CONSTRUCTION COSTS , PAYBACK TIMES , AND THE LEAF ECONOMICS OF CARNIVOROUS PLANTS 1

Understanding how the investment (or allocation) of carbon and mineral nutrients varies in different plant organs and among species, plant functional types, and the vegetation of different biomes is a major goal for plant ecology ( Wright et al., 2004 ). A synthesis of global data on plant traits — leaf nitrogen (N), phosphorus (P), and potassium (K) content (%), leaf mass area (LMA: g/cm 2 ), leaf lifespan, and maximal leaf photosynthetic rates ( A mass : nmol CO 2 ⋅ g − 1 ⋅ s − 1 ) — of 2548 species (the Glopnet data set; Wright et al., 2004) illustrated and quantifi ed the bivariate scaling relationships between traits (e.g., A mass as a function of %N) and revealed a “ universal spectrum of leaf economics ” ( Wright et al., 2004 , 2005 ). This spectrum defi nes trade-offs in plant allocation when changes in one trait co-vary consistently with changes in a second trait. Additional research has identifi ed carnivorous plants ( Ellison and Farnsworth, 2005 ; Farnsworth and Ellison, 2008 ), mangroves ( Ellison, 2002 ), and lianas in tropical forests ( Santiago and Wright, 2007 ) as notable outliers relative to the universal spectrum of leaf traits. These exceptions to the rule have reinforced the importance of tradeoffs in our understanding of why leaf traits consistently scale with one another. Shipley et al. (2006) proposed that the trade-off between allocation of nutrients to structural tissues and long-term storage vs. their immediate use in boosting photosynthetic rates is a potential mechanism for the observed coordination among leaf traits. Such trade-offs can be observed easily with carnivorous plants because carnivorous plants inhabit open environments (e.g., bogs and other wetlands) where light and water are not limiting but nutrients are in extremely short supply, and therefore it is relatively easy to separate out experimentally the effects of nutrient limitation from effects of limitation of light or water ( Butler and Ellison, 2007 ; Farnsworth and Ellison, 2008 ). Plants respond to resource imbalances by allocating new biomass (carbon) to acquisition of resources that most strongly limit plant growth ( Bloom et al., 1985 ). Economic models have been used successfully to examine resource allocation and performance of plants ( Givnish, 1986 ), and in these models carbon most often is the currency used because it is straightforward to measure the cost in grams or energy-equivalents of carbon needed to produce and maintain a structure. However, it is not just the total construction cost (CC, usually as g glucose/g dry mass [DM]), but the marginal costs and benefi ts of an investment in any particular structure that must be determined. The payback time (hours or days) to recover the carbon investment can be calculated easily as CC mass / A mass when both cost and net photosynthesis ( A mass ) are expressed as nmol C/g DM, and A mass is measured per unit time. Payback time can be thought of as the time span that a leaf must photosynthesize to recover (amortize) the carbon investment used in its construction ( Poorter et al., 2006 ). Carnivorous plants have elaborate traps (e.g., pitfall traps, sticky pads, snap-traps) that they use to catch prey; associated glandular hairs and secretory cells subsequently dissolve the prey and release nutrients that are absorbed by the plant ( Darwin, 1875 ; Lloyd, 1942 ). These traps are thought to be physiologically costly structures, and it has been hypothesized that constructing elaborate traps would be selected for only if they provide a net marginal benefi t to the plant by capturing prey that provides essential nutrients (e.g., N, P) required for photosynthesis ( Givnish et al., 1984 ; Benzing, 2000 ). An explicit test of this hypothesis requires simultaneous measurements of both marginal costs and marginal benefi ts (photosynthesis) of carnivory. Here, we provide for the fi rst time a simultaneous assessment of both the costs and benefi ts of botanical carnivory. 1 Manuscript received 19 February 2009; revision accepted 21 April 2009. The authors thank R. Sage for use of his microbomb calorimeter; K. Griffi n, L. Patrick, and N. M. Holbrook for assistance with the heat-ofcombustion method; P. Kuzeja, K. Savage, and J. Butler for technical assistance; and H. Poorter, W. Carlson, and S. Stark, two anonymous reviewers, and an Associate Editor for comments on an earlier draft of the manuscript. This research was supported by NSERC of Canada postdoctoral fellowship to J.D.K. and NSF grants 02-35128, 04-00759, 04-52254, and 05-46180 to A.M.E. 2 Author for correspondence (e-mail: aellison@fas.harvard.edu)

Understanding how the investment (or allocation) of carbon and mineral nutrients varies in different plant organs and among species, plant functional types, and the vegetation of different biomes is a major goal for plant ecology ( Wright et al., 2004 ).A synthesis of global data on plant traits -leaf nitrogen (N), phosphorus (P), and potassium (K) content (%), leaf mass area (LMA: g/cm 2 ), leaf lifespan, and maximal leaf photosynthetic rates ( A mass : nmol CO 2 ⋅ g −1 ⋅ s − 1 ) -of 2548 species (the Glopnet data set; Wright et al., 2004) illustrated and quantifi ed the bivariate scaling relationships between traits (e.g., A mass as a function of %N) and revealed a " universal spectrum of leaf economics " ( Wright et al., 2004( Wright et al., , 2005 ) ).This spectrum defi nes trade-offs in plant allocation when changes in one trait co-vary consistently with changes in a second trait.Additional research has identifi ed carnivorous plants ( Ellison and Farnsworth, 2005 ;Farnsworth and Ellison, 2008 ), mangroves ( Ellison, 2002 ), and lianas in tropical forests ( Santiago and Wright, 2007 ) as notable outliers relative to the universal spectrum of leaf traits.These exceptions to the rule have reinforced the importance of tradeoffs in our understanding of why leaf traits consistently scale with one another.Shipley et al. (2006) proposed that the trade-off between allocation of nutrients to structural tissues and long-term storage vs. their immediate use in boosting photosynthetic rates is a potential mechanism for the observed coordination among leaf traits.Such trade-offs can be observed easily with carnivorous plants because carnivorous plants inhabit open environments (e.g., bogs and other wetlands) where light and water are not limiting but nutrients are in extremely short supply, and therefore it is relatively easy to separate out experimentally the effects of nutrient limitation from effects of limitation of light or water ( Butler and Ellison, 2007 ;Farnsworth and Ellison, 2008 ).
Plants respond to resource imbalances by allocating new biomass (carbon) to acquisition of resources that most strongly limit plant growth ( Bloom et al., 1985 ). Economic models have been used successfully to examine resource allocation and performance of plants ( Givnish, 1986 ), and in these models carbon most often is the currency used because it is straightforward to measure the cost in grams or energy-equivalents of carbon needed to produce and maintain a structure.However, it is not just the total construction cost (CC, usually as g glucose/g dry mass [DM]), but the marginal costs and benefi ts of an investment in any particular structure that must be determined.The payback time (hours or days) to recover the carbon investment can be calculated easily as CC mass / A mass when both cost and net photosynthesis ( A mass ) are expressed as nmol C/g DM, and A mass is measured per unit time.Payback time can be thought of as the time span that a leaf must photosynthesize to recover (amortize) the carbon investment used in its construction ( Poorter et al., 2006 ).
Carnivorous plants have elaborate traps (e.g., pitfall traps, sticky pads, snap-traps) that they use to catch prey; associated glandular hairs and secretory cells subsequently dissolve the prey and release nutrients that are absorbed by the plant ( Darwin, 1875 ;Lloyd, 1942 ).These traps are thought to be physiologically costly structures, and it has been hypothesized that constructing elaborate traps would be selected for only if they provide a net marginal benefi t to the plant by capturing prey that provides essential nutrients (e.g., N, P) required for photosynthesis ( Givnish et al., 1984 ;Benzing, 2000 ).An explicit test of this hypothesis requires simultaneous measurements of both marginal costs and marginal benefi ts (photosynthesis) of carnivory.Here, we provide for the fi rst time a simultaneous assessment of both the costs and benefi ts of botanical carnivory.September 2009] Karagatzides and Ellison -Economics of botanical carnivory To our knowledge, this is the fi rst study to measure, on the same plant and at the same time, both the construction costs and photosynthetic rates for carnivorous plants.These measurements permit the calculation of marginal gain as payback time and integrate traits used to calculate leaf trait scaling relationships ( Wright et al., 2004 ) with construction costs.This integration allowed us to test a fourth hypothesis: that there should be clear and signifi cant trade-offs between A mass and CC mass , as proposed by Shipley et al. (2006) .× miranda , a modern hybrid of N. maxima × [ N. northiana × N. maxima ;] Nepenthaceae); the sundew Drosera fi liformis (Droseraceae), and the Venus fl y trap Dionaea muscipula (Droseraceae).Sarracenia , Darlingtonia , Drosera fi liformis , and Dionaea are native to North America, whereas the Nepenthes species are hybrids of species native to the Southeast Asian tropical lowlands.The modifi ed leaves of Dionaea , Drosera , and North American Sarracenia and Darlingtonia both photosynthesize and trap prey, although some Sarracenia spp.also produce photosynthetically more effi cient phyllodia (fl at, nontrapping leaves; Ellison and Gotelli, 2002).Nepenthes spp., by contrast, have a fl at lamina (a modifi ed petiole) and an attached cylindrical trap that is modifi ed from a leaf or leafl et ( Arber 1941 ).Previous work has shown that A mass of the laminae of Nepenthes is the primary source of photosynthate and that A mass of Nepenthes pitchers is near zero ( Pavlovi č et al., 2007 ).Thus, we measured construction costs of both laminae and pitchers of Nepenthes and phyllodia when present on Sarracenia .All plants used in our study had reached reproductive maturity; by excluding juvenile plants, we minimized the potential confounding effects of nonfunctional traps that are too small to capture prey (common in juvenile plants) and heterophylly relative to adult plants ( Franck, 1976 ).

MATERIALS AND METHODS
Photosynthesis -Among the pitcher plants, we measured plants with at least three mature, fully expanded pitchers.There were six replicate plants for each species with the exception of N .× coccinea ( N = 4) and Darlingtonia californica , N .× miranda , and S .rosea ( N = 2 each).The Venus fl y trap Dionaea muscipula and the sundew Drosera fi liformis were fl owering at the time of measurement, but the other species were not.Maximum photosynthetic rate ( A area , as µ moles CO 2 ⋅ m − 2 ⋅ s − 1 ) of one trap (and lamina for Nepenthes ) on each plant was determined using a LI-COR 6400 IR gas analysis system (LI-COR, Lincoln, Nebraska, USA) fi tted with a 3 cm × 2 cm cuvette that was clasped onto the central portion of a pitcher or leaf.We also measured A area of phyllodia that were produced by S .fl ava , S .leucophylla , and S .oreophila during our study.All measurements were taken between 0900 and 1400 hours at a photosynthetic photon fl ux density (PPFD) of 1200 µ mol ⋅ m − 2 ⋅ s − 1 during 23 -25 July 2006.Measurement of A area for Dionaea included both its snap-trap and the attached petiole.In those few instances when the sample did not cover the entire surface area of the cuvette (i.e., Dionaea , Drosera , N .× coccinea ), photosynthetic rates were adjusted for the proportion of the cuvette covered by leaf tissue.In those cases where only two plants were available, 2 -4 pitchers were sampled from a plant and averaged for that individual.
Harvest -Plants were harvested immediately after photosynthetic rates were measured.Pitchers were cut longitudinally with a stainless steel razor blade and washed with tap water to remove any prey, detritus, or extrafl oral nectar.Pitchers were subsequently rinsed with distilled-deionized water, patted dry with a paper towel, and spread on the conveyer belt of a Li-Cor 3000 to measure leaf area ( ± 1 mm 2 ).Leaf areas and associated masses were used to calculate leaf mass per unit area (LMA: g/m 2 ) and to re-express A area as A mass (nmoles CO 2 ⋅ g − 1 ⋅ s − 1 ).Roots and rhizomes were washed separately with tap water and rinsed with distilled-deionized water.Traps, phyllodia, roots, and rhizomes were dried separately at 70 ° C to constant mass, weighed ( ± 0.001 g dry These data allow us to further assess trade-offs between allocation of nutrients to structural tissues and their immediate use for enhancing photosynthesis; i.e., to test Shipley et al. ' s (2006) proposed mechanism underlying the universal spectrum of leaf economics.
The majority of previous studies of costs and benefi ts of botanical carnivory have used feeding experiments to examine the response of plant growth or photosynthesis to addition of prey or nutrients (e.g., N, P; reviewed by Ellison, 2006 ;Ellison and Gotelli, 2009 ).Our previous work has shown that carnivorous plants have unexpectedly low A mass for their leaf N and P content ( Ellison and Farnsworth, 2005 ;Ellison, 2006 ;Ellison and Gotelli, 2009 ), and these scaling relationships do not change signifi cantly when plants are given supplemental prey ( Wakefi eld et al., 2005 ;Farnsworth and Ellison, 2008 ).Low rates of A mass by carnivorous plants have been suggested to be one cost of carnivory ( Ellison and Farnsworth, 2005 ;Ellison, 2006 ;Pavlovi č et al., 2007 ).For pitcher plants ( Sarracenia spp.), this latter result is thought to result from allocation of excess nutrients to storage and subsequent growth, rather than to immediate syntheses of enzymes required for photosynthesis ( Butler and Ellison, 2007 ).Osunkoya et al. (2007) examined the cost of building carnivorous traps for eight species of Nepenthes pitcher plants in Borneo.These species have a fl at photosynthetic lamina (modifi ed from the leaf base: Lloyd, 1942 ) and an attached cylindrical trap (pitcher) that is a modifi ed, epiascidiate leaf ( Arber, 1941 ;Owen and Lennon, 1999 ).Pitchers should have lower construction costs than leaves or evolution would not have favored this extra pathway for nutrient uptake/assimilation ( Osunkoya et al., 2007 ).Pavlovi č et al. (2007) found similar rates of respiration for traps and conjoined laminae of Nepenthes alata and N .mirabilis but lower rates of photosynthesis for traps, and they concluded that reduced photosynthesis is a cost of multiple functions (digesting prey, absorbing nutrients and transferring nutrients to other plant parts).However, as Givnish et al. (1984) outlined, it is the marginal gain that needs to be measured because high costs can have high benefi ts or low costs can have low benefi ts.That is, the measurements of construction cost (e.g., Osunkoya et al., 2007 ) and photosynthesis (e.g., Pavlovi č et al., 2007 ) need to be undertaken on the same individuals.This analysis is currently lacking for carnivorous plants and is the focus of the work presented here.
We asked four questions in this study.First, what is the CC mass of traps of 15 carnivorous plant species?Second, how does CC mass of carnivorous plants compare with CC mass of noncarnivorous species?Third, how does CC mass of traps compare with CC mass of roots and rhizomes (underground stems) for carnivorous plants that produce rhizomes?Fourth, what is the payback time (i.e., CC mass / A mass in hours or days) for a carnivorous trap?The answers to these questions allowed us to address the following three hypotheses.
First, for carnivorous plants in which the trap is modifi ed from a leaf and when the trap simultaneously fi xes carbon (through photosynthesis) and acquires nutrients (through carnivory), we hypothesized that traps should be relatively expensive structures with high construction costs.Second, for carnivorous species, such as those in the genus Nepenthes , that have separate photosynthetic leaves and traps, we hypothesized that the CC mass of a trap should be lower than the CC mass of the associated leaf.Third, we hypothesized that roots would be less expensive than leaves because of lower concentrations in the roots of expensive compounds such as lipids and proteins ( Poorter and Villar, 1997 ).

RESULTS
Trap traits -With the exception of A mass , there were significant differences among species for all variables measured ( Table 1 ).Mean ash concentration of carnivorous traps ranged more than sixfold from 1.4 to 9.5%.Mean N concentration, by comparison, was more constrained (range 0.58 -1.31% N).On average, the greatest ash and N concentrations were measured in N .× coccinea pitchers.All the carnivorous plants examined in this study had low mean maximal net photosynthesis when expressed on a mass basis ( A mass range = 1.1 -64.0 nmol CO 2 ⋅ g − 1 ⋅ s − 1 ).Pitchers of both Nepenthes species had the lowest, and S. rubra had the highest A mass .Although N concentrations observed among species varied only about threefold, rates of photosynthetic nitrogen-use effi ciency (PNUE N : µ mol CO 2 ⋅ mol − 1 N ⋅ s − 1 ) varied over sixfold from 13.4 for N. × coccinea lamina to a maximum of 89.3 for S .fl ava pitchers ( Table 1 ).The mean energy content (kJ/g AFDM) of the carnivorous traps ranged from 12.8 in S .minor to 22.5 in S .purpurea .Traps of S .fl ava had energy content similar to S .purpurea , and traps of all other species had an energy content less than 17.4 kJ/g AFDM.The lowest LMA (40 g/m 2 ) was measured for traps of N. × coccinea compared to a maximum LMA of 118 g/m 2 for sticky pads of Drosera fi liformis .The ratio of dry mass to fresh mass was lowest in the species with the smallest pitchers and lowest LMA ( N. × coccinea ).Low dry mass to fresh mass ratios were also found for N .× miranda and for some of the smaller carnivorous traps including Drosera fi liformis and Dionaea muscipula (range dry : fresh mass = 0.10 -0.14).Sarracenia purpurea and Darlingtonia californica had intermediate ratios of dry to fresh mass, while the remaining species of Sarracenia had ratios ≥ 0.20 up to a maximum of 0.26.
Construction costs -Differences among species for energy, nitrogen, and ash content led to highly signifi cant differences among species for CC mass of traps, roots, and rhizomes.The overall nested analysis of variance model (whole-plant analysis pooling traps, roots, and rhizomes with structures nested within species) was highly signifi cant ( F 26, 149 = 9.385; P = 2.49 × 10 − 20 ).Pooled across all species, CC mass of traps, roots, and rhizomes were similar (1.15 ± 0.28, 1.15 ± 0.13, and 1.16 ± 0.16 g glucose/g DM, respectively; F 2,149 = 1.353;P = 0.262).There were, however, signifi cant differences in construction costs of particular structures among species ( Table 2 ).Similar to the trend for energy content, signifi cantly greater CC mass was measured for traps of S .purpurea ( P < 0.0003) and S .fl ava ( P < 0.006) than for the other species measured in this study.A group comprised of S .minor , S .alabamensis , S .oreophila , S. leucophylla , and S .jonesii had the lowest CC mass for traps.There were fewer signifi cant differences among species for the construction of belowground structures.For roots, CC mass of S .alabamensis was signifi cantly greater than S .minor ( P = 0.0001), S .oreophila ( P = 0.0001), and Dionea muscipula ( P = 0.04; Table 2 ).Construction costs of rhizomes for S .fl ava and S .minor were signifi cantly lower ( P < 0.0014) than for S .oreophila and S. purpurea ( Table 2 ).
Carnivorous traps had signifi cantly lower ( t = 3.35, df = 12, P = 0.006) CC mass (1.29 ± 0.20 g glucose/g DM) than did the mass [DM]) and ground to a fi ne powder with a stainless steel capsule and ball bearing in a Wig-L-Bug grinder (Bratt Technologies, LCC, East Orange, New Jersey, USA).
Estimation of construction cost -Tissue construction costs (CC mass ; g glucose/g DM) were estimated for roots, rhizomes, and the leaf tissue on which A area had been measured using the heat-of-combustion method ( Williams et al., 1987 ): where Δ H c is the heat of combustion (energy as kJ/g ash-free dry mass [AFDM]), Ash is the ash content (g ash/g DM), k is the oxidation state of the nitrogen substrate (nitrate = +5, ammonium = -3), N is the organic nitrogen content (g N/g DM), and E g is the growth effi ciency (the proportion of energy used to produce reductant that is consumed during the formation of tissue but not contained within the biomass).An E g = 0.87 incorporates cost of transport and gives a good fi t against the detailed biochemical analysis used as the standard ( Griffi n, 1994 ).
Heat of combustion was determined using a microbomb calorimeter (construction details available online at http://harvardforest.fas.harvard.edu/personnel/web/aellison/research/stoichiometry/calorimetry/Micro-bomb%20Home%20 Page.htm) calibrated with benzoic acid pellets of known calorifi c values.The calibration was verifi ed with a spinach reference standard (NIST 1570a; National Institute of Standards and Technology, Gaithersburg, Maryland, USA) with a noncertifi ed calorifi c value of 3500 cal/g DM.Analysis of N = 35 spinach pellets during our assay yielded an average calorifi c value of 3536 cal/g DM (i.e., +1% of the expected value).In almost all cases, we had suffi cient carnivorous plant tissue so that each sample could be analyzed in triplicate as 2 -12 mg pellets pressed from the ground sample.The H c values obtained for the triplicate pellets of each sample were then averaged.Because of the large number of analyses ( > 1000), we used Ni-Cr ignition wire (which contributes a small amount of heat during the reaction) rather than the more expensive Pt wire that does not give off heat from combustion.Therefore, fi ve samples of Ni-Cr wire and no sample pellet were combusted to obtain the heat given off by the Ni-Cr wire and to determine the intercept of the calibration line.
Total nitrogen was substituted for organic N ( Nagel et al., 2005 ) and measured on a Carlo -Erba Model 2500CN elemental analyzer.Nagel et al. (2005) found that the substitution of total N for organic N overestimated CC by only 0.03 -0.06%.Ash content was determined by combusting a 10 -100 mg subsample of the powdered plant tissue in a muffl e furnace at 550 ° C for 6 h.Construction costs were calculated using both k = +5 and -3, and the average value reported on a dry mass basis.
We estimated CC per gram DM rather than per plant biomass because of the differing sizes of plants used in the analysis.Dry mass per structure is provided in Appendix S1 (see Supplemental Data with online version of article) to allow for scaling up the results to the whole-plant level.Payback time was calculated as CC mass / A mass after conversion of CC mass from g glucose/g DM to nmol C/g DM and conversion of A mass from nmol CO 2 ⋅ g − 1 DM ⋅ s − 1 to nmol C ⋅ g − 1 DM ⋅ h − 1 .Calculations for payback times of pitcher construction for Nepenthes also were made using A mass of the attached lamina.We estimated payback time on an hourly rather than daily basis because of the differing light levels during a daily period and across the growing season.Thus, our estimate of payback time represents the minimum amortization.
Statistical analysis -We tested for differences in CC mass of traps, roots, and rhizomes among species using a nested analysis of variance (plant structure nested within species) using the program SPSS, release 14.0.0 for Windows ( SPSS, 2005 ).Signifi cant ( P < 0.05) differences for the ANOVA model were followed by Tukey ' s (HSD) post hoc test to compare CC mass of organs among species.A paired t -test was used to test for differences in CC mass for species with traps and phyllodia/laminae.We used an unpaired t -test to test for differences in CC mass of traps against leaves of 267 noncarnivorous species compiled from a search of the published literature ( Griffi n, 1994 ;Isagi, 1994 ;Mitchell et al., 1995 ;Baruch and Gomez, 1996 ;Dai and Wiegert, 1996 ;Isagi et al., 1997 ;Marquis et al., 1997 ;Niinemets, 1997 ;Spencer et al., 1997 ;Wullschleger et al., 1997 ;Baruch and Goldstein, 1999 ;Eamus et al., 1999 ;Baruch et al., 2000 ;Nagel and Griffi n, 2001 ;Villar and Merino, 2001 ;George et al., 2003 ;Suarez, 2003Suarez, , 2005 ; ;Nagel et al., 2004Nagel et al., , 2005 ; ;Oikawa et al., 2004Oikawa et al., , 2007 ; ;Osunkoya et al., 2004Osunkoya et al., , 2007 ; ;Barthod and Epron, 2005 ;Brunt et al., 2006 ;J. D. Karagatzides, unpublished data).We tested for a relationship between A mass and CC mass using reduced major axis regression on logarithmically transformed data using custom code written for the R statistical software package, version 2.6.1 (R Devel-Table 1. Mean ( ± 1 SD; unless pooled into one composite sample or not available [n/a]) for traits of carnivorous plants with (A) traps or (B) lamina/phyllodia.Different lowercase letters indicate signifi cant differences among species for each trait ( P < 0.05, Tukey ' s HSD post hoc test for multiple comparisons among means).No post hoc comparisons were done on the data for laminae/ phyllodia because the overall ANOVA was not signifi cant for any of the traits ( F 5,16 < 3.0; P > 0.1) Values presented are N : sample size; Ash: ash content following combustion (g ash/g dry mass); N: nitrogen content (g N/g dry mass), PNUE N : photosynthetic nitrogen use effi ciency ( ), Energy: energy content measured by bomb calorimetry (kJ/g ash-free dry mass); LMA: mass of leaves/m 2 leaf area); and the ratio of dry mass to fresh mass.2) when separate structures for photosynthesis and prey capture occur on the same plants, CC mass of traps should be lower than CC mass of laminae and phyllodia, (3) construction costs of roots and rhizomes should be less than leaves (traps, phyllodia, lamina), and ( 4) there should be a clear tradeoff between A mass and CC mass .To our knowledge, this is the fi rst time that both construction costs and photosynthetic rates for carnivorous traps have been measured simultaneously, and our results provide further insights into hypothesized mechanisms underlying the universal spectrum of leaf traits ( Wright et al., 2004 ;Shipley et al., 2006 ).Our data did not support our fi rst hypothesis.As a group, traps of carnivorous plants had signifi cantly lower average CC mass than did leaves of noncarnivorous plants ( Fig. 2 ).But our data associated lamina or phyllodia (1.41 ± 0.14 g glucose/g DM) for the fi ve carnivorous plants in this study and eight species of Nepenthes in Borneo ( Osunkoya et al., 2007 ) that have both traps and phyllodia or laminae ( Fig. 1 ).Construction costs of traps were also signifi cantly lower than construction costs of leaves of 267 noncarnivorous species compiled from a search of the published literature ( t = 6.32, df = 288, P = 9.87 × 10 − 10 ; Fig. 2 ).Six carnivorous species were at the lower extreme of the overall distribution, including plants with snap-traps ( Dionaea muscipula ), sticky pads ( Drosera fi liformis ), and pitfall traps ( Sarracenia alabamensis , S .jonesii , S .minor , N .× miranda ).The two species with the greatest CC mass for traps measured in this study ( S .purpurea, S .fl ava ) were ~20% below the maximum of 2.10 g glucose/g DM found in the literature for leaves.Construction costs of roots of carnivorous plants were also significantly lower than CC mass of roots of 20 noncarnivorous species ( t = 3.20, df = 34, P = 0.003; Fig. 2 ).Construction costs of rhizomes were also lower than CC mass of bamboo, the only noncarnivorous species for which rhizome CC mass has been reported ( Fig. 2 ).

Payback time -We found no signifi cant relationship between
A mass and CC mass ( r = 0.17, df = 54, P = 0.51), but differences in payback time to recover (amortize) the carbon cost of constructing carnivorous traps were signifi cantly different among species ( F 14,54 = 4.980, P = 8.02 × 10 − 6 ; Table 3 ).The longest times were for S .purpurea (1551 h) and Darlingtonia californica (1370 h).Sarracenia fl ava had a payback time (1262 h) similar to Darlingtonia californica , and four other species ( S .leucophylla , S. rosea , Drosera fi liformis , and Dionaea muscipula ) had payback times exceeding 900 h.The Nepenthes and six remaining Sarracenia species recovered pitcher CC mass in about one-third to one-half the time (range 495 -849 h) required by S .purpurea .

DISCUSSION
The goal of this study was to use carnivorous plants to examine mechanisms underlying the universal spectrum of leaf traits ( Wright et al., 2004 ;Shipley et al., 2006 ) by testing four specifi c hypotheses about construction costs, photosynthetic rates, and payback times: (1) carnivorous traps that both photosynthe-   2007) measured near-zero A mass for pitchers of Nepenthes alata and N .mirabilis and higher rates of A mass for laminae.Together, these results suggest long payback times based on pitcher CC mass measured by Osunkoya et al. (2007) , although Pavlovi č et al. (2007) did not measure A mass of the Nepenthes spp.studied by Osunkoya et al. (2007) .Overall, the low CC mass of carnivorous traps is associated with low A mass ; traps have small marginal gains and long payback times.
Our data neither clearly supported nor failed to support our third hypothesis.Contrary to the fi ndings of previous studies that measured costs for individual compounds in structures of herbaceous plants (e.g., Poorter and Villar, 1997 ), we found that traps (modifi ed from leaves) were not consistently more costly to build than roots or rhizomes ( Table 2 ).We note that Poorter and Villar (1997) found that construction costs of stems in herbaceous plants were similar to those of roots.It may be that traps of carnivorous plants have high concentrations of total structural carbohydrates that are relatively cheap compounds ( Poorter and Villar, 1997 ), but have low concentrations of the expensive compounds used for photosynthesis (hence the low A mass measured for carnivorous plants).Similar to traps, average CC mass for carnivorous plant roots ( Table 2 ) was lower in all but one case ( S. alabamensis ) than the average CC mass of roots of 20 noncarnivorous species ( Fig. 2 ).Rhizome CC mass for carnivorous plants in the current study is similar to tubers of Potamogeton pectinatus but 10 -37% lower than for bamboo ( Phyllostachys bambusoides and P .pubescens , both with a cost of 1.49 g glucose /g DM), the only other studies we found reporting CC mass for rhizomes ( Fig. 2 ).
Payback time of traps ranged threefold (495 -1551 h) and differed signifi cantly among the carnivorous plants we studied.Energy content was signifi cantly greater in the two Sarracenia species with the highest CC mass and payback times ( S .purpurea and S .fl ava ; Table 1 ), suggesting that these two species invest in expensive compounds (e.g., lipids, soluble phenolics, protein, lignin; Poorter et al., 2006 ).This observation lends credence to the hypothesized trade-offs between investments in liquid-phase processes such as photosynthesis and structural processes required to construct leaves, roots, and rhizomes ( Shipley et al., 2006 ).Several carnivorous plants in the current study had relatively similar (and low) CC mass but substantially different LMA (e.g., Drosera fi liformis , Dionaea muscipula , S .minor , S .alabamensis , S .jonesii ).Leaf mass area increases with added cell layers, but CC mass remains unchanged if these layers are of similar biochemical composition ( Griffi n, 1994 ).Leaf mass area also increases with an investment in less costly compounds (e.g., structural and nonstructural carbohydrates; Poorter et al., 2006 ).The generally observed pattern in our data of decreasing N concentration with increasing LMA ( Wright et al., 2004 ) further supports the notion of an investment in non-N-based compounds as leaf density increased.
All carnivorous plants examined in the current study had long payback times to recover carbon invested in traps.As the number of functions of an organism or organs increases, the effi ciency of performance of any particular function may decline ( Read and Stokes, 2006 ).Pavlovi č et al. (2007) suggested that reduced photosynthesis of Nepenthes traps was a cost of multiple functions -digesting prey, absorbing nutrients, and transferring nutrients to other plant parts.The long payback times we measured for carnivorous plants ( Table 3 ) also may refl ect the ineffi ciency of a modifi ed leaf in fulfi lling these multiple roles in addition to photosynthesizing.Kikuzawa and did support our second hypothesis.Traps were less costly to build than phyllodia in Sarracenia and laminae in Nepenthes (our data and data of Osunkoya et al. [2007] ; Fig. 1 ).Similar to Fig. 2. Tissue construction cost for 15 carnivorous plants (this study), eight species of Nepenthes in Borneo ( Osunkoya et al., 2007 ) and 267 noncarnivorous plants (data from Griffi n, 1994 ;Isagi, 1994 ;Mitchell et al., 1995 ;Baruch and Gomez, 1996 ;Dai and Wiegert, 1996 ;Isagi et al., 1997 ;Marquis et al., 1997 ;Niinemets, 1997 ;Spencer et al., 1997 ;Wullschleger et al., 1997 ;Baruch and Goldstein, 1999 ;Eamus et al., 1999 ;Baruch et al., 2000 ;Nagel and Griffi n, 2001 ;Villar and Merino, 2001 ;George et al., 2003 ;Su á rez, 2003Su á rez, , 2005 ; ;Nagel et al., 2004Nagel et al., , 2005 ; ;Oikawa et al., 2004Oikawa et al., , 2006 ; ;Osunkoya et al., 2004Osunkoya et al., , 2007 ; ;Barthod and Epron, 2005 ;Brunt et al., 2006 ; J. D. Karagatzides, unpublished data).  1 and Ellison , 2006 ), then a longer lifespan could offset differences in carbon cost and gain ( A mass ) in the current study.This trade-off between lifespan, carbon cost, and carbon gain is an example of the " many-to-one mapping relationship in functional design " that refl ects different ways to achieve a constant lifetime carbon gain for individual leaves ( Kikuzawa and Lechowicz, 2006, p. 381 ).However, for 20 of the 25 species in the study by Kikuzawa and Lechowicz (2006), they used only the mean value of CC mass (1.5 g glucose /g DM) based on 79 species compiled in Griffi n (1994); therefore the constant lifetime carbon gain hypothesis needs further testing with species-specifi c values of CC mass .Finally, our data did not support our fourth hypothesis, Shipley et al. ' s (2006) prediction concerning trade-offs between A mass and CC mass .Our data do show that carnivorous plants overall have low construction costs, but their very low photosynthetic rates led to long payback times.However, it is difficult to compare our estimates of payback time (in hours) with those for noncarnivorous plants reported in the literature (in days); accurate estimates of payback time require information on the daily hours of maximal A mass for these species (e.g., mean labor time; Kikuzawa and Lechowicz, 2006 ) because light is not available 24 h per day.For example, a generous assumption of 10 h/d of maximal A mass for the carnivorous plants in our study that are found in open environments would require ~50 -150 d to recover the carbon used to construct a trap.This range exceeds estimates for noncarnivorous plants growing in open, well lit areas (e.g., 4 -30 d for Piper spp.( Williams et al., 1989 ); 15 -20 d for sun leaves of six adult tree species ( Poorter et al., 2006 )).Actual payback times could be lower, however, because early returns on foliar investment can offset later losses ( Westoby et al., 2000 ), and carnivorous plants translocate nutrients from trap to trap ( Butler and Ellison, 2007 ).Thus, traps may have higher rates of A mass early in the growing season, particularly before they open and actively trap and digest prey, which may reduce the payback time estimated from our calculations.Pavlovi č et al. (2007) , for example, found that phyllodia (emerging in spring) of Sarracenia psittacina had signifi cantly greater A mass than pitchers (which formed later in the growing season).Ellison and Gotelli (2002) also found that phyllodia of S. purpurea had photosynthetic rates 25% higher than those of pitchers.Additionally, prolonged tissue life is a major mechanism by which the effi ciency of resource use is maximized in resource-poor environments ( Bloom et al., 1985 ), and measurements of leaf lifespan of carnivorous plants would provide the additional data needed to complete their carbon budgets.
Our results of low CC mass for carnivorous traps are contrary to the common expectation that the construction of elaborate carnivorous traps should be costly ( Givnish et al., 1984 ).Furthermore, carnivorous plants are poorly represented in the universal spectrum of leaf economics data set (four of 2548 observations; Wright et al., 2004 ) and are outliers because they have very low photosynthetic rates ( Ellison and Farnsworth, 2005 ;Farnsworth and Ellison, 2008 ).Our use of payback time, integrating traits used to assess leaf-scaling relationships with construction costs, yields better estimates for total costs and benefi ts of carnivorous structures to place carnivorous plants at the " slow and tough " end of the universal spectrum of leaf economics.

Table 2 .
Mean ( ± 1 SD; unless pooled into one composite sample or not available [n/a]) construction costs (g glucose/g dry mass) for phyllodia/laminae (noncarnivorous leaves), traps, roots, and rhizomes of 15 carnivorous plants.The overall nested analysis of variance model (structures nested within species) was highly signifi cant ( F 18,133 = 14.97,P = 2.49 × 10 − 20 ), but there were no signifi cant differences among traps, roots, and rhizomes ( F 2,149 = 1.353;P = 0.262).Different lowercase letters indicate differences among species in the construction cost of a given plant part ( P < 0.05, Tukey ' s HSD post hoc test for multiple comparisons among means).

Table 3 .
Mean ( ± 1 SD) payback time (h/g DM) for traps.Different lowercase letters indicate signifi cant differences among species in payback time ( P < 0.05, Tukey ' s HSD post hoc test for multiple comparisons among means).Note that payback for Nepenthes is for the mass-weighted cost of the lamina plus the trap, but is based on A mass by the lamina alone ( A mass of Nepenthes pitchers ≈ 0).Lechowicz (2006)found a near constant lifetime carbon gain for leaves of 25 species.If carbon gain similarly is equal across all carnivorous plants (and A mass is low across all carnivorous plants studied to date: Table