Xylem function and climate adaptation in Pinus†
This research was supported by a Discovery grant and a postgraduate scholarship from the Natural Sciences and Engineering Research Council of Canada and Infrastructure grants from the Canada Foundation for Innovation and the Ontario Innovation Trust. The authors thank C. M. Caruso, G. Poon, M. E. Sherrard, B. A. Sikes, N. Sokol, E. Wassink, S. Weber, and two anonymous reviewers for helpful comments on the manuscript; and T. King, N. Deravi, and J. Creese for assistance with various phases of the project.
Abstract
• Premise of the Study: The distribution of species is determined in part by their functional traits. One important function is the ability of xylem to supply water to leaves and withstand water-stress-induced cavitation. These hydraulic traits are hypothesized to have evolved in response to selection by precipitation and temperature.
• Methods: We grew 26 species in the genus Pinus in a common environment and used phylogenetic comparative methods to examine whether the evolution of seedling hydraulic and wood density traits were associated with the climate of the extant geographic range of the species. We also examined whether these traits were correlated with each other, with integrated water-use efficiency (WUE), and with plant growth.
• Key Results: Contrary to predictions from a hydraulic model, we found no association between stem hydraulic conductivity (KS) and precipitation, even though there was substantial variation for KS in the genus. Nevertheless, KS was positively correlated with temperature, plant biomass, and WUE. Wood density was infrequently associated with climate or correlated with other traits except for plant biomass.
• Conclusions: Reduced KS in cold climates, if associated with reduced conduit diameter, likely evolved to increase resistance to freezing-induced xylem cavitation. The absence of a KS–precipitation relationship among Pinus seedlings suggests that associations between hydraulic traits and precipitation found in adult trees arise through plastic responses to moisture availability and/or develop over ontogeny. The weak association among wood density, climate, and other traits suggest that this trait does not contribute to climate adaptation in Pinus.
Plant functional trait variation is associated with stress tolerance and specialization to particular habitats and can explain the distribution of species (50; 27; 61; 72). One major class of traits involves xylem function; these traits determine water supply to leaves (78; 69) and are related to growth rate, drought adaptation, and the biomechanical support of leaves (5; 42; 9). Increased xylem hydraulic conductivity, for example, can facilitate faster growth rate because it reduces stomatal limitations of photosynthesis (66). By contrast, increased xylem (wood) density is associated with increased conduit resistance to implosion during water stress (31) and resistance to mechanical failure (9).
The degree to which stem-xylem hydraulic-conductivity and wood-density traits explain growth and survival in contrasting micro- and macroclimates varies among comparative studies (5; 10; 9; 58). These findings may reflect actual differences in functional trait relationships across environmental and geographic gradients (72) or be caused by variability in the phylogenetic scale of the species included in the sample (1). Specifically, traits that are associated with climate or phylogeny across broad taxonomic surveys of multiple lineages may not show the same pattern within a genus (8). One way to determine whether patterns of trait evolution identified at larger phylogenetic scales accurately reflect adaptation is to carry out comparative studies of closely related species that also span broad environmental gradients (2; 10; 75). If the traits are important for adaptation, they should be correlated with environmental variation even within a group of closely related species.
One broadly distributed yet closely related group of species is Pinus, the most speciose and ecologically dominant conifer genus (62). The distribution of Pinus along broad gradients of precipitation and temperature allows for a strong comparative test of the relationships among hydraulic and other plant functional traits, as well as a test of the degree to which variation in these traits reflects adaptation to climate. Predictions about mechanisms of climate adaptation in Pinus (47; 38; 45) can be derived from a simple hydraulic model (73; 45),
where AL/AS is leaf:sapwood area ratio, KS is specific hydraulic conductivity, ΔΨ/l is the water potential gradient, gS is canopy-weighted stomatal conductance, and D is the time-averaged vapor-pressure deficit of the air. The model is restricted to predictions of aboveground water transport traits because belowground water uptake and transport in roots are not formally included.
The model suggests two ways in which pine structure and function can vary with decreasing soil moisture availability (45). First, lower AL/AS or higher KS would increase whole-plant hydraulic conductivity, reducing the water potential gradient required to obtain water from drying soil (69; 40). Second, higher resistance to water-stress-induced xylem cavitation would allow the water potential gradient to increase to extract water from drying soil in the absence of any changes in hydraulic conductivity (57; 42). Empirical support is stronger for the first prediction; AL/AS decreases and KS increases with site water deficit in several pine species (47; 14; 17; 40; 46; 11; 3; 44). By contrast, there is relatively little intra- and interspecific variation in the vulnerability of pine xylem to water-stress-induced cavitation (39; 45, 44; 37).
Theories of freezing adaptation suggest that hydraulic traits should also vary with temperature in Pinus. Upon freezing, dissolved gases in xylem sap are forced out of solution and form air bubbles that can block water transport after thawing (78). Larger xylem conduits allow larger bubbles to form during freeze–thaw cycles, which make them more vulnerable to freezing-induced cavitation (18; 70; 13). If the incidence of freezing-induced cavitation selects for smaller-diameter conduits, then pines growing in colder climates should have lower KS than those from warmer climates. The evolution of small tracheids in cold climates, as well as an accompanying reduction in growth rate (34), could also result in higher wood density, all else being equal (56).
We examined the relationships among hydraulic, wood-density, and growth-related traits, as well as their relationships to precipitation and temperature variation among seedlings of 26 pine species grown in a common environment. We tested four predictions about how trait variation would be correlated with climate of origin. First, we predicted that species from lower precipitation environments would have higher KS and lower AL/AS to facilitate water extraction from drying soils without a rise in the water-potential gradient. Second, we predicted that species from warmer environments would have higher KS because of selection for reduced conduit diameter at low temperatures. Third, because the results of previous studies indicate that cavitation resistance in conifers is reflected by variation in wood density (56), we predicted that wood density would increase with decreasing precipitation. Fourth, we predicted that species from lower-precipitation environments would be more likely to restrict water losses through stomata, which would be reflected by an increase in integrated water-use efficiency (WUE). We used phylogenetic relationships to determine whether traits were conserved (i.e., whether extant variation was associated with divergences at ancient nodes; 49), to test for correlated evolution among traits and to test whether trait–environment relationships were consistent with adaptation to climate (e.g., 42).
MATERIALS AND METHODS
Species selection and growth conditions
To examine covariation between traits and climate of origin, we grew plants in a single greenhouse environment. This was done to minimize the effects of environmentally induced variation on the interspecific differentiation of traits (i.e., 28). We obtained seeds of Pinus species from Sheffield's Seed Company (Locke, New York, USA). We selected species on the basis of their availability in collections that spanned a large geographic gradient in precipitation and temperature and representation from the hard-pine (Pinus) and soft-pine (Strobus) subgenera of Pinus. Nevertheless, poor seed germination for species in Strobus resulted in a sample of 26 species containing a majority from the Pinus subgenus (Table 1). Although provenance tests have indicated that there is ecotypic differentiation among populations within species of Pinus (e.g., 60), multiple populations from each species were not available for inclusion. To minimize the effects of intraspecific variation on the comparative analysis among species, we sampled a single population of each species. Seeds were soaked in water and stratified at 4°C according to species requirements (USDA Forest Service Woody Plant Seed Manual; http://www.nsl.fs.fed.us/wpsm/Pinus.pdf) in staggered times so that they could be planted on the same date.
| Species | Abbreviation | Seed source (country) | MAP (mm) | MAT (°C) |
|---|---|---|---|---|
| Subgenus Pinus | ||||
| Pinus banksiana | Pba | Canada | 8251 | –0.51 |
| P. caribaea-hondurensis | Pch | Honduras | 23301 | 23.51 |
| P. contorta murrayana | Pcm | USA | 12002 | 15.54 |
| P. coulteri | Pco | USA | 3501 | 9.51 |
| P. densiflora | Pde | Japan | 12701 | 131 |
| P. elliottii | Pel | USA | 13251 | 19.51 |
| P. glabra | Pgl | USA | 14001 | 163 |
| P. massoniana | Pma | China | 13001 | 17.51 |
| P. mugo | Pmu | Slovenia | 14001 | 2.51 |
| P. muricata | Pmr | USA | 11901 | 131 |
| P. oocarpa | Poo | Guatemala | 18501 | 201 |
| P. pinaster | Ppi | Italy | 8001 | 16.51 |
| P. ponderosa | Ppo | Canada | 10151 | 7.51 |
| P. pseudostrobus | Pps | Mexico | 12501 | 15.51 |
| P. pungens | Ppu | USA | 13953 | 10.3753 |
| P. radiata | Pra | New Zealand | 14251 | 12.51 |
| P. resinosa | Pre | USA | 10001 | 61 |
| P. rigida | Pri | USA | 10501 | 10.51 |
| P. roxburghii | Pro | India | 9251 | 161 |
| P. thunbergii | Pth | China | 12851 | 14.51 |
| P. virginiana | Pvi | USA | 11451 | 11.253 |
| P. yunnanensis | Pyu | China | 10501 | 15.51 |
| P. taeda | Pta | USA | 15501 | 191 |
| Subgenus Strobus | ||||
| P. koraiensis | Pko | Russia | 8001 | 31 |
| P. sibirica | Psi | Russia | 11251 | –3.51 |
| P. strobus | Pst | Canada | 12701 | 8.51 |
- 1CAB International, 2005; 2Grotkopp et al., 2004; 3Burns and Honkala, 1990; 4Parker, 1986.
After stratification, 200 seeds per species were planted into 115-mL “conetainers” (model RLC7, Stuewe and Sons, Corvallis, Oregon, USA) with Sunshine Mix LA4 (SunGro Horticulture, Vancouver, British Columbia, Canada). Conetainers were arranged in random order on a greenhouse bench, and plants were grown with supplemental light to ensure a 16 hour photoperiod. Greenhouse climate was set to reflect prevailing conditions in Guelph, Ontario, Canada, during the summer, where mean daytime temperature ranged from 23 to 28°C, mean nighttime temperature ranged from 16 to 19°C, and plants experienced a mean ambient humidity of 53%. Twenty-five to 30 d after germination, 10–20 seedlings per species were transplanted in random order into 2.83-L tree pots (model TPOT1, Stuewe and Sons) containing 1:1:1 Sunshine Mix #2 (Sun Gro Horticulture, Vancouver), Turface MVP (Profile Products LLC, Buffalo Grove, Illinois, USA) and sand (Nu-Gro IP, Hillview, Ontario). Plants were watered daily by pressure-equalizing irrigation (Zwart Systems, Beamsville, Ontario) to maintain volumetric water content at ∼15%. Each plant was fertilized with 0.6 g of slow-release Nutricote (14–14–14 N:P:K, Plant Products, Brampton, Ontario) and two micronutrient foliar sprays (Plant Products Chelated Micronutrient Mix—25 ppm Fe, 7 ppm Mn, 1.4 ppm Zn, 0.35 ppm Cu, 4.5 ppm B, 0.21 ppm Mo) at 80 d and 135 d after transplanting.
Physiological measurements
When plants were 150 d old, we randomly selected four individuals per species to harvest for hydraulic-conductivity measurements (65). We measured one individual per species each week during the four-wk measurement period to avoid the confounding effects of seedling age on interspecific variation in hydraulic measurements. After harvest, seedlings were submerged in distilled water while shoots were separated from roots. Shoots were placed in a plastic bag filled with distilled water and stored at 4°C overnight to dissolve emboli and reduce resin emission from wood (39) and then allowed to reach room temperature before measurement. Shoots were prepared for measurement by placing them under water and stripping all bark, then excising the basal 5–10 mm portion of the stem. The exposed xylem of the shoot segment was wrapped tightly in parafilm and fitted with a rubber tubing gasket (Fisher Scientific, Ottawa, ON, Canada). The segment was inserted into vinyl Tygon tubing (Saint-Gobain Performance Plastics, Beaverton, MI, USA) that was attached to a Xyl'em meter (Xylem Embolism Meter, Bronkhorst, Montigny les Cormeilles, France; 12) which utilizes a high resolution liquid mass flowmeter to measure volumetric flow rate (F, kg s–1). After attachment, the distal portion of the stem was excised, leaving a ∼2 cm long unbranched stem segment for measurement. Hydraulic measurements were made with a pressure head of 9–14 kPa using distilled, filtered (0.2 µm) and degassed water.
Initial flow rate recorded within the first two minutes provided a measure of maximum water flux and was strongly correlated with stable flow rates obtained for a subset of samples after a longer period of 20–30 min (r2 = 0.99, P < 0.00001, n = 48). To expedite measurements we used the initial flow rate, corrected to water viscosity at 20°C to calculate hydraulic conductivity. Hydraulic conductivity (KH; kg m–1 MPa–1 s–1) was calculated as volume flow rate (kg s–1) divided by the pressure gradient per unit stem length (MPa m–1). Sapwood-specific conductivity (KS; kg m–1 MPa–1 s–1) was calculated by dividing KH by the stem xylem area (AS; m2).
After hydraulic measurements, leaves attached to the distal end of each stem segment were harvested, dried to constant mass at 65°C, and weighed. To estimate AL we calculated specific leaf area (SLA; m2 g–1) per individual as the total projected leaf area of 6 fresh needles divided by their dry mass. Projected leaf area was calculated from length and width measurements. AL was calculated as the product of SLA and total leaf dry mass. Stem cross-sectional area was calculated from stem diameter averaged from measurements of the proximal and distal ends of the stem. Stem xylem area (AS) was calculated by subtracting the pith area from total stem cross-sectional area. The dimensions of stem and pith cross-sections were measured on free hand sections using a dissecting microscope (Leica MZ 95, Germany) and OpenLab 4.0.3 software (Improvision Ltd., Lexington, Massachusetts, USA). Wood density (g cm–3) was calculated by dividing the dry mass of the stem segment used for hydraulic conductivity measurement by its fresh volume. Fresh volume was calculated from segment dimensions.
Because sampling for hydraulic measurements caused some loss of tissue, we measured growth on a separate sample of seedlings. Total biomass was determined by harvesting five randomly selected individuals per species 220 d after planting. Shoots were separated from roots directly above the first lateral root. Roots were rinsed of soil and all material was dried to constant mass at 65°C and weighed. To examine the regulation of gas exchange of each species, we estimated the ratio of photosynthesis to transpiration by estimating WUE through measurements of carbon isotope discrimination (δ13C; 19) on a subset of dried leaf tissue from all plants used for hydraulic measurements. To determine whether variation in δ13C was caused by variation in photosynthesis or stomatal regulation of water loss, we measured leaf nitrogen content (%N), which is an indicator of photosynthetic capacity (21). If δ13C and %N are not correlated, then variation in δ13C can be ascribed to stomatal behavior. We determined δ13C with a Thermo Finnigan Delta Plus mass spectrometer (Thermo Fisher Scientific, Waltham, Massachusetts, USA), and %N with a combustion elemental analyzer (ECS 4010, Costech Analytical Technologies, Valencia, California, USA) at the Stable Isotopes in Nature Laboratory (SINLAB, University of New Brunswick, Fredericton, NB, Canada; http://www.unb.ca/research/institutes/cri/labs/sinlab/index.html).
Climate data
Because we obtained seeds from a commercial source, specific geographic coordinates for seed collection sites were not available for most species, which also prevented us from obtaining site-specific climate information. In addition, seed for some species was obtained from populations planted outside of their native or historical ranges. To accommodate uncertainty associated with the lack of information on specific geographic locations of seed sources and the expanded geographic ranges of some species, we obtained average mean annual precipitation (MAP; includes snowfall) and mean annual temperature (MAT) information for each species, based on its extant geographic range according to several published sources (52; 6; 29; 7).
Comparative statistical analyses
We analyzed relationships among traits and between traits and climate both without and with phylogenetic information. The sign, strength, and statistical significance of the interspecific relationships (i.e., without phylogenetic information) between functional traits and climate variables were examined with regression (PASW Statistics 18.0; SPSS, Chicago, Illinois, USA). KS and plant mass were not normally distributed and were log transformed to meet assumptions of the statistical tests. Because MAP and MAT could both influence trait variation, we analyzed the combined effects of these two variables on functional traits using multiple regression. We tested whether including both MAP and MAT in the regression model violated multicollinearity assumptions by calculating variance inflation factors (VIFs; SPSS 18.0). All traits in the model had VIFs <10.0, which indicates that multicollinearity was low (51). Interspecific relationships among traits were evaluated with Pearson correlations.
To examine patterns of character evolution, we assembled a phylogenetic tree derived from published Pinus phylogenies (29; 24, 25; 16; 76), pruned to include only the species in our data set. We used the BLADJ function in PHYLOCOM 4.0.1 (71) to calculate branch lengths from fossil-calibrated Pinus node ages (25). To determine whether functional traits were correlated with each other and with climate data within a phylogenetic context, we calculated phylogenetic independent contrasts (PICs) using the AOTF function in PHYLOCOM. Contrasts were calculated for each trait and for each climate variable as the difference in x and y trait values between sister lineages along each branch of the phylogeny (20). Contrasts were standardized by dividing them by the standard deviation of the expected amount of change along each branch (22). A single contrast was calculated at each polytomy (71). We calculated the Pearson correlation among PICs and evaluated its statistical significance using linear regression with the intercept forced through the origin (22). As with the interspecific relationships, we analyzed the combined effects of MAP and MAT on functional traits using multiple regression with the intercept forced through the origin (e.g., 48).
To test whether functional traits differed among species, we used one-way analysis of variance (ANOVA) with species identity as the factor and KS, AL/AS, wood density, and plant mass as the dependent variables. We could not use ANOVA to test for interspecific differences in %N and δ13C because these traits were measured by pooling samples across individuals of each species. To test whether functional traits were conserved, we calculated contribution indices (CIs) for each node in the phylogeny and a tree-wide phylogenetic signal using the AOTF function in PHYLOCOM. A trait is considered conserved if more variation is explained by relatively ancient than by recent divergences in the phylogeny. Contribution indices vary between zero and 1 and estimate the degree to which individual nodal divergences along the phylogeny contribute to extant trait variation (49). Phylogenetic signal is derived from the tree-wide variance of standardized independent contrasts (4). If closely related lineages have similar traits, then the magnitude of the independent contrasts should be low, resulting in low tree-wide variance. To determine whether CIs and tree-wide phylogenetic signal were statistically significant (P ≤ 0.05), they were compared with a distribution of 1000 values calculated by randomly swapping trait values across the tips of the phylogeny (71).
RESULTS
Plant functional trait variation in Pinus was more frequently associated with mean annual temperature (MAT) rather than mean annual precipitation (MAP). Hydraulic, wood density and plant growth traits did not covary with MAP (Fig. 1), whereas more plant functional traits were significantly correlated with MAT (Fig. 2). In particular, KS and plant mass increased with increasing temperature. These relationships were also significant and positive when analyzed among phylogenically independent contrasts (PICs) (insets, Fig. 2A, C). Variation in δ13C was not correlated with MAT (R = –0.349, P = 0.08; PIC R = 0.235, P = 0.29) or with MAP (R = –0.188, P = 0.36; PIC R = 0.051, P = 0.82).
Relationships between mean annual precipitation and (A) specific hydraulic conductivity (KS), (B) leaf:sapwood area ratio (AL/AS), (C) plant mass, and (D) wood density for 26 Pinus species. Plots of phylogenetically independent contrasts (PICs) for each pair of traits and the corresponding PIC correlation coefficients are shown in the insets (ns = nonsignificant).
Relationships between mean annual temperature and (A) specific hydraulic conductivity (KS), (B) leaf:sapwood area ratio (AL/AS), (C) plant mass, and (D) wood density for 26 Pinus species. Plots of phylogenetically independent contrasts (PICs) for each pair of traits and the corresponding PIC correlation coefficients are shown in the insets (*P ≤ 0.05, **P ≤ 0.01, ns = nonsignificant).
In the multiple regression analysis, functional traits were also more frequently associated with MAT than with MAP in both the cross-species and PIC analyses. Specifically, KS (β = 0.763, P = 0.002; PIC β = 0.604, P = 0.01) and plant mass (β = 0.822, P < 0.0001; PIC β = 0.649, P = 0.01) increased with MAT, whereas wood density KS (β = –0.534, P = 0.02; PIC β = –0.654, P = 0.01) decreased with MAT. AL/AS (β = 0.208, P = 0.39; PIC β = 0.005, P = 0.99) and δ13C (β = –0.343, P = 0.15; PIC β = 0.410, P = 0.14) were not associated with MAT. Though plant mass was negatively associated with MAP among species (β = –0.412, P = 0.03), this relationship was not significant among PICs (PIC β = –0.139, P = 0.55). By contrast, wood density was not associated with MAP among species (β = 0.390, P = 0.08), but this relationship was positive and significant among PICs (PIC β = 0.595, P = 0.02). As in the univariate analysis, KS (β = –0.161, P = 0.37; PIC β = –0.068, P = 0.75), AL/AS (β = 0.064, P = 0.79; PIC β = –0.057, P = 0.84) and δ13C (β = –0.01, P = 0.96; PIC β = –0.295, P = 0.29) were not associated with MAP.
Plant mass was most frequently correlated with other plant functional traits (Table 2). For example, plant mass increased with KS and AL/AS and was negatively correlated with wood density and δ13C. Of these relationships, those between plant mass and KS and wood density remained significant among PICs. By contrast, the relationship between plant mass and AL/AS was not significant among PICs, and the negative interspecific relationship between plant mass and δ13C was positive among PICs. In addition to a correlation with plant mass, KS was negatively correlated with wood density and positively correlated with AL/AS, but neither of these correlations was significant among PICs (Table 2). There was a weak negative interspecific correlation between KS and δ13C, but the PICs of these traits were positively correlated. Other than plant biomass and KS, wood density was not correlated with any other functional trait among species. Moreover, wood density was only correlated with plant biomass among PICs. AL/AS and δ13C were not correlated with each other, either among species or among PICs. In addition, δ13C and leaf %N were not correlated, either among species (R = 0.067, P = 0.75) or among PICs (PIC R = –0.356, P = 0.11).
| Ks | Wood density | Plant mass | AL/AS | δ13C | |
|---|---|---|---|---|---|
| Ks | – | –0.503** | 0.821*** | 0.435* | –0.368 ns |
| Wood density | –0.390 ns | – | –0.578*** | –0.019 ns | 0.306 ns |
| Plant mass | 0.743 ** | –0.479* | – | 0.417* | –0.399* |
| AL/AS | 0.279 ns | –0.042 ns | 0.412 ns | – | –0.272 ns |
| δ13C | 0.414 * | –0.404 ns | 0.442* | 0.204 ns | – |
Functional traits differed significantly among Pinus species (Appendix S1; see Supplemental Data with the online version of this article). Interspecific variation was associated with the Pinus phylogeny (Fig. 3). We detected a marginally significant tree-wide phylogenetic signal for KS (contrast variance = 0.009, P = 0.056), which was conserved because the largest CI was associated with the deepest node in the phylogeny (node U; Fig. 3A). This node represents the divergence between Pinus and Strobus subgenera and accounted for 32% of extant trait variation (P = 0.053). No other node accounted for statistically significant variation in extant KS. WUE was also conserved (contrast variance = 0.309, P = 0.028) because the largest CI was also associated with the divergence between Pinus and Strobus subgenera, which explained 51% of extant variation (P = 0.007). Species in the Pinus subgenus had more negative δ13C (–29.58‰) than species in subgenus Strobus (–26.65‰), and no other node explained significant variation in this trait. Tree-wide phylogenetic signals for AL/AS (contrast variance = 0.001, P = 0.237), wood density (contrast variance = 0.0001, P = 0.152; Fig. 3C), and plant mass (contrast variance = 0.03, P = 0.114; Fig. 3D) were not statistically significant. The amount of variation differed among traits. In particular, the coefficient of variation for wood density (0.13) was lower than that for AL/AS (0.35), δ13C (0.61), KS (0.62), and plant mass (0.87).
Phylogeny with labeled nodes used for comparative analysis of physiological variation among 26 Pinus species along with mean (±1 SE) trait values for (A) specific hydraulic conductivity (KS), (B) leaf:sapwood area ratio (AL/AS), (C) wood density, and (D) plant mass. For presentation, branch-length information has been omitted, but this information was used to calculate independent contrasts and patterns of trait conservation.
DISCUSSION
Functional traits hypothesized to confer adaptation to water stress did not vary with climate in Pinus. Contrary to predictions, hydraulic trait (KS and AL/AS) variation among Pinus species was not associated with variation in annual precipitation from their extant ranges (Fig. 1). Pinus distribution along the precipitation gradient was also unlikely to be associated with vulnerability to xylem cavitation, because wood density did not consistently vary with precipitation and was on the low end (range: 0.20–0.35 g cm–3; Fig. 3B) of variation reported for conifers (0.3–0.8 g cm–3; 56). Moreover, the single instance of correlated evolution between wood density and precipitation identified in the multiple regression analysis was positive, and thus opposite to the prediction derived from the hydraulic model. If greater wood density is a predictor of increased conduit resistance to implosion in conifers (56), then the prevalence of low wood density in Pinus seedlings suggests that they are generally vulnerable to water-stress-induced cavitation. The potential for high vulnerability to cavitation in seedlings is consistent with previous estimates of this trait in adult trees of the genus (40; 42; 45, 44; 37) and with many reports of relatively high minimum leaf-water potentials in the field (Ψ > –2.0 MPa; 36; 15; 26; 17; 40; 3).
The absence of relationships between stem hydraulic traits and annual precipitation in Pinus seedlings conflicts with correlations between stem hydraulics and precipitation in field-grown populations (47; 14; 40; 44). One explanation for this discrepancy is that correlations between hydraulic traits and precipitation in the field are caused by plastic responses to water limitation in individual trees. This type of response could not be captured in our controlled-environment study. Nevertheless, stem hydraulic traits in Pinus have been shown to acclimate to water limitation in ways that conform to the predictions specified by Eq. 1. For example, AL/AS decreases with aridity when the same Pinus genotypes are grown in contrasting precipitation environments (47; 17; 43). This change increases leaf-specific hydraulic conductivity, which reduces the water-potential gradient required to extract water from soil and the likelihood of water-stress-induced cavitation (47). The frequent observation of hydraulic acclimation to water limitation (47; 17; 30; 11) along with a lack of ecotypic variation for these traits among populations (43) suggests that hydraulic plasticity to precipitation is an important mechanism that explains the broad distribution of Pinus (i.e., 37). Alternatively, we note that our ability to quantify the precipitation environment of each species was limited by a lack of information on the specific location of seed sources. It is therefore possible that weak relationships between precipitation and stem hydraulic traits were caused by error associated with characterizing the precipitation each population experienced in its home environment.
In contrast to annual precipitation, variation in stem KS was strongly correlated with variation in mean annual temperature from extant geographic ranges of Pinus (Fig. 2A). Three nonexclusive hypotheses can explain the positive relationship between KS and temperature. First, because hydraulic conductivity increases with conduit size (78), increased hydraulic capacity at high temperatures may have evolved through selection for increased conduit size in environments with high atmospheric evaporative demand (74; 39). This change may be advantageous in warm climates because increased KS in the absence of decreased AL/AS (Fig. 2B) raises plant hydraulic conductivity without the negative effects of leaf shedding on carbon gain (40). Second, species growing in cold climates tend to have lower growth rates because of physiological costs associated with increased resistance or tolerance to freezing (68; 34). Because xylem conduit diameter—and, thus, KS—scales with plant size (23; 39), the positive relationship between KS and temperature could be caused by the positive relationship between plant growth and temperature, rather than a direct effect of temperature on KS. Third, small tracheid diameters in Pinus (10–20 μm; 39; 44) could be advantageous for growth in cold temperatures. This is because freezing stress in many pines is accompanied by low soil water potentials (64), which can dramatically increase the vulnerability of conifer xylem to freezing induced cavitation (55). Small conduits would reduce the size of bubble formation in xylem during freezing events, making them less likely to experience air blockage by embolism (13; 54).
Our results suggest that wood density is less critical to climate adaptation in Pinus than other functional traits. Variation for wood density in the genus was rarely correlated with variation in climate (Figs. 1 and 2) or with variation in functional traits other than plant biomass (Table 2). It is possible that wood density in Pinus seedlings was low and not associated with climate because plants were grown in a greenhouse environment rather than in the field. Although wood densities would be higher in adult trees because of increased biomechanical stress, the wood densities in Pinus seedlings (Fig. 3C) were similar to the range of values reported for branches of adult pine trees from the same phylogenetic clades (0.35–0.45 g cm–3; 56). Nevertheless, our findings are consistent with a recent field study of co-occurring hardwood and conifer trees in New Zealand (63), where variation in wood density was not correlated with demographic and life-history variation.
Positive correlated evolution between stem hydraulic traits and plant biomass (Table 2) was consistent with relationships reported for both conifers (45, 44) and angiosperms (9; 58; 63), which indicates that hydraulic conductivity and growth rate are tightly linked. Our results also suggest that in addition to KS, stomatal regulation of water loss influences the evolution of increased growth in Pinus. For example, there was positive correlated evolution between KS and δ13C (i.e., WUE) as well as between δ13C and plant biomass (Table 2). Because %N and δ13C were not correlated, increased WUE can be attributed to reduced stomatal opening rather than changes in photosynthesis. As a result, fast-growing species with high KS appear to engage in isohydric stomatal regulation by closing stomata at relatively high water potentials to avoid tissue dehydration. This hypothesis is consistent with reports of isohydric stomatal regulation in adults of several pines (15; 26; 17; 40; 3). However, we note that because the Strobus subgenus had lower KS and growth, but higher δ13C than the Pinus subgenus, the relationships between growth and KS and between growth and δ13C were negative among species, but positive among PICs. The strong effect of the divergence between the two lineages, as well as the dominance of the Pinus subgenus in the sample, indicates that the association between fast growth, high KS, and isohydric stomatal regulation applies primarily to Pinus and not Strobus species.
A negative correlation between hydraulic efficiency and wood density is expected in comparative studies (67; 77) because the total lumen area in xylem increases KS (78) but decreases wood density (59). Although this negative correlation could be weak in angiosperms because intervessel components such as fibers and cell walls account for the majority of variation in wood density (30; 33; 77), it may be stronger in conifers, which lack the specialized xylem tissues of angiosperms (56; 75). Although we observed a negative interspecific correlation between stem KS and wood density in Pinus, there was no correlated evolution between these traits (Table 2). Thus, if increased KS evolves in Pinus (Eq. 1), it may not be accompanied by a costly increase in the vulnerability of xylem to implosion during water stress that is associated with lower wood density. This result may reflect the phylogenetic scale of the sample. The previously reported strong relationship between stem hydraulics and wood density was observed in a comparison across a wide taxonomic range of conifers (56), rather than within a single genus such as Pinus, where overall variability in wood density was lower. It is possible that significant interspecific relationships reported elsewhere may change when phylogenetic information is included (e.g., 41; 75).
A limitation of our study was that a majority of our species sample came from the Pinus subgenus. Although this subgenus is the most speciose and broadly distributed group in the lineage (Table 1; 25), we note that our conclusions may not necessarily represent the entire genus. In particular, the omission of the Cembroides (‘pinyon’) subsection of the Strobus subgenus, which exclusively occupy the arid regions western North America (62), suggests that we could have underestimated trait variation and phylogeneic conservation for hydraulic traits, as well as their covariation with climate. Another potential limitation is that absolute trait values for seedlings may differ from adult values. However, even though hydraulic traits change with development, the relative variation in seedling traits of Pinus can be indicative of adaptation in adults. This is because mortality is highest at the seedling stage (53), which suggests that natural selection on seedling traits may be particularly strong and could shape the functional traits of adults (35). Moreover, xylem traits and wood density are typically correlated across ages in Pinus (32), which suggests that the relative differences (or lack thereof) among species at the seedling stage would also be apparent for adult trees.
Although theory and previous research suggest that stem hydraulic traits would be correlated with both precipitation and temperature (e.g., 18; 47), we found that traits varied almost exclusively with temperature. In particular, decreased stem hydraulic capacity in cold climates may be a consequence of the evolution of reduced vulnerability of xylem to freezing induced cavitation. The absence of a relationship between stem hydraulics and precipitation in a common environment, but frequent detection of such relationships in comparisons among populations and species in the field (47; 14; 44), suggests that many pine species adjust to decreased moisture through plastic changes in hydraulic conductivity (37). Wood density, a trait associated with climatic adaptation in broad samples (67; 9), was not strongly associated with climate in Pinus. Thus, our work highlights how comparative studies at fine phylogenetic scales are necessary to qualify insights derived from comparisons at larger phylogenetic scales.
