HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (15) Citing Articles via Google Scholar Google Scholar Articles by Givnish, T. J. Articles by Goldstein, G. Search for Related Content PubMed Articles by Givnish, T. J. Articles by Goldstein, G. Agricola Articles by Givnish, T. J. Articles by Goldstein, G.

Adaptive radiation of photosynthetic physiology in the Hawaiian lobeliads: light regimes, static light responses, and whole-plant compensation points¹

²Department of Botany, University of Wisconsin, Madison, Wisconsin 53706 USA; ³Department of Biology, University of Miami, P.O. Box 249118, Coral Gables, Florida 33124 USA

Received for publication February 25, 2003. Accepted for publication August 14, 2003.

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

 
Six endemic genera/sections of lobeliads (Campanulaceae) occupy nearly the full range of light regimes on moist sites in the Hawaiian Islands, from open alpine bogs and seacliffs to densely shaded rainforest interiors. To determine whether this clade has undergone a corresponding adaptive radiation in photosynthetic adaptations, we studied the natural light habitats and physiological characteristics of 11 species representing each sublineage. Across species in the field, average photon flux density (PFD) varies from 2.3 to 30.0 mol · m^–2 · d^–1, and maximum assimilation rate (A_max) ranges from 0.17 to 0.35 µmol CO₂ · g^–1 · s^–1. Across species, A_max, dark respiration rate (R), Michaelis-Menten constant (k), light compensation point, specific leaf area (SLA), maximum carboxylation rate (V_cmax), maximum rate of electron transport (J_max), photosynthesis at saturating CO₂ (A_satCO2), and carboxylation efficiency () all increase significantly and in tightly coupled fashion with PFD, in accord with classical economic theory. Area-based rates have a higher degree of physiological integration with each other and tighter coupling to PFD than the corresponding mass-based rates, despite the energetic importance of the latter. Area-based rates frequently show adaptive cross-over: high-light species outperform low-light species at high PFD and vice versa at low PFD. A_max-mass has little relationship to leaf mass per unit area (LMA), leaf N content, or leaf lifespan individually, but a multiple regression explains 96% of the variance in A_max-mass across species in terms of SLA, leaf N content, and average PFD. Instantaneous leaf compensation points range from 0.1 to 1.2% full sunlight, far lower than the ecological (whole-plant) compensation points (ECPs) of 1.1 to 29.0% sunlight calculated based on photosynthetic parameters, leaf longevity, and allocation to leaf vs. nonleaf tissue. The ECPs are much closer to the lower limits of PFD actually experienced by lobeliads, suggesting they may play an important role in restricting species distributions. Taken together, these data provide evidence for an adaptive radiation in photosynthetic traits that is strongly correlated with—and indeed may help determine—the light regime that each species inhabits.

Key Words: adaptive cross-over • adaptive radiation • Brighamia • Clermontia • Cyanea • Delissea • Lobelia • photosynthetic light response • Trematolobelia

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

 
The Hawaiian lobeliads (Campanulaceae: Lobelioideae) consist of five genera and two sections of Lobelia, all endemic to the Hawaiian archipelago. This group is one of the most spectacular examples of adaptive radiation in flowering plants and has undergone striking diversifications in habitat, growth form, leaf shape, flower morphology, and mode of seed dispersal (Rock, 1919 ; Carlquist, 1965 , 1970 ; Lammers, 1990 ; Givnish et al., 1994 , 1995 ). With ca. 110 species, they comprise one-ninth of the Hawaiian flora (Wagner et al., 1990 ) and include alpine bog rosettes, seacliff succulents, and trees, treelets, and shrubs of mesic and wet forest edges and interiors. A few species are epiphytes and vines. A cpDNA phylogeny indicates that this remarkably divergent group arose from a single colonization (Givnish, 1998 ; T. J. Givnish et al., unpublished manuscript).

Surprisingly, nothing is known about the lobeliads' diversification in photosynthetic adaptations, even though they have invaded nearly the entire range of light environments in moist to wet habitats in Hawaii. Such habitats include (a) open alpine bogs and gaps in subalpine forest for Lobelia sect. Galeatella and Trematolobelia; (b) partly shaded seacliffs, interior rock faces, dry forests, and mesic- and wet-forest edges for Brighamia, Lobelia sect. Revolutella, Delissea, Clermontia, and purple-fruited Cyanea; and (c) densely shaded wet-forest interiors for orange-fruited Cyanea (Givnish, 1998 ).

Photosynthetic adaptations to contrasting light environments have long been a core concern of physiological ecologists (e.g., Björkman et al., 1972 ; Björkman, 1981 ; Chazdon and Pearcy, 1986a , b ; Chazdon, 1988 , 1992 ; Givnish, 1988 , 1995 ; Pearcy, 1988 , 1990 ; Mulkey et al., 1993 , 1996 ; Kitajima, 1994 ; Sims and Pearcy, 1994 ; Chazdon et al., 1996 ; Walters and Reich, 1996 , 2000 ; Feild et al., 2001 ; Valladares et al., 2002 ). Despite this, such adaptations have only rarely been studied in large numbers of closely related taxa (Euphorbia and Scaevola, Robichaux and Pearcy 1980 , 1984 ; Acer, Lei and Lechowicz 1997a , b , 1998 ; Psychotria, Valladares et al., 1997 , 2000 ) and never for plants whose distributions along light gradients have been quantified and whose phylogenetic relationships to each other are known in detail.

The Hawaiian lobeliads are a model system with which to study the evolution of photosynthetic adaptations to light availability, given their diversity in leaf form, distribution from nearly full sun to dense shade, and existence of a DNA phylogeny with which to interpret the roles of ecology vs. phylogeny in shaping the characteristics of present-day species. We have therefore initiated a long-term investigation of the ecology and evolution of lobeliad photosynthetic adaptations to different light regimes, focusing on 11 species that represent each major Hawaiian lineage. This first report details the light regimes experienced by these species in the field and their in situ photosynthetic light responses, leaf longevities, specific leaf area, N content, and various measures of mesophyll photosynthetic capacity. We analyze relationships among photosynthetic parameters, leaf traits, and photon flux density across species to test whether these trends accord with theoretical predictions. We ask whether species show evidence of differential adaptation to light availability in the form of adaptive cross-over in photosynthetic light response—that is, do shade species outperform sun species at low light levels, with the reverse at high light levels? Finally, we test whether the lowest light level at which each species is found corresponds to its ecological compensation point, based on the balance of the instantaneous rate of photosynthesis against current leaf respiration, nighttime leaf respiration, leaf construction costs amortized on leaf lifetime, and allocation to nonleaf tissue (Givnish, 1995 , 1998 ).

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

 
Study sites and species
We studied 11 species representing each major clade of Hawaiian lobeliads (Table 1). A cpDNA restriction-site phylogeny places Cyanea sister to Clermontia, Brighamia sister to Delissea, and Trematolobelia sister to Lobelia sect. Galeatella; Cyanea-Clermontia is sister to Brighamia-Delissea, and short branches connect this clade to Lobelia sect. Revolutella and Trematolobelia-Galeatella (Givnish et al., 1995 ; Givnish, 1998 ). Within the large genus Cyanea (ca. 65 spp.), there is a basal split between the purple-fruited species (exemplified by Cy. leptostegia) that typically grow in more open forests, and the orange-fruited species (exemplified by Cy. floribunda, Cy. hirtella, and Cy. pilosa var. longipedunculata) that typically inhabit densely shaded understories (Givnish et al., 1995 ). In Clermontia, Cl. fauriei is sister to all other taxa (except one intergeneric hybrid), while Cl. parviflora is one of the most recently derived species based on more recent study using ISSR variation (Givnish et al., 2000 ).

Light regimes
We measured daily courses of photon flux density (PFD, in micromoles per square meter per second) immediately above 5–12 plants per population of each species, using GaAs photodiodes (Hamamatsu, New York, New York, USA) connected to CR-10X dataloggers equipped with multiplexers (Campbell Scientific, Logan, Utah, USA). Photodiodes were calibrated against a LI-COR 190S quantum sensor (LI-COR, Lincoln, Nebraska, USA). We logged 10-s average of 1-s measurements for 2 wk each in winter and summer in 28 study populations. Rarity and difficulty of access permitted study of only two populations of Lobelia villosa, L. yuccoides, and Trematolobelia kauaiensis, and only one of Cyanea hirtella. Average values of instantaneous and total daily PFD were tabulated for each individual; means ± SE (N = number of individuals) were then calculated for each species. Within each species, we compiled similar statistics for the focal individuals whose photosynthetic performance was measured and tabulated the lowest average PFD for an individual plant within each species. Across species, the mean PFD for focal plants is strongly correlated with that for all conspecifics (y = 1.11x + 0.21, r² = 0.84, P < 0.0001 for 9 df). Given the strength of the relationship between focal and global PFDs, and the much larger and more representative sample across all plants within a species, we related interspecific trends in photosynthetic parameters to global PFD rather than local PFD. This is a conservative approach: many of the trends reported are as strong, and often slightly stronger, when related to focal PFD instead.

Photosynthetic measurements
Static light responses
We measured steady-state (i.e., fully induced) photosynthetic light responses (assimilation rate [A] vs. PFD curves) of each species in the field, using attached, fully expanded leaves from five individuals whose light regimes had been quantified. Leaves were clamped into the cuvette of an LI-6400 portable photosynthesis system (LI-COR) equipped with a red/blue LED light source and a CO₂ mixing system. We measured photosynthesis at 10–15 PFD levels between 0 and 1000 (to 1500) µmol · m^–2 · s^–1. At each PFD level, gas exchange values were recorded after stomatal conductance and CO₂ uptake stabilized, beginning at a PFD of 500 µmol · m^–2 · s^–1 to photoactivate ribulose-1,5-biphosphate carboxylase/oxygenase (Rubisco) and open the stomata (Pearcy, 1990 ; Valladares et al., 1997 ). We maintained relative humidity, leaf temperature, and cuvette CO₂ concentration within narrow limits close to those experienced by most species (ca. 75% relative humidity, 20°–25°C, 360 µmol/mol CO₂). All measurements were made between 0600 and 1300 hours. There was little relationship between precipitation and season during the period of this study, so we did not stratify sampling between "dry" and "wet" seasons. Specific leaf area (SLA, in square centimeters per gram), its inverse leaf mass per unit area (LMA, in grams per square meter) and leaf N content (in milligrams per gram) were determined from leaf punches, excluding veins >1.5 mm diameter. Net photosynthetic rates per unit leaf mass and area were each fit to the Michaelis-Menten model

(1)

(where A_max = maximum assimilation rate, k = Michaelis-Menten constant, and R = dark respiration rate) (Fig. 1) using a third-order Newton-Rapheson algorithm. Data for individual leaves fit this model quite well, with values of r² generally exceeding 0.98. The slope of the linear portion of the light response curve (the first three or four points, depending on goodness of fit) was estimated using linear regression, and quantum yield () was then calculated as that slope divided by absorptance.

View larger version (14K): [in this window] [in a new window]   Fig. 1. Schematic diagram of Michaelis-Menten photosynthetic light response, showing asymptotic convergence of net carbon uptake on Amax – R at high PFD, rising from –R (dark respiration) at zero PFD and traversing half the potential range in net uptake by a PFD of k (see Eq. 1 ). Note that Amax is substantially above the actual maximum net photosynthetic rate. The instantaneous compensation point (ICP) is achieved where A(PFD) = 0, that is, at the photon flux density at which the light-response curve crosses the x-axis

Mesophyll photosynthetic capacity
We measured CO₂ response curves (A vs. c_i) on the same five, fully induced leaves of each species used for light response curves. We varied c_i from nearly 0 to 1000 (to 1500) µmol/mol in seven steps, all at saturating PFD, by varying cuvette (CO₂) with other conditions constant. We used these data to estimate carboxylation efficiency (initial slope of A vs. c_i), maximum photosynthetic rate at saturating CO₂ (A_satCO2), maximum rate of electron transport (J_max), maximum carboxylation rate (V_max), and CO₂ compensation point () using Photosyn Assistant (Dundee Scientific, Dundee, UK), based on the photosynthetic models of Farquhar et al. (1980) , von Caemmerer and Farquhar (1981) , Sharkey (1985) , Harley and Sharkey (1991) , and Harley et al. (1992) .

In situ leaf characteristics
In addition to SLA and leaf N content, we measured the absorptance of each leaf studied for photosynthesis, using an appropriately filtered quantum sensor and an integrating sphere (LI-COR). Leaf longevity was estimated by marking the petioles of newly expanded leaves with a census-specific colored ink at bimonthly intervals and then monitoring the presence or absence of each marked leaf and fraction of surviving tissue at subsequent censuses. We followed all leaves produced on at least 10 twigs on different individuals of each species for at least 1 yr. The resulting data were analyzed using an equilibrium conveyor-belt model to obtain leaf longevity (LL) as

(2)

where N is the average total number of leaves held per twig, T is the average intercensus time interval, and n is the average number of new leaves per census. Estimates obtained using this approach were similar to those calculated by applying more complex demographic techniques to the dates of leaf births, partial losses, and deaths (results not shown).

Data analysis
The Fisher least significant difference (LSD) multiple comparison of means ANOVA was used to determine which species significantly differed in average daily PFD. Correlations and simple LMS regressions among measured variables across species were calculated using standard techniques. Regressions involving PFD used light data expressed on a daily basis (in moles per square meter per day). Log transformation was used to linearize relationships where necessary. Phylogenetically structured analyses (e.g., Felsenstein, 1985 ; Harvey and Pagel, 1991 ; Pagel, 1994 , 1999 ) were not employed because we believe they are inappropriate for the kind of data reported here, which reflect both genetic differences among related species and differences among those species in light habitat. Given that photosynthetic light responses are highly plastic in response to conditioning light levels (Björkman, 1981 ) and that different species/lineages of Hawaiian lobeliads occur in substantially different light environments, using a phylogenetically structured analysis on field data would conflate the effects of phylogeny and ecology.

Backward-elimination multiple regressions were used to identify variables with significant effects on A_max. Potential predictors of A_max included raw and log-transformed values of PFD, leaf N content, SLA, LMA, stomatal conductance, and leaf longevity.

Given the large number (27) of morphological, physiological, and environmental parameters quantified for each species, and the strong likelihood of cross-correlations among them, we used principal components analysis (PCA) to identify the most important axes of covariation among parameters (PC-ORD; McCune, 2002 ). We conducted PCA on three sets of parameters, involving those based on leaf area (e.g., in grams of CO₂ per square meter), leaf mass (e.g., in grams of CO₂ per gram), or both, including in each case the four traits that depend on both area and mass or neither (SLA, , and leaf lifespan), and excluding PFD and SLA's inverse LMA. We assessed the strength and pattern of covariation among parameters using joint plots of each parameter (including PFD) against the ordination axes. Joint plots portray the strength and direction of a parameter's association with the ordination axes by superimposing vectors whose components are the correlation coefficients (r) of that parameter with each axis (McCune, 2002 ).

We counted the number of species pairs that displayed adaptive cross-over, in which the light response curve of the species native to higher PFD starts lower than that of the low-light species, but ends higher. Björkman et al. (1972) and Givnish (1988 , 1995) argue that low-light species should outperform others at low PFD, and high-light species should outperform others at high PFD, especially when performance is measured as photosynthesis per unit leaf mass or other energetic investment or as whole-plant carbon gain per unit investment in leaf and nonleaf tissue. To test this proposition, we first constructed a diagonal matrix tabulating adaptive cross-overs, nonadaptive cross-overs, and non-cross-overs for all 55 ordered species pairs, with the order of individual columns/rows reflecting the rank of each species by average daily PFD. We then randomly permuted the rows and columns of this matrix in tandem, effectively shuffling the PFDs associated with each species. This process was repeated 100 000 times. The resulting distribution of "adaptive" vs. "nonadaptive" cross-overs above the diagonal was used to calculate whether the actual number of adaptive cross-overs significantly exceeded the null expectation.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

 
Light environments
Across species, average instantaneous PFD varies 13-fold in essentially continuous fashion, from 53 ± 5 µmol · m^–2 · s^–1 for Cyanea floribunda in shaded wet-forest understories to 695 ± 69 µmol · m^–2 · s^–1 for outplanted Brighamia insignis on an open hillside (Tables 1 and 2). The corresponding daily totals ranged from 2.3 ± 0.2 to 30.7 ± 3.0 mol · m^–2 · d^–1. The minimum average PFD experienced by a species is strongly correlated with its average PFD (y = 0.132x^1.31; r² = 0.90, P < 0.0001 for 9 df); the minimum/mean ratio is 31 ± 16%, with the expected value rising from 12 to 40% with increasing PFD.

Photosynthetic light responses
Classical economic theory (Björkman et al., 1972 ; Björkman, 1981 ; Givnish, 1988 , 1995 ) predicts that species adapted or acclimated to sunny habitats should display higher maximum rates of photosynthesis and higher rates of dark respiration and require more light to saturate photosynthesis and balance respiration than shade species. These predictions are nicely illustrated by the static light responses of Clermontia parviflora and Cyanea pilosa, which occur under gaps and dense canopies, respectively, in the same wet forests of windward Hawai'i (Fig. 2). As expected, photosynthesis per unit leaf mass in Clermontia parviflora reaches higher maximum levels, saturates at higher light levels, entails a higher rate of dark respiration, and requires more light to balance leaf respiration (i.e., has a higher leaf compensation point) than does Cyanea floribunda (Fig. 2). The static light response curves cross, with Clermontia having an advantage above 290 µmol · m^–2 · s^–1 and Cyanea having an edge under darker conditions (see Discussion for further details). Cyanea has an instantaneous leaf light compensation point of 2.8 ± 0.2 µmol · m^–2 · s^–1, while Clermontia requires 10.4 ± 1.3 µmol · m^–2 · s^–1 to achieve zero net photosynthesis.

View larger version (20K): [in this window] [in a new window]   Fig. 2. Static photosynthetic light responses of Cyanea pilosa subsp. longipedunculata and Clermontia parviflora in the Ola'a Tract, Hawaii Volcanoes National Park. Dots and error bars represent the mean ± SE of measurements on five individuals of each species; curves are Michaelis-Menten functions based on the mean values of Amax, k, and R across individuals. Net CO2 assimilation shows adaptive cross-over, with understory Cyanea having a higher rate of net photosynthesis below ca. 290 µmol·m–2·s–1 PFD and gap-dwelling Clermontia having a higher rate above that light level. Compared with Cyanea, Clermontia has a higher maximum rate of carbon uptake, needs more light to saturate photosynthesis, and possesses a higher respiration rate and instantaneous light compensation point

View larger version (13K): [in this window] [in a new window]   Fig. 3. Maximum photosynthetic rate (Amax) as a function of average photon flux density experienced by 11 representative species of Hawaiian lobeliads, expressed per unit (A) leaf area and (B) leaf mass. Dots and error bars represent the mean ± SE of photosynthetic data for five individuals of each species, and the mean ± SE of PFD measurements for all datalogged individuals of each species. From left to right, the species are Cyanea floribunda, Cy. pilosa var. longipedunculata, Cy. hirtella, Cy. leptostegia, Clermontia parviflora, Lobelia yuccoides, Delissea rhytidosperma, Cl. fauriei, Trematolobelia kauaiensis, L. villosa, and Brighamia insignis; the same sequence is used for all plots of photosynthetic data as a function of PFD (see Table 3 ). Curves represent LMS regressions: (A) Amax = 5.09PFD0.326, r2 = 0.64 for all species (P < 0.005 for 9 df); and (B) Amax = 0.0146PFD + 0.114, r2 = 0.75 excluding four species of boggy sites or possessing substantial water storage tissue (hollow dots) (P < 0.001 for 5 df)

View larger version (12K): [in this window] [in a new window]   Fig. 4. Respiration for 11 species of Hawaiian lobeliads, expressed (A) per unit leaf area as a function of average daily PFD (R = 0.054PFD + 0.348, r2 = 0.77, P < 0.0005 for 9 df); (B) as a fraction of Amax and function of PFD (R/Amax = 0.0027PFD + 0.056, r2 = 0.65, P < 0.005 for 9 df); and (C) as a function of Amax alone (R = 0.099.Amax, r2 = 0.64, P < 0.005 for 9 df).

View larger version (11K): [in this window] [in a new window]   Fig. 5. Leaf mass per unit area (LMA) and the Michaelis-Menten constant k as a function of average daily PFD for 11 representative species of Hawaiian lobeliads. Lines represent LMS regressions: (A) LMA = 2.67PFD + 33.1, r2 = 0.47 (P < 0.02 for 9 df); (B) k = 3.75PFD + 56.7, r2 = 0.72 (P < 0.001 for 9 df). See Fig. 2 for sequence of species from left to right

View larger version (22K): [in this window] [in a new window]   Fig. 6. Light compensation points–instantaneous (ICP), leaf (LCP), and whole-plant or ecological (ECP)–as a function of average daily PFD. Dots represent ECPs and actual mean PFDs for individual species (see Fig. 2 for sequence from left to right); error bars represent the standard errors of mean PFDs. For clarity, data for individual species are not shown for calculated ICPs and ECPs. The dashed line running 45° across the figure is CP = PFD. Note that the traditional ICPs are far below the actual light levels inhabited by species, while ECPs are much closer to (but still below) the mean PFD occupied by each species

Mesophyll photosynthetic capacity and related traits
As average PFD rises across species, there are significant increases in carboxylation efficiency (), the area-based maximum rate of carboxylation (V_cmax), maximum rate of electron transport (J_max), and maximum rate of photosynthesis at saturating c_i (A_satCO2). Variation in these characteristics appears to be closely coordinated, with strong correlations among most traits (Table 4), the strongest being that between electron transport capacity J_max and A_satCO2 (r = 0.96***). A_max is most strongly correlated to (r = 0.95***), k to A_satCO2 (r = 0.90***), and R to V_cmax (r = 0.80***) (Fig. 7). Leaf absorptance shows no significant relationship to PFD, but is strongly tied to leaf N_area (r = 0.97***) and LMA (r = 0.87***). The values of N_area and LMA, as well as quantum yield and stomatal conductance, have no significant relationship to PFD (Table 4).

View larger version (13K): [in this window] [in a new window]   Fig. 7. Amax, k, and R in Hawaiian lobeliads are closely related to parameters of mesophyll photosynthetic capacity based on A vs. ci curves. Lines represent LMS regressions. (A) Amax = 25.9carboxylation efficiency + 0.68, r2 = 0.90 (P < 0.0001 for 9 df); (B) k = 0.48AsatCO2 + 15.9, r2 = 0.85 (P < 0.0001 for 9 df); (C) R = 0.37Vcmax – 0.04, r2 = 0.64 (P < 0.005 for 9 df).

View larger version (12K): [in this window] [in a new window]   Fig. 8. Axis 1 score of the PCA ordination of photosynthetic parameters as a function of average daily PFD received by each species

View larger version (17K): [in this window] [in a new window]   Fig. 9. (A) PCA ordination of Hawaiian lobeliad species by photosynthetic parameters. (B) Joint plot of photosynthetic parameters and PFD; direction and length of the vectors corresponding to each measurement correspond to the strengths of the correlation (r) between that measurement and scores on axis 1 and 2 of the PCA ordination. Note the near colinearity of several area-based parameters with each other, PFD, and axis 1, and the near colinearity of several mass-based parameters with each other and axis 2

The PCA of the mass-based variables, SLA, leaf lifespan, and also produces two significant axes, with the first explaining 57.6% of the variance and the second 20.9%. Axis 1 is strongly associated with daily PFD (r = 0.83**), although more weakly than the relationship seen for area-based parameters. Across mass-based parameters, 13 (62%) of 21 pairwise correlations are statistically significant. Maximum stomatal conductance per unit leaf mass is aberrant in having no significant correlations with any of the other mass-based parameters. Even though axis 1 of the mass-based PCA is significantly correlated with PFD, no individual mass-based parameter is significantly correlated with daily PFD. Based on these results, there appears to be a greater degree of physiological integration based on leaf area than on leaf mass. This tendency is especially striking if one focuses only on A_max, R, k, the various light compensation points, A_satCO2, V_cmax, and J_max. In that case, PCA axis 1 recovers 83.8% of the total variance, is essentially collinear with each of the included photosynthetic parameters, and correlates nearly perfectly with daily PFD (r = 0.94***).

Adaptive vs. nonadaptive cross-over of static light responses
As expected from the fact that A_max, k, and R increase with average daily PFD, many pairs of species show adaptive cross-over, with species from sunnier conditions having a higher photosynthetic rate at high PFD and a lower rate at low PFD (see Fig. 2, Table 5). Of the 55 distinct pairwise comparisons among species, 38 display cross-over in their area-based photosynthetic light responses; 36 of these cross-overs are adaptive, a highly significant deviation from the null expectation and toward the prediction by economic theory (P < 4.5 x 10^–4, permutation test). There are 43 cross-overs for mass-based rates, of which 30 are adaptive, a marginally nonsignificant tendency in the same direction (P > 0.066, permutation test).

Among the Hawaiian lobeliads, leaf longevity is not significantly related to A_max,mass, N_mass, or SLA. However, A_max,mass has a strong positive correlation to N_mass and SLA (Fig. 10). The proportions of variance explained by the log-log regressions and their slopes were generally much less than those reported by Reich et al. (1999) , as would be expected if the Hawaiian lobeliads simply resample the data distribution across vascular plants within a much narrower range of each variable. Within the Hawaiian lobeliads, leaf lifespans vary from 3.8 ± 0.3 mo in Clermontia parviflora to 10.1 ± 0.4 mo in Cyanea floribunda, representing a 2.7-fold variation, 30 times less than that in the global data set. By contrast, leaf N content and SLA vary by 2.3- and 4.6-fold in the Hawaiian lobeliads, only 4.2 and 6.7 times less than the variation in the global data set. The much lower proportional variation in leaf lifespan in lobeliads, combined with the lower slopes of the power-law regressions on leaf lifespan vis-à-vis N_mass and SLA in the global data set, probably account for the lack of significant relationships of lobeliad leaf lifespan to A_max, leaf N content, and SLA (see also Diemer et al., 1992 ; Reich, 1993 ; Diemer, 1998 ).

View larger version (9K): [in this window] [in a new window]   Fig. 10. Values of Amax-mass as a function of (A) leaf N content per unit leaf mass and (B) specific leaf area. Lines represent LMS regressions: Amax = 0.0075[N]– 0.029, r2= 0.51 (P < 0.02 for 9 df); Amax = 0.00071SLA + 0.06729, r2= 0.48 (P < 0.02 for 9 df)

(3)

(P < 0.005 for 8 df). Neither parameter had a significant effect in this regression, however. Given the strong relationships of several physiological parameters to A_max and to average daily PFD (see earlier), we asked whether including daily PFD in a multiple regression—as a proxy for the effects it has on other physiological factors, and hence, on A_max—would increase our ability to predict A_max. Indeed, including daily PFD allows us to account for fully 96% of the variance in A_max across species:

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

 
Adaptive radiation is a recurring evolutionary process on oceanic islands, where extreme isolation permits but a few colonists to arrive from mainland source areas, and intense competition among the closely related, functionally similar descendants selects for ecological divergence (Lack, 1940 , 1947; Carlquist, 1965 , 1970 ; Grant, 1986 ; Givnish, 1997 , 1998 ; Schluter, 2000 ; Losos et al., 2003 ). The Hawaiian Islands, with their tremendous present-day diversity in elevation and rainfall, unsurpassed isolation from continental areas, and a geological history spanning at least 80 Myr, have a fauna and flora whose composition and ecology have been shaped fundamentally by adaptive radiation (Carlquist, 1970 ; Robichaux et al., 1990 ; Wagner et al., 1990 ; Gillespie et al., 1994 , 1997 ; Carson and Clague, 1995; Givnish et al., 1995 ; Baldwin, 1997 ; Kambysellis and Craddock, 1997 ; Givnish, 1998 ; Weller et al., 1998 ; Lovette et al., 2002 ; Remsen and O'Grady, 2002 ; Shaw, 2002 ; Carlquist et al., 2003 ). Examples include the Drepanidae, with a range of bill morphologies comparable to that seen across almost all other passerine families; Drosophilidae, Laupala, and Megalagrion among the insects; Tetragnatha among spiders; and the Hawaiian lobeliads, silversword alliance, mints, spurges, and pinks among the vascular plants. Fully half of the native vascular flora is derived from 20 initial colonists (Wagner et al., 1990 ), and the Hawaiian lobeliads alone account for nearly a quarter of the species derived from these leading radiations.

The last decade has seen revolutionary advances in our understanding of the origins and phylogeny of several Hawaiian groups, but little is yet known about the functional and ecological significance of phenotypic divergence within putative radiations. Exceptions include work on the light responses of Euphorbia species from habitats with different light regimes and rainfalls (Robichaux and Pearcy, 1980 ; Pearcy et al., 1982 ), water-use efficiencies of Euphorbia and Scaevola species from areas with different rainfall (Robichaux and Pearcy, 1984 ), turgor maintenance and hydraulic elasticity in the silversword alliance (Robichaux, 1984 ; Robichaux and Canfield, 1985 ; Robichaux et al., 1986 , 1990 ), shifts in leaf longevity, nutrient use efficiency of P vs. N, and cold tolerance of populations of Metrosideros polymorpha at different elevations and on substrates of different age (Cordell et al., 1998 , 2000 , 2001a , b ), and outcrossing mechanisms in Alsinodendron/Schiedea (Weller et al., 1995 , 1998 , 2001 ; Sakai et al., 1996); research on drought tolerance in the latter group has also begun (T. Dawson, University of California at Berkeley, personal communication).

In all but the second of these cases, there is some evidence that phenotypes are correlated with habitat in ways that seem likely to enhance plant performance within a species' native habitat. The most notable study in our context is that of Robichaux and Pearcy (1980) , who showed that forest-dwelling Euphorbia forbesii had lower rates of photosynthesis and respiration per leaf area, and lower light saturation and compensation points, than open-habitat Euphorbia celastroides and E. degeneri, when all were grown at ca. 500 µmol · m^–2 · s^–1 with ample water and nutrients. Euphorbia hillebrandii from partly open slopes was intermediate in all respects. This was an important pioneering study, demonstrating adaptive cross-over of area-based photosynthetic rates in five of six species pairs, even though it lacked data on species photosynthesis and whole-plant growth in the field, on the quantitative distribution of species along light gradients, on the relationship between various photosynthetic parameters and native PFD, and on the phylogenetic relationships of the taxa surveyed. Our paper reports the first stage of a parallel but more extensive investigation, involving the largest adaptive radiation among Hawaiian plants, and lineages that have invaded the entire light gradient under similar, moist conditions. Taken together, the data presented here provide evidence for an adaptive radiation in photosynthetic physiology that is strongly correlated with—and indeed may help determine—the light regime that each species inhabits.

Light regimes and static light responses
Species representing each lineage of Hawaiian lobeliads occupy dramatically different light habitats, with daily PFDs varying from 2.3 mol · m^–2 · d^–1 in Cyanea floribunda to 30.7 mol · m^–2 · d^–1 in Brighamia insignis, nearly the entire range for native moist and wet habitats (Table 2). Some orange-fruited species of Cyanea occupy the shadiest understories on Hawaii, but these are still roughly nine times brighter than those in the understories of continental tropical forests in Central and South America and Malesia (Chazdon and Fetcher, 1984 ; Bellingham et al., 1996 ; Valladares et al., 1997 ; Montgomery and Chazdon, 2001 ).

The static photosynthetic light responses of the 11 lobeliad species studied provide striking support for several predictions of classical economic theory. Maximum photosynthetic rate per unit area increases with the average PFD incident on species, as does dark respiration, the Michaelis-Menten constant k, leaf mass, and several measures of photosynthetic capacity per unit area (Figs. 2–7). Species in brightly lit environments thus respire at higher rates at low PFD, require more light to achieve a zero net rate of carbon uptake and to saturate photosynthesis, and achieve higher rates of photosynthesis at high PFD—all on a leaf-area basis—than species found in shadier habitats. A similar pattern is seen in net photosynthesis per leaf mass if species of boggy habitats or with substantial water storage tissue in their leaves are excluded. Taken together, these data are unparalleled in documenting adaptive patterns of variation in photosynthetic traits across species distributed along quantified light gradients.

Leaf lifespan and structural determinants of photosynthetic rate
Across the Hawaiian lobeliads, leaf lifespan shows only weak, nonsignificant relationships to A_max and leaf N content on a mass basis or to SLA (Table 4). The weakness of these relationships, relative to that seen across plants from several biomes (Reich et al., 1997 , 1999 ), may simply reflect the limited variance in leaf lifespan across the Hawaiian lobeliads (see Results). Alternatively, these weak relationships might mean that leaf lifespan is more strongly affected by A_max, leaf N content, and LMA than vice versa, given the strong correlations among these last three variables in the same limited number of species and the strong correlations they have with other measures of photosynthetic capacity and PFD. High rates of photosynthetic return per unit leaf mass guarantee rapid overtopping of old leaves by new leaves and favor short leaf lifespans (Reich et al., 1992 ; Ackerly, 1999 ; Givnish, 2002 ; Kitajima et al., 2002 ). This paradigm, however, must be modified to take into account the apparent increase in allocation to nonleaf tissue among the slowest growing species with the longest-lived leaves, often found on infertile sites (Givnish, 2002 ). The low fraction of variance in A_max explained by ln N and ln SLA, compared with the nearly complete explanation when ln PFD is included (Eq. 4), suggests that there are one or more important structural determinants of leaf photosynthetic capacity not captured in leaf N content and LMA. Identification of these additional determinants should be a high priority, given the potential utility of the relationship of A_max to leaf N content and SLA already established (Reich et al., 1997 ).

Instantaneous vs. whole-plant compensation points and partitioning of the light gradient
Instantaneous leaf compensation points range from 0.1 to 1.2% full sunlight, much lower than the ecological (whole-plant) compensation points of 1.1–29.0% sunlight calculated based on photosynthetic parameters, leaf longevity, and allocation to leaf vs. nonleaf tissue (Fig. 6). The ECPs are much closer to the lower limits of PFD actually experienced by lobeliads, suggesting they—unlike the traditional, instantaneous compensation points—may play an important role in restricting species distributions. This is the first study to test Givnish's (1988 , 1995) ECP theory and demonstrate a relationship between whole-plant compensation points and the distributions of species along light gradients.

Ecological compensation points should constrain species' occurrences at the low end of their PFD range. Greater growth by competitors may limit their occurrence at high PFDs, as suggested by the adaptive cross-over of static light responses in many pairs of taxa, in which the species from shadier conditions has a higher rate of photosynthesis at low PFD, while the species from sunnier conditions has a higher rate at high PFD (Fig. 2; Table 5). Of 55 species pairs, 36 display adaptive cross-over in their area-based light responses, while 30 show adaptive cross-over in their mass-based responses. The fraction of adaptive vs. nonadaptive cross-overs is significantly greater than expected for area-based rates (P < 4.5 x 10^–4) and marginally nonsignificant (P > 0.066) for mass-based rates. Greater leaf thickness and LMA, as well as higher leaf N content and associated metabolic capacity, should each tend to increase maximum photosynthesis, dark respiration, and the amount of light needed to saturate photosynthesis (Björkman et al., 1972 ; Björkman, 1981 ; Givnish, 1988 , 1995 ). If A_max, k, and R were thus to increase in lock-step with each other as PFD increases, we would expect all species pairs to exhibit adaptive cross-over in their light response curves. The best measure of deviation from an "expected" number of cross-overs of any form may thus be the tightness of fit of the regressions of A_max, k, and R against PFD. For area-based measures, the values of r² for these regressions are significantly greater than the null expectation in each instance, and A_max, k, and R are all strongly correlated with each other (0.83 < r < 0.91).

Integration and adaptive cross-over of photosynthetic parameters on an area vs. mass basis
It is surprising that many components of photosynthetic capacity show greater integration with each other (Fig. 9) and apparent adaptation to PFD (Figs. 2–8) based on leaf area rather than mass. Givnish (1988 , 1995) argued that competition in a given habitat should favor plants that maximize energetic returns per unit investment in leaf mass, leaf N content, or (best) whole-plant energy content. Returns per unit of energy invested should be more important than returns per unit leaf area, because competitive ability should be tied more closely to whole-plant growth and, hence, to returns on investment. Givnish (1988) showed that the paradigmatic study of Björkman et al. (1972) on Atriplex triangularis grown at three different PFDs supported the claim of adaptation only if photosynthetic rates per unit leaf area were reanalyzed on the basis of leaf mass or soluble protein content. Only then did plants grown at each of the conditioning PFDs exhibit adaptive cross-over, with each group having a photosynthetic advantage only at or near the light levels at which they were grown.

Several studies over the past decade have shown a similar shift in the ranking of species by photosynthetic rate or whole-plant growth in moving from low to high PFD. Kaelke et al. (2001) found a reversal in the relative growth rates (RGR, in grams per gram per day) of field-grown seedlings of Populus tremuloides vs. Acer saccharum and Quercus rubra in moving from 2.6 to 69% sunlight in Wisconsin, with sun-adapted Populus outgrowing Acer and Quercus above 2.5 mol/d PFD. Growth rates were related weakly to photosynthesis per leaf area, strongly to photosynthesis per leaf mass, and most strongly to photosynthesis per total plant mass. Similar reversals in species' growth ranks of temperate tree seedlings have been reported by Pacala et al. (1994) , Walters and Reich (1996 , 2000) , and Reich et al. (2003) . In tropical rain forest, Montgomery and Chazdon (2002) recently demonstrated cross-over in whole-plant growth (in grams per gram) by seedlings of Dipteryx panamensis, Virola koschnyii, and Brosimum aliscastrum transplanted along a light gradient from 0.2 to 6.5% full sunlight at La Selva, Costa Rica. Virola and Brosimum outgrow Dipteryx below 2.5% full sunlight, while Dipteryx outgrows Brosimum and Virola above that level. All tree species survive at similar rates above 2.5% sunlight, but in denser shade the survival of Dipteryx and Brosimum falls below that of Virola. When Kobe (1999) took both survival and growth into account, he was able to show that seedlings of each of four closely related tree species (Trophis racemosa, Castilla elastica, Pourouma aspera, Cecropia obtusifolia), transplanted along almost the entire light gradient at La Selva (<1% to ca. 85% full sunlight), have an advantage in realized growth relative to competitors along a specific part of that gradient.

In each of the eight studies mentioned in the two preceding paragraphs, species (or acclimated clones) show adaptive cross-over in energetic return on investment, whole-plant growth, or realized whole-plant growth as a function of light availability. This pattern of behavior should permit species to partition the light gradient, with each having a range of conditions where it has an advantage in growth and, presumably, compe