The experiments were carried out with juvenile plants of five herbaceous eudicotyledonous species, two shade-tolerant (Geum urbanum L. and Impatiens parviflora DC.), and three shade-intolerant species (Chenopodium album L., Helianthus annuus L. and Rumex palustris Sm.) Seeds were collected in their natural habitat in the vicinity of Utrecht, except for Rumex, which was collected in a floodplain of the river Waal near Nijmegen and Helianthus, which was obtained commercially. Impatiens seeds were stratified at 4°C for 2 months. Seeds were germinated on sand in the growth room at an irradiance of 200 μmol m⁻² s⁻¹. When the first true leaves were formed, the plants were transferred to 33 L containers with an aerated nutrient solution, having a concentration of 2 mM NO⁻₃ and other nutrients in proportion as in Hoagland and Snijder (1933). The pH was adjusted regularly at 5.6 and the solution was changed weekly. Conditions in the growth room were a constant air temperature of 20°C, a relative air humidity 70% and a photoperiod of 16 h. Five levels of irradiance (provided by Philips HPI 400 W lamps) were achieved by creating compartments with reflective walls and neutral shade screen on top. The lower part of each compartment was largely open for ventilation. Irradiance levels were ~20, 50, 100, 220, and 500 μmol m⁻² s⁻¹ photosynthetically active radiation, equivalent to daily irradiances of 1.2, 2.9, 5.8, 12.7, and 28.8 mol m⁻² day⁻¹, respectively. Irradiance was checked regularly during growth of the plants.

To minimize the effect of plant-to-plant variation on the growth parameter estimates, plants of each treatment and species were divided by eye in two size groups before each harvest (Poorter, 1989). Two plants from each size class were then randomly sampled at each of three harvest occasions. The experiment was repeated once and the data were combined, thus resulting in eight plants per harvest, light condition and species, 600 plants in total. Each of the sampled plants was divided into leaf blades, stems plus petioles and roots, after which leaf area and organ fresh mass were determined. Dry mass was measured after drying at 70°C for 48 h. The first harvest was at a whole-plant fresh mass of about 1.5 g, but tended to be somewhat lower at low irradiance due to low initial growth rates and somewhat higher for the large-seeded Helianthus. In order to further minimize size differences over the harvest intervals, time until the third harvest varied between 7 and 20 days, depending on growth rate.

Right before the second harvest, plants were measured for their rate of net photosynthesis and the respiration of shoots and roots. The plants were taken from the growth room, their above-ground parts were enclosed in gas exchange cuvettes and measured for CO₂ uptake at the growth irradiance after reaching steady state (A_growth). After 2 h in the dark, the CO₂ release of the shoots was measured (shoot R). Thereafter, the roots were detached, enclosed in light-tight cuvettes with an oxygen electrode for the measurement of O₂ consumption in water (root R). Details of the technique are given by Poorter et al. (1990). A short description is given below.

In the CO₂ exchange setup where A_growth and shoot R were measured, light was provided by similar lamps as in the growth chamber. Measurement irradiance and temperature were made identical to the growth conditions. CO₂ concentration in the incoming air was maintained at 400 μmol mol⁻¹, and water vapor partial pressure was set at 1200 Pa. CO₂ uptake and transpiration modified plant cuvette values to a CO₂ concentration of ~380 μmol mol⁻¹ and a relative humidity of circa 65%. The difference in CO₂ and H₂O concentration between inlet and outlet air was measured with an IRGA (Licor—6262, Lincoln, NE, USA).

For the measurement of root respiration, fresh nutrient solution was equilibrated with air at the measurement temperature of 20°C. The full root system was detached from the shoots and enclosed in a custom-made air-tight cuvette completely filled with nutrient solution. The rate of decrease of the O₂ concentration was measured over a period of 15 min with a Clark electrode (Yellow Group Instruments, OH, USA). Due to the small root systems of the plants from the lowest irradiances, the resolution was insufficient for precise measurements. In those cases roots of two individuals were combined for better sensitivity. Unforeseen technical problems with the system caused that only for two species reliable root respiration measurements are available for the full range of growth irradiances.

For separate plants from each species and irradiance treatment, leaf-level gas exchange was measured on recently matured leaves (n = 3). Leaves were enclosed in leaf chambers and measurements were done at growth irradiance and at light-saturation. The setup for these leaf-level measurements has been described by Pons and Welschen (2002).

Chemical composition of dry matter was measured on two independent bulk samples per light level and species, for leaf blade material and the rest of the plants separately. C and N concentrations were measured with an elemental analyzer (Carlo Erba, Milan, Italy). Nitrate was determined colorimetrically after extraction with boiling water, and mineral content determined in ash in combination with ash alkalinity. A full description of procedures and calculations is given at Prometheus wiki (http://prometheuswiki.publish.csiro.au).

Growth parameters were calculated as follows. For each of the plants harvested we kept track of whether they were a-priori classified into the “small or large” group. From each category within each experiment we randomly linked one plant of a given harvest to a randomly chosen plant of the other two harvests, giving eight time-series represented by triplets of plants. For each of these time series a linear regression of log_e-transformed dry masses over time was calculated, with the slope being the RGR. Morphological (SLA, LAR), allocation (LMF, SMF, RMF) and tissue density variables were calculated as average values per triplet. Finally, for each triplet ULR was calculated from RGR/LAR. Each of the eight growth parameter values was thus based on information from an independent set of plants.

The growth response coefficients (GRCs) of components of the RGR summarize the contribution of the variation in the respective growth parameters of equation 1 to the variation in RGR. To this end, a scaling slope analysis was carried out by calculating the regression coefficients of log-transformed growth parameter data versus log-transformed RGR (Poorter and van der Werf, 1998; Renton and Poorter, 2011). The same analysis was done for the ULR and its components, and the contribution of leaf density and leaf thickness to the increase of LMA with irradiance.

A_growth was calculated from the CO₂ uptake in the light, and shoot R from the CO₂ release in the dark. Values were expressed per unit leaf area and dry mass as appropriate. Root R was calculated from root O₂ consumption. For the calculation of whole-plant R in CO₂ units, we assumed a respiratory quotient of 1.2, as is often the case in nitrate-fed plants (Poorter et al., 1990). Whole-plant R/A, integrated over 24 h, was calculated on the basis of gross photosynthesis by adding shoot R to measured shoot net A, assuming identical R in light and dark. Hence, the R in R/A refers to 24 h of whole-plant R and the A to 16 h shoot gross photosynthesis, both calculated on the same expression basis.

A Two-Way ANOVA was performed on log_e-transformed data using the aov procedure in R (R Core Team, 2013), with Species and Irradiance as main factors. Specific a-priori contrasts between the shade-tolerant and shade-intolerant species were made for the main effect Species, as well as for the Species × Irradiance interaction.

Light had a strong and statistically highly significant effect on most of the growth variables (Figure 1, Table 2). All species had a positive RGR even at the lowest irradiance (Figure 1A), and therefore a positive C-balance. The increase in RGR with irradiance was strong in the lower light range and much less in the higher range. Interestingly, as far as there were differences in RGR between species at the low-irradiance range, there was no clear association with shade-tolerance. However, above an irradiance of ~100 μmol m⁻² s⁻¹ the two shade-tolerant species increased RGR substantially less compared to the intolerant ones (Figure 1A, Table 2). This resulted at the highest irradiance in RGR's of on average 244 and 354 mg g⁻¹ day⁻¹ for the shade-tolerant and intolerant species, respectively. The ULR across the low irradiance range (20–100 μmol m⁻² s⁻¹) was very similar for all species (Figure 1B). For the shade-intolerant species, ULR increased in an almost linear fashion with irradiance. For the shade-tolerant ones, the ULR increase was less strong, explaining to a large extent their lower RGR under these conditions.

All plants showed a decrease of SLA with irradiance, strongly in the lower and less so in the higher range (Figure 1C). Two of the shade-intolerant species, Helianthus and Chenopodium, deviated in the lower light range from the general trend because they did not develop a high SLA there (Figure 1C). The decrease of SLA with irradiance counteracted the almost linear increase of the ULR (Figure 1B), resulting in a curvilinearly increasing RGR. In contrast, the LMF, which is the other morphological component that determines RGR (equation 1, Appendix 1), remained relatively constant (Figure 1D). The exception was Rumex, which showed a decreasing trend of LMF with irradiance. All species showed a moderate increase in RMF (Figure 1D) and a decrease in SMF (Supplement Table S1) with irradiance.

A further analysis of traits underlying SLA can best be done using its inverse, leaf dry mass per unit area (LMA = 1/SLA) (Supplement Table S1). LMA can be factorized as the product of leaf thickness and leaf tissue density, for which we used as proxies fresh mass per area and dry mass per fresh mass, respectively. The importance of light-driven variation in leaf thickness and density for variation in LMA is summarized by the numbers shown in Figure 2. Given that the vales of these scaling slope analyses all centered around 0.5, we concluded that both components were on average equally responsible for the increase of LMA with irradiance. At a given LMA, the leaves of the two shade-tolerant species Geum and Impatiens were relatively thin and had a higher density compared to the three shade-intolerant species. For these two species, the contribution of tissue density to the irradiance effect on LMA was also somewhat larger (Figure 2). The density of stem and/or petiole tissue also increased with irradiance. However, this was not clearly the case for the density of root tissue (Supplement Table S1).

The concentration of organic N in leaf dry matter (N_m) was on average 48 mg g⁻¹ and generally not much different between growth irradiances (Supplement Table S2). The N content per leaf area (N_a), however, increased almost linearly as a result of the similar increase in LMA (Supplement Tables S1, S3). Remarkable is the low N_a of Geum, particularly at high irradiance, which was caused by its low N_m. The C-concentration in plant dry matter ([C]) increased with irradiance and was consistently higher for the shade-tolerant species (Figure 3A). The increase in tissue density with irradiance and shade-tolerance was thus associated with increases in [C] (Figures 2, 3A).

Whole-plant photosynthesis measured at growth irradiance and expressed per unit leaf area (A_growth) increased almost linearly with increasing irradiance from an average across all species of 1 μmol m⁻² s⁻¹ at the lowest irradiance to 16 μmol m⁻² s⁻¹ at the highest (Figure 4A). At the highest irradiance, there was substantial variation between the species, with Geum showing the lowest value (12 μmol m⁻² s⁻¹) and Helianthus the highest (23 μmol m⁻² s⁻¹). Photosynthetic capacity, measured at the leaf level as the light-saturated rate of photosynthesis per unit leaf area (A_sat) also increased with growth irradiance (Figure 4B). A_sat was higher than A_growth at all light levels, increasing from on average 5 to 25 μmol m⁻² s⁻¹. Only Geum grown at the highest irradiance had similar values for A_sat and A_growth (12 μmol m⁻² s⁻¹; Figure 4).

A_growth expressed per unit shoot dry mass (A_m) showed a qualitatively similar curvilinear response to irradiance as RGR, with a strong increase in the lower range but less so in the higher range (Figure 5A). Shoot R_m also increased with irradiance and was on average 17% of shoot A_m (Figures 5A,B). However, the relative increase in shoot R_m between the lowest and the highest irradiance was only about 3-fold, whereas shoot A_m increased about 6-fold (Figures 5A,B). Consequently, the instantaneous shoot R/A ratio decreased with increasing irradiance (Figure 5D).

Root R_m was successfully measured across the full range of growth irradiances for Helianthus and Geum only. For the other species data are available for just part of the light range. Averaged over species and conditions, root R_m was on average 2.9 times higher than shoot R_m. Helianthus showed a gradual increase of root R_m with irradiance, but Geum, and also Impatiens for which data are available for most of the range, were rather constant across most of the light range, with lower values at the lowest irradiance only (Figure 5C).

Daily whole-plant R/A of Helianthus and Geum, for which the R_m data are available for the full range, showed a tendency to increase with decreasing irradiance. However, there was a significant difference in irradiance dependence between the two species (Table 2). In Helianthus R/A increased relatively strongly below an irradiance of 100 μmol m⁻² s⁻¹, to 0.51 at 20 μmol m⁻² s⁻¹, whereas Geum showed a more stable daily R/A, increasing to only 0.33 at 20 μmol m⁻² s⁻¹ (Figure 5F). Taking also the available data for the other species into account, the evidence points to a rather stable R/A of about 0.3 on average across a large part of the irradiance gradient with a tendency to increase only at the lowest irradiance (Figure 5E). In relative terms, the C-balance was therefore essentially constant across a large part of the growth irradiance gradient (Figure 5F). In absolute terms, however, the C-balance showed an increasing positive difference between A and R with irradiance.

For the five species investigated in this experiment, the general form of the irradiance response for RGR and its components was similar to what has been reviewed in for example Poorter and van der Werf (1998), with the exception that RGR did not fully saturate at 20 mol m⁻² day⁻¹. The latter phenomenon is often reported for shade-tolerant species, such as Impatiens (Evans and Hughes, 1961; Corré, 1983a) and Geum (Blackman and Wilson, 1951; Corré, 1983a). In the high-light range shade-tolerant species may even show decreases in RGR (Huxley, 1967; Veenendaal et al., 1996; Poorter, 1999). Growth experiments in an irradiance gradient generate large differences in demand for water and nutrients (Poorter et al., 2012a). A lack of increase or even a decrease in RGR at the high irradiance end in pot-grown plants could therefore be a secondary effect of limitations in the root environment, rather than a result of an intrinsic character of the species. Indeed, a saturated RGR across the high irradiance range was also observed for Geum in an outdoor experiment where watering of the plants was not optimal, whereas replication of the experiment with ample water supply resulted in an increasing RGR up to full daylight (Pons, 1977a), similar to what was found here (Figure 1A).

In their meta-analysis, Poorter and van der Werf (1998) showed a predominance of SLA in explaining species specific variation in RGR for herbaceous plants, and this was also the case among the three shade-intolerant species, with Rumex showing the fastest growth and highest SLA at each irradiance level (Figures 1A,C). However, at higher irradiances the result is strongly different when we include the shade-tolerant species: in that case, variation in ULR is the dominant variable that scales with interspecific variation in RGR. The evidence is summarized in Figure 6A, where we calculated the so-called GRC. These GRC values indicate how much the variation in RGR scales with variation in the multiplicative components of equation 2 in Appendix 1 (Renton and Poorter, 2011). They can be used to summarize in a highly efficient way what the relative importance is for each of the components in causing the variation in RGR. A GRC value for ULR of 0, for example, would indicate RGR increases without a concomitant increase in ULR, whereas a value of 1 would imply that a 10% increase in RGR goes with a 10% increase in ULR. In our experiment, the GRC values for SLA and LMF decrease with increasing irradiance, whereas the GRC for the ULR rises from 0 to 0.8, indicating that the higher the irradiance, the stronger interspecific variation in RGR is determined by ULR differences. A similar response was found in the meta-analysis of Shipley (2006), who included both herbaceous and woody species, of which several were shade-tolerant. In other studies with herbaceous species—using partly the same shade-tolerant species as the present experiment—a lower ULR in conjunction with a lower RGR compared to their shade-intolerant counterparts was also found (Pons, 1977a; Corré, 1983a). Similar results were reported for tropical trees, where the differences in RGR at higher irradiances were due to the low ULR of shade-tolerant species as well (Veneklaas and Poorter, 1998; Poorter, 1999). However, when the GRC-analysis is restricted to the three shade-intolerant species, the SLA is much more important for explaining differences in RGR at high irradiance (Figure 6B; Poorter and van der Werf, 1998). Clearly, the ecological background of species included in the comparison is important for the conclusion whether variation in assimilation rate or allocation and morphology are important for explaining interspecific variation in RGR. This is particularly an issue when shade-tolerant species grown at high irradiance are included in the comparison.

The low ULR of Geum at the highest irradiance is associated with a low A_growth (Figures 1B, 4A). Its A_sat was only slightly higher, thus A_growth was limited by its low photosynthetic capacity (Figure 4B, Supplement Table S3). Why do shade species often have a low photosynthetic capacity when grown at high irradiance? Geum is known to form only a single layer of palisade parenchyma (Pons, 1977a), which limits the development of a high A_sat (Terashima et al., 2001). A lack of the capability to develop a multilayered palisade parenchyma was also reported for Impatiens together with a relatively low photosynthetic capacity (Groen, 1973). Although Impatiens had not a particularly low A_sat in our experiment it did not increase in the highest irradiance interval (Figure 4B). A single layer of palisade parenchyma and/or a low photosynthetic capacity in high-irradiance grown shade-tolerant plants is often reported, such as in temperate herbaceous species (Osborne et al., 1994; Murchie and Horton, 1997), tropical herbaceous plants (Chow et al., 1988), tropical shrubs (Valladares et al., 2000), temperate deciduous trees (Jackson, 1967; Hanba et al., 2002) and tropical trees (Houter and Pons, 2012). Shade-intolerant species, on the other hand, generally develop multilayered palisade parenchyma at high irradiance, which is associated with their high photosynthetic capacity (Jackson, 1967; Groen, 1973; Pons, 1977a; Hanba et al., 2002). We therefore conclude that one of the likely reasons for the low photosynthetic capacity at high growth irradiance in shade-tolerant species is their incompetence to develop a multilayered palisade parenchyma.

A very high A_sat of 34 μmol m⁻² s⁻¹ at the leaf level was indeed found for Chenopodium grown at the highest irradiance (Figure 4B), as was reported elsewhere (Sage and Pearcy, 1987). Such a high A_sat was also reported for Helianthus (e.g., Fredeen et al., 1991). We found a lower value at the highest irradiance than expected, which is likely due to the rather low N_m of the leaves used for the A_sat measurements. This was not representative for the plants used for the whole-plant gas exchange (Figure 4, Supplement Tables S1–S3). The high A_sat of Chenopodium and supposedly also Helianthus facilitated the high A_growth (Figure 4A) and as A_growth is strongly related to ULR (Figure 7), also the latter. However, it should be noted that the high A_sat of the shade-intolerant species is not fully utilized in the growth conditions with a constant relatively low irradiance during daytime. Chenopodium and Helianthus had a low investment of leaf N per unit A_sat compared to Geum, and thus a high photosynthetic nitrogen use efficiency (PNUE_sat; Supplement Table S3). However, as their A_sat is not fully utilized, the PNUE of the species was similar at the growth irradiance (Supplement Tables S1, S3). In field conditions, high-light exposed plants experience widely fluctuating irradiance often exceeding saturation. A high A_sat is then utilized to a much larger extent and correlates better with daily assimilation (Zotz and Winter, 1993). Larger differences in daily assimilation and consequently ULR between species with different A_sat may thus be expected at variable irradiance as in field conditions compared to the constant light regime often used in growth rooms.

At the lower irradiances, the differences in RGR between species are more determined by differences in SLA—and to a lesser extent LMF—than by ULR (Figure 6A). Such a predominance of variation SLA for explaining species specific differences in RGR at low irradiance was more often reported (Veneklaas and Poorter, 1998; Shipley, 2006). This can be explained by the strong light limitation of photosynthesis at low irradiance resulting in a more similar A_growth, which is also evident in our data (Figure 4A). A higher SLA is then of crucial importance for increased growth.

Contrary to the situation at high irradiance, we did not observe a systematic difference of RGR between shade-tolerant and shade-intolerant species at low irradiance (Figure 1A, Supplement Table S2). Chenopodium and Helianthus showed little further increase of SLA with decreasing irradiance (Figure 1C). This lack of competence to develop a high SLA at low irradiance was earlier reported for Helianthus (Hiroi and Monsi, 1963) and also for the shade-intolerant Cirsium palustre (Pons, 1977a). Helianthus and Chenopodium may thus show a stronger decrease in net C-gain when the trend continues with a further decrease in irradiance compared to species that are able to increase their SLA further.

There was no evidence for important differences between the shade-tolerant and intolerant species in acclimation to irradiance at the chloroplast level, as derived from the chlorophyll a/b ratio and photosynthesis per unit chlorophyll (Supplement Table S3). However, at the leaf level, anatomical leaf traits are likely to make acclimation different between species, as discussed above.

Although a low LMA (and thus a high SLA) maximizes growth potential at low irradiance (Evans and Poorter, 2001), it does not necessarily increase fitness. A low LMA can weaken the leaves (Onoda et al., 2008), which may reduce leaf longevity and therefore diminish return on carbon investment (Lusk et al., 2008). A higher LMA would increase longevity when based on investment in defense components such as lignin and tannins (Lusk and Warton, 2007; Kitajima et al., 2013). However, the relatively high LMA of Helianthus and Chenopodium at low irradiance (Figure 1C) is associated with a relatively high A_sat (Figure 4B, Supplement Tables S1, S3) and thus a relatively large investment in protein-rich chloroplasts. Similar to the situation at high irradiance, the high A_sat of these species grown at low irradiance was also not utilized in the growth conditions, resulting in a low PNUE_growth (Supplement Tables S1, S3). Such shade leaves with a relatively high N_a (Supplement Tables S1, S3) can furthermore be attractive for herbivores. It is thus not likely that the high LMA of these species would add to their leaf longevity in shade, but rather makes them inefficient and vulnerable under these conditions.

The construction costs of plant tissue increased with irradiance (Figure 3B). This was mainly the result of lower concentrations of minerals in the dry matter, including nitrate (Supplement Table S2). The [C] increased for the same reason (Figure 3A). The clearly higher construction costs and [C] of the shade-tolerant species compared to the shade-intolerant ones (Figure 3) was not only due to lower mineral and nitrate concentrations, but also to a lower organic acid concentration in their dry matter (Supplement Table S2). These traits of the shade-tolerant species were in turn associated with high density of leaf tissue (Figure 2) and—in the case of Geum—also of petioles and roots (Supplement Table S1). They could, in the case of Geum including a low N_m, be associated with better defense resulting in increased leaf longevity in shade. This is important at low irradiance, as photosynthetic rates are inherently low. The time that it takes to generate the construction costs (pay-back time) is therefore unavoidably longer (Poorter et al., 2006). The high construction costs and [C] of the shade-intolerant species are at the expense of RGR (equation 1), but when these traits do indeed increase longevity, they are essential for survival in shade.

The growth parameter ULR can be factorized into A_growth, CUE (= 1 − R/A) and the [C] (equation 3 in Appendix 1). Using the same scaling slope analysis as explained above for the GRC, we can quantitatively asses the relative importance of variation in each of these components. The analysis shows that A_growth scaled linearly and strongly with ULR when irradiance increases, with an average slope across species close to 1.0 (Figure 7). Variation in R/A and [C] contributed little to variation in ULR (Figures 3A, 5E,F, 7). The growth calculated on the basis of gas exchange (NAR_ge) was on average 11% higher than from harvest data (ULR). A small difference is not surprising as A_growth and R measured at one moment in time are not necessarily fully representative for the rates over the whole experimental period. Nevertheless, it shows that for juvenile plants grown at various light levels, ULR can be an effective estimator of photosynthesis under growth conditions.

Photosynthesis also increased proportionally with growth rate when both variables are expressed per unit dry mass (resp. A_m and RGR) (Figure 8A). Although R_m of both shoots and roots increased linearly with RGR as well, it did not scale fully proportionally with RGR as a result of the positive y-intercept (Figures 8B,C). This intercept, R_m at zero growth, is considered to be an estimate of the maintenance respiration. The part that is proportional to RGR represents the growth-related respiration (Lambers et al., 2002). From the relationships of A_m and R_m with RGR (Figure 8) it would follow that the R/A ratio should decrease with increasing RGR and thus with irradiance. That was indeed found for the instantaneous values as measured on shoots (Figure 5D). However, as root R_m was found to be 2.9 times higher than shoot R_m (Figures 5B,C), the decrease in RMF with decreasing irradiance (Figure 1D) has a diminishing effect on daily whole-plant R, which explains the relatively constant daily whole-plant R/A ratio across a broad range of irradiances (Figures 5E,F). Adjustments of whole-plant R can quickly occur when A changes after transfer to another irradiance (McCree and Troughton, 1966; Pons, 1977b), probably as a result of altered demand for ATP (Noguchi et al., 2001). The adjustments are apparently such that whole-plant R/A remained more or less constant at around 0.3 across a wide range of irradiances.

The situation is somewhat different at light levels close to the light compensation point. At 20 μmol m⁻² s⁻¹, R/A showed a tendency to be higher than at the other irradiances (Figure 5F). As the data set for root R_m is incomplete, we estimated R/A in two other ways. In the first, we assumed root R_m to be 2.9 times higher than the shoot R_m, which is the average of the measurements. Shoot R_m is available for all species and irradiances and R/A calculated in this way thus also (Figure 5E). In the second we calculated daily whole-plant R/A from the linear relationship of A_m, and shoot and root R_m with RGR (Figure 8), and the linear relationship of RGR and RMF with log-transformed irradiance for all species together (Supplement Figure S1). In both cases the outcome was an almost constant R/A with indeed increasing values below about 50 μmol m⁻² s⁻¹ (black continuous line in Figure 5F). Extrapolation of the relationships yielded an R/A of unity (i.e., the whole-plant light compensation point sensu Givnish (1988) at an irradiance of 6 μmol m⁻² s⁻¹(0.35 mol m⁻² day⁻¹). This value is higher than the measured light compensation points for growth reported by Mahmoud and Grime (1974) and Pons (1983), which were 3.2 and 1.7 μmol m⁻² s⁻¹, respectively. Our calculation assumes a constant maintenance respiration at all irradiances. The fact that lower compensation points were measured than calculated suggests that the assumption that maintenance respiration remains constant is not correct. Alternatively, maintenance respiration may be down-regulated at very low irradiances, as is also found after longer periods in darkness (Gifford, 2003).

A constant R/A was found for Pisum sativum between 100 and 400 μmol m⁻² s⁻¹ (McCree and Troughton, 1966), and Holcus lanatus and Plantago major at 150 and 300 μmol m⁻² s⁻¹ (Poorter, 2002). When calculated from shoot A and R data from Pons (1977a), assuming the same 2.9-fold higher root R_m than shoot R_m, whole-plant R/A was also largely constant between 5 and 100% full daylight in Geum urbanum. In conjunction with previous observations, we therefore conclude that—with the exemption of very low light levels—the growth dependence of R_m and the decreasing RMF can establish a stable whole-plant R/A across a broad range of growth irradiances.

A conservative R/A ratio was also found when plants were grown at different temperatures in the range that plant normally encounter in their natural habitat (Gifford, 1995; Loveys et al., 2002). Whole-plant R and A acclimated to the growth temperature within this range, but higher temperatures caused an increase in R/A as a result of increasing R and decreasing A (Atkin et al., 2007). Elevated CO₂ also has only marginal effects on the R/A ratio (Poorter, 2002; Gifford, 2003). However, a decrease in nutrient availability caused an upward shift in R/A (van der Werf et al., 1992; Poorter et al., 1995). This was the result of a strong increase in RMF, in combination with the fact that roots have a so much higher R than shoots. Hence, plants tend to maintain a homeostatic R/A ratio when irradiance temperature and CO₂ vary, but R increases relative to A when nutrient availability declines as costs for nutrient acquisition increase.

Are there species-specific differences in the C-balance at low irradiance? And are these associated with shade-tolerance? Reliable measurements of root R_m across the whole range of irradiances are available for only two species, the shade-tolerant Geum and the intolerant Helianthus (Figure 5C). Geum had a more stable R/A than Helianthus, which showed a strong increase in R/A at the lowest irradiance (Figure 5F, Table 2). However, the available data points at low irradiance for the other species do not support a systematic difference between shade-tolerant and intolerant species (Figure 5F). The R/A at the shoot level shows higher values for Helianthus and Chenopodium at low irradiance (Figure 5D) and so does the whole-plant R/A based on shoot R_m measurements only (Figure 5E). The differences are significant for the shade-tolerant–shade-tolerant contrast (Table 2), but the growth analysis data do not show clear evidence of reduced net C-gain at 20 μmol m⁻² s⁻¹ for the shade-intolerant species (Figure 1A, Supplement Figure S1A). The available evidence that we have therefore does not conclusively point to a systematic difference in C-balance between our shade-tolerant and intolerant species at the lowest irradiance, which is representative for deep canopy shade. Sterck et al. (2013) arrived at the same conclusion when comparing tropical shrubs of different shade-tolerance, although they assumed a constant shoot and root R_m. The number of species in our study is not large enough for broader generalizations, but when taking into account the literature data as cited above, the conclusion emerges that there is not much evidence for a systematic difference in C-balance between shade-tolerant and intolerant species at the low light levels found in deep shade under a dense canopy.

Mahmoud and Grime (1974) and Pons (1983) reduced irradiance to very low values and found no systematic difference between shade-tolerant and intolerant species in light compensation points. However, mortality was much higher for shade-intolerant species at and below the compensation point, and as far as surviving plants permitted, their RGR was estimated to be more negative. This suggests that shade-tolerant species can reduce R further under C-starvation than shade-intolerant species.

Many shade-intolerant species, including the ones from this study, show a pronounced shade-avoidance response at the low red: far-red ratios in canopy shade light, which involves among others increased stem and petiole growth (Morgan and Smith, 1976; Kurepin et al., 2007; Pierik et al., 2011). This is not only likely to increase R, but may also go at the expense of LMF (Poorter et al., 2012b), thus reducing RGR (equation 1). Shade-tolerant species, including Geum and to a lesser extent Impatiens, show a reduced shade-avoidance response in canopy shade light (Morgan and Smith, 1979; Corré, 1983b; Gommers et al., 2013). The reduced growth is likely to reduce R. This resembles the quiescent strategy found in submergence-tolerant species or genotypes (Bailey-Serres and Voesenek, 2008) that stop growth and reduce respiration under water where photosynthesis is negligible, as opposed to enhanced elongation growth under a negative C-balance of species that have an escape strategy such as Rumex palustris (Groeneveld and Voesenek, 2003). The latter shows similarities with the shade-avoidance response of shade-intolerant species, which furthermore negatively affects stem mechanics and strength (Anten et al., 2005), and can go on the expense of defense as is among others also documented for Chenopodium (Kurashige and Agrawal, 2005). Under natural conditions shade-intolerant species do normally not survive the lowest irradiance used in our experiment, which was equivalent to dense canopy shade. As we have little evidence for a more favorable C-balance of shade-tolerant species at the lowest irradiance under controlled conditions, tolerance to other biotic or a-biotic stresses (see discussion above) in combination with the low irradiance stress is likely to be more important for survival in canopy shade.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.