The origin of tetrapod herbivory: effects on local plant diversity

Abstract

The origin of herbivory in the Carboniferous was a landmark event in the evolution of terrestrial ecosystems, increasing ecological diversity in animals but also giving them greater influence on the evolution of land plants. We evaluate the effect of early vertebrate herbivory on plant evolution by comparing local species richness of plant palaeofloras with that of vertebrate herbivores and herbivore body size. Vertebrate herbivores became diverse and achieved a much greater range of body sizes across the Carboniferous–Permian transition interval. This coincides with an abrupt reduction in local plant richness that persists throughout the Permian. Time-series regression analysis supports a negative relationship of plant richness with herbivore richness but a positive relationship of plant richness with minimum herbivore body size. This is consistent with studies of present-day ecosystems in which increased diversity of smaller, more selective herbivores places greater predation pressures on plants, while a prevalence of larger bodied, less selective herbivores reduces the dominance of a few highly tolerant plant species, thereby promoting greater local richness. The diversification of herbivores across the Carboniferous–Permian boundary, along with the appearance of smaller, more selective herbivores like bolosaurid parareptiles, constrained plant diversity throughout the Permian. These findings demonstrate that the establishment of widespread vertebrate herbivory has structured plant communities since the late Palaeozoic, as expected from examination of modern ecosystems, and illustrates the potential for fossil datasets in testing palaeoecological hypotheses.

1. Introduction

Organisms within communities interact with each other, influencing survival and fitness, and therefore evolution and diversification [1]. Interactions between plants and animals have been a fundamental aspect of terrestrial ecosystems since these groups first moved onto land. Plants are primary producers, influencing the structure of higher trophic levels and architecturally defining the environment in which animals live, modifying moisture, light and space regimes [2]. However, plants are also influenced by interactions within these ecosystems, responding to the selection pressures placed on them by animals.

Terrestrial vegetative ecosystems were established at least by the beginning of the Silurian [3,4]. An architecturally modern terrestrial flora was established by the Middle Devonian, with large, lignified, tree-like plants, a shrubby understory and herbaceous plants at ground level [5]. Alongside this, a fauna of arthropods, and later tetrapods, was being established. At first, this consisted only of carnivores and detritivores [6,7], with no animals feeding directly on living plants. The earliest evidence of arthropod herbivory occurs in the Devonian and indicates removal of spores from sporangia and piercing of stems [8]. Direct consumption of leaves, roots or seeds (high-fibre herbivory) did not appear until the late Carboniferous in either arthropods or tetrapods [2,7–10].

The earliest tetrapod herbivores appear in the fossil record in the Pennsylvanian, although the precise time of origin for this behaviour is uncertain due to the difficulty of assessing diet in extinct organisms. Potentially herbivorous taxa appear in the Bashkirian (earliest Pennsylvanian)-aged Joggins Formation: pantylid microsaurs with robust palatal dentition, a possible adaptation for grinding plant material or crushing the thick exoskeletons of arthropods (or both) [11]. More probable herbivores are known from the Kasimovian stage later in the Pennsylvanian: specimens of the diadectid Desmatodon from the Conemaugh Group of Pennsylvania [12].

The potential impacts of herbivory on Palaeozoic plant evolution have been discussed extensively. Much of this discussion concerns morphological innovations in plants, both to defend against herbivory and to encourage biotic dispersal of seeds [2,13–16]. However, most work on early plant–animal interactions has been focussed on arthropod herbivory [13–16], overlooking the potential importance of tetrapod herbivores in influencing plant evolution [17]. Moreover, thus far, there has been little discussion on how the origin and early evolution of herbivory, either in arthropods or tetrapods, affected patterns of community richness in plants.

Here, we examine how the origin and evolution of herbivorous tetrapods may have affected plant species richness in local communities, using sampling-corrected estimates of α-diversity and estimates of body size in tetrapods.

2. Material and methods

(a). Diversity and size estimates of herbivorous tetrapods

(i). Dataset

A dataset of herbivorous tetrapod occurrences between the Carboniferous and early Triassic was derived from a review of the published literature, museum collections and the Paleobiology Database (PBDB), downloaded in March 2020. The herbivorous taxa were determined based on prevailing dietary opinions in the literature (electronic supplementary material, data S1).

(ii). α-Diversity

The impact of environmental and biotic drivers and processes is scale dependant, so the scale of the observations affects the inference of biological and ecological patterns [18–22]. Competitive interactions, such as herbivores feeding on plants, will have their effect at a local (community) scale [23], and so where possible are better tested by comparing the α-diversities of herbivores and plants.

Here, geological formations were used to represent contemporaneous long-term communities of organisms (see electronic supplementary material, text, for discussion of this protocol). Where well-established biostratigraphic subdivisions existed within formations, e.g. assemblage zones, these were used. To account for sampling heterogeneity, the species richness within each formation was estimated using shareholder quorum subsampling (SQS) [24], implemented in R 3.4.3 [25], using the estimateD() function in the package iNEXT. A coverage quorum of 0.9 was applied, following research suggesting that values lower than this produce imprecise results [26,27]. Formations containing only a single amniote taxon were removed prior to analysis, as were those from which fewer than five localities had been sampled. This left a dataset of 164 herbivorous species and species of uncertain diet in 42 formations.

Species for which diet categorization was uncertain were included or excluded at random before calculating the diversity within each formation. This was repeated 100 times, and the median of the diversities obtained from each formation was calculated.

(iii). Size

As cranial remains are the most commonly reported anatomical region for Palaeozoic tetrapods, skull length is used as a proxy for body size (electronic supplementary material, data S2). Cranial size is a relatively poor estimator of body mass, but this is the only widely available index of size for Palaeozoic tetrapods [28] and we temper our interpretations with an understanding of the possible errors.

Skull lengths in the dataset represent the largest known complete specimens for the 160 species for which they are available. Specimens were measured from the anterior tip of the snout to the junction between the skull roof and occipital plate along the cranial midline. The majority of measurements were taken from specimen photographs processed in ImageJ 1.52a [29] and checked against published measures and direct calliper measurements where possible. Scale bars in photographs were placed as closely as possible to the plane of the skull roof in a horizontal position to minimize the effects of parallax; comparison with direct calliper measures indicates measurement error associated with scale bar position is less than 2.0%. Other specimens were included based on measurements from the published literature or measured from images in the literature, again using ImageJ.

(b). Diversity estimates of plants

(i). Dataset

The plant data were downloaded from the PBDB in March 2020 (electronic supplementary material, data S3). The dataset includes global plant fossil occurrences from the Carboniferous, Permian and Early Triassic. The plant data were assessed at the genus level, as species-level assignments are only very rarely reported in the PBDB data (PBDB plant data from this time is primarily derived from the Paleogeographic Atlas Project, compiled at the genus level).

(ii). α-Diversity

The same procedure as used in herbivorous tetrapods was used to calculate α-diversity within each formation in plants. The same sampling criteria for inclusion of a formation were also applied, leaving a dataset of 124 formations containing 267 taxa. Sensitivity tests to assess the impact of various issues that may affect the reliability of diversity estimates of plants are presented in the electronic supplementary material, text.

(c). Multivariate comparisons of diversity

The time bins used were informal substages, produced by splitting each stage in two (early and late) at the midpoint. To produce time series of α-diversity, the median diversity of all formations whose range overlaps with each time bin was calculated; for herbivore size, the median, minimum, maximum and variance of skull lengths of the species present in each time bin was used. These time series were compared using generalized least-squares regression (GLS). This method allows models comparing multiple independent variables to the dependant variable [30], enabling testing of hypotheses that may not be mutually exclusive, where otherwise apparently weak correlations between time series may simply be due to the tested variable representing a composite signal [31].

The GLS analyses were carried out in R using the function from the package nlme, using a first-order autoregressive correlation structure in all cases to account for autocorrelation [30]. All comparisons cover the period of time from the Gzhelian until the Olenekian, the span for which there are data for all variables for all substages. Ten models were tested as explanations of the α-diversity of plants: (1) a null model (random walk); (2) α-diversity of herbivores only; (3–6) median, variance, maxima and minima of herbivore size respectively; (7–10) multivariate models comparing both herbivore diversity and, respectively, median, variance, maxima and minima of size respectively to plant diversity. Akaike weights [32] were used to determine the best-fitting models; due to the shortness of the time-series (23 bins) these were based on the sample-size corrected Akaike information criterion (AICc). The relative importance of the independent variables was calculated by averaging R² contributions over ordering of regressors [33], implemented in the R package relaimpo [34].

3. Results

(a). α-Diversity of plants

Median α-diversity of plants is at its highest during the Carboniferous (figure 1). Across the Carboniferous/Permian boundary, the median diversity drops noticeably. None of the Permian plant-bearing formations contain species richness equal to the most diverse Carboniferous formations, and the median plant α-diversity remains lower than that of the Carboniferous throughout the Permian, undergoing another abrupt decrease across the Permian/Triassic boundary.

Figure 1.

α-Diversity of plants (grey) and tetrapod herbivores (black), calculated by SQS. Each thin line represents the diversity in a fossil-bearing formation. The thick lines represent the median diversity of all formations in each time bin.

(b). α-Diversity of herbivores

α-Diversity of herbivores (figure 1) is low during the Gzhelian (the last stage of the Carboniferous)). Median α-diversity increases rapidly across the Carboniferous/Permian boundary and remains at a similar level for the rest of the Permian. Diversity decreases across the Permian/Triassic boundary.

(c). Herbivore size

The earliest herbivores appear at relatively large sizes (median skull length of about 175 mm). Maximum skull length increases slightly across the Carboniferous/Permian boundary, and larger sizes appear throughout the Permian (figure 2). However, the appearance of numerous small herbivore species in the earliest Permian causes both the median and minimum sizes to drop across the Carboniferous/Permian boundary, and the variance to increase. Large herbivores become extinct over the Permian/Triassic boundary and small herbivore species increase in richness, so that median herbivore size is lower in the Early Triassic than at any time in the Palaeozoic.

Figure 2.

Size of Palaeozoic herbivorous tetrapods. Thin lines represent the log transformed skull length of each taxon; those in grey are those whose diet is uncertain. The thick black line represents the median size in each time bin, and the bars represent the standard error. (a) Diasparactus zenos (Field Museum of Natural History (FMNH) UC 679), a diadectid; (b) Bolosaurus striatus (American Museum of Natural History (AMNH) FARB 4321), a bolosaurid; (c) Embrithosaurus schwarzi (South African Museum (SAM) PK 8034), a pareiasaur; (d) Casea broilii (FMNH UC 656), a caseid and (e) Cistecephalus microrhinus (Evolutionary Studies Institute, BP/1/33), a dicynodont. (Online version in colour.)

(d). Generalized least squares

The two best-fitting GLS models explaining α-diversity of plants receive similar support based on Akaike weights. The best fit explains plant α-diversity in terms of herbivore α-diversity and minimum body size (AICc weight = 0.51; R² = 0.55), whereas the second explains it in terms of herbivore diversity and size variance (AICc weight = 0.31; R² = 0.39), both substantially higher than the other models, including the null (table 1). In both cases, the α-diversity of herbivores explains less variance than herbivore body size. The relationship of plant diversity with minimum herbivore size was positive, whereas that with between herbivore size variance and herbivore diversity was negative.

Table 1.

GLS models tested, their log likelihood (LnL), Akaike information criterion (AICc), Akaike weights, coefficients and the proportion of variance explained by the variables. The two best-fitting models are italicized.

model	LnL	AICc	Akaike weights	coefficient of variable 1	coefficient of variable 2	variance explained by model (%)
plant α-diversity ∼ null	−1.97	7.94	9.3 × 10−8	NA	NA	NA
plant α-diversity ∼ herbivore α-diversity	11.90	−15.80	0.013	−0.09	NA	19.6
plant α-diversity ∼ median herbivore size	12.69	−17.40	0.030	0.38	NA	20.2
plant α-diversity ∼ herbivore size variance	14.06	−18.11	0.043	−0.25	NA	28.4
plant α-diversity ∼ maximum herbivore size	11.84	−15.67	0.013	−0.22	NA	2.0
plant α-diversity ∼ minimum herbivore size	14.62	−19.25	0.075	0.41	NA	29.6
plant α-diversity ∼ herbivore α-diversity + median size	11.86	−13.72	0.0047	−0.14	0.37	7.44
plant α-diversity ∼ herbivore α-diversity + size variance	1504	−22.09	0.31	−0.14	−0.44	39.8
plant α-diversity ∼ herbivore α-diversity + maximum size	11.02	−12.05	0.0021	−0.24	−0.29	9.18
plant α-diversity ∼ herbivore α-diversity + minimum size	15.54	−23.08	0.51	−0.15	0.58	55.34

4. Discussion

The presence, diversity (species richness), body size and dietary selectivity of herbivores influences plant diversity in modern-day ecosystems [23,35–37]. These effects depend on the environment under study and the spatial scale at which the system is examined and can include both increases and decreases in plant diversity (reviewed by Olf and Ritchie [23]). The Carboniferous–Permian transition documents the first evolutionary appearance of widespread vertebrate herbivory, leading us to ask: what was the effect of this on the species richness of plant communities?

Studies of invasive herbivores show that increased predation pressure from novel herbivore species can reduce plant diversity [38]. By contrast, increased predation pressure can also reduce the dominance of highly abundant plants leading to increased diversity, as shown by studies comparing grazed areas to areas from which herbivores are excluded [39–41]. This is particularly the case when herbivores are less selective.

Among modern-day herbivores, selectivity is more pronounced in insects and small-bodied tetrapods [36,41–48] although it is also seen in some large mammals [49–52]. Browsing herbivorous mammals tend to be more selective than grazers [53–55] and generally larger species are less selective than smaller [36,47]. These two patterns are perhaps interrelated; larger herbivores can survive on plant matter of lower nutrient quality, due to lower energy requirements relative to their body size: the Jarman–Bell principle [53,55,56]. As such, a greater presence of large-bodied herbivores reduces the dominance of plant species and increases species richness [36,42], whereas an increased presence of smaller, more selective species leads to reduced species richness by targeting more palatable taxa, as demonstrated by selectively excluding herbivores of varying sizes from sites [36,47,48].

We find that the early Permian diversification of herbivorous tetrapods constrained the α-diversity of plants for the rest of the Palaeozoic. However, the variation in herbivore body size has a greater impact, and by itself is responsible for more than 30% of the variance in plant α-diversity. Leaving aside the uncertain diet of the pantylid microsaurs, the Carboniferous tetrapod herbivores appear to have originated at, or rapidly evolved to, relatively large sizes [57,58]. The fall in plant diversity across the Carboniferous–Permian boundary, while mirroring the increase in herbivore diversity at this time, also coincides with the origin of smaller, presumably more selective herbivores such as the bolosaurids [59], and a lowering of minimum body size relative to the Carboniferous. The influence of herbivore body size variance on plant diversity found by the GLS analysis might simply reflect this same effect of the radiation of small herbivores. Alternatively, it may reflect the fact that size variance is an indicator for the diversity of functional groups [60,61]; the presence of a greater ecological diversity of herbivores places greater predation pressure on plants.

The impact of arthropod herbivory on plant diversity is difficult to ascertain. Patterns of terrestrial arthropod diversity, in particular of herbivorous arthropods, are uncertain, due in part to difficulty assessing the diet of extinct forms. The terrestrial arthropod record, particularly in the Permian, is heavily biased by the rock record and dominated by isolated wings [62,63]. Even in specimens preserving mouthparts, diet is not always clear [64]. Herbivorous arthropods underwent a decrease in diversity of functional feeding groups at the beginning of the Permian (assessed by types of damage to plant fossils [64]), potentially indicating a relaxation of predation pressure from arthropods on plants. Diversity of arthropod herbivory types similar to that of the Carboniferous is not again seen until the Middle Triassic [65].

The relative contributions of the origin of arthropod and vertebrate herbivores to plant evolution is clearly a complicated issue. It is possible that the origin of vertebrate herbivory had a greater impact on plant diversification than arthropod herbivory. Studies on modern patterns of herbivory suggest that vertebrates are responsible for most of the attacks on seedlings, at which stage the attack is more often fatal to the plant [66]. Arthropods cause more damage to adult plants, but these are using fewer resources for growth and more for defence and tolerance [66,67]. These inferences remain uncertain and require thorough reviews of the biases affecting preservation of feeding damage on plants.

Past studies of Palaeozoic plant diversity have attributed a drop in global (rather than local as tested here) species richness in the latest Carboniferous to the climatic changes occurring at this time [68–70], in particular a ‘rainforest collapse’: the shift from everwet ‘coal-swamp’ environments characteristic of the Carboniferous the more open summerwet environments characteristic of the equatorial Permian. This climatic shift could account in part for the drop in plant α-diversity observed here independently of the evolution of herbivorous tetrapods; while tetrapod evolution is found to be responsible for the majority of the observed variance in plant diversity, it does not explain all. However, this possibility is not consistent with timing of palaeofloral changes based on recent examinations of the transition, which appears to have been largely complete by the end of the Moscovian (307 Mya, preceding the end of the Carboniferous by 7 Mya) [70]. Examinations of temperature and carbon dioxide levels suggest that these were relatively stable, with no substantial changes from the end of the Mississippian until the late Sakmarian [71,72]. Although a shift in climate might have caused turnover of plant lineages, it cannot explain why plant diversity remained low throughout the Permian. In fact, particularly when considering diversity at small spatial scales, examinations of modern ecosystems suggest that summerwet biomes may have greater plant richness than everwet biomes [73,74]. Although these observations of the extant flora might not be directly applicable to the radically different floras of the Palaeozoic, it does make it less likely that the shift in environment was primarily responsible for the persistence of a low-diversity flora, rather than the radiation of smaller bodied herbivores.

5. Conclusion

The earliest diversification of vertebrate herbivores across the Carboniferous–Permian boundary coincides with a substantial drop in plant diversity. Although it is difficult to infer direct macroevolutionary interactions between clades, our findings suggest that plant richness was to some extent structured by vertebrate herbivory from its earliest origins more than 300 Mya. Studies of modern ecosystems suggest that this should be the case, but this has generally been overlooked in previous examinations of plant diversity. This study provides an illustration of the potential for fossil data to test predictions of ecological interactions first observed in extant ecosystems.

Data accessibility

All data are included in the electronic supplementary material.

Authors' contributions

N.B. conceived the study; N.B. and C.F.K. collected the data; N.B. analysed the data; N.B., C.F.K. and R.J.B. wrote the paper.

Competing interests

We declare we have no competing interests.

Funding

N.B.'s research is funded by Deutsche Forschungsgemeinschaft grant no. BR 5724/1-1. C.F.K.'s research is funded by Deutsche Forschungsgemeinschaft grant no. KA 4133/1-1.