Effect of substrate fertility on tank-bromeliad performances

Members of the plant family Bromeliaceae can uptake nutrients directly from their leaves via leaf absorbing trichomes and their roots have long been reduced to anchorage function, thus overlooked. Recently, evidence has accumulated for a significant role for the roots of some species of tank bromeliads in both water and nutrients absorption. However, to date, little attention has been paid to the importance of the substrate fertility for the structure of the roots and the growth and performances of tank bromeliads. This study investigated the effect of substrate fertility on Aechmea aquilega regarding leaf and root traits, nutrient content, and growth. Seeds of this tank bromeliad were sowed in a greenhouse in French Guiana in three different substrates: a nutrient-poor, a nutrient-rich and a mixed substrate. The performances of 15-month-old A. aquilega were assessed by measuring leaf and root traits related to nutrient acquisition and resources capture. We show that plants growing in nutrient-poor substrate grew twice slower and were smaller than plants grown on the nutrient-rich substrate with fewer leaves and roots, lower total dry mass, and smaller leaves and root length. Overall, 70% of measured traits responded significantly to the experimental treatments indicating that the response of A. aquilega to nutrient availability is a combination of physiological processes, leaf and root structure, and chemistry. This study is the first to show that the fertility of the substrate on which the bromeliad A. aquilega grows has a strong and lasting effect on the plant performances and may be a relevant factor for bromeliad ecology.

water for their growth and survival. Soil is the main source of water and nutrients for most plants, but as nutrients are unevenly spatially distributed, the availability of nutrients for plants also varies (Cain et al. 1999) as do the plants response strategies (Bloom et al. 1985;Kraiser et al. 2011). Ecologists use the functional traits of leaves, stems and roots to obtain information on the resource (i.e., light, water, nutrients) acquisition strategies of species (Sterck et al. 2011). Cross-species comparisons have revealed major axes of trait variations reflecting strategies (i.e. leaves and roots economic spectra) which range from rapid acquisition to conservation of resources (Roumet et al. 2016;Wright et al. 2004). The same range can also be found within a single plant species across environmental gradients Delpiano et al. 2020;Fajardo and Siefert 2018;Hajek et al. 2013;Hayes et al. 2019;Isaac et al. 2017). Plants facing harsh conditions (i.e. low resources availability) are expected to have more conservative traits than plants growing in nutrient-rich environments (Lambers and Poorter 1992;Wright et al. 2004). Functional trait approaches have mainly been used in trees and herbaceous species (e.g., Fort et al., 2013;Freschet et al., 2013) while such studies of epiphytes are rare (Oliveira et al. 2021;Richards and Damschen 2021).
Epiphytes, plants that live non-parasitically on trees, have to cope with heterogeneous and intermittent nutrient and water availability (Zotz 2016). These plants have consequently developed numerous morphological adaptations (e.g., leaf-absorbing trichomes, velamen radicum, water-storage tanks, pseudo-bulbs) to optimise water and nutrient uptake and conservation (Males 2016;Zotz 2016). Epiphytic plant species can be obligate or facultative and can be found growing on different substrates including bark, rocks, canopy soil, or the ground (Wu et al. 2020;Zhang et al. 2021). Such different substrates can modify the morphological, anatomical, physiological and stoichiometric traits of conspecific individuals (Chen et al. 2019;Lu et al. 2015;Wu et al. 2020;Zhang et al. 2021). Hoeber & Zotz (2021) recently found that accidental epiphytic individuals performed better than terrestrial conspecifics due to the beneficial growth conditions in the forks of branches filled with organic-rich arboreal soil. The nutrient use strategies of facultative epiphytes have been shown to be flexible, thus enabling facultative epiphytes to exploit different substrate interchangeably (Wu et al. 2020;Zhang et al. 2021).
Plants belonging to the Bromeliaceae family display many remarkable morphological, anatomical and physiological adaptations to facilitate nutrient uptake and conservation (Givnish et al. 2014;Leroy et al. 2016;Males 2016). Bromeliads grow on different substrates (terrestrial, lithophyte or epiphyte) and differ in their ability to retain water and nutrients (i.e., tank-forming, or tankless) and in the photosynthetic pathway (i.e., C 3 or CAM). The leaves of tankless Tillandsioideae, of tank-forming Bromelioideae, and of some Brocchinioideae are the most important vegetative organ because they perform essential physiological functions including photosynthesis, water and nutrient assimilation and water conservation (Benzing 2000). In tank bromeliads, the basal part of the leaf is devoted to water and nutrient absorption through absorbing trichomes and to nutrient uptake thanks to transporters, plus nitrate reductase activity similar to that found in the roots of terrestrial plants (Gonçalves et al. 2020b;Kleingesinds et al. 2018). The middle and apical portions of the leaf are devoted to photosynthesis and glutamine synthetase activity (Gonçalves et al. 2020b).
The functional importance of leaves in epiphytic tank-bromeliad nutrition led scientists to overlook the roots, which were thought to only be used for anchorage (Benzing 2000;Takahashi et al. 2022). However, a few recent studies showed that the roots of some facultative epiphytic tank bromeliads were involved in nutrient and water absorption (Carvalho et al. 2018;Gomes et al. 2021;Leroy et al. 2019a;Silva et al. 2018;Vanhoutte et al. 2017). These recent studies investigated mature tank bromeliads, either collected in the field or from commercial nurseries and relied on shortterm experimental approaches. Because they used fully grown plants, these studies are unable to identify the real quantitative role of the roots and the consequences for plant growth and performances. In addition, at the seedling stage, some tank-forming bromeliads belonging to the genus Aechmea were shown to be totally devoid of leaf absorbing trichomes and thus depended entirely on their roots for water and nutrient absorption (Leroy et al. 2019b(Leroy et al. , 2017Petit et al. 2014). In that case, the seedlings performed better when growing on an organic-rich substrate, pointing to a significant nutritional role for the roots (Leroy et al. 2017(Leroy et al. , 2019b.

3
Vol.: (0123456789) Despite this increasing evidence for a significant role for roots in bromeliad nutrition, little is known about the root structure and the importance of the nutrient content of the substrate for the development and growth of tank-bromeliads. The lasting effects of substrate fertility on tank-bromeliad development (from the seedling to the mature stage) may have a key ecological impact on their performances.
The two objectives of this study were thus to (i) assess the extent to which the fertility of the substrate affects the growth and overall performances of the tank-bromeliad Aechmea aquilega and (ii) to identify specific physiological and morpho-anatomical leaf and root responses to substrate fertility. Aechmea aquilega (salisb.) Griseb is a tank-bromeliad which belongs to the subfamily Bromelioideae and has crassulacean acid metabolism (CAM) photosynthesis (Crayn et al. 2004). The root system of A. aquilega contributes significantly to both nutrient uptake (Leroy et al. 2019a) and seedling performance (Leroy et al. 2019b). Furthermore, in nature, this species can be found growing on terrestrial, lithophytic and epiphytic supports (Leroy et al. 2013). Because leaf absorbing trichomes appeared several weeks after germination of A. aquilega (Leroy et al. 2017(Leroy et al. , 2019b, we hypothesised that the substrate on which the bromeliad grows since germination plays a key role and lasting effect on trait expression (i.e., growth and overall performance of plants growing on nutrient-rich substrates are better than that of plants growing on nutrient-poor substrates). In addition, because environmental pressures have profound impacts on leaf and root trait values (Wang et al. 2021) and because A. aquilega individuals are found in different habitats (Talaga et al. 2017), we hypothesised that leaf and root trait values vary widely across different substrates allowing high capacity for adaptive phenotypic plasticity. To test these hypotheses, we grew A. aquilega for fifteen months in different substrate fertilities (a nutrient-poor, a nutrient-rich and a mixed substrate) and we measured leaf and root traits related to nutrient acquisition and resources capture.

Plant material and growth conditions
In October 2018, seeds from one mother epiphytic plant growing in natural environment were sowed in a greenhouse at the campus agronomique in Kourou (French Guiana) in horticultural seedling trays in three different substrate: 100% white sand (nutrientpoor substrate, WS), 100% potting soil (nutrient-rich substrate, P) and an equal volume of white sand and potting soil (v:v 50:50, intermediate substrate, WSP). The sand was collected from a location close to Paracou research station (5°16′26″N, 52°55′26″W), while the potting soil was purchased. Physico-chemical analyses were performed on triplicate samples of each type of substrates to determine the mean particle size and carbon, nitrogen, and phosphorus contents (PAPC, Toulouse). Results showed significant differences of mean particle sizes with WSP being intermediate between WS and P. Substrate CNP contents were also significantly different between the three treatments but with WSP closer to WS (supplementary information, Fig. S1).
During their development, individual plants were transplanted in their corresponding substrate into 0.5-l, 1-l and then 2-l pots as the plants grew bigger. The pots were placed randomly on horticultural tables in similar environmental conditions. Plants were irrigated with tap water for ten minutes twice a day at 8:00 am and 6:00 pm to maintain soil moisture at field capacity. Greenhouse temperature, light intensity and relative humidity were monitored with HOBO probes (model UA-002-64, HOBO Pendant Temp/Light -64 k and model U23-001, HOBO Pro V2 Temp/RH Data logger, Amanvillers, France). The mean relative humidity was 82.9 ± 0.1%, mean air temperature was 27.8 ± 0.1 °C, and light intensity was 21,016.7 ± 580.6 lx (ca. 30% of full external irradiance, corresponding to a mean PAR (photosynthetically active radiation) of 496.5 ± 34.4 µmol m −2 s −1 ).

Sampling
Measurements and samplings were performed on 15-month-old plants. Ten plants per treatment were randomly selected and all measurements and sampling were performed on three to four mature leaves, depending on plant size and the amount of plant material required for chemical analysis.

Carbon metabolism traits
Chlorophyll fluorescence-Two chlorophyll fluorescence parameters, the maximum quantum yield 1 3 Vol:. (1234567890) of photosystem II (F v /F m ) and the maximum electron transport rate (ETR max , µmol photon m −2 s −1 ) were measured with a portable fluorometer (MINI-PAM II, Walz, Effeltrich, Germany). Measurements were made on the adaxial surface of the leaf between 8:00 am and 12:00 am.
To measure F v /F m , a portion of the leaf was dark acclimated for 30 min with a dark leaf clip (DLC-8, Walz). The minimal fluorescence F 0 was measured by exposing the leaf to a low intensity light (< 0.1 µmol photon m −2 s −1 ), then a 0.8 s saturating pulse (5000 µmol photon m −2 s −1 ) was produced to assess maximal fluorescence F v /F m was calculated as: ETR max was calculated using the rapid light curve (Manzi et al. 2022;Rascher et al. 2000) procedure on the same leaf as that used for F v /F m . For quasi-dark acclimation, the leaf was placed in an opaque plastic bag for 30 s (Manzi et al. 2022;Rascher et al. 2000) and left in the bag for the rapid light curve. The leaf was then gradually exposed to increasing PAR values in 12 steps from 50 to 3000 µmol photon m −2 s −1 each lasting for 30 s. The ETR was calculated using the fluorometer and the WinControl-3 software (Walz, Effeltrich, Germany) according to the photoinhibition REG1 function of Platt et al. (1980). ETR max was then extracted from the resulting curves as the highest measured ETR.
Chlorophyll content-Leaf chlorophyll a and b content (CHL, mg g −1 ) were estimated with a SPAD-502 chlorophyll meter (KONICA MINOLTA, U.S.A). Eight measurements were taken on the apical and median leaf portion and averaged to obtain the average SPAD value of the leaf. SPAD values were then converted into µg cm −2 according to Coste et al. (2010): , then converted into mg g. −1 as Gas exchange-Net photosynthesis assimilation (A, µmol CO 2 m −2 s −1 ) and stomatal conductance (Gs, µmol H 2 O m −2 s −1 ) were measured on 18 additional plants (N = 6 per treatment) with a CIRAS-3 analyser (PP Systems, Amesbury, U.S.A). All the measurements were made at ten-minute intervals throughout the night from 6:00 pm to 9:00 am the following morning. The CO 2 concentration in the leaf chamber was set to 400 ppm, the temperature to 27 °C, and the air flow to 250 µmol s −1 while relative humidity and light were left at ambient conditions.
To compare treatments, we calculated the maximum net photosynthesis assimilation (A max , µmol CO 2 m −2 s −1 ) by averaging the five highest consecutive values and the corresponding G S values were averaged to obtain the maximum stomatal conductance (G Smax , µmol H 2 O m −2 s −1 ). In addition, we quantified integrated net photosynthesis assimilation (A int , mmol CO 2 m −2 over a 15 h period) by integrating the area under the assimilation curve for the whole night (15 h) using the AUC function in the DescTools R package (Signorell et al. 2021) (Fig. S2).

Plant size and growth
Prior to all measurements, leaves were counted (Nb leaves) and the water volume in the tank (Tank capacity, mL) was measured. The total leaf and root dry mass (see below) were used to assess the root-toshoot ratio as follows: Root − to − Shootratio = DM root DM shoot . We calculated the relative growth rate (RGR, mg g −1 month −1 ) as the increase of dry mass of the plants relative to the initial dry mass over 15 months. RGR of the plant biomass was calculated as follow: RGR = (lnW2 -ln W1)/t2-t1, where W2 and W1 are the dry mass of final and initial plant biomass, and t2 and t1 are time. 85 mature seeds that were oven dry at 60 °C for 48 h were the initial plant biomass (W1, i.e., 0.987 mg) and the total leaf and root dry mass after 15-month growth was the final plant biomass (W2).

Leaf structural traits
The youngest mature leaf was sampled, and its length measured (Leaf length, cm). Eight 10-mm diameter disks were collected with a cork borer from the apical portion of the leaf of plants growing on nutrient-rich and intermediate substrates (P and WSP) while four 8-mm diameter disks were collected from bromeliads growing on nutrient-poor substrate (WS). The thickness of each disk was measured with a micrometric calliper (Digit Outside Micrometre 193-101, Mitutoyo, Japan) and averaged to determine leaf thickness (Leaf thickness, mm). Fresh mass (FM, g) was determined by weighing on an electronic balance (AB 204-S Mettler Toledo, Switzerland), while to obtain the turgid mass (TM, g), the disks were stored in distilled water at 4 °C in the dark for 48 to 72 h. To determine the dry mass (DM, g), the disks were then dried at 60 °C for 72 h. The leaf mass area (LMA, g m −2 ) was calculated as DM/sum of leaf disk area, while the leaf dry matter content (LDMC, g g −1 ) was calculated as DM/FM.
A 1 × 4 cm portion of each leaf was collected from both the apical and basal part of the leaf and fixed in FAA (5% formalin, 5% glacial acetic acid, 70% ethanol and 20% water) for two weeks and then stored in 70° ethanol. Stomatal density (Nb stomata mm −2 ) was measured on the adaxial face of the apical portion and trichome density (Nb trichomes mm −2 ) and diameter (mm) were measured on both sides of the basal part of the leaf. Stomata and trichomes were observed from imprints made using transparent nail varnish. The imprints were observed with an inverted microscope (Olympus BX51). Four pictures per imprint were acquired with a digital camera (Lumenera LW1135C-IO, Ottawa, Canada) and processed using ImageJ software (Schneider et al. 2012). The number of stomata and trichomes per mm 2 were recorded for each imprint and averaged. The diameter of 20 trichomes (5 per picture) was measured on ImageJ and averaged to estimate the mean trichome diameter. A trichome area index (TAI, %), giving a proxy for leaf trichome coverage, was calculated as: TAI (%) = (trichome diameter × 0.5) 2 × × trichome density leaf area

Root structural traits
Once all the measurements of the aerial part were completed, the bromeliads were gently unpotted. The root system was carefully washed with tap water and any remaining soil particles were very carefully removed from each adventitious root. The roots were dried with a paper towel and weighed on an electronic balance to get the fresh mass (FM root , g). The root system was scanned using an office scanner (Xerox DocuMate 4700 5.1) and the basic WinRHIZO software (Instrument Regent, Quebec City, QC, Canada). The roots were then placed in distilled water at 4 °C for 48-72 h to get the turgid mass (TM root , g) and oven dried at 60 °C for another 48 h to get the dry mass (DM root , g). From the scanned images we obtained the following morphological root traits: total root length (TRL, cm), number of adventitious roots (Nb roots), number of root tips (Nb tips), average root diameter (ARD, mm), root volume (cm 3 ), and the total surface area of the root system (cm 2 ). The root tissue density (RTD, g cm −3 ) was calculated by dividing DM root by its volume, specific root length (SRL, m g −1 ) TRL by DM root and specific tip root average (STRA, tips g −1 ) Nb tips by DM root .

Leaf and root chemical traits
Carbon, nitrogen and phosphorus were quantified in both leaf tips and roots. The leaf tips and roots were dried in an oven at 60 °C for 72 h and ground into fine powder in a mill. About 9-11 mg of leaf and root powder were used to quantify carbon (C, mg g DM ; spectrometer, Uvi Light XT5 Secomam, spectrometric method with ammonium molybdate at 880 nm after H 2 SO 4 acid hydrolysis and persulfate oxidation, adapted NFEN 6878).

Statistical analysis
All statistical analyses were carried out in R version 4.0.4 (R Core Team 2021). Graphs were produced using the R package ggplot2 (Wickham 2016). In order to estimate multivariate differences in A. aquilega leaf and root trait coordination, we used standardised multiple factor analysis (MFA, Escofier and Pagès 1990). The MFA method enables examination of common structures in datasets with many variables that can be separated into different groups of variables (i.e., leaf and root traits). MFA was performed with the Factominer package (Lê et al. 2008) on two set of variables: leaf traits (LMA, LDMC, Stomatal density, TAI, C, N, P, Chl, F v /F m , ETR max ) and root traits (STRA, SRL, ARD, RTD, C, N, P). We then plotted all individuals and variables on the two first MFA dimensions and showed the 95% confidence ellipses for all three treatments. To test whether bromeliad traits were significantly affected by the fertility of the substrate, we performed permutational multivariate analysis of variance (perMANOVA) with the adonis function of the vegan package (Oksanen et al. 2022) using Euclidean distances and 10,000 permutations. A post-hoc test was then conducted with the pairwise.adonis wrapper function in the pairwiseAdonis package (Martinez Arbizu 2022) with p-values adjusted with Holm's method.
In addition, we examined differences in all traits in the three treatments with the Kruskal-Wallis rank test, which is a non-parametric alternative to one-way analysis of variance (ANOVA) when the assumptions of homogeneity of variance and normality are not met. The Kruskal-Wallis test was performed with the kruskal.test function in base R. When the Kruskal-Wallis associated p-value was significant (P < 0.05) the post-hoc pairwise Wilcoxon test was used to identify the effects of the treatments on each trait. P-values were corrected for multiple comparisons using Holm's procedure. Finally, we calculated the coefficients of variation (CV, %) of each trait among treatments as CV = SD mean × 100 to describe the extent of phenotypic plasticity of each trait. A summary table with mean ± SD, CV and statistics values and significance is available in the supplementary (Tab. S1).

Effects of substrate fertility on A. aquilega size and growth
Plants growing on nutrient-rich substrate (P) had approximately twice as many leaves, three times more adventitious roots, 16-fold higher total biomass, 35 times greater tank water capacity, ca. 1.5-fold higher RGR, a five times lower root-to-shoot ratio, longer and thicker leaves, and total root length was up to six times greater than that of plants growing on nutrient-poor substrate (WS; Fig. 1). Plants grown on the intermediate substrate (WSP) had either intermediate numbers of roots, total dry mass, tank capacity, growth rate, leaf thickness and total root length compared to plants grown on the two other substrates (Fig. 1B-E, H, I) or the number of leaves, root-to-shoot ratio and leaf length did not significantly differ from those of plants grown on the nutrient-rich substrate (Fig. 1A, F, G). Number of roots, total dry mass, tank capacity, root-to-shoot ratio and root length were the traits with the largest coefficient of variation (Fig. 1B, C, D, F, I).
Trait correlations and overall effect of substrate fertility on A. aquilega The two first dimensions of the MFA explained 51.9% of the variability of the data (PC1 explained 31.2% and PC2 explained 20.7%). The MFA showed two orthogonal groups of traits ( Fig. 2A). Most of the traits were correlated with PC1 while RTD, STRA and SRL were correlated with PC2. Multivariate analysis of trait correlations showed that root structural traits were orthogonal to leaf structural and chemical traits. Projection of all individuals separated the three treatments along PC1 whereas no segregation appeared on PC2 (Fig. 2B). PerMANOVA revealed a significant effect of the substrate on the trait values (F = 7.2195, Df = 2, p < 0.0001). Specifically, all individuals grown on each of the three substrates differed significantly from one another (Pairwise.adonis, Holm adjusted P < 0.01 in all cases).

Effects of substrate fertility on plant functioning
Substrate fertility had no significant effect on Gs max (Fig. 3A). Net photosynthesis assimilation (A max and A int ) differed significantly between the three substrates with higher values for plants growing on nutrient-rich (P) and intermediate (WSP) substrates than for plants growing on nutrient-poor (WS) substrate (but with marginal non-significant differences for A max , Fig. 3B-C). There was no significant effect of treatment on chlorophyll fluorescence as F v /F m and ETR max values were similar (Fig. 3D-E). The latter was very stable with a CV of 2.74%. Chlorophyll content was similar in the three treatments (Fig. 3F).

Effects of substrate fertility on leaf structural and chemical traits
Substrate fertility had significant effects on LMA but not on LDMC ( Fig. 3G-C). Plants growing on nutrient-rich substrate (P) had significantly higher LMA than plants growing on nutrient-poor substrate (WS), whereas there was no significant difference between the plants growing on intermediate substrate (WSP) and those growing on the two other substrates (Fig. 3G). Stomatal density and TAI were ca. 1.5 and 0.5 times higher in plants growing on nutrientrich substrate than in plants growing on nutrient-poor substrate, respectively (Fig. 3I-J). Higher TAI was mainly due to change in trichomes density rather than size (Fig. S3). Plants growing on the intermediate substrate had similar stomatal density and significantly lower trichome leaf coverage than plants growing on rich substrate. Finally, substrate fertility had 1 3 Vol.: (0123456789) Fig. 1 Effects of substrate fertility on (A) number of green leaves, (B) number of adventitious roots, (C) total dry mass (g), (D) tank water capacity (mL), (E) RGR (mg g −1 month −1 ), (F) Root-to-shoot ratio, (G) leaf length (cm), (H) leaf thickness (mm), and (I) total root length (cm). Different letters indi-cate significant differences between treatments based on pairwise Wilcoxon tests (α < 0.05) after significant Kruskal-Wallis (α < 0.05). CV, coefficients of variation of each trait among treatment; WS, white sand; WSP, white sand/potting soil; P, potting soil 1 3 Vol:. (1234567890) significant effects on leaf C and N contents but not on the leaf P contents (Fig. 3K-M). The leaves of plants growing on the nutrient-rich substrate were characterised by higher leaf C but lower leaf N contents than the leaves of plants growing on the nutrient-poor substrate ( Fig. 3K and L, respectively). The C and N contents of the leaves of plants growing on the intermediate substrate were similar to those of the leaves of plants growing on the nutrient-rich substrate. Overall, all the leaf traits displayed moderate variation (12.7%-30.1%), except for leaf C content, which had a particularly low coefficient of variation (3.47%).

Effects of substrate fertility on root structural and chemical traits
Substrate fertility did not affect STRA, SRL or RTD (Fig. 4A, B and D respectfully). Substrate fertility had a significant effect on the average root diameter with plants growing on the intermediate substrate had significantly larger roots than plants growing on the nutrient-rich substrate (Fig. 4C). Root C, N, and P contents were significantly higher in plants growing on the nutrient-rich substrate than in plants growing on the nutrient-poor substrate (Fig. 4E-G). Plants growing on the intermediate substrate had intermediate values compared to the two other substrates. STRA and SRL had high coefficients of variation (around 50%) while other root structural (RTD, and ARD) and chemical traits had lower CVs.

Discussion
Substrate fertility had a strong effect on the size and growth of A. aquilega. Plants growing on the nutrient-poor substrate (white sand, WS) had a 50% lower RGR and were 16 times smaller than plants growing on the nutrient-rich substrate (potting soil, P) had fewer leaves and roots, less total dry mass, smaller leaves, and shorter total root length. This overall smaller size reduced the capacity of the tank. Biomass partitioning was also strongly affected by substrate fertility. The root-to-shoot ratio was ca. Small dots are individuals' projections, big dots the centroid for each treatment. The 95% confidence ellipse is shown for each treatment. Substrates are WS, white sand; WSP, white sand/potting soil; P, potting soil. Leaf traits are C, carbon content (mg g −1 ); CHL, Chlorophyll content (mg g −1 ); ETR max , maximum electron transport rate (µmol m −2 s −1 ); F v /F m , PSII maximum quantum efficiency; LDMC, leaf dry mass content (g DM g FM −1 ); LMA, leaf mass area (g m −2 ); N, nitrogen content (mg g −1 ); P, phosphorus content (mg g −1 ); stomatal density (Nb mm −2 ), and TAI, trichome area index (%). Root traits are ARD, average root diameter (mm); C, carbon content (mg g −1 ); N, nitrogen content (mg g −1 ); P, phosphorus content (mg g −1 ); RTD, root tissue density (g m −3 ); SRL, specific root length (m g −1 ), and STRA, specific tip root average (root tips g −1 ) 1 3 Vol.: (0123456789) 4 times higher in A. aquilega growing on the nutrient-poor substrate, indicating a higher proportion of biomass in the roots, thus enhancing foraging. Such biomass partitioning is a well-known mechanism by which plants of a wide range of growth forms cope in nutrient-poor environments (e.g., Hermans et al. 2006;Mašková and Herben 2018;Sainju et al. 2017). In addition to differences between rich and poor substrates, plants growing on the nutrient-rich substrate (P) also out-performed the plants growing on the intermediary substrate (WSP) with larger tank capacity, higher total biomass, higher RGR, and higher and longer roots. The fertility of the substrate on which A. aquilega was grown from seed to 15 months strongly affected both the morphology of the rosette and biomass allocation. Such phenotypic Fig. 4 Effects of substrate fertility on (A) specific tip root average (STRA, tips g −1 ), (B) specific root length (SRL, cm g −1 ), (C) average root diameter (ARD, mm), (D) root tissue density (RTD, g cm −3 ), (E) root carbon, (F) root nitrogen, and (G) root phosphorus contents (mg g −1 ). Different letters indi-cate significant differences between treatments based on pairwise Wilcoxon test (α < 0.05) after significant Kruskal-Wallis (α < 0.05). The Kruskal-Wallis p-value and coefficient of variation (CV, %) are indicated for each variable within the plot. WS, white sand; WSP, white sand/potting soil; P, potting soil Vol.: (0123456789) plasticity' is relatively frequent in bromeliads, which grow in a broad range of light and water regimes, and different nutrient availability conditions (de Freitas et al. 2003;González et al. 2011;Scarano et al. 2002;Zotz and Asshoff 2010).
Most of the structural and chemical leaf traits were related to substrate fertility. Plants growing on the nutrient-rich substrate were characterised by higher LMA, thicker leaves, higher stomatal density and trichome coverage, along with higher leaf C and Chl and lower leaf N contents. Concerning LMA, our results disagree with those generally reported in the literature. Indeed, plants growing in nutrient-poor habitats often display higher LMA (Givnish 1979;Poorter et al. 2009). Givnish (1979) suggested that change in LMA was related to change in leaf thickness while Poorter et al. (2009) reported that most nutrient-driven change in LMA are due to alteration in leaf tissue density. Moreover, as LDMC is closely related to leaf tissue density (Shipley and Vu 2002), a variation in LMA with no variation in LDMC is likely to be the result of variations in leaf thickness (Vile et al. 2005). Hence, in our study, higher LMA in the nutrient-rich substrate is likely to result from leaf thickening because we did not find significant variation in LDMC. As leaves of bromeliads show strong allometric relationships between plant size and leaf thickness (Zotz et al. 2004;Meisner et al. 2013), we assume that the increase of LMA with substrate fertility is mainly the consequence of the strong treatment effect on plant size.
In general, epiphytes have lower N and P content in their tissues than ground-rooted herbaceous plants and trees (Hietz et al. 2021;Reich and Oleksyn 2004). Epiphytes are considered slow growing plants whose traits are associated with "slow" species, because of their low foliar nutrient concentrations and long leaf lifespan (Zotz 2016;Hietz et al. 2021). In our study, A. aquilega had overall low leaf N values compared to most of other plants, which is consistent with the literature (Hietz et al. 2021;Wagner et al. 2021;Wanek and Zotz 2011). We showed that plants growing on the nutrient-poor substrate had the highest leaf N content, but similar leaf P contents as compared to plants growing on the nutrient-rich substrate. These findings are highly surprising as in nutrient-poor habitats, species are generally known to have low leaf nutrient concentrations (Lambers and Poorter 1992;Wright et al. 2004) making the high leaf nitrogen concentration of nutrient-poor substrate plants confusing. Given that bromeliads from the nutrient-poor substrate had similar chlorophyll content as compared to the two other treatments, it is likely that nitrogen was allocated to other parts of the photosynthetic apparatus (e.g., RuBisCo), or to defence compounds and other non-photosynthetic processes (Evans 1989;Takashima et al. 2004). Additionally, we found similar ETR max and F v /F m across treatments. This result suggests that the substrate fertility did not affect photosystem functioning in A. aquilega. Nutrient deficiency has been shown to strongly influence the structure and functions of the photosynthetic apparatus with some damage to PSII, resulting in a reduction in F v /F m and ETR (Kalaji et al. 2014(Kalaji et al. , 2018Wu et al. 2008). Conversely, in other studies, N supply was reported to have no effect on photochemical efficiency (Cruz et al. 2003;Shrestha et al. 2012). These divergent results could be due to several factors (e.g., plant form and species, age of the plant, growing conditions, N recycling mechanisms) but Gonçalves et al. (2020a) showed that water and nutrient deprived Guzmania monostachia were able to maintain F v /F m similar to those of well-watered and nourished plants confirming that, at least at short terms, bromeliads can cope with low resource conditions via physiological adjustments. Nevertheless, concerning carbon assimilation, we found higher net photosynthesis assimilation (A int ) in A. aquilega growing on the nutrient-rich substrate than in plants growing on the nutrient-poor substrate. Our results show that the light harvesting (photosystem and electron transport) and CO 2 fixation (Calvin cycle and Rubisco) processes of photosynthesis responded differently to substrate fertility. The light harvesting process was not sensitive to our treatments while it did affect CO 2 fixation. Such contrasting responses may result from a trade-off in N allocation to different components (e.g., thylakoids which are important for the electron transport capacity or soluble proteins which are important for the Calvin cycle, Evans, 1989). Additionally, bromeliads from nutrient-rich substrate with more and larger leaves may likely had higher overall carbon assimilation which is in line with the ca. 16-fold size biomass differences observed between rich and poor substrate.
Concerning P content, A. aquilega had a higher concentration than that usually reported for other bromeliad species (e.g., Wanek and Zotz 2011;González et al. 2011). In addition, root P content was linked 1 3 Vol:. (1234567890) to substrate fertility with higher P storage in plants growing on nutrient-rich substrate. The increase in root but not leaf P contents as the substrate got richer can result from a preferential allocation of P to leaves rather than roots in nutrient-poor bromeliads. However, the fact that leaf P did not increase with substrate richness is surprising, especially considering that leaf N decreased as well. Indeed, as the substrate got richer the leaf N:P ratio dropped from 5.39 (nutrient-poor substrate) to 3.16/3.29 (intermediate and nutrient-rich substrates, respectively, Fig. S4). This points to a severe N limitation (Güsewell 2004;Zotz and Asshoff 2010;Wanek and Zotz 2011) and makes even more confusing the low leaf N content. However, leaf and root P, N, and N:P ratio do not always respond to substrate fertility and can be affected by plant functional groups (Hong et al. 2015), size and/ or leaf age (Schreeg et al. 2014), RGR or ammonium versus nitrate availability making interpretation less straightforward (Güsewell 2004). Further studies with bromeliads and epiphytes in general are needed to disentangle the respective contribution of N and P to the efficiency of photosynthesis and plant growth as well as to identify their specific critical threshold if we are to understand our results and the nature of nutrient-growth relationship in epiphytes.
Quantitative data on structural and chemical root traits in bromeliads and in epiphytes in general are only very rarely reported in the literature. In our study, we found that chemical root traits were linked to the fertility of the substrate, whereas this was not the case for structural root traits. Plants growing on the nutrient-rich substrate had higher C, N, and P contents in their roots than plants growing on the nutrient-poor substrate. In ground-rooted species, specific tip root average, root length and root tissue density are expected to be linked with soil resources and properties (Freschet et al. 2021). The second axis of the MFA showed clear co-variation between root structural traits (SRL, STRA, ARD and, RTD) but failed to distinguish between our three treatments, which was confirmed by the non-significance of the pairwise comparisons. The marked variability explained by this axis (21%) seems to result from soil properties or other variables that were not accounted for in our study rather than substrate fertility. The negative relationships between SRL and ARD was observed in another bromeliad species (Lutheria splendens, Leroy et al., 2022) and such negative relationships have been widely observed in the context of interspecific variations of mature plants (Bergmann et al. 2020;Kong et al. 2014;Roumet et al. 2016;Spitzer et al. 2021;Wagner et al. 2021). While not different between our treatments, SRL and RTD were negatively correlated in our study as shown by the second axis of the MFA. This results were in contrast to what has been reported in vascular epiphytes ) and in L. splendens regarding ontogenetic effects (Leroy et al. 2022). SRL and RTD co-variation across species is not clear, as some studies found a negative correlation (Bergmann et al. 2020;Garbowski et al. 2021), some found a positive correlation (Holdaway et al. 2011;Kong et al. 2014) while still others found no correlation (Kramer-Walter et al. 2016;Valverde-Barrantes and Blackwood 2016). Such variability in root trait combinations may be due to trade-offs between different root functions. Indeed, roots fulfil a wide range of functions comprising physical anchoring, resource storage, and resource capture via the interface with soil bacterial and fungal symbionts (Freschet et al. 2021). Anchorage is likely to be a dominant function in the roots of epiphytes ) which could mask the structural root trait response to substrate fertility. In addition, in our study, A. aquilega were potted in different substrates and this might have affected the structural root traits even if the root systems were not constrained by the pot (pers. obs.).
While the role of the roots in nutrient absorption has already been demonstrated for some bromeliad species (e.g., Gomes et al. 2021;Leroy et al. 2019a), this study is the first to show that the fertility of the substrate on which the bromeliad A. aquilega grows following germination has a strong and lasting effect on overall plant morphology and performance. Because of the lack of leaf-absorbing trichomes and of a water tank, A. aquilega seedlings can only rely on the root system for nutrient absorption (Leroy et al. 2017(Leroy et al. , 2019b. In a close congeneric species (A. mertensii), leaves became larger to form wells and the density of leaf-absorbing trichomes increased as the plant grew (Petit et al. 2014). In A. aquilega, leaf-absorbing trichomes and wells appeared at an age of 4-6 months (pers. obs.). With broader leaves, higher absorbing trichome leaf coverage at the base of the lamina, and a 30-fold higher tank capacity, A. aquilega plants growing on a nutrient-rich substrate may benefit from higher nutrient supply compared to plants growing on a nutrient-poor substrate. We have shown that the fertility of the substrate during the germination and establishment stages in A. aquilega is of the utmost importance as it also affects performances at later ontogenetic stages. Given the high diversity of substrate (soil, bark or rocks) used by wild A. aquilega, this could have important ecological repercussions. Tank bromeliads provide a habitat for aquatic, semi-aquatic and terrestrial organisms and contribute to many ecosystem services such as maintenance of biodiversity, nutrient cycling, and the provisioning of food and water (Ladino et al. 2019). Large plants with a high tank capacity were characterised by a greater diversity and abundance of aquatic organisms that echo the bromeliad nutrition (Leroy et al. 2016) and a higher external water storage to resist drought stress (Males 2016). Hence, when growing in nutrient-rich substrate tank bromeliads may sustain higher biodiversity and resistance to climatic change, resulting in ecological advantages.
Overall, we found that 70% of measured traits linked to plant performance responded significantly to the experimental treatments, indicating that A. aquilega response to nutrient availability is a combination of physiological processes and leaf and root structure and chemistry. The strong effect of substrate fertility on overall plant size and performance confirmed the importance of the root system in the establishment and growth of A. aquilega. This study and others showed that the roots of some bromeliad species are able to absorb nutrients in addition to the leaf trichomes. In our study, we further demonstrate that the nutrients absorbed by the roots determine the growth, size and performance of the plants which likely reflects bromeliad ecology. It thus appears that we need to reconsider the functional role of roots in bromeliad nutrition and the existence of different degrees of dependence on the substrate in the bromeliad family with probably significant variations both between and within species along ontogeny. Given the extraordinary diversity of bromeliads, future studies need to cover a much wider range of species, spanning both phylogenetic and ecological diversity.