Borago officinalis (Boraginaceae) is an annual, entomophilous bee-pollinated plant species. The flowering period extends from June to September in temperate areas, and about 100 flowers are produced per plant. This species produces abundant floral resources (about 5 μl nectar and 12,700 pollen grains), which makes it highly attractive to pollinators (Thom et al., 2016; Descamps et al., 2018). The anthers and nectaries are easily accessible on the surface of the corolla.

Borago officinalis seeds were provided by Semailles nursery (Faulx-les-Tombes, Belgium). Seedlings at the three-leaf stage were transplanted into 2-L pots filled with a 1:1 (v/v) mix of sand (size 0/5, M PRO, Netherlands) and universal peat compost (DCM, Amsterdam, Netherlands) and grown in the glasshouses at the University campus (SEFY platform, Louvain-la-Neuve, Belgium). Plants were watered daily with rainwater. Treatments were applied at floral transition, 4 weeks after sowing. At this stage, bolting occurred, flowering stems developed, and the first floral buds were visible. Plants were subjected to three temperature regimes (21, 24, and 27°C) and two watering regimes (well-watered vs. water-stressed). The well-watered plants received daily watering (soil humidity about 30%), whereas the water-stressed plants were watered twice a week (soil humidity lower than 15%, Descamps et al., 2018). Soil water content was quantified with a ProCheck sensor handheld reader (Decagon Devices, Pullman, United States). In total, six treatments were applied to 12 plants per treatment: 21°C well-watered (21WW), 21°C water- stressed (21WS), 24°C well-watered (24WW), 24°C water-stressed (24WS), 27°C well-watered (27WW), and 27°C water-stressed (27WS). Seventy-two plants were monitored in three growth chambers under three temperature regimes (day/night): 21°C/19°C, 24°C/22°C, and 27°C/25°C. The photoperiod was 16 h light:8 h dark, and relative humidity was maintained at 80 ± 10%. Each growth chamber was divided into two parts to accommodate two watering regimes. 21WW was considered the “control” treatment with optimal temperature growth and watering regime, while 27WS was considered the most stressful treatment, with combined high temperature and water stress conditions (Descamps et al., 2018). Growth chamber experiments lasted 5 weeks. Water stress was applied after 1 week of acclimation to the growth chambers (this week was considered week 0). Experiments were performed in June–July 2020.

During weeks 2 and 3, nectar was collected with 10-μl glass capillary tubes (Hirschmann Laborgeräte, Eberstadt, Germany). The nectar volume per flower was estimated for seven flowers per treatment by measuring the length of the nectar column in the capillary tube. The total sugar concentration (C,°Brix, g sucrose/100 g solution) was measured on the same flowers with a low-volume hand refractometer (Eclipse handheld refractometer; Bellingham and Stanley, Tunbridge Wells, United Kingdom). Nectar sugar content per flower (s, mg) was calculated as s = 10 × d × v × C, where d is the density of a sucrose solution at concentration C (d = 0.0037921 × C + 0.0000178 × C² + 0.9988603) and v is nectar volume (ml) (Prys-Jones and Corbet, 1991).

Ten samples of 10 μl of nectar were taken per treatment (from a minimum of two flowers from two different plants) and analyzed for sugar composition and amino acid content. The sugar composition of the nectar was determined using gas chromatography–flame ionization detection (GC-FID, Thermo Scientific, Germany) using a Restek RTX5-MS column (30 × 0.25 mm i.d. 0.25 μm) as described by Somme et al. (2016). The nectar extracted from the capillaries was weighed (Sartorius-Basic scale 0.1 mg, Germany) and 1 ml water was added. Internal standard (50 μl of a 1,000 ppm aqueous mannitol solution) was added to 50 μl diluted sample. This mixture was then dried under N₂ before forming trimethylsilyl (TMS) derivatives [using hydroxylamine hydrochloride and hexamethyldisilazide (HMDS) in pyridine]. A 1-μl volume of TMS derivatives was injected into GC-FID and separated via a Restek RTX5-MS column (30 × 0.25 mm i.d. 0.25 μm) in split mode (1/10). Nitrogen at 2 ml/min was used as carrier gas. The injector and detector were set to 270 and 300°C, respectively. The temperature program started at 105°C for 4 min, before increasing at a rate of 15°C/min to 280°C. This temperature was maintained for 20 min. All analytical measures were performed in duplicate.

Amino acids were quantified using high performance liquid chromatography (HPLC), in duplicate after derivatization using phthaldialdehyde (OPA) reagent in combination with 9-fluorenylmethyl chloroformate (FMOC, Meussen et al., 2014; Aubert et al., 2021). The nectar extracted from the capillaries was weighed (Sartorius-Basic scale, Germany) and 100 μl of an aqueous solution of 25 μM norvaline (used as an internal standard) was added. As no acid hydrolysis took place, we could not measure the tryptophan content. The determination of cysteine and methionine required a prior oxidation step (H₂O₂/formic acid, 110°C for 18 h). A double derivatization process was performed in pre-columns using (i) 2-mercaptoethanol 4% + 25 mg OPA dissolved in 0.5 ml methanol in a total volume of 5 ml borate buffer pH 10.4 [for detection of all amino acids except proline (PRO) and hydroxy-proline (OH-PRO)] and (ii) FMOC 0.25% in acetonitrile (for detection of cyclic amino acids: PRO and OH-PRO). Hydroxy-proline is an amino acid derived from proline by hydroxylation. Samples were injected on a Zorbax Eclipse Plus column (Agilent; 3.5 μm particle size; 150 × 21 mm) maintained at 40°C. The mobile phase was composed of (A) phosphate buffer 40 mM pH 8.4 and (B) acetonitrile/methanol/water (45:45:10 v/v/v) at a flow rate of 0.42 ml/min (100% A–0% B 0.5 min; progressive increase from 0 to 57% B 0.5–25 min). OPA-derivatized and FMOC-derivatized amino acids were, respectively, detected at 340 and 266 nm excitation and 450 and 338 nm emission wavelengths.

During weeks 2 and 3, pollen was collected from at least 10 flowers per treatment (minimum of two different plants), by squeezing and opening anthers with pliers over a microfuge tube. We estimated pollen production (mg) per flower by dividing the weight of each sample by the number of flowers from which pollen was collected.

Polypeptide content (molecular weight > 10 kDa) of pollen was determined from 5 mg dry pollen in triplicate for each treatment following the method described in Vanderplanck et al. (2014a). Total polypeptides were quantified using the bicinchoninic acid (BCA) Protein Assay Kit (Pierce, Thermo Scientific), with BSA as standard.

Amino acid profile and concentrations were determined from 2 mg pollen in triplicate after acid hydrolysis (with or without prior oxidation) and derivatization using OPA reagent in combination with FMOC (Meussen et al., 2014; Aubert et al., 2021). For acid hydrolysis, samples were exposed to 150 μl 6 M hydrochloric acid containing 1% (w/v) of phenol and incubated for 18 h at 110°C. For acid hydrolysis with prior oxidation, 50 μl performic acid reagent was added to the samples. The reagent consisted of 10 mg phenol crystal in 10 ml water with 1 ml 33% H₂O₂ and 9 ml 88% formic acid. After addition of the reagent, tubes were incubated for 3 h at 4°C, and 8 mg metabisulfite was then added to decompose the performic acid. Tubes were then stirred for 1.5 s to liberate any SO₂ produced. Derivatization and quantification were as described for the nectar samples. The essential amino acids were determined by de Groot (1952) for Apis mellifera and include: arginine, histidine, isoleucine, leucine, lysine, methionine, phenylalanine, threonine, tryptophan, and valine. This information is not known for wild bee species (Kriesell et al., 2017; Woodard and Jha, 2017; Nikkeshi et al., 2021).

All chemical analyses were performed in the MOCA platform and PEPA lab (UCLouvain, Louvain-la-Neuve, Belgium).

Normality of the data was visually assessed using quantile–quantile plots. Linear mixed models and analysis of variance (type II) were performed using a significance level of p < 0.05 to evaluate the effects of temperature, water stress, and their interaction. For nectar volume, sugar concentration, and pollen dry weight, analyses of variance (type II) were performed. Linear mixed models were fit including the two fixed factors, their interaction (temperature × water) and a random effect for sampling number due to repeated measurements on the same sample (sugar proportion, polypeptide concentration, and amino acid concentration). Results for all amino acids were presented as relative differences compared with the control treatment (21WW) for each amino acid. The relative difference was obtained by subtracting the value of the 21WW treatment concentration from the value of each treatment concentration, divided by the value of the 21WW treatment concentration. Amino acids are only shown when their concentrations significantly varied with temperature and/or water stress. Concentrations for each amino acid for each treatment with statistical results are available in Supplementary Material. To obtain a global view of the influence of temperature rise and water stress on all measures, a principal component analysis (PCA) was performed. All analyses were performed with R 3.6.1 (R Core Team, 2020), using the package car for F tests, the package lme4 for linear mixed models, the package FactomineR for PCA and packages ggplot2 and yarr for plots. Data are presented as mean ± standard error (SE).

We performed a principal component analysis of all nectar and pollen measurements according to the different treatments. The first two principal components explained 89.6% of the variance, with axis 1 accounting for 74.8% of the variance (Figure 1). The difference between the control treatment (21WW) and the most stressful treatment (27WS) was evident along axis 1. Floral resource quantities (nectar volume, and pollen dry weight) were higher for plants grown at 21WW than for plants grown at 27WS, while pollen polypeptide concentration and nectar amino acid concentration were higher for plants grown at 27WS than for plants grown at 21WW. Axis 2 discriminated plants grown at 21WS and 24WS from plants grown at 24WW according to the amino acid concentration of pollen.

We quantified the modification of pollen quantity and composition (polypeptides and amino acid profil) induced by temperature rise and water stress.

Temperature rise and water stress decreased the overall quantity of nectar with an 80% reduction in nectar volume. These results were consistent with previous observations that temperature rise and water stress decreased nectar volume per flower in B. officinalis (6.4 ± 1 μl per flower under 21WW, 0.7 ± 0.6 μl per flower under 27WS, Descamps et al., 2018) and in other entomophilous species (Scaven and Rafferty, 2013; Descamps et al., 2021b). The volume of nectar is an evolutionary trade-off between a high volume that is energetically costly and a low volume that would not be sufficient to attract pollinators and ensure an efficient visitation rate of flowers (Faegri and Pijl, 1979; Vaudo et al., 2015; Parachnowitsch et al., 2019). The reduction in volume that we observed in B. officinalis due to stress may affect pollinator visitation rate and have repercussion on plant reproduction. Physiologically, temperature rise and water stress reduce the quantity and availability of both water and sugars (Lemoine et al., 2013; Lamaoui et al., 2018). The amount of nectar sugars per flower was 3.5 times lower in plants grown in stressful conditions (27WS, 0.57 ± 0.08 mg) compared to control plants (21WW, 3.57 ± 0.28 mg). Decreased sugar quantity affect negatively flight performance of bees (Hendriksma et al., 2014). We also observed that sugar nectar concentration increased with temperature rise, from on average 50°Brix at 21°C to 60°Brix at 24 and 27°C. This increase might perhaps negatively impact nectar uptake by insects as it was recently shown that the opening of the papillae of bee tongue (Bombus terrestris) was maximized up to a sugar concentration in the nectar of 55% concentration (Lechantre et al., 2021). A reduced nectar volume per flower decreases also the amount of water that the insects collect through nectar which could influence their water balance, although bees seem less subject to desiccation than most terrestrial insects (Nicolson, 2009).

Floral resource composition, in addition to quantity, determines flower attractiveness for insect pollinators. In our study, the decrease in total sugar content was accompanied by a change in the composition of sugars in the nectar. We observed that water stress decreased the sucrose to hexoses (fructose and glucose) ratio. Relatively few studies have examined the effects of abiotic stresses on the sugar composition in nectar. Similarly to our results, Rering et al. (2020) observed that water stress decreased the proportion of sucrose compared to hexoses in the nectar of Fagopyrum esculentum. However, no change in the nectar sugar composition was detected under water stress in other studies (Villarreal and Freeman, 1990 on Ipomopsis longiflora and Clearwater et al., 2018 on Leptospermum scoparium). The change in sucrose/hexoses ratio in response to water stress could be explained by a modification in the invertase activity. This enzyme, which catalyzes the cleavage of the disaccharide sucrose into fructose and glucose, is sensitive to nectar sugar concentration (Nepi et al., 2012). According to Rering et al. (2020), nectar sucrose could also be converted into glucose and fructose due to a greater accumulation of reactive oxygen species (ROS) in the nectar of water-stressed plants.

We did not observe a decrease in pollen weight for B. officinalis in response to water stress, contrarily with other studies (Waser and Price, 2016; Borghi et al., 2019; Rering et al., 2020). By contrast, we observed a 50% decrease of pollen weight due to temperature rise, between 21 and 27°C. The sensitivity of pollen to temperature is relatively well established in the literature (Hedhly, 2011; Mesihovic et al., 2016). For example, a temperature increase of 8°C (22–30°C) reduced the amount of pollen produced by about 30–50% in Glycine max (Koti et al., 2005). According to the optimal foraging theory, reduced floral resource quantity per flower increases the visitation cost for pollinators, since they have to visit a greater number of flowers to obtain the same amount of resources (MacArthur and Pianka, 1966; Latty and Trueblood, 2020). The availability of floral resources is a major limiting factor for bee survival (Vaudo et al., 2015; Carvell et al., 2017; Abrahamczyk et al., 2020). Particularly, during summer periods, a shortage of floral resources is observed which could be a cause of bee species extinction (Requier et al., 2017; Balfour et al., 2018; Timberlake et al., 2019; Langlois et al., 2020; Simanonok et al., 2020). Temperature rise combined with drought can reinforce nutritional resource gaps at the landscape level.

While nectar is the main source of carbohydrates for pollinators, pollen is the main source of proteins and amino acids. We observed that temperature rise increased the concentration of polypeptides in the pollen of B. officinalis. The observed increase in polypeptide concentration in response to temperature rise supports the hypothesis of van der Kooi et al. (2019) which posits that protein and lipid synthesis are stimulated by a temperature increase. The modification of pollen composition in response to abiotic stresses seems, however, to be species-specific. Indeed, this result contrasts with previous observations on another entomophilous species, Impatiens glandulifera, grown under the same controlled conditions, where we observed a decrease in pollen polypeptide concentration in response to temperature rise and water stress (Descamps et al., 2021a). To our knowledge, the other study investigating the impact of a temperature increase of 0.5°C compared to ambient temperature under field conditions did not detect any differences on the protein content for Carduus nutans (Russo et al., 2020). Although we observed an increase in protein concentration for B. officinalis pollen under temperature rise (+6°C), this increase was mitigated by a decrease in pollen weight per flower.

Both nectar and pollen could be a source of amino acids for pollinators, although the total quantity of amino acids in pollen was more than 1,000 times higher than the quantity of total amino acids in nectar per flower (at 21WW, 0.23 mg in pollen vs. 0.19 μg in nectar). The impact of temperature rise and water stress on the amino acid profiles of flower resources differed between nectar and pollen. Total amino acid concentration increased for nectar but not for pollen, while the proportion of essential amino acids decreased in response to temperature rise for nectar but not for pollen. We observed that the total amino acid concentration in the nectar increased while the nectar volume decreased, which is consistent with the observations of Lohaus and Schwerdtfeger (2014) for several species (Maurandya barclayana, Lophospermum erubescens, and Brassica napus). Furthermore, the proportion of essential amino acids decreased in response to temperature rise in nectar but not in pollen.

We observed that the relative proportions of amino acids were modified by temperature rise and water stress for both nectar and pollen in B. officinalis. Amino acid profiles differ among species, although some phylogenetic trends are detected (Weiner et al., 2010; Ruedenauer et al., 2019b; Jeannerod et al., Submitted). Our results show that intraspecific variations can be observed in response to abiotic stresses. We found that the concentration of proline, in particular, increased with plant stress in both pollen and nectar. Proline is an osmoprotectant that accumulates in plants in response to environmental stresses and acts as a signal to improve stress tolerance (Hayat et al., 2012). Proline is also attractive for insects (Teulier et al., 2016). It contributes to a preferred taste (Alm et al., 1990; Bertazzini et al., 2010). Proline is required for egg-laying by the honey bee queen (Hrassnigg et al., 2003) and it is also used as fuel for the initial phase flight (Carter et al., 2006; Teulier et al., 2016). No other amino acid can be metabolized as rapidly as proline and releases as much ATP without complete metabolism (Carter et al., 2006). The increase in proline in the nectar and pollen of stressed plants could be detected by pollinating insects and, potentially, have positive impacts on them but this needs to be confirmed (Ruedenauer et al., 2019a).

The concentrations of other amino acids were also impacted by plant stress, although the impact of temperature rise was more pronounced than that of water stress. Differences in amino acid composition affect flower visitor behavior, influencing olfactory learning and memory of Apis mellifera and bumblebees (Simcock et al., 2014; Ruedenauer et al., 2019a). Recently, it was shown that bumblebee can perceive different amino acid (asparagine, cysteine, glutamic acid, hydroxyproline, lysine, phenylalanine, and serine) by chemotactile information through antennae and even different concentration of a same amino acid (Ruedenauer et al., 2019a). Although the nutrient requirements of honeybees and bumblebees have been partially investigated (Brodschneider and Crailsheim, 2010; Moerman et al., 2017), little is known about the nutrient requirements of wild bees (Woodard and Jha, 2017; Nikkeshi et al., 2021). For pollen, a decrease in the concentration of isoleucine and leucine, two essential amino acids, was observed at elevated temperatures. Bees prefer pollen with high concentration of essential amino acids, particularly isoleucine, leucine, and valine (Cook et al., 2003). We also observed an increase in the phenylalanine proportion (+10% in nectar, +150% in pollen) in response to water stress. Phenylalanine is induced in metabolic pathways of response to water stress (Wan et al., 2021). It was shown that this amino acid was predominant in the nectars of 73 Mediterranean plants (Petanidou et al., 2006), contrary to another study on 30 British species, where this amino acid was present in much lower quantities (Gardener and Gillman, 2001). The predominance of phenylalanine-rich nectar is explained by Petanidou et al. (2006) as a consequence of selection by long-tongue bees due to strong phagostimulatory effect of this amino acid (Inouye and Waller, 1984; Hendriksma et al., 2014). According to our observations, this presence of phenylalanine could also be a consequence of water stress, that plants may experience under a Mediterranean climate. Not all amino acids have the same importance for insects, and there may be species-specific differences in preference. However, there is a lack of information about the specific contribution of each amino acid to insect nutritional requirements.

Our results showed that temperature rise and water stress affect both the quantity and composition of floral resources in a bee-pollinated species. We observed a decrease in the quantity of nectar and pollen and a change in their composition. If the mutualistic relationship is compromised by changes in floral resources, this can affect both partners, plants, and pollinators. For pollinators, the reduction of resources at the flower level may require greater foraging efforts by pollinators for their food supply. Moreover, a modification of the floral resource composition may affect the nutrition and foraging behavior of pollinators. For plants, changes in pollinator foraging behavior can limit the seed set and decrease reproduction. However, both effects of floral resources modifications on pollinator nutrition and behavior and on plant reproduction need to be further investigated.