Plant defence traits among discrete vegetation assemblages in a mesic savanna landscape in Kenya

Abstract

A trade-off between structural and chemical defences against herbivory in woody plants is alleged to depend on edaphic factors in African savannas. We studied anti-herbivory traits, in an edaphic mosaic of fertile and infertile soils within a savanna landscape in East Africa, towards elucidating herbivory defence traits expressions in woody plants of African savannas. We used data of 81 plants for 8 species from 8 sites — four sites from fertile soils (42 plants) and another four sites from infertile soils (39 plants). We did not find a general divide between structural and chemical strategies in our data. Instead, we found a range of defence traits combinations. Our results highlight that in woody plants of African savannas, chemical and structural defences can augment each other, and not necessarily trade-off. The diversity of herbivores, ranging from insects to mesobrowsers, may have driven the evolution of multiple defence strategies within the African savannas.

Introduction

Plants are primary producers in most ecosystems (Woodwell and Whittaker 1968). It is therefore no surprising that one of the most prominent sets of adaptations in their life history is defence against natural enemies. Essentially, plants employ several different lines of defence strategies against herbivory, including structural (Hanley et al. 2007, War et al. 2018) and chemical (War et al. 2018) ones. Structural defence traits, such as leaf pubescence and leaf sclerophylly, affect herbivores by decreasing both palatability and digestibility (Hanley et al. 2007). Spinescence, another form of structural defence trait where parts of the plant shoot are modified into sharp tips, also affects herbivory by reducing feeding rates (Gowda 1996). Among the chemical defence types, tannins serve as a potent defensive secondary metabolite, as they bind proteins, inhibit enzymatic activity, and render protein present in a food nutritionally unavailable for herbivory (Mazid et al. 2011, Swain 1977).

Until present, a trade-off between structural and chemical defences is alleged for fertile soil fine-leaved and infertile soil broad-leaved African savannas woody plants (Scholes and Walker, 2004). The basis for this hypothesised dichotomy is that African savannas fine-leaved woody plants, growing in nutrient-rich areas, tend to have high nutrient content leaves that are attractive to herbivores. As such, for the trees to defend themselves from intense herbivores browsing, they invest in structural defence strategy. Broad-leaved trees growing in nutrient-poor areas in African savanna landscapes, on the other hand, are purported to employ a chemical defence strategy (e.g., low leaf nutrient content and high secondary metabolites), which makes them unattractive to herbivores. Strong empirical support for a trade-off between the two defences (structural and chemical) in plants is however weakly reported in the field (see Moles et al. 2013).

In this study, using nutrient-rich and nutrient-poor soils common trees from a savanna landscape in Kenya, we further understanding of anti-herbivore traits dynamics in woody trees of African savannas. We test a support for the chemical and structural defences trade-off hypothesis in nutrient-rich versus -poor soils African savanna wood plants.

Materials and Methods

Description of study sites

The study was conducted at southeastern part of Kenya (0.67°S to 2.62°S, 37.70E° to 38.09°E; Figure 1a), in East Africa (Figure 1b). Study sites comprise a set of four sites from broad-leaved woody plant communities (Figure 1c), and another set of four sites from fine-leaved woody plant communities (Figure 1d). In the area, fine- and broad-leaved woody plant communities grow in relatively nutrient rich and nutrient poor soils, respectively (see Kimeu et al. 2020). Domestic herds (largely goats) and small African mammalian browsers (e.g., dik-diks and antelopes) are the browsing ungulates predominant in the study vegetation. However, the vegetation is part of the larger East African savannas, which in the past formed one continuous ecosystem that harboured a large population of African mega-herbivores (Marchant et al. 2018). The study vegetation is thus likely evolved in the presence of large densities of mega-herbivores, and perhaps under influence of fire — the other major disturbance to the vegetation in African savanna landscapes (see Bond 2008, Sankaran et al. 2008 and also Midgley et al. 2015).

Figure 1. Sampling sites and exemplar of vegetation types. (a) Location of fertile soil fine-leaved (circle, orange) and infertile broad-leaved (triangle, green) vegetation types, (b) position of study area in tropical Africa, (c) broad-leaved Combretum dominated woodlands, and (d) fine-leaved Acacia dominated woodlands.

Sampling strategy

We sampled woody plant species from eight sites — four sites from fertile soil fine-leaved Acacia dominated vegetation and another four sites from infertile soil broad-leaved Combretum dominated vegetation — for leaf samples. Sampling was conducted within a 2-week period in December 2019 to January 2020. Fully expanded mature leaves were sampled from three mature individuals (10 leaves per individual) of each study species at every site the species occurred. Only indigenous and the commonest woody species were sampled, and these include spiny Acacia and non-spiny Combretaceae species, which had previously showed to be the most dominant and frequent species to the study area (Kimeu et al. 2020).

Plant traits measurement

Four leaf traits including specific leaf area (SLA), nitrogen (N), carbon (C), and condensed tannins (CTs) were measured. SLA index for every sample plant was calculated from averaged leaf areas and masses taken from a set of 10 mature and undamaged leaves collected per individual sample plant. Areas for the leaf samples were determined using the open-source software ImageJ (Abràmoff et al. 2004); and images used in the ImageJ encompassed a photograph of a whole leaf (i.e., including both petiole and rachis) photographed shortly after its collection (see Figure 2). The leaf weights were measured on air-dried leaf samples using a 0.01 gram digital electronic portable measuring scale. Leaf weighing was carried out at the University of Cape Town in the Biological Science department ecology laboratory.

Figure 2. (a–h) Representative leaf specimens of fertile soil Acacia woodland species (a) Acacia mellifera, (b) A. nilotica, (c) A. senegal, and (d) A. tortilis. (e–h) Representative leaf specimens of infertile soil Combretum woodland species (e) Combretum apiculatum, (f) C. collinum, (g) C. zeyheri, and (h) Terminalia brownii.

Measurement of leaf chemical traits used the same leaf samples examined for SLA. Specifically, the 10 leaves collected for each individual plant sample were pooled into one composite sample. Each composite was then mill ground to pass through a 2-mm sieve using a Hammer mill (United Scientific, Cape Town, USA). Leaf nitrogen (N) and carbon (C) contents were analysed using Thermo Scientific FLASH 2000 CHN Elemental Analyser (Thermo Fisher Scientific Inc., Massachusetts, USA). We used the protocol by Porter et al. (1986), as modified by Hagerman (2002), for quantifying CTs in plant samples leaves, and used purified Sorghum tannin as the protocol calibration standard (Hattas and Julkunen-Tiitto 2012).

Statistical analyses

To assess how well the predicted groups (nutrient-rich soils fine- and nutrient-poor soils broad-leaved woody plants) were able to assign each sample species to the correct group based on the sampled leaf traits (SLA, nitrogen, carbon, and CTs), we ran linear discriminant analysis (LDA) using the ‘Mass’ package in R. Further, we used the boxplot tool (in the R ‘stats’ package) to evaluate spread of values for the sampled leaf traits across the sample species. To test evidence for a trade-off in allocation to leaf traits associated with structural defence versus chemical defence for nutrient-rich soils, fine-leaved and nutrient-poor soils, broad-leaved woody species, we correlated SLA (which encompasses a plant’s sclerophylly index aspect) against each of the three sampled leaf chemical traits (nitrogen, carbon, and CTs). We used the ‘cor.test’ function of the R ‘stats’ package, and we plotted the plants for the groups in scatterplots spaces, with different symbols used for each group.

Results

A total of 81 plants (42 and 39 of from sites of Acacia dominated and Combretum dominated woody plant communities, respectively) were recorded for this study (Table 1 and Appendix 1).

Table 1. Commonest woody plant species in fertile soil fine-leaved and infertile soil broad-leaved plant communities at the mesic savanna of southeastern Kenya

Defences between fine-leaved and broad-leaved woody plant communities

LDA revealed 86.98% of variance of the studied leaf traits distributed along LDA1 (50.45%) and LD2 (36.53%). Fine-leaved, spinescent Acacia and broad-leaved, non-spinescent Combretaceae species loading along axis LDI were principally correlated with SLA and nitrogen, and along LD2 with carbon and CTs (Table 2).

Table 2. Coefficient of linear discriminants for the study leaf traits along axes LD1 and 2 derived from the study linear discriminant model.

When we compared measurements for each of the four sampled leaf traits between fine-leaved and broad-leaved plant functional groups, both SLA and leaf nitrogen differed markedly between the two plant groups. All fine-leaved spiny Acacia species had higher SLA than broad-leaved non-spiny Combretaceae species (Figure 3a). Fine-leaved spiny Acacia species again had generally higher leaf nitrogen content than Combretaceae species (Figure 3b). This pattern however was not repeated for leaf carbon and CTs — that is, both traits (carbon and CTs) had high and low peaks, which were species-specific rather than vegetation-type-specific (Figure 3c and d).

Figure 3. Boxplots showing SLA, leaf nitrogen, total carbon, and condensed tannin traits profiles for the common species (Acacia mellifera (A.me), A. nilotica (A.ni), A. senegal (A.se), A. tortilis (A.to), Combretum apiculatum (C.ap), C. collinum (C.co), C. zehyeri (C.ze), and Terminalia brownii (T.br)) of spiny, fine-leaved (orange) and non-spiny, broad-leaved (green) vegetation systems of the mesic savannas of southeastern Kenya.

Defence strategies trade-offs between fine- and broad-leaved woody plant communities

Despite some overlap, the current data revealed evidence for a trade-off of leaf nitrogen content profiles between Acacia and Combretaceae species (Figure 4a). The correlation coefficient for leaf nitrogen, between fine- and broad-leaved vegetation leaf categories, is significant (r = 0.67, p < 0.001). Carbon and CTs, however, did not show evidence of a trade-off between fine- and broad-leaved leaf samples, where the correlation coefficients recovered for carbon and CTs between fine- and broad-leaved vegetation leaf categories both are low and not significant (r = 0.12, p = 0.240 for carbon (Figure 4b), and r = 0.13, p = 0.210 for CTs (Figure 4c).

Figure 4. Pearson correlations of SLA plotted against leaf chemical traits nitrogen, carbon, and condensed tannins for spiny fine-leaved and non-spiny broad-leaved vegetation species of the mesic savanna of southeastern Kenya.

Discussion

This study set to test a trade-off between structural and chemical defence strategies alleged for fertile soil fine-leaved and infertile soil broad-leaved African savanna woody plants (Scholes and Walker 2004). The data presented here did not consistently support a trade-off between structural and chemical defence strategies in fertile soil fine-leaved and infertile soil broad-leaved woody plant samples. For example, results from the LDA showed that SLA and leaf nitrogen contributed the most to the first axis, while carbon and CTs contributed the most to the second axis, suggesting that there is no trade-off between the two traits sets in the studied vegetation systems. Had there been evidence for a strong trade-off between them, they would be expected to separate along a single axis. Furthermore, both leaf SLA and N indices were generally different between fertile soil fine-leaved and infertile soil broad-leaved plant samples, with a trend of higher quality (i.e., higher SLA and N) in fertile soil fine-leaved plant samples compared to the infertile soil broad-leaved plant samples). This trend however is not replicated for the other sampled two leaf traits (leaf carbon [C] and CTs).

The concept that chemical and structural defence syndromes can augment each other and do not necessarily trade-off has emanated from this work. This pattern is curious given that a previous work by Tomlinson et al. (2016) reported a support for the Scholes and Walker (2004) chemical and structural defences trade-off hypothesis in nutrient-rich versus -poor soils African savanna wood plants — although also there are other data that show in African savanna landscapes a section of spiny woody species growing in nutrient-rich soils could invest in chemical anti-herbivore defences equally to non-spiny broad-leaved plants growing in nutrient-poor soils (Wigley et al. 2018, 2019). The current data perhaps is highlighting Agrawal (2011) intuition that a simple trade-off model is unlikely as multiple defence traits in concert would be more effective, that is, a diverse suite of herbivores with different responses to specific chemicals or defences may attack a particular species.

In the simple trade-off model, fine-leaved species should not invest in high-carbon concentration (Figure 3c), when fine-leaved species are assumed to be defended by structural defences rather chemical defences (Scholes and Walker 2004) — although carbon is also a critical structural element that provide biomechanical support for plant tissues (Niinemets and Tamm, 2005). It is also not clear why some fine-leaved species invest in high CTs (Figure 3d), and yet this is a trait expected in broad-leaved savanna trees in the simple trade-off model (see Scholes and Walker, 2004, and also Tomlinson et al., 2016). CT concentration has been shown to influence herbivores’ diet choice as it makes nutrients less available after ingesting (Ward and Young 2002, Scogings et al. 2004). Nonetheless, similar to our results, both low and high CTs indices have been previously reported by other studies focusing on fine-leaved spiny woody plant species in African savannas. In north-central Kenya, fine-leaved Acacia drepanolobium browsed by antelope and megaherbivores contained 1.0 to 17.1% CTs (Ward and Young 2002). Furthermore, in the same study area, A. etbaica and A. brevispica contained 1.3% and 20.8% CTs, respectively (Ford et al. 2014; Ward and Young 2002).

Aspects of phylogenetic history are known to shape expression of suites of defence traits in species, and especially when species are closely related (Agrawal and Fishbein 2006). In the current study, for each vegetation type we sampled species from a singular family (Fabaceae and Combretaceae for fine- and broad-leaved vegetation types, respectively). While we take seriously the recognizant of a consequence of functional attributes having deep historical origins, we argue that the trait patterns recovered in our data cannot all be explained by biases due to phylogenetic signals. In the study area, for example, there are two Combretaceae species (Combretum aculeatum and Terminalia spinosa, see Kimeu et al. 2020) with relatively fine-leaved and spines, but none of the sampled Combretaceae species here has these traits.

Overall, this study results have shown that plant anti-herbivory defences are complex, and perhaps operating in tandem. The dominant species in each vegetation type in the study showed the two defence strategies may overlap, for example, at least with some species combining structural defence strategy traits (nutritious leaves [high SLA and nitrogen]) with chemical defences (CTs and carbon). The results affirm the logical thought that it is inappropriate to follow a single line of defence in plants as plants themselves can employ multiple sets of defences (see Agrawal and Fishbein 2006). Lastly and more importantly, our study greatly contributes to understanding plants’ herbivory niche partition that could account for the high herbivore diversity supported by wood species within landscapes of East Africa savannas (Appendix 1).