Photosynthetic quantum efficiency in south-eastern Amazonian trees may be already affected by climate change

2.1 Study site

The study site is located in a large Amazonian rainforest fragment in Vera Cruz farm (14.833°S 52.168°W; plot code VCR-02 in the RAINFOR forest inventory network), Nova Xavantina, Mato Grosso, Brazil. The plot is located 0.8 km from the nearest forest edge. The vegetation of the site located at the forest-savanna ecotone is a transitional Amazonian rainforest (Marimon et al., 2014) that is mostly (over 80%) characterised by typical Amazonian tree species but also containing some species more commonly found in the Cerrado biome (see Morandi et al. (2016) for floristic description). Mean canopy height is 13.6 m, but the height of some trees can exceed 25 m. The soil is dystrophic, acidic and shallow plinthosol.

The site is at the southernmost dry limit of the Amazon rainforest, is highly seasonal and is characterised by an average annual rainfall of 1,369 mm/year (recent 20-year average). Most of the rain falls during a 6-month wet period (mid-October until April), which also experiences lower mean temperatures compared to the 6-month-long dry season (May–October). The peak of the dry season (August–October) coincides with the hottest time of the year, where maximum air temperatures (T_max) frequently exceed 40°C. The daily maximum temperatures have risen considerably over the last two decades (Figure 1a). The recent decade recorded 19 heatwave events (consecutive 3-day spells where maximum daily air temperatures exceed the 90th percentile level for the entire 1999–2008 period) compared to 13 events during the previous decade (Figure 1b); maximum air temperatures have exceeded 40°C 95 times during 2009–2018 compared to 29 times during the 1999–2008 period. The absolute highest T_max recorded was 43.9°C (October 15, 2017, coinciding with the hot-dry sampling period reported in this study).

2.2 Plant material

The most dominant canopy and subcanopy evergreen tree species were chosen for the study. For each species, individual trees were randomly selected from within the VCR2 plot. Branches from the same set of individual trees were sampled during both the seasons. Leaf samples were collected during two seasons: (a) the hottest and driest period of the year during October 2017, referred to as ‘end of dry’ period and (b) the end of the wet season during March–April 2018, referred to as the ‘wet’ period (Figure S1). Plant material was collected in both seasons from three individuals from each of five commonly occurring evergreen tree species: Hymenaea courbaril L. (Fabaceae), Brosimum rubescens Taub. (Moraceae), Amaioua guianensis Aubl. (Rubiaceae), Cheiloclinium cognatum (Miers) A.C.Sm. (Celastraceae) and Mouriri apiranga Spruce ex Triana. (Melastomataceae). Plant material from three individuals of two further species, Chaetocarpus echinocarpus (Baill.) Ducke (Peraceae) and Tetragastris altissima (Aubl.) Swart (Burseraceae), was only collected in the wet period; and one tree was sampled during the end of the dry period for C. echinocarpus (Table S1). Among the species studied, H. courbaril is the most dominant emergent tree with an average height of 29.9 ± 0.9 m. C. echinocarpus (12.7 ± 0.5 m), T. altissima (11.6 ± 0.5 m), B. rubescens (10.4 ± 0.6 m) and A. guianensis (9.5 ± 0.3 m) are mid-storey trees while C. cognatum (7.1 ± 0.2 m) and M. apiranga (6.8 ± 0.4 m) being the understorey trees.

Trained climbers harvested sunlit canopy branches of the trees between 07:00 to 08:00 a.m. The excised branches were separated into small twigs and transported to the lab (~35 km) in closed Styrofoam boxes with ice packs. Heat treatment assays were done typically within 12 hr of branch harvest. Fully expanded mature leaves were sampled in all cases, but during the end of the dry period campaign, H. courbaril trees had a new flush of leaves and hence, unavoidably for this species younger leaves were also sampled.

2.3 Heat tolerance assay

Following the thermal tolerance measurement protocols adapted from Krause et al. (2010), heat treatment assay was conducted on fully mature healthy leaves were transferred into a water trough in which leaf discs of ∅ 20.5 mm were cut underwater using a cork borer, consistently choosing the central part of the lamina, and avoiding the midrib area. One disc was cut from a single leaf except for a small number of cases where leaf material for analysis was limited when two discs per leaf were excised. Leaf discs were then randomly selected and wrapped in two layers of moist tissue paper on both sides covering the cut edge of the leaf discs. Care was taken not to touch the leaves, and the leaf discs always remained in water or covered in damp tissue paper to avoid anaerobiosis (Harris & Heber, 1993). The discs were transferred into separate plastic bags (30 × 100 mm) keeping them flat and covered in a thin film of water. The bags were subjected to heat treatment in sets of at least three discs per temperature point. Heat treatment was conducted in temperature-controlled thermos flasks with pre-heated water. The water temperature was recorded in the central area of the flask using an HI 9063 K-type thermocouple probe. Separate sets of leaf discs were treated to one of the following temperatures points: 30, 35, 40, 42, 44, 46, 48, 50, 52, 54, 56, 58, 60 and 70°C and a set of untreated discs were used as controls (about 25°C). Hence, one assay run, in most cases, includes measurements from at least 45 leaves. After the heat treatment, the leaf discs were dark-adapted for at least 30 min before measuring F_v/F_m with a FluorPen FP100 (Photon System Instruments, Czech Republic). FP100 uses a 455 nm light source at 1 kHz modulation frequency to irradiate a ∅ 5 mm surface aperture. F₀ is read after 40 μs exposure to a 0.09 μmol m⁻² s⁻¹ light per pulse and F_m after a 1 s exposure to 3,000 μmol m⁻² s⁻¹ light pulse. The response was measured on a PIN photodiode with a 667–750 nm bandpass filter for 10 μs. The leaf discs were incubated in lab conditions (25°C, ~20 μmol m⁻² s⁻¹ light) in Petri plates with a thin film of water. F_v/F_m recovery after 24 hr was measured following a 30-min period of dark adaptation.

2.4 Determination of thermal tolerance parameters

Four-parameter logistic curves were fitted for each F_v/F_m—temperature response series (one curve pooling all the leaves per individual run) as follows:

(1)

where T = treatment temperature, b = steepness of the curve, c = lower asymptote, d = upper asymptote or QY_max and e is the inflection point. We extracted three measures from the fitted F_v/F_m temperature response curves: (a) a breakpoint temperature referred to as T₅ calculated as the temperature at which F_v/F_m starts to decline below 95% of the QY_max level, (b) T₅₀, the temperature where F_v/F_m is half of QY_max level and finally (not same as parameter e from Equation 1), (c) T₉₅ the temperature at which F_v/F_m is at 5% of the QY_max level. The difference between T₅ and T₉₅ is referred to as decline width (DW = T₉₅–T₅), the window where F_v/F_m declines from 95% to 5% of the QY_max level (Figure 2). Response curves were fitted using the ‘drc’ function of the ‘drm’ package (Xia, Ritz, Baty, Streibig, & Gerhard, 2015) in R, version 3.5.0 (R Core Team, 2018).

2.5 Analysis of field data

The analysis of field data from Nova Xavantina focused on testing for differences in, T₅₀, T₅, T₉₅ and the decline width, across species and seasons. Two-way repeated measure linear mixed-effect analysis of variance (ANOVA) models were fitted using the ‘lme’ function in the ‘nlme’ package (Pinheiro, Bates, DebRoy, R Core Team, 2018). Season and species were included as fixed effects and individuals (trees) were included as a random effect in the models. Tukey's post-hoc tests were used to test for differences between specific pairs of species in the significant models. Data analysis was performed using the R program, version 3.5.0 (R Core Team, 2018); all means are presented with ±SEs and significance results are presented for 95% confidence interval (CI = 95%, α = .05).

2.6 Comparison with data from other tropical sites

The thermotolerance data gathered in the study presented here were compared with published data for adult tropical evergreen trees from other sites, that is, excluding seedling/greenhouse studies. The final compilation includes data from 106 species occurring in nine tropical sites (Table 1). Some of these studies determined thermotolerance using the T_crit parameter derived from the F₀ rise approach and others used T₅₀ derived from the F_v/F_m approach used in this study. Data for each of these two metrics were treated separately in our analyses. Based on climatological records from CRU (Osborn & Jones, 2014), we classified the season during the sample collection period as ‘hot’ (three hottest months of the year) and ‘cool’ (three coldest months of the year). Similarly, from the precipitation data, we treated the sampling period that took place during the months with rainfall <100 mm as ‘dry’ and periods above 100 mm as ‘wet’. Hence, the classification of ‘dry’ is only a relative description of the site's precipitation status and does not necessarily indicate no precipitation. Data points for seasonal sites such as Iquitos were classified as ‘wet’. Details of the forest regions and study sites can be found in Table S2. The MuMIn package (Bartoń, 2018) was used to extract conditional pseudo R² for the complete model fits.

3.1 Photosynthetic thermal tolerance

Measured F_v/F_m ratios (average of 0.69 ± 0.02 arbitrary unit, n = 35) were stable with increasing temperatures up to a maximal temperature point and declined thereafter to near-zero levels. The breakpoint temperature T₅ averaged at 43.5 ± 1.7°C across species during the end of the dry period and 41.9 ± 0.9°C during the wet period (Figure 3a). The paired difference of 1.06°C over the seasons was not statistically significant. A. guianensis recorded the highest T₅ at 53.9 ± 1.3°C (p < .0001, t₃ = 10.7) during the end of dry period while the rest of the species recorded ~13.4°C lower T₅ at 40.6 ± 3.3°C. However, during the wet period, T₅ for A. guianensis was statistically indistinguishable from that of the other species. Furthermore, T₅ for A. guianensis was higher during the dry period compared to the wet period—while on the contrary, the remaining species showed lower or indifferent T₅ during the dry period. The mixed effects model to explain T₅ variation showed a weak species effect (F = 3.48, p = .01) but no influence of season was found. The species difference was mainly due to the response of A. guianensis which was significantly different from C. cognatum (difference of 11.6°C, p = .005) and M. apiranga (difference of 8.7°C, p = .029).

For T₅₀, we found significant species (F = 7.98, p = .0007) and season (F = 6.69, p = .022) effects (Table S3). A. guianensis was significantly different from the rest of the species except T. altissima. End of dry period T₅₀ for all species combined was ~1.6°C higher at 51.6 ± 0.8°C compared to the wet period that averaged at 49.4 ± 0.6°C. A. guianensis showed the highest seasonal plasticity in T₅₀ while B. rubescens the least. A. guianensis recorded the highest T₅₀ during both at the end of dry (56.4 ± 0.5°C) and wet (53.2 ± 0.6°C) periods (Figure 3b). Compared to the rest of the species studied, T₅₀ for A. guianensis was 6.4°C (p = .001, t₁₄ =5.08) and 4.1°C (p = .016, t₂₁ = 2.61) higher during the end of dry and wet seasons, respectively. Excluding A. guianensis, there was no significant difference across species in T₅₀ during both seasons. It is important to note that all species had mature, fully expanded leaves except for H. courbaril, which had a young flush of leaves during the end of the dry period. However, T₅₀ for H. courbaril was not different from the pool of species excluding A. guianensis.

T₉₅ averaged 60.1 ± 0.2°C across species during the end of the dry period and 57.3 ± 0.3°C during the wet period. Although the ranges differed, the difference in means across the season of 2.9°C was not statistically significant (Figure 3c). The recovery of F_v/F_m, measured 24 hr after heat treatment following a 30-min duration dark period and found no significant recovery upon incubation under the irradiance conditions used (< 20μmol m⁻² s⁻¹ and ~25°C room temperature) in T₅₀ (t₅₀ = 1.23, p = .22), T₅ (t₅₀ = −0.81, p = .42) or T₉₅ (t₄₈ = 1.97, p = .059).

3.2 Gradual versus rapid losses in PSII maximum QY

A negative relationship between the F_v/F_m breakpoint temperature, T₅ and decline width was observed (R² = .68 p = .02; Figure 4a). Moreover, species that sustain QY_max levels to a high breakpoint temperature show sudden and steeper QY decline with a narrow DW. Conversely, in cases where the QY decline started at lower temperatures, the decline was gradual (wider DW). A. guianensis and H. courbaril showed the highest seasonal variation in DW compared to other species and showed similar response over both seasons. C. echinocarpus showed the lowest T₅ and widest DW. To further expand this analysis, we also included as yet unpublished dataset for four evergreen species from a Cerradão (savannah) forest about 50 km from our study site and data from three Panamanian tropical forest species (Slot et al., 2018). The results from this additional analysis (Figure 4b) showed a similar negative relationship between T₅ and decline width (R² = .96 p < .001) consistent with results from our dataset. Calophyllum inophyllum and Inga spectabilis from Panama, showed a similar response to A. guianensis in our measurement with much narrower DW; A. guianensis however, remained the species with highest T₅ in our combined analysis.

3.3 Pan-tropical variation in leaf thermal tolerance

Only five studies in the literature reported data for adult evergreen trees in tropical forests. O'Sullivan et al. (2017) and Zhu et al. (2018) report the F₀ rise inflexion point metric T_crit, whereas the remaining tropical studies, namely, Krause et al. (2010), Sastry and Barua (2017), Slot et al. (2018) and this study report T₅₀ based on F_v/F_m decline. Although the number of studies is limited, this dataset includes rainforests in Australia, India and the NeoTropics (Table 1). Both T₅₀ (Figure 5) and T_crit showed very closely located modes around 47–48°C. While most of the T₅₀ was >45°C and the mode was around 47°C; the range of T₅₀ (45–52°C) was much narrower than T_crit, which showed very high interspecific variability within sites. Mixed-effect models showed that the choice of metric (F = 24.32, p < .001) with season (F = 52.89, p < .001) and species (F = 13.04, p < .001) explained significant variation in the data indicating that both metrics are different given different underlying mechanisms. The two metrics were therefore compared separately.

The T₅₀ values for the Central America and Southern Amazon species in the wet period were 50.4 ± 0.36°C and 49.4 ± 0.73°C, respectively, whereas for the Western Ghats it was significantly (t₃₁ = 2.40, p = .02) lower at 46.9 ± 0.31°C. T_crit was found to vary much more across species and seasons than T₅₀. While T_crit ranged from 37.2 to 66.7°C (n = 96), T₅₀ ranged from 45.5 to 56.4°C (n = 54). A linear mixed-effect model fitted to explain T₅₀ showed season (F = 12.96, p < .0001) and biogeographical region (F = 15.26, p < .0001) as significant variables (pseudo R² = .885). In contrast, T_crit showed no effect of season, however, the variability and the limited number of sites represented, limits any meaningful comparison.

4.1 Thermal tolerance at the hottest Amazonian forest site

Tree species measured in Nova Xavantina showed a very high thermotolerance, with one species, Amaioua guianensis, recorded the highest T₅₀ documented for any tropical evergreen tree thus far (52.7 ± 1.05°C). The data reported here represent the first F_v/F_m measurements made on adult Amazon trees. Thus, it is not possible to directly compare these results with published data from other Amazonian sites. However, the T₅₀ values reported here are similar to values reported for four Panamanian species (Slot et al., 2018) and also to T_crit values reported for two Amazonian sites (O'Sullivan et al., 2017). However, although F_v/F_m and F₀ are related, they are not equivalent. Hence, care must be taken when making comparisons across metrics. However, the species in Nova Xavantina (Table 1) were considerably more thermotolerant than the Western Ghats forests in India, where the most extensive datasets by far are available for leaf thermotolerance (Sastry & Barua, 2017).

The differences in the long-term T_max mean values for the site across the two seasons was ~2.5°C (T_max = 35.7 ± 0.37°C for October and 33.2 ± 0.23°C for the months March–April). Consistent with the literature, T₅₀ measurements across seasons showed significant differences, with T₅₀ values for the hot/dry season being ~1.6°C greater than in the cooler/wetter season. A. guianensis, a characteristic mid-storey species found across a wide range of ecosystems in the region (Morandi et al., 2016), had the highest T₅₀ and the highest seasonal plasticity in T₅₀. A. guianensis is the slowest growing species among the species sampled (Mews, Marimon, Pinto, & Silvério, 2011). These results are consistent with studies that indicate that slow-growing species have a greater capacity for flexible heat dissipation and thermal protection mechanisms than fast-growing plants (Adams & Demmig-Adams, 1994). Maintaining high thermal tolerance during dry periods could be energetically expensive (Wahid, Gelani, Ashraf, & Foolad, 2007), thus requiring down-regulation during wet periods. Stomatal regulation varies across seasons and species. Moreover, lower water availability limits the transpiration-dependent cooling capabilities in dry periods.

The observed leaf-level differences could be due to small variations in the prehistory of the leaves within the canopy for example in terms of light or temperature exposure (Colombo & Timmer, 1992) during the peak dry period, prior to sampling. For example, the acquired tolerance induced by a pre-exposure to heat is associated with the synthesis of heat shock proteins (HSPs) and with other low molecular weight proteins that protect PSII (Gifford & Taleisnik, 1994). The variations in T₉₅ values were higher than those observed in T₅ or T₅₀. This may indicate that there is greater variation in the high-temperature threshold for loss of photosystem integrity (T₉₅) between species than in the initiation temperature of sensitivity (T₅; Figure 6).

The mechanisms that underpin the thermal stability of photosynthesis in tropical trees have not been fully characterised. The data presented here clearly show that species such as A. guianensis are able to maintain PSII functions up to high temperatures (53.9 ± 0.75°C during the end of dry period, that is, the highest reported in literature for C3 plants). These trees could have mechanisms to not only protect the PSII from irreversible thermal inactivation but also have a high degree of plasticity in relation to temperature fluctuations. In such trees, loss of photosynthetic functions may be prevented by a more rapid repair cycle than that occurring in other species. However, the precise nature of the specific mechanisms involved remains unclear. Further molecular and metabolic studies on tropical tree species are required to understand how photosynthesis can withstand high temperatures. Some studies indicate that redundancy and diversity of light-harvesting complexes (Tang et al., 2007) could play a role in buffering the high light, high temperature-induced changes in photosystem functions. Hence, heat sensitive systems could be replaced with more thermally stable forms. Similarly, large variations in heat shock factors/proteins, oxidative stress/signalling and thermal energy dissipation could exist across taxa, hinting at differential thermal sensitivity of tropical evergreen trees to high-temperature stress.

4.2 Thermal sensitivity thresholds for PSII

A key question concerns how close tree species are to their thermal sensitivity thresholds in the warmest region of the Amazon? The data presented here show that T₅₀ for seven focal species is at least 4°C above the highest air temperatures ever experienced at the study site (43.9°C). However, for most species, T₅, the breakpoint temperature_, is already surpassed in the peak of the warm/dry season (Figure 6). Thus, the site temperatures are already reaching levels where the incipient irreversible loss of PSII activity occurs during the dry/warm periods. Such periods are predicted to become more frequent with future warming (Yao, Luo, Huang, & Zhao, 2013). However, these inferences are based on air temperatures, while leaf temperatures are arguably a more meaningful measure of proximity to thermal sensitivity thresholds. During the dry/warm season, limited access to soil moisture restricts stomatal opening. This is likely to lead to leaf temperatures that exceed air temperatures (Krause et al., 2010; Slot & Winter, 2016), at least for part of the day. Under these circumstances, plants in our study area, especially those exposed to full sunlight, may operate closer to thermal sensitivity thresholds than the simple examination of air temperatures suggests.

4.3 Variations in the responses of PSII to heat

The negative relationship between T₅ and decline width (Figure 4) indicates a range of response strategies in trees. The two extremes of PSII thermal sensitivity could be described as ‘sensitive’ and ‘tolerant’ (see Figure 7). Tolerators were found to sustain their PSII QY values up to a remarkably high-temperature point (high T₅). They were characterised by a rapid decline to near-zero levels at higher temperatures beyond T₅. Conversely, heat ‘sensitive’ responses are characterised by a sensitivity of PSII to much lower (low T₅) temperatures. The decline in PSII QY to near-zero levels in these trees, however, occurs gradually over a very wide temperature range. Hence, in heat sensitive trees, when the leaf temperatures are at this ‘decline width’ range, PSII is already negatively affected by temperature stress. The slow decline could potentially indicate PSII protection mechanisms that ensue mediated by HSPs and/or xanthophyll mediated thermal protection mechanisms. Given the change in PSII activity, it is reasonable to hypothesise that electron flow could switch from the non-cyclic pathway to cyclic electron transport around PSI, a mechanism that is known to protect PSII against thermal inactivation (Essemine, Xiao, Qu, Mi, & Zhu, 2017; Sun, Geng, Du, Yang, & Zhai, 2017).

High-temperature tolerant trees, on the other hand, have better protected PSII which can continue to function up to a very high-temperature point. By sustaining PSII activity under extremely high leaf temperature conditions, tolerators could maintain better photosynthetic electron flow rates than thermosensitive species. The effective thermal sensitivity of heat tolerant trees could be linked to stomatal controls and transpiration cooling. However, evaporative cooling can reduce the need for investment in thermal tolerance mechanisms. If transpiration cooling is strong, the mechanisms that facilitate high PSII thermal tolerance might not be required. Hence, thermal tolerance of leaf metabolism must consider the molecular physiology of the leaf.

4.4 Ecophysiological relevance of PSII QY temperature sensitivity

During the dry and hot period of the seasonal cycle, leaf to air temperature differentials (T_leaf–T_air), in evergreen trees in some tropical sites have been measured to be as high as 18°C (Fauset et al., 2018; Pau, Detto, Kim, & Still, 2018; Tribuzy, 2005). Furthermore, during the parts of the day when transpirational cooling capacity is limited by stomatal closure (Slot et al., 2018), leaf temperatures reach as high as 48°C in some tropical trees (Slot & Winter, 2016). Comparing these leaf temperatures with the range of T₅ measured, consistent with the ideas presented in Mau, Reed, Wood, and Cavaleri (2018) and Doughty and Goulden (2008), it is likely that some sensitive species are already experiencing temperatures that are affecting PSII functioning. Given the species level variability and largely unexplored mechanisms of thermal protection strategies of trees, it is probable that there will be a varied range of response to high-temperature conditions across species. Early leaf senescence is triggered by exposure to high temperatures (De la Haba, De la Mata, Molina, & Agüera, 2014; Way, 2013). Strategies such as increased leaf turnover rates could, therefore, be triggered by high temperatures with implications for carbon acquisition by the trees. Transpirational cooling mechanisms under high temperatures could be limited and vary between species, making species with less water available for cooling more susceptible to thermal stress. However, some species continue to transpire even at extremely high temperatures (Drake et al., 2018).

The F_v/F_m or F₀ based fluorescence data only indicate the responses of PSII (see Figure S3 for F_v/F_m and F₀ relationship). However, the limits of other processes such as thermal protection, stomatal and metabolic controls of CO₂ assimilation are largely unknown except in a few model plant species. Hence, such integrated definitions of thermal thresholds could be more meaningful and provide a better understanding of the thermal sensitivity of trees or plants in general. Thus, concepts such as thermal safety margins (O'Sullivan et al., 2017; Sunday et al., 2014) could be expanded further to derive ecophysiologically meaningful conclusions (see Supporting Information 7 and Table S4).

Overall, the data presented here show that there is significant diversity in the temperature sensitivity of photosynthesis across tropical evergreen trees (Cseh et al., 2005; Holm, Várkonyi, Kovács, Posselt, & Garab, 2005), much of which remains uncharacterised. It is probable that with increasing temperatures, especially during the summer, the leaf temperature differences between species, together with differences in transpiration cooling capacity, will reveal large interspecific variations in tree sensitivity to extreme heat conditions.