Variation of non-structural carbohydrates across the fast– slow continuum in Amazon Forest canopy trees

1. Tropical tree species span a range of life-history strategies within a fast– slow continuum. The position of a species within this continuum is thought to reflect a negative relationship between growth and storage, with fast-growing species allocating more carbon to growth and slow-growing species investing more in storage. For tropical species, the relationship between storage and life-history strategies has been largely studied on seedlings and less so in adult trees. 2. We evaluated how stored non-structural carbohydrates (NSC) vary across adult trees spanning the fast– slow continuum in the Peruvian Amazon by: (a) analysing whole-tree NSC in two species of contrasting growth and (b) investigating the relationships with key life-history traits across a broader set of species. 3. Our results are consistent with a growth– storage trade-off. The analysis of whole-tree NSC revealed that the slow-growing Eschweilera coriacea stored about 2.7 times as much NSC as the fast-growing Bixa arborea due to markedly higher storage in woody stems and roots. B . arborea also had higher seasonality in NSC, reflecting its strong seasonality in stem growth. Across a range of species, stem starch was negatively related to species growth rate and positively related to wood density.

1. Tropical tree species span a range of life-history strategies within a fast-slow continuum.The position of a species within this continuum is thought to reflect a negative relationship between growth and storage, with fast-growing species allocating more carbon to growth and slow-growing species investing more in storage.For tropical species, the relationship between storage and life-history strategies has been largely studied on seedlings and less so in adult trees.
2. We evaluated how stored non-structural carbohydrates (NSC) vary across adult trees spanning the fast-slow continuum in the Peruvian Amazon by: (a) analysing whole-tree NSC in two species of contrasting growth and (b) investigating the relationships with key life-history traits across a broader set of species.
3. Our results are consistent with a growth-storage trade-off.The analysis of whole-tree NSC revealed that the slow-growing Eschweilera coriacea stored about 2.7 times as much NSC as the fast-growing Bixa arborea due to markedly higher storage in woody stems and roots.B. arborea also had higher seasonality in NSC, reflecting its strong seasonality in stem growth.Across a range of species, stem starch was negatively related to species growth rate and positively related to wood density.

| INTRODUC TI ON
Amazon forest tree species vary markedly in their life-history strategies, exhibiting an approximately 50-fold variation in stem growth rates (Coelho de Souza et al., 2016a).This fast-slow growth continuum fundamentally shapes Amazon forest ecology, driving basin-wide patterns of tree mortality by means of a pervasive growth-survival trade-off (Esquivel-Muelbert et al., 2020), which in turn controls regional biomass storage (Johnson et al., 2016).The position of a species along the fast-slow growth continuum may have important consequences for resistance to climate stressors (Oliveira et al., 2021) and to disease (Fine et al., 2006), but the underlying mechanisms which underpin different life-history strategies remain unclear.Photosynthetic assimilate is predominantly used for respiration, growth or storage.Whereas the respiration:photosynthesis ratio is narrowly constrained across species, growth rate and storage vary more freely, giving rise to a trade-off between carbon allocation to growth on the one hand (fast-growing species), and carbon allocation to storage and defence on the other hand (slowgrowing species; King et al., 2006;Van Oijen et al., 2010).This tradeoff has been postulated to drive differences in life-history strategies of tropical trees (Kitajima, 1996;Poorter & Kitajima, 2007).
Combinations of plant functional traits result in particular lifehistory strategies (Salguero-Gómez et al., 2016).For example, hardwooded species which invest in wood of high mechanical strength have slow growth rates and are expected to have increased investment in storage (Herrera-Ramírez et al., 2021;King et al., 2006).This is often an advantageous strategy in habitats with low nutrient availability and other environmental stress, as it enhances plant capacity to survive pathogen attack, biomass loss or abiotic stress (Aleixo et al., 2019;O'Brien et al., 2014;Oliva et al., 2014;Quentin et al., 2015).In contrast, soft-wooded fast-growing species have high water demand and photosynthetic rates, enabling faster growth, but reducing resistance to drought-induced embolism (Chave et al., 2009;Hacke et al., 2006;Mcculloh et al., 2012;Oliveira et al., 2021) and may be associated with lower carbon storage, that is non-structural carbohydrates (NSC; Falchi et al., 2020;Herrera-Ramírez et al., 2021;Poorter & Kitajima, 2007).However, there have been very few empirical studies exploring the relationship between storage and life-history strategies, particularly in adult tropical trees.
Non-structural carbohydrates are the major form of energy storage in many plant species and have been shown to play a critical role in maintaining hydraulic function and enhancing survival during drought (Guo et al., 2020;Liu et al., 2018;O'Brien et al., 2014;Tomasella et al., 2017).Starch and soluble sugars comprise the most significant portions of NSC reserves in most tree genera (Martínez-Vilalta et al., 2016), but other compounds such as lipids also play an important role as storage (Herrera-Ramírez et al., 2021;Hoch et al., 2003).Soluble sugars are involved in multiple functions such as supporting new growth, for respiration and defence, regulating stress-related genes and acting as osmoprotectants mitigating the negative effect caused by water stress (Krasensky & Jonak, 2012;MacNeill et al., 2017;Rosa et al., 2009;Sapes et al., 2019;Signori-Müller et al., 2021a).Starch is a storage compound for future use that under challenging environmental conditions can be remobilized to soluble sugars to sustain the necessary carbon supply for metabolic functions (Aubrey & Teskey, 2018;MacNeill et al., 2017;Signori-Müller et al., 2021a;Sulpice et al., 2009;Thalmann & Santelia, 2017).
Our current understanding of how NSC interacts with the fastslow tree species continuum is fragmented and has largely been informed by studies on seedlings.Studying species with different life-history strategies, Poorter and Kitajima (2007) found that stem sugar concentrations and pools of seedlings of species occurring in moist tropical forests in Bolivia were positively correlated with survival but negatively correlated with stem growth.Recently, O'Brien et al. (2020) used 13 C pulse labelling to track sugar and starch movement through seedlings of two contrasting conservative and acquisitive species from Borneo.Their results showed clear differences in carbon allocation and dynamics between species under experimental drought, as the conservative species allocated more sugars to stems and roots.However, patterns observed in seedlings may not be readily extrapolated to mature trees (Hartmann et al., 2018).For example, Signori-Müller et al. (2021a) measured leaf and branch NSC concentrations in 82 Amazonian tree species and found no relationship between these and species-level growth and mortality rates.
However, that study did not consider the major plant woody storage organs, including stems and coarse roots (Würth et al., 2005).The elucidation of the functional role of storage in the fast-slow continuum requires assessing NSC dynamics across multiple plant organs in adult trees of species with contrasting life-history strategies.
A priori, slow-growing species would be expected to have greater NSC concentrations/pools in longer term storage tissues (e.g.stem, coarse roots) than fast-growing species which primarily allocate carbon to growth over storage (Herrera-Ramírez et al., 2021;Smith & Stitt, 2007).Fast-growing species may also have more seasonal NSC dynamics as these have been found to concentrate most of 4. Given the role of NSC in mediating plants' response to stress, our results suggest that slow-growing species with greater storage reserves may be more resilient to drought than fast-growing species.

K E Y W O R D S
fast-growing, slow-growing, soluble sugars, starch, tropical their growth in the wet season, compared to slow-growing species (Rowland et al., 2014).
In this study, we tested the hypotheses that (a) slow-growing species have higher NSC concentrations, especially in woody storage organs; (b) fast-growing species have greater seasonal variability in NSC content; and (c) concentrations of NSC and its components are positively correlated with wood density but negatively correlated with growth and mortality rates.To address these hypotheses, we worked within a mature tropical moist forest in south-western Amazonia.To test hypotheses (a) and (b), we analysed total NSC concentrations (NSC T ) and its components (i.e.starch and soluble sugars) in 17 different tree compartments in two contrasting evergreen species over two seasons.We worked with the slow-growing, widespread Neotropical species Eschweilera coriacea (DC.) S. A.
Mori, the most common non-palm Amazonian tree species, and the fast-growing widespread Neotropical species Bixa arborea Huber.To test hypothesis (c), we collected stem samples from 11 species and used available leaf and branch NSC data from 21 species in our study site spanning a broad range of wood density and growth rates.

| Field collection and sampling strategy
We performed our sampling in the Tambopata National Reserve, in Madre de Dios, southern Peru (12°49′S, 69°16′W).
Permit for sampling collection was granted for Dirección de Gestión de las Áreas Naturales Protegidas (SERNANP) no.
039-2016-SERNANP-RNTAMB-PRD, and permit to import samples from Peru to Brazil to carry out the NSC analysis was conceded by Ministério da Agricultura, Pecuária e Abastecimento, Oficio nº 334/2016/SSV-SP -MAPA.This site has a mean annual precipitation of 2,451 mm/year and a dry season length, that is ≤100 mm/month of approximately 3 months, from June to August (Signori-Müller et al., 2021a).At this site we selected two contrasting evergreen canopy tree species, Eschweilera coriacea (Lecythidaceae) and Bixa arborea (Bixaceae).E. coriacea is one of the most dominant species in the Amazon Basin (Fauset et al., 2015) and is a shade-tolerant species with low mean diameter growth rate (mean basin-wide growth rate of 0.17 cm/ year) and high wood density (0.85 g/cm 3 ; Chave et al., 2009;Coelho de Souza et al., 2016a;Zanne et al., 2009).B. arborea is a lightdemanding tree species with high diameter growth rates (mean basin-wide growth rate of 0.56 cm/year) and low wood density (0.37 g/cm 3 ; Chave et al., 2009;Coelho de Souza et al., 2016a;Zanne et al., 2009).These species also differ in their seasonal stem growth patterns, with B. arborea exhibiting markedly higher growth rates in the wet season while E. coriacea does not display seasonality in growth (Figure 1).We collected samples from three individuals per species, all mature trees with canopies exposed to full sunlight and without liana infestation or visible structural damage.For both species, all sampled trees have DBH between 20 and 30 cm, common diameters for adult trees of these species in our study plot [Rowland et al., 2014;B.arborea mean ± SE 18.87 ± 1.46 cm (n = 18); E. coriacea 19.36 ± 2.52 cm (n = 8)].The selected trees were sampled in the wet (January 2017) and in the dry (July 2017) seasons, and since NSC concentrations may vary depending on the time of the day, especially in leaves (Tixier et al., 2018), we standardized the sampling time to be between 08:30 and 10:00 a.m. for both species.For each tree, we sampled leaves, branches, stem wood, coarse roots and fine roots, in different tree strata.In the field and during the transport to the laboratory, samples were kept on ice, with the time taken to the laboratory from the plot being approximately 45 min.Upon arrival at the laboratory, we microwaved the samples for 90 s at 700 W to stop enzymatic processes and oven-dried at ~60°C for at least 48 hr or until they were completely dry (no more than 72 hr).
As there is no consensus about the within-organ variability in NSC concentration across studies (Furze, Huggett, et al., 2018;Smith, Miller, et al., 2018;Würth et al., 2005), we performed a detailed sampling within organs to ensure that the within-organ variation in NSC concentrations was well-captured in the two contrasting species.We sampled leaves and branches from three different posi-  (n = 2-3 trees per species).Of the 11 species we sampled for stem wood NSC, eight were also sampled for leaf and branch NSC concentrations (Table 1).
We relate the NSC concentrations to species-level wood density, growth and mortality rates compiled in previous studies (Coelho de Souza et al., 2016b) and in situ branch wood density measurements.
The species used for these analyses span a wide range of wood density, mean growth rates and mortality rates (Table 1).Species for

| Branch wood density
To estimate in situ branch wood density, we sampled 2-3 individuals of each species, we used fully sun-exposed top canopy branches from the same trees NSC were analysed.Branch wood density was measured using the displacement method (Pérez-Harguindeguy et al., 2016).We first cut a branch segments of ~2.5 cm length and 1.2 cm diameter and removed the bark, then we placed the samples in a recipient with filtered water to rehydrate for 24 hr.Subsequently, with the help of a small needle we completely submerged the samples under filtered water in a recipient on a scale (three decimal) to register variations in the weight (volume cm 3 ).Samples were then oven-dried for 48-72 hr at ~60°C.After drying, we measured the segments' dry weight (g) with a scale (three decimal).Wood density is expressed as the ratio of wood dry mass and wood fresh volume (g/cm 3 ).

| Laboratory preparations and NSC analyses
Non-structural carbohydrates are defined here as free, low molecular weight sugars (glucose, fructose, sucrose, etc) plus starch.The aliquot that we initially separated was incubated overnight to react with amyloglucosidase from Aspergillus niger (Sigma-Aldrich) to break down the total NSC to glucose.Thereafter total glucose (corresponding to NSC) was determined as described above and starch was calculated as total NSC minus soluble sugars.All NSC values are expressed in mg/g dry mass.

| Statistical analysis
We performed all statistical analysis using the R software (R Core Team, 2021; www.R-proje ct.org; version 3.4.4)and its base packages.We tested for normality (Shapiro-Wilk) and homogeneity of variances (Fligner-Killeen) for each NSC fraction (NSC T , soluble sugars and starch) and plant compartment.As NSC T , starch and soluble sugars were not normally distributed and often contained values close or equal to zero, they were either log1p [y i = log e (x i + 1)] transformed prior to parametric analysis or used without transformation in nonparametric analyses, as described below.
To test for differences between different levels within specific organs (e.g.sunlit, partially shaded and shaded leaves or stem wood samples collected at 5 cm, 1.30 m, 10 m and maximum height), ANOVA was used following log1p transformation.As there were no differences between levels within organs, species differences and the seasonal variation were evaluated only at the organ scale (i.e.leaf, branch, stem wood, radial profile, root and fine roots) and not separately for each level within an organ (e.g.sunlit leaves).For these analyses, we used the mean organ concentrations across all levels for which we had data (e.g.stem wood samples at 5 cm, 1.30 m, 10 m and maximum height).
Seasonal changes in NSC T , starch and soluble sugar concentrations within species were tested separately for each organ using nonparametric paired-sample Wilcoxon tests.To evaluate differences between E. coriacea and B. arborea, we used the nonparametric Kruskal-Wallis test.Relationships between NSC (NSC T , starch and soluble sugars) and species traits (wood density, growth and mortality rates) were investigated using bivariate linear regression models lm() (R base function), following log1p transformation for both NSC and traits.For all analyses, we assume a significance level of 0.05.
To calculate whole-tree biomass-weighted mean NSC T concentrations, we assumed that roots represented 21% of the total biomass for both species (Houghton et al., 2001).Above-ground biomass was allocated across branches, leaves and stems following Marra et al., 2016, who presented biomass allocation ratios separately for pioneer and late-successional Amazonian tree species (Table 2).As we did not sample fine roots for B. arborea, we use only coarse root NSC T data to scale up to the whole tree.
For stem biomass concentrations, we took the mean of all sampled stem components (top to bottom and outer and inner stem segments).To assess the sensitivity of our whole-tree estimates, we also computed biomass-weighted mean NSC concentrations using the biomass distribution values assumed in Würth et al. (2005) (Table S1).

| Whole-tree NSC patterns in E. coriacea and B. arborea
We did not find differences in NSC T , starch and soluble sugar concentrations across light levels (sunlit, partially shaded and shaded) in leaves or branches for either of our focal species (Figure 2; Figures S1   and S2).Similarly, no differences were observed across stem samples obtained at different heights (close to the ground, 1.30 m, 10 m and max.height) or across coarse root samples obtained at different depths (5, 55 and 105 cm).We found no difference across radial profile segments for both species in most NSC fractions evaluated, with the exception that soluble sugar concentrations in E. coriacea were two times lower in the innermost segments compared to the outermost segments (Figure S2; p < 0.05).
Patterns within organs revealed that NSC concentrations in the slow-growing species are less seasonal than in the fast-growing species (Figure 3).In the slow-growing E. coriacea, only branch NSC varied seasonally, with all fractions decreasing considerably from wet to dry season (declines by 70%, 85% and 54% for NSC T, starch and soluble sugars respectively).In the fast-growing B. arborea, NSC in all organs displayed seasonality in at least one fraction (Figure 3), with pronounced dry season decreases in NSC T in leaves (−49%) and branches (−64%) but with increases in the stem wood (+216%) and in the stem radial profile (+101% a WD branch : wood density (g/cm 3 ) measured in branches of trees from the Tambopata Nacional reserve.
b WD stem : wood density (g/cm 3 ) measured in the stem.Data referred to the mean species WD for the Amazon Basin.Data from the Global Wood Density database (Chave et al., 2009;Zanne et al., 2009 F I G U R E 3 Seasonal differences in non-structural carbohydrates (NSC) for each species and organ.Total NSC (NSC T ), starch and soluble sugar concentrations in the fast-growing Bixa arborea (left) and slow-growing Eschweilera coriacea (right).Asterisks denote significant differences between concentrations in wet (blue) and dry (red) seasons calculated using a paired-sample Wilcoxon test.Vertical bars denote 1 standard error of the mean from wet to dry season in branches (−68%) and roots (−59%), but increased in stem wood (+293%).Soluble sugars increased in coarse roots (+73%), stem wood (+149%) and in the stem radial profile (+112%) during the dry season, while in leaves soluble sugar concentrations decreased by 51% (Figure 3).

| Relationship between NSC and traits across a broader set of species
We found that leaf NSC T , starch and soluble sugar concentrations are not significantly related to branch wood density, mean growth rate and mortality rate (Figure 4; Figures S4 and S5) across the locally occurring species for which we had data.Branch starch concentrations are negatively related to branch wood density (Figure 4; p = 0.04, R 2 = 0.21), but not with other traits or NSC fractions (Figures S4 and S5).In contrast, stem starch concentrations across the 11 species for which we had data were positively related to stem wood density (p = 0.007; R 2 = 0.57) and negatively related to mean growth rate (p = 0.001; R 2 = 0.48).

| DISCUSS ION
Overall, our results provide support for our three overarching hypotheses as we found that: (a) mean organ-level NSC concentrations in the slow-growing E. coriacea was 2.7 times greater than in the fast-growing B. arborea, with much of the difference between them attributable to storage in woody stem and root organs (Table 2); (b) NSC dynamics in B. arborea was markedly more seasonal than in E. coriacea; and (c) stem NSC T and starch were positively related to wood density but negatively related to growth rate across a broader set of species.Taken together, these findings are consistent with the alignment of NSC dynamics with life-history strategies and are in line with the existence of a growth-storage trade-off in adult tropical trees (Kitajima, 1996;O'Brien et al., 2020;Poorter & Kitajima, 2007;Van Oijen et al., 2010).

| Within-tree NSC concentrations
The apparent lack of variation in NSC within organs (Figure 2; Figures S1 and S2) is reassuring for future studies that seek to quantify community patterns in NSC storage, as it suggests that measurements for a single level of a given organ (e.g. the outer part of the stem or the uppermost coarse roots) represent overall organ NSC concentrations well.In the case of vertical profiles of roots and stems, the observed homogeneity with depth/height may indicate a uniform demand for NSC across the profile-for example through similar respiration requirements on a per mass basis throughout the profile.The lack of difference between sunlit and shaded leaves was more unexpected, given the differences in metabolic rates that have been observed between sunlit and shaded leaves (Mathur et al., 2018).However, our results are consistent with a previous assessment along an elevational transect which found that sunlit and shaded leaves differed in leaf mass per area and other leaf traits but not in foliar NSC concentrations (Martin et al., 2020).
The largest differences between species were in starch concentration in the roots and in the inner portions of the stem wood (radial profile), as these were 31 and 56 times higher in the slowgrowing E. coriacea than in B. arborea (Figure 2; Figure S3).The high stem storage of NSC in slow-growing shade-tolerant species is consistent with previous work on seedlings in moist Amazon forests (Poorter & Kitajima, 2007), implying that a similar growth-storage trade-off axis also applies to adult trees within moist tropical forests.Slow-growing, shade-tolerant trees endure strong light limitation as they grow through the canopy and high NSC concentrations are needed to ensure survival during periods of low light (Poorter & Kitajima, 2007).Being longer lived, slow-growing species are also more likely to experience additional stress such as drought.Our results are consistent with a recent study in southern Amazonia which also found that stem NSC storage was higher in slow-growing species (Herrera-Ramírez et al., 2021).Interestingly, a previous study on a semi-deciduous forest in Panama (Würth et al., 2005) did not find a clear relationship between life-history strategy (early and midsuccessional species) and NSC storage patterns.However, the differences in life-history strategy considered in that study were not as pronounced as between the two focal species in our study and the semi-deciduous nature of that Panamanian site (Kunert et al., 2021) may have made any growth-storage relationship more difficult to detect.
Very few studies in tropical forests have evaluated NSC concentrations in roots (Newell et al., 2002;Würth et al., 2005).We found that coarse root NSC and starch concentrations in E. coriacea were the highest recorded across any organ.Whether this is indicative of a specific life-history strategy or a passive consequence of the slow growth rates in this species is unclear.However, many studies suggest that roots NSC have an important role in plant tolerance to stress (Kannenberg et al., 2018;Mei et al., 2015); and a study based on 14 C isotopic analysis shows that relatively old carbon stores can be used to produce new fine roots after environmental disturbance in tropical species (Vargas et al., 2009).This implies that long-term reserves, especially below-ground storage of NSC, may be particularly important for plant survival of slow-growing, shade-tolerant species with significant NSC reserves.However, further sampling is needed to establish the generality of this finding across other species.
Despite the large differences in NSC and starch concentrations between species, soluble sugar concentrations during the dry season were similar between species.These results reinforce the view that soluble sugars, which are used for immediate metabolic activity and not as long-term stores, should be kept above certain thresholds for the maintenance of plant function (Sala et al., 2012;Signori-Müller et al., 2021a).
F I G U R E 4 Relationship between starch (mg/g) and wood density (WD; g/cm 3 ), mean growth rate (Growth cm/year) and mortality rate (Mortality %).From top to bottom, relationships are presented for leaves, branches and stems.For relationships between stem starch and WD, we used stem WD, and for relationships with leaf and branch starch, we used branch WD.

| Seasonal patterns
A previous study in our sampling site evaluated seasonal patterns of tree growth and found that dry season declines in precipitation and soil water content resulted in reduced tree growth, but that the seasonality of growth varied greatly across tree species (Rowland et al., 2014).Seasonality in growth was found to be most pronounced for fast-growing species, suggesting that while plants with a fast strategy have the capability for greater growth in wet season conditions, it is more constrained during the dry season (Rowland et al., 2014).Slow-growing species, on the other hand, were found to have less seasonal growth patterns (Rowland et al., 2014).The differences in seasonality of NSC concentrations between species reflect these contrasting growth patterns (Figure 1) as the fastgrowing B. arborea was characterized by strong seasonality in almost all components while the slow-growing E. coriacea was characterized by an overall lack of seasonality in NSC, except in canopy organs (Figure 3).
Both species showed dry season decreases in NSC concentration in the canopy (leaves and branches).These results were especially consistent for branches, which presented a dry season decrease of ~65% in starch and soluble sugar concentrations in both fast-and slow-growing species.The consumption of branch NSC in the dry season in both species suggests that this may be a phenomenon that is independent of life-history strategy and may represent a mechanism to satisfy increased maintenance demands in the dry seasonfor example a greater requirement for osmoprotection.Branch NSC consumption is frequently reported in trees, both in temperate or tropical species (Furze, Trumbore, et al., 2018;Martínez-Vilalta et al., 2016;Newell et al., 2002;Würth et al., 2005), suggesting that canopy reserves act as an important carbon buffer during unfavourable periods.Indeed, Signori-Müller et al. (2021a) showed that in sites with more marked dry season almost all species exhibit declines in leaf and branch starch during the dry season, although this effect was less pronounced in more mesic sites such as Tambopata.
The strong depletion of stem starch and soluble sugar concentrations in the wet season in B. arborea and not in E. coriacea likely reflects the strong differences in the seasonality of growth between the two species (Figure 1; Martínez-Vilalta et al., 2016;Smith, Miller, et al., 2018).Some studies have found that under water limitation conditions growth may cease before photosynthesis, leading to NSC accumulation (Fatichi et al., 2014;Körner, 2003;Muller et al., 2011).
The build-up of NSC in B. arborea stem wood in the dry season may therefore arise from decreases in carbon demand whereby photosynthate is not utilized for growth.On the other hand, the lack of seasonality in stem NSC observed in E. coriacea is consistent with the aseasonality in growth observed for that species.In a recent experimental study with seedlings of two tropical species of contrasting growth rates and drought sensitivity, O'Brien et al. ( 2020) found that growth of seedlings of the fast-growing species was more restricted under water stress and that this led to a passive accumulation of NSC in the stem.Our results suggest that such a mechanism may also play out in mature trees, but further studies investigating the seasonality of NSC in stem tissues across a spectrum of life-history strategies are needed to establish the generality of this pattern.
We found that root starch concentrations in the fast-growing B.
arborea decreased by 59% during the dry season in synchrony with an accompanying increase in root-soluble sugars, thus suggesting starch to soluble sugar remobilization (Figure 2).This pattern may be associated with an increased requirement for regulation of osmotic potential and maintenance of water transport from the roots to the canopy (Sánchez et al., 1998).The reduction in starch could potentially also be associated with the production of new roots to enhance water absorption in the dry season or to sustain root respiratory demands.'Fast' species have higher respiration rates than 'slow' species (Reich, 2014) and maintaining these rates during the unfavourable season may require consumption of root starch reserves.
There have been very few studies to date examining seasonal patterns in root reserves in tropical trees.However, the limited number of studies to date has not found strong evidence of seasonal root starch consumption in mature tropical trees under baseline seasonality conditions (Newell et al., 2002;Würth et al., 2005), or during exceptionally dry years (Mauro Brum, personal communication).
Further studies are needed to understand whether the dry season root starch consumption in B. arborea is also observed in other fastgrowing species.

| Relationship with species traits
Our analyses of NSC concentrations versus species traits are consistent with the hypothesis of a growth-storage trade-off in adult tropical trees, as has been reported previously for seedlings (Poorter & Kitajima, 2007).As expected, we found a positive relationship between starch concentrations in the stem and wood density and a negative relationship with annual growth rates (Figure 4).This is consistent with other studies that associate higher NSC storage with species that have lower growth and mortality rates (Herrera-Ramírez et al., 2021) and hard-wooded tissue with greater resistance to attack by pathogens (Larjavaara & Muller-Landau, 2010).The maintenance of high stocks of starch may be part of a set of ecological strategies, such as high hydraulic safety (Anderegg et al., 2016) that confer lower mortality risks to slow-growing trees (Esquivel-Muelbert et al., 2020).The larger NSC storage in slow-growing trees could increase their ability to tolerate unusually stressful situations through osmoregulation or recovery after periods of water deficit (Nardini et al., 2017;Pagliarani et al., 2019;Savi et al., 2016;Tomasella et al., 2017) and enhance their chances of survival through resprouting, as has been broadly demonstrated to occur in species with high NSC content (Hoffmann et al., 2004;O'Brien et al., 2020;Poorter & Kitajima, 2007;Shibata et al., 2016;Smith, Arndt, et al., 2018).
Our results show that fast-growing species do not maintain sizeable long-term NSC stores in the stem but do maintain NSC reserves in the canopy (i.e.leaves and branches).The positive relationship between branch starch and branch wood density is in line with a study from Panamá (Dickman et al., 2019), suggesting branches tions of the canopy: (a) sunlit, (b) partially shaded and (c) fully shaded.Sunlit was defined as the outermost branches of the top third of F I G U R E 1 Mean increment in diameter (mm/month) during the wettest quarter (January to March) and driest quarter (July to August) for Bixa arborea and Eschweilera coriacea, measured with stem dendrometer bands from January 2006 to December 2010 in the Tambopata National Reserve (plot TAM-05).Data are from Rowland et al. (2014), B. arborea n = 18 trees and E. coriacea n = 8 trees.Each box encompasses the 25th to 75th percentiles; the median is indicated by the horizontal line within each box while external horizontal lines indicate the 10th and 90th percentiles; dots indicate outliers.To test for differences between seasons, we used paired-sample Wilcoxon tests p partially sunlit was defined as the outer branches of the middle third of the canopy and shaded was defined as the most shaded branches found in the interior of the canopy.We used a 4.3mm increment borer to sample coarse roots and stem wood (Haglöf Company Group).Roots were sampled by excavating close to the junction between the roots and the main trunk, and following and sampling the same coarse roots in three different depths: 5, 55 and 105 cm.At the deepest point, we also collected fine roots (diameter <3 mm).The fine-root sampling was performed only in the slowgrowing E. coriacea, as in the fast-growing B. arborea we did not find fine roots at any depth up to 105 cm (the deepest point sampled).Stem wood samples were obtained at four different heights: (a) close to the ground (5 cm height), (b) at breast height (1.30 m), (c) at approximately 10 m and (d) at maximum stem height above the ground.The maximum height was defined as the highest point in the stem before forking was observable and thus varied from tree to tree.At 5 cm, 10 m and max.height, we cored the trees and took samples ~5 cm long; at breast height, we cored the trees to the centre of the stems.As the two focal species have different growth rates, we did not use the same stem wood core length for the NSC analyses as this would represent a different number of growth years and environmental conditions for each species.As neither species produces annual growth rings, we calculated the growth rate of each species using inventory data from our study plot (TAM-05; ForestPlots.net;Lopez-Gonzalez et al., 2011).The annual mean growth in DBH (1.3 cm) for the slow-growing E. coriacea in this site was 0.26 cm/year, compared to 0.42 cm/year for the fast-growing B. arborea in our study plot.For NSC analysis, we removed the bark and phloem and used segment lengths corresponding to ~5-year growth increments (0.26 cm × 5 years = 1.3 cm for E. coriacea and 0.42 cm × 5 years = 2.1 cm for B. arborea).Stem wood cores obtained at breast height were divided into multiple 5-year growth segments (1.3 cm in length for E. coriacea and 2.1 cm in length for B. arborea).As DBH differed across individuals, the total number of radial segments (thereafter 'radial profile') varied among individuals.Therefore, we used only the first five segments (representing ~25 years of growth in total), which were common among all trees, to statistically evaluate radial variation in NSC storage (Figure2).

2. 2 |
Stem, leaf and branch NSC from other local speciesLogistical and financial constraints meant that it was not possible to conduct an intensive multi-compartment sampling of NSC across more than two species.To further investigate whether there is a relationship between NSC and life-history strategy in our study site, we: (a) sampled stem wood from nine additional species (n = 11 species; n = 5-6 trees per species, except for B. arborea and E. coriacea n = 3); and (b) made use of leaf and branch-level NSC data collected in Tambopata by Signori-Müller et al. (2021a, 2021b) for 21 species Whole-tree NSC T concentrations (mg/g, mean ± SE) in the fast-growing Bixa arborea (left) and in the slow-growing Eschweilera coriacea (right).The top box depicts the biomass-weighted mean whole-tree NSC T concentrations for the fast (green box) and slow-growing species (purple box).Components with relatively high storage in each species are shown in dark orange (canopy for B. arborea and stems and roots for E. coriacea).Mean NSC values were calculated using samples from both the dry and wet seasons (n = 3 trees per species) which we have stem NSC concentration (n = 11) represent 30% of the total plot basal area while species sampled for leaf and branch NSC concentrations (n = 21) represent 47% of the total plot basal area.All stem wood, branch and leaf samples were collected within RAINFOR permanent plots in the Tambopata National Reserve (ForestPlots et al., 2021; Lopez-Gonzalez et al., 2011) during the wet season in January 2017.Stem wood samples for the broader set of species were obtained at breast height (1.30 m).Unlike for B. arborea and E. coriacea, we did not core up to the centre of the tree but took cores of 3-6 cm in length.For these species, we used the same methodology as for B. arborea and E. coriacea, whereby we calculated the mean annual diameter growth of each species based on local inventory data and took cores corresponding to 5 years of growth.The branch and leaf NSC data available inSignori-Müller et al. (2021a, 2021b) were collected earlier in the morning (before 6:00 a.m.) than the B. arborea and E. coriacea samples collected specifically for this study (08:30-10:00 a.m.).As leaves and branches may exhibit significant diurnal variation in NSC(Tixier et al., 2018), we only used the canopy NSC data fromSignori-Müller et al. (2021a, 2021b) to test for relationships with life-history attributes across the broader spectrum of species.
NSC was analysed as described inHoch et al. (2002) with minor modifications, as inRowland et al. (2015) andSignori-Müller et al. (2021a).First, we diluted 15 mg of the ground plant material with 1.6 ml of distilled water and then incubated in a water bath at 90-100°C for 60 min to solubilize sugars.Then we took an aliquot of 700 μl from each sample.We used the remaining aliquot volume (900 μl) to determine soluble sugar concentrations using invertase from Saccharomyces cerevisiae (Sigma-Aldrich) to break down sucrose and fructose to glucose.Additionally, for both reaction routines, we used GAHK (Glucose Assay Hexokinase Kit -Sigma-Aldrich) together with phosphoglucose isomerase from Saccharomyces cerevisiae (Sigma-Aldrich).The concentration of free glucose was measured photometrically in a 96-well microplate spectrophotometer at 340 mm (EPOCH -Biotek Instruments INC).
Sampled species and their traits ). Starch concentration decreased TA B L E 1 Houghton et al. (2001)lculated using ordinary linear regression.Best-fit lines and p-values are displayed only for significant relationships.The relationship between stem starch and mortality rate are marginally significant (p = 0.08).Purple dots represent the slow-growing E. coriacea and green dots represent the fast-Biomass-weighted whole-tree NSC T concentrations, calculated from the mean NSC T concentrations in each organ and weighted by the fractional contribution of each organ to total plant biomass.Biomass allocations are obtained fromMarra et al. (2016), for pioneer species (Bixa arborea) and for late-successional species (Eschweilera coriacea).The contribution of below-ground biomass is based on the estimates ofHoughton et al. (2001)