Seeds from two species P. coronopus and P. lanceolata were collected from 50 to 200 plants per accession in 2015, totaling 20 accessions in six regions covering four countries: Canary islands (Ci, Spain), Andalucía (An, Spain), Minho and Trás-os-Montes (PT, Portugal), Asturias (As, Spain), Sussex (Eg, United Kingdom), and Provence-Alpes-Côte d’Azur (Fr, France). The accessions were sampled at the time of natural seed dispersal (late-summer, Supplementary Table S1). Upon collection, seed of all accessions were cleaned and kept in a controlled environment [15°C, 15% relative humidity (RH)] until germination experiments were performed.

For each accession, we obtained the climatological conditions by cross-referencing their geographic localities with the WORLDCLIM database (Hijmans et al., 2005) using DIVA-GIS software (Hijmans et al., 2004). With the obtained interpolated climatological data, we calculated the average annual temperature (AT) plus the maximum (T_max) and minimum (T_min) temperatures. We also recorded average annual precipitation (AP), the monthly average from November to April [or “winter precipitation” (WP)] and May to October [or “summer precipitation” (SP)], and the sum of the annual precipitation (AP_sum; Supplementary Table S1). A map of the collecting sites was made using The Environmental Stratification of Europe dataset (QGIS Version 2.10.1; Metzger, 2018) with QGIS software version 2.10 (Open Source Geospatial Foundation Project) (Supplementary Figure S1).

The seed mass of each accession was recorded as the mean of 96 seeds selected randomly and weighed individually with a precision balance (UMT2 d = 0.1 μg, METTLER TOLEDO, Leicester, United Kingdom).

Laboratory germination (G) experiments were performed in temperature-controlled incubators (LMS, Sevenoaks, United Kingdom). Seeds were sown on 9-cm-diameter Petri dishes on two layers of filter paper (Whatman^® Grade1, Sigma Aldrich, Gillingham, United Kingdom) and saturated with 3.5 ml of sterile Milli-Q^® water (Merck Millipore, Watford, United Kingdom) and placed in closed plastic bags to prevent evaporation. Each replicate comprised a single Petri dish containing 50 seeds, and three replicate experiments were performed independently. Germination temperature regimes ranged from 5 to 30°C at 5°C intervals at a constant temperature and a 12-h light/12-h dark photoperiod. Germination was scored twice daily for 5 weeks, and germinated seeds were distinguished by protruded radicles (>1 mm). Germinated seeds were monitored every 3 days to score the date of cotyledon opening and were considered seedlings when the developing shoot meristem was visible and any abnormality was absent (NS). At the end of the test period, seeds that had failed to germinate were cut open and classified by appearance as viable or dead depending of embryo presence/absence (Teixeira et al., 2020).

Germination percentages were calculated, and statistical analyses were undertaken using only viable seed (i.e., total germinated plus ungerminated living seed).

The capacity of the seed to germinate under simulated conditions of water deficit was assessed using polyethylene glycol (PEG) solutions of varying concentrations as osmoticum. These tests were performed in three independent replicate experiments using 3.5 ml of PEG solution ranging from 0 to −1.0 MPa with 0.2-MPa intervals. Seeds were incubated at constant temperature in filter paper-containing Petri dishes (as described above) and at a constant temperature of 15°C. Each solution was prepared in triplicate according to Equation (5) described in Hardegree and Emmerich (1990), Equ. 1 (0.129 [PEG]² T – 14.0 [PEG]² – 0.40 [PEG]). Where “T” is the temperature for germination and “PEG” is MW 8,000 (Sigma Aldrich, Gillingham, United Kingdom). The Petri dishes were placed in closed plastic bags; one dish per treatment per experiment. Each Petri dish containing 50 seeds sown was scored daily, and seeds with protruded radicles >1 mm were considered to have completed germination. In order to keep the osmotic solutions at the designated water potential, the germinated seeds were carefully transferred to germination paper in transparent resealable plastic boxes (17 cm × 12 cm × 5 cm), which was moistened with 15 ml of the same osmotic solution. The boxes were also placed inside sealable plastic bags. The growing seedlings (S) were observed every 3 days to score cotyledon opening. As before, and at the end of the test, non-germinated seeds were cut open and classified by appearance as viable or dead depending on embryo presence/absence, germination percentages were calculated, and further statistical analyses (see below) were undertaken only with viable seeds (germinated plus remaining viable).

Statistical tests were performed using R v3.6.2 (R Core Team, 2016). The medians and quartiles of the boxplots were calculated using the R-Basic functions. “Germination rate” (GR_T in response to differing temperatures and GR_W for different osmotica) was inferred from the reciprocal of the time to 50% germination (t50), which was estimated via a sigmoidal curve fit of the cumulative germination for each dish using the Boltzmann equation (Supplementary Figure S2a). Using the range of constant temperatures of each replicate experiment per accession, they were regressed with a linear model to estimate the “base temperature” at which the germination rate was equal to zero (T_bG), and “thermal time” (θ_TG) was estimated with the reciprocal of the (temperature × 1/t50) slope value (Supplementary Figure S2b). The “seedling rate” (SR) was inferred from the reciprocal of the time until 50% of the germinated seeds had open cotyledons (at the constant temperature). The same procedure was used to estimate the base temperature (T_b), at which the rate of development to the seedling stage (for temperature and osmoticum, SR_T and SR_W, respectively) was equal to zero (T_bG and T_bS, respectively). The thermal time for this transition (θ_TS) was also estimated as already described above. To estimate the base water potential (Ψ_b) at which the GR was equal to zero and the “hydrotime” (θ_H; time for transition to seedling stage), a similar procedure to that described above was applied using a range of water potentials (i.e., osmotica) to give estimates of θ_HG and θ_HS, respectively (Supplementary Figures S2c,d). To estimate the temperatures and osmotica at which the germination and development to the seedling stage percentages were the highest, data were fitted with quadratic polynomial equations using the Origin software (OriginLab, Northampton, MA, United States). This identified the optimum temperatures for germination (T_oG) and for seedlings occurrence (T_oS).

The GLM tests for maximum germination and development to the seedling stage were performed using a binomial distribution with logit link function, with species as predictor factor using the maximum likelihood estimation method, Wald Chi-Square statistics, and contrast pairwise species. The GLM tests of the base temperatures, thermal and hydrotime, optimum temperatures, and base water potential were performed using the above method with a normal distribution and the identity link function after normality and homogeneity of the variances checking using the Shapiro–Wilk and Levene tests, respectively. All GLM tests were performed with the SPSS Statistics for Windows software (Version 21.0). Germination rates and normal seedling development rates were tested using the Kruskal–Wallis test, comparing temperature conditions and osmoticum using the SPSS Statistics for Windows software (Version 21.0). The base temperatures and thermal time for germination (T_bG and θ_TG, respectively) were regressed against the base temperatures and thermal times for normal seedling development (T_bS and θ_TS, respectively), the base water potentials and hydrotime for germination (Ψ_bG and θ_HG, respectively) were regressed against the base water potential and hydrotime for normal seedling development (Ψ_bS and θ_HG, respectively), using linear regression with the mean values of the three replicates for each accession and species using the R basic functions. Base temperatures and water potentials were regressed against thermal and hydrotime, and the T_bG, T_bS, Ψ_bG, and Ψ_bS were regressed against the climate traits T_min, T_max, WP, and SP. The two-way ANOVA tests comparing species’ T_bG with T_bS and Ψ_bG with Ψ_bS were performed using the SPSS Statistics for Windows software (Version 21.0). The factor analysis of mixed data (FAMD) was performed using the mean values of the three replicates in each physiological trait and the above-described data for geographical and climate conditions as numerical variables with species and region as categorical variables. All numerical variables were centered and scaled to homogenize high discrepancies. The FAMD was performed using the R FactoMineR package (version 2.0). The Spearman’s rank correlation was performed for individual selected species with the means of the three replicates of all treatments. The double correlation matrix of Spearman’s rank and p-values was produced with the R package corrplot (version 0.84) using the correlation matrix (adjusted for ties) and the exact probabilities values previously calculated with the GenStat software (VSN International, Hemel Hempstead, United Kingdom).

Graph theory, a mathematical object corresponding to a network that recently gained much attention, was used to analyze ecological attributes such as biological interaction webs, gene–protein interrelations, and flower and pollinator interactions (Proulx et al., 2005). The approach considers a set of units, called nodes or vertex, connected by edges where nodes represent a component of the network and edges indicate a relationship between them (Windram and Denby, 2015). The graph network analysis of traits was performed with the significant Spearman’s rank p-values in order to obtain a network graph of physiologic and environmental traits using the significant rho traits, where the significant p values in the graph correspond to the edges and the traits correspond to the nodes in an adjacency matrix. The rho p values of geographical vs. geographical traits and geographical vs. climate traits were excluded from the analysis. The weighted and undirected graph was generated with the R package igraph (version 1.0.1).

Significant differences were observed in the trait parameters between the two species (Table 1). P. coronopus reached an observed germination median of 88% at a temperature of 10°C, the experimental temperature closest to the T_oG of 11.5°C, while P. lanceolata reached an observed germination median of 51% at a temperature of 15°C, the experimental temperature closest to the T_oG of 14.4°C (Figure 1A and Supplementary Table S3). Across geographical regions, P. coronopus showed similar germination responses to temperature, except the accession from Ci (Supplementary Figure S4f) that showed high germination from 5 to 25°C and from As that showed high germination from 10 to 20°C (Supplementary Figure S4e). P. lanceolata also followed similar temperature responses in most regions, except Eg, which showed the lowest germination across temperatures (Supplementary Figure S5d).

Normal seedlings showed significant differences between species (Table 1) with a high response to a broad range of temperatures for P. coronopus and a narrower range of temperatures with a high normal seedling response for P. lanceolata near the optimum germination temperature (Figure 1B and Supplementary Figures S3, S4 and Supplementary Table S3).

Final germination for each of the species showed a gradual decrease of germination with decreasing water potentials (Figure 1C). A mild osmotic stress of −0.2 MPa moderately decreased the germination in P. coronopus and P. lanceolata compared with the control (0.0 MPa) by 25% and 18%, respectively, while a water potential of −0.8 MPa almost fully reduced germination in both species (Figure 1C and Supplementary Table S4). The development of normal seedlings in osmotic solutions reflects the result for germination of each species (Figure 1D and Supplementary Figures S3, S4 and Supplementary Table S4), showing significant differences between species (Table 1) with a strong decrease in normal seedling development at −0.4 MPa in P. coronopus when compared with the control. Across geographical regions, the inhibition by osmotic stress on normal seedling development was particularly evident in all P. lanceolata accessions (Supplementary Figure S4).

Germination rates and seedling development rates (GR_T, SR_T) differed between temperatures and osmoticum treatments (GR_W, SR_W). Contrary to P. coronopus where the GR_T and SR_T increased steadily from 5 to 30°C (Supplementary Figures S5a,b), P. lanceolata showed a transient increase in GR_T, with the fastest germination and seedling development at 20°C. Moderate osmotic stress only slightly decreased P. coronopus GR_W, but a strong drop was observed at −0.8 MPa. Normal seedling development rates in osmoticum (SR_W) were generally lower than GR_W, reaching an SR_W of zero at −0.8 MPa for both species (Supplementary Figure S5d).

The germination base temperatures (T_bG) showed a small difference for the two species, (P = 0.008) (Table 1) with values of 2.0°C for P. coronopus and 3.0°C for P. lanceolata (Supplementary Table S2). Similarly, base temperatures for normal seedling development (T_bS) also differed (P = 0.001) (Table 1). P. coronopus showed a lower T_bS (1.7°C) when compared with the T_bG (P = 0.005) where P. lanceolata showed similar values (Figure 2A and Supplementary Table S2). Despite this, positive correlations were found for both species between the T_bG and the T_bS (rho = 0.564, P = 0.022; rho = 0.745, P = 0.004) (Figure 2E). The base water potential for normal seedling development (Ψ_bS) differed between species (P = 0.037), in contrast with the base water potential for germination (Ψ_bG) (Table 1). P. coronopus and P. lanceolata showed a Ψ_bS that was higher than the Ψ_bG (P = 0.0001) (Figure 2B) with values of −0.8 and −1.0 MPa for the former and −0.7 and −1.1 MPa for the latter species (Supplementary Table S2). Unlike the positive correlations for T_bG and T_bS, both species did not show significant correlations between Ψ_bG and Ψ_bS (Figure 2F). Differences were found in normal seedling development thermal time (θ_TS) between the species but not for germination thermal time (θ_TG), although strong positive correlations were observed between these traits in both species (Figure 2G). Differences were found for both germination hydrotime (θ_HG) and normal seedling development hydrotime (θ_HS), but only P. coronopus revealed a significant correlation between both traits (Figure 2H). Both species showed a higher θ_T and θ_H for normal seedling development than for germination (Figures 2C,D).

An FAMD for P. coronopus and P. lanceolata was performed using the recorded physiological, geographical, and meteorological traits (Figure 3). The first two dimensions of the FAMD for P. coronopus explained 35.0% and 22.2% of the overall variance and for P. lanceolata 34.1% and 26.7%.

The geographical and climate conditions showed a strong contribution to the overall variation for both species (Figures 3A,C). The individual factor maps showed similar and clear clustering by geographical region in both species (Figures 3B,D). In the FAMD for P. coronopus, the first dimension showed the association of altitude with the base temperature, thermal time, and hydrotime due its high contribution to the variance, but T_bS and θ_HG revealed a strong correlation (up to 0.73) with this dimension. In this species, the second dimension explained the effect of the latitude positively correlated with the optimum temperatures, opposite to θ_TG that was negatively correlated (Figure 3A). In the FAMD for P. lanceolata, the first dimension showed the optimum temperatures and seed mass similarly affected by climate conditions, particularly summer precipitation. The second dimension showed that thermal times, θ_TG and θ_TS, were similarly affected by altitude and T_bG and θ_HG were affected by longitude (Figure 3B).

At the collecting sites, the climate conditions vary widely, with only 129 mm annual precipitation in the Canary Islands to 1,432 mm in the north of Portugal and 713 mm in the south of England. The mean annual temperatures range from 20°C in the Canary Islands to around 10°C in the north of Spain and the south of England (Supplementary Table S1).

The bivariate correlation rank coefficients allowed the evaluation of significant correlations between climate and geographical traits with physiological traits for P. coronopus and P. lanceolata (Figure 4). P. coronopus showed significant correlations of altitude with base temperatures and hydrotime for both germination and normal seedling development, although only base water potential for germination shows the same correlation. This contrasts with P. lanceolata where only correlations of altitude with the hydrotime for germination and thermal time for germination and normal seedling development were found. In both species, the optimum temperatures (T_oG and T_oS) were correlated with the latitude, while in P. coronopus (but not P. lanceolata), the latitude was also correlated with the base water potential (Ψ_bG and Ψ_bS); in P. lanceolata (but not P. coronopus), latitude showed correlations with base temperatures (T_bG and T_bS) and with thermal time and hydrotime (θ_TG and θ_HG). P. coronopus showed positive correlations of T_bG and T_bS with environmental temperature traits, while in P. lanceolata, only the T_bS showed such significant negative correlations. Surprisingly, the T_min negatively correlated with P. coronopus θ_HG (Figure 4). In both species, T_oG and T_oS correlated with environmental precipitation traits, but only P. coronopus showed positive correlations between precipitation and hydrotime traits and negative correlations of all three precipitation traits with the base temperatures. P. lanceolata showed negative θ_HG and positive θ_TS correlations with WP. Contrary to P. coronopus, the Ψ_bG of P. lanceolata was negatively correlated with all precipitation traits (Figures 4, 5G,H). In both species, the Ψ_bS showed correlations with environmental temperature traits: in P. coronopus with AT and T_max and in P. lanceolata with T_min (Figures 4, 5B). The P. coronopus seed mass (Sm) was correlated negatively with the environmental precipitation values and with T_oG and T_oS (Figure 4). Meanwhile, P. lanceolata seed mass was positively correlated with SP and AP, as well as with T_oG and T_oS. A positive correlation for P. lanceolata Sm was also found with latitude, and negative correlations were found with the environmental temperature conditions (Figure 4). The correlation analysis demonstrated clear differences in the strategies of these two Plantago species to cope with environmental conditions.

Plantago coronopus and P. lanceolata differed in the established community modules and the connected nodes within and among communities (Figure 6). The network with two P. coronopus community modules highlighted that the base temperature for normal seedling development and the germination hydrotime played a key role due to its higher connectivity degree with other traits. The θ_HG trait was more affected by precipitation than temperature conditions, whereas T_bS was similarly affected by both climacteric traits (Figure 6A). In contrast, P. lanceolata showed that the θ_HG trait connected only with winter precipitation but not summer precipitation (Figure 6B). P. lanceolata was organized in three community modules, highlighting a key role for the seed mass due to its higher connectivity degree with other traits within and between communities. This trait was shown to be strongly affected by temperature and precipitation associated with both optimum temperatures for germination and seedling development, as became clear also in the correlation analysis. Contrary to P. coronopus, the base temperature for normal seedlings in P. lanceolata was associated with both latitude and longitude, and the germination base water potential was mostly associated with the latitude and precipitation values (Figure 6B). These results underline that in these two species, the physiologic traits responded differently to the environmental stimuli.