomputer Power. Ueno, K., Ishiguro, Y., Yoshida, M., Onodera, S., Shiomi, N., 2011. Cloning and functional characterization of a fructan 1-exohydrolase (1-FEH) in edible burdock (Arctium lappa L.). Chem. Cent. J. 5:16. http://dx.doi.org/10.1186/1752-153X-5-16. Van Arkel, J., Vergauwen, R., Sévenier, R., Hakkert, J.C., van Laere, A., Bouwmeester, H.J., Koops, A.J., van der Meer, I.M., 2012. Sink filling, inulin metabolizing enzymes and carbohydrate status in field grown chicory (Cichorium intybus L.). J. Plant Physiol. 169:1520–1529. http://dx.doi.org/10.1016/j.jplph.2012.06.005. Van den Brink, P.J., Ter Braak, C.J.F., 1999. Principal response curves: analysis of time-de- pendent multivariate responses of biological community to stress. Environ. Toxicol. Chem. 18, 138–148. Van den Ende, W., 2013. Multifunctional fructans and raffinose family oligosaccharides. Front. Plant Sci. 4:247. http://dx.doi.org/10.3389/fpls.2013.00247. Van den Ende, W., Michiels, A., De Roover, J., Verhaert, P., Van Laere, A., 2000. Cloning and functional analysis of chicory root fructan 1-exohydrolase I (1-FEH): a vacuolar en- zyme derived from a cell-wall invertase ancestor? Mass fingerprint of the 1-FEH I en- zyme. Plant J. 24, 447–456. Vieira, C.C.J., Figueiredo-Ribeiro, R.C.L., 1993. Fructose-containing carbohydrates in the tu- berous root of Gomphrena macrocephala St.-Hil. (Amaranthaceae) at different pheno- logical phases. Plant Cell Environ. 16, 919–928. Vijn, I., Smeekens, S., 1999. Fructan: More Than a Reserve Carbohydrate? 120 pp. 351–359 Ward, J.H., 1963. Hierarchical grouping to optimize an objective function. J. Am. Stat. Assoc. 58, 236–244. Würth, M.K.R., Peláez-Riedl, S., Wright, S.J., Körner, C., 2005. Non-structural carbohydrate pools in a tropical forest. Oecologia 143, 11–24.