Leaf fossil record suggests limited influence of atmospheric CO₂ on terrestrial productivity prior to angiosperm evolution

Abstract

Declining CO₂ over the Cretaceous has been suggested as an evolutionary driver of the high leaf vein densities (7–28 mm mm⁻²) that are unique to the angiosperms throughout all of Earth history. Photosynthetic modeling indicated the link between high vein density and productivity documented in the modern low-CO₂ regime would be lost as CO₂ concentrations increased but also implied that plants with very low vein densities (less than 3 mm mm⁻²) should experience substantial disadvantages with high CO₂. Thus, the hypothesized relationship between CO₂ and plant evolution can be tested through analysis of the concurrent histories of alternative lineages, because an extrinsic driver like atmospheric CO₂ should affect all plants and not just the flowering plants. No such relationship is seen. Regardless of CO₂ concentrations, low vein densities are equally common among nonangiosperms throughout history and common enough to include forest canopies and not just obligate shade species that will always be of limited productivity. Modeling results can be reconciled with the fossil record if maximum assimilation rates of nonflowering plants are capped well below those of flowering plants, capturing biochemical and physiological differences that would be consistent with extant plants but previously unrecognized in the fossil record. Although previous photosynthetic modeling suggested that productivity would double or triple with each Phanerozoic transition from low to high CO₂, productivity changes are likely to have been limited before a substantial increase accompanying the evolution of flowering plants.

Role of CO₂ in Plant Evolution and Productivity over Geological Time Scales

Atmospheric CO₂ concentrations have fluctuated greatly over the past 400 million years: CO₂ levels are thought to have decreased by a factor of 10 or more as a result of the Devonian evolution of deep rooting plants and, subsequently, to have varied between levels somewhat less than and fivefold greater than preindustrial levels (1, 2). Just as land plant evolution has been a primary driver of these changes, in turn, CO₂ has often been assigned a dominant role in plant evolution. CO₂ changes have been implicated for the radiation of vascular plants, seed plants, and flowering plants; for the spread of C₄ photosynthesis and grasslands; and for the evolution of arborescence and both laminate leaves in general and the high vein density leaves of angiosperms in particular (3–12). Furthermore, models of plant function have repeatedly predicted that large swings in terrestrial productivity of 200–300% accompany these changes in atmospheric CO₂ (13–16).

Plants require CO₂, and productivity can be adversely affected by the CO₂ minima of the recent geological past that have approached the CO₂ compensation point for plant growth (17, 18), but did the positive relationship with CO₂ carry to a doubling or tripling of current productivity during the Mesozoic highs in CO₂ concentration? This suggestion raises a number of complications. For example, it is in conflict with a variety of hypotheses regarding the angiosperm radiation that are dependent on high productivity being unique to the flowering plants and not a general trait of all plants in a high-CO₂ world (19–22). Furthermore, if productivity were highly CO₂-dependent, that dependence would carry through even to the availability of different architectural and ecological possibilities. For example, occupation of frequently disturbed habitats requires high enough productivity to complete a life cycle between successive disturbance events (23); thus, the accessibility of disturbed habitats as a viable environment would have fluctuated tremendously through time for all plant lineages. Finally, if the correlation between CO₂ concentrations and productivity is tightly and strongly positive, that would even have an impact on our understanding of atmospheric CO₂ itself, because the standard modeling of vegetation, atmospheric, and geological processes to determine CO₂ concentrations includes only modest increases in productivity as CO₂ increases. The temporal variability of CO₂ would be dampened if CO₂ fertilization effects were greater because of the feedbacks of the additional plant productivity on CO₂ drawdown through increased silicate weathering (figure 10 in ref. 24). Paradoxically, the models of plant function suggesting that large swings in productivity accompany the large changes in CO₂ concentrations would also make those large changes in CO₂ difficult to achieve because of these feedbacks.

Models of plant function in the geological past are difficult to test experimentally because modern plants have evolved under low-CO₂ conditions and greenhouse tests at elevated CO₂ levels cannot be carried out over macroevolutionary time scales (15), but a solution may be available for the testing of their accuracy and completeness by comparing predictions against the fossil record. Here, we use leaf fossils to test morphological predictions derived from models of plant function under elevated CO₂ regimes to evaluate previous hypotheses regarding the role of CO₂ fluctuations in driving angiosperm evolution and primary productivity.

Leaf Vein Density

The high density of veins found in the leaves of many flowering plants is unique. Nonangiosperms average about 2.5 mm of vein length per square millimeter of leaf area and rarely reach higher than 5 mm mm⁻², but angiosperms average around 10 mm mm⁻² and can reach higher than 25 mm mm⁻², with high vein densities appearing independently in at least three angiosperm lineages: the magnoliids, monocots, and eudicots (14, 25, 26). This abundance of vasculature was shown to correlate with much higher physiological activity, including a fourfold increase in transpiration capacity and more than a doubling of assimilation capacity, that resulted in important implications both for tropical climate and vegetation feedbacks and for the rise to ecological dominance of angiosperms over the Cretaceous and early Paleocene (14, 27–29). However, those correlations of leaf vein density with transpiration and assimilation capacities are based on empirical measurements of living plants grown under the low ambient CO₂ concentrations of the modern world (14, 25, 30).

Modeling of leaf physiology at the 1,000-ppm CO₂ concentrations that are believed to have been exceeded during the Cretaceous suggests that nearly all plants would have been much more productive than they are today (13–15) and that the advantage of increased vein density would saturate at low vein density levels that are accessible by all plant groups (14) (results reproduced in Fig. 1B). Thus, it has been argued that the Cretaceous initiation of an overall trend in atmospheric CO₂ reduction toward modern levels was essential for the emergence of angiosperm leaf characteristics because high vein density provides no advantage with high CO₂ (14).

Fig. 1.

Expected relationships between vein density and assimilation capacity and fossil leaf vein density distributions relative to CO₂ concentrations. (A) Comparison of empirical measurements (circles) of relation between leaf vein density and stomatal conductance [digitized from the study by Brodribb and Feild (14)], with alternative models relating vein density and assimilation capacities, including a previously used model with fixed leaf thickness (14) and a previously undescribed corrected model allowing thickness to vary with vein density (36). (B) Modeled effects of vein density on assimilation capacities at high (1,000 ppm) and low (350 ppm) atmospheric CO₂ concentrations for uncorrected and vein density/leaf thickness-corrected models. (C) Photosynthetic rate gain for change in vein density at high and low CO₂ calculated with the corrected model. (D) Distribution of vein densities at different times in Earth history plotted at GEOCARBIII (24) and GEOCARBSULF (55) estimates of contemporaneous CO₂. Myr, million years. Specific data are provided in Table S1.

Emphasis has been on the lack of advantage for high vein density in a high-CO₂ world (14), but perhaps the more important conclusion to draw from this physiological model is that possession of leaf vein densities much lower than 3 mm mm⁻² would present enormous disadvantages because it is associated with dramatic loss of CO₂ assimilation capabilities. Although the modern low-CO₂ world can be permissive of a broad morphological diversity because only moderate gains in assimilation rate accompany increasing vein density, this should not be expected in a world of 1,000 ppm CO₂, where plants with a vein density of 2 mm mm⁻² were modeled to assimilate almost twice as fast as plants with a vein density of 1 mm mm⁻² (figure S4 in ref. 14). This modeling of plant function in high CO₂ thus suggests that an absolute difference in assimilation capacities as large as that between modern ferns and angiosperm tropical trees and crops would be compressed between two vein densities that would both be exceedingly low by modern angiosperms. This range of very low vein densities is equally achievable by any lineage of vascular plants and in any environment: High vein densities may require ample water supply, but low vein densities are available anywhere from the rainforest understory to desert succulents.

Testing the hypothesized relationship between CO₂ and angiosperm leaf evolution is complicated by high leaf vein density being unique to the angiosperms; however, all other plants should respond similarly within their potential range even if they lack the developmental and physiological capacities for the very high vein densities exhibited by angiosperms. Thus, at times of high CO₂ when very low vein densities should be maladaptive, few fossil leaves should have had vein densities less than 3 mm mm⁻² regardless of phylogenetic affinity.

Results and Discussion

Fossil History of Leaf Vein Densities.

When leaf fossils are sampled from high-CO₂ intervals, the prediction of decreasing prevalence of low vein densities below 3 mm mm⁻² proves to be false (Table S1; full ANOVA statistical results are provided in SI Text). Vein densities change little through time, and high proportions of nonangiosperms have vein densities lower than this threshold throughout Earth history and over a wide range of CO₂ concentrations (Fig. 1D). This abundance of low vein density leaves is too great to be accounted for by plants that are independently constrained to low productivity by deep shade or other environmental stressors. First, the vein densities in question are low regardless of environment: Shade-requiring angiosperms can have vein densities greater than 5 mm mm⁻² (26). More generally, the paleobotanical record is biased toward preservation of woody plants of forest canopies and more open environments because herbs and shrubs of the shaded forest floor produce fewer leaves and their leaves have less opportunity for transport to waterway-based sites of deposition and are more prone to decay (31, 32). The cases in which all plants are preserved equally are rare (33, 34). Thus, a few of the lower bound outliers of vein density may well represent tolerators of deep shade, but the median vein density for each time interval is sometimes less than 2 mm mm⁻² and is never more than 2.6 mm mm⁻² for each high-CO₂ time interval (Fig. 1D). Thus, if more than 50% of the taxa have very low vein densities, it can be safely inferred that a high proportion of the canopy plants had such vein densities. Furthermore, the bias toward preservation of plants growing along lakes and rivers that has been problematic for leaf-based paleoclimate proxies (35) does not explain the prevalence of low vein density plants through time because riparian and nonriparian plants have similar vein densities (Fig. S1 and Table S2). Our results indicate that plants with very low densities grew side by side with plants of higher vein density regardless of CO₂ concentrations, contrary to the prediction that those plants would be subject to enormous disadvantages when atmospheric CO₂ was high.

Reconciling Physiology and the Fossil Record.

The prediction derived from physiological modeling that very low leaf vein densities should be rare during times of elevated CO₂ is not met by the fossil record, and no trivial correction is available to bring these two approaches into concordance (full details on model corrections are provided in SI Text). First, the original model (14) only allowed for very thin leaves, which is adequate only for high vein densities. The distance between veins should roughly equal the distance from vein to stomata for optimal water delivery (36), such that low vein density plants in exposed environments would require much thicker leaves to avoid partial desiccation of the leaf lamina. A more accurate scaling of leaf thickness to vein density provides a better fit to experimental data (Fig. 1A), but this more accurate depiction increases the distance of high-resistance hydraulic transport through the mesophyll in low vein density leaves, and thereby accentuates their potential disadvantages (Fig. 1 B and C). Second, although poorly constrained, stomatal conductivity should have been lower with the reduced stomatal density seen at times of high CO₂ in the geological past (37–40). Reducing stomatal conductivity does increase the vein density at which productivity plateaus but without altering the overall conclusion that the penalty of low vein density is far larger when CO₂ is high (Fig. 2). Third, the O₂/CO₂ ratio can have an impact on maximum photosynthetic rates through photorespiration, but elevated oxygen does not alter expectations (Fig. 2), and neither does consideration of a more complete range of potential CO₂ concentrations (Fig. 3). The expectation would remain that very low vein densities should be strongly disfavored during high-CO₂ regimes because even small increases in the vein density would dramatically improve photosynthetic capacity (Fig. 3), an expectation rejected by the fossil record even though these small changes would be equally available to all plant lineages.

Fig. 2.

Impact on assimilation of stomatal conductance and atmospheric oxygen. (A) Changes in predicted assimilation rate in relation to vein density caused by introducing decreased stomatal conductance in response to high-CO₂ concentrations (16) and/or changes attributable to the elevated O₂ concentrations expected for the Cretaceous (52, 55) applied to the model corrected for leaf thickness. (B) Photosynthetic rate gain for change in vein density as predicted for inclusion of model corrections for stomatal conductance and O₂ concentrations. g_s, stomatal conductance to water

Fig. 3.

Model expectations for the relationship between productivity and the full range of potential leaf vein densities and CO₂ concentrations. Modern O₂ concentrations, as well as corrections to the original model for variable leaf thickness and stomatal conductance, were used.

The steep disadvantages to very low vein density suggested by photosynthetic models derive from the assumption that given enough CO₂, any fern exposed to adequate light could be more productive than angiosperm crops, such as sunflower or maize, are now with modern CO₂ concentrations (e.g., figure S4 in ref. 14). Only lower maximum assimilation rates for nonangiosperms could explain the lack of a correlation between vein density and CO₂ concentration: A lower plateau for assimilation capacities would also lower the vein density at which that plateau is achieved, potentially encompassing the entire range of available vein densities (Fig. 4 A and B). Vein density would then be free to vary in response to environmental and ecological specialization with little impact on assimilation capacities.

Fig. 4.

Impact of maximum rate of Rubisco carboxylation, V_cmax, on assimilation capacity at 1,000 ppm of CO₂. (A) For the fully corrected model, changes in predicted assimilation rate in relation to vein density are shown for different values of V_cmax (μmol⋅m⁻²⋅s⁻¹). (B) Photosynthetic rate gain with changing vein density for different values of V_cmax. (C) Empirical measurement of lineage-specific relationships between vein density and assimilation at 1,000 ppm of CO₂ in living ferns, gymnosperms, and angiosperms plotted along with modeled relationships between vein density and assimilation capacity at two V_cmax levels. A_max, light-saturated photosynthesis. A plant species list is provided in Table S3.

In this scenario, angiosperms would have advantages over other plant groups regardless of CO₂ concentration, as would be consistent with other differences in angiosperm vegetative biology, including leaf traits and stomatal conductivity through time (16, 25, 41–43). In the modern low-CO₂ world, individual nonangiosperms, particularly some conifers, can overlap with less productive flowering plants, but both the typical and maximum photosynthetic capacities of angiosperms are substantially higher (42–45). Although it is always unclear how literally such experiments should be extrapolated up to macroevolutionary time scales, angiosperm advantages do persist and substantial lineage-specific differences in CO₂ fertilization are seen when modern plants are exposed to 1,000 ppm CO₂ (Fig. 4C).

Further support for physiological distinctions between angiosperms and other plants is that angiosperm leaf fossils do conform to the model expectations that nonangiosperms violate. The earliest angiosperms share the same range of very low vein densities as nonangiosperms; however, as they evolve the higher vein densities indicative of higher photosynthetic capacities over the high-CO₂ Late Cretaceous, they also vacate the vein densities less than 3 mm mm⁻² in a way that is not seen among nonangiosperms (figure 1 in ref. 26), where very low vein densities account for almost 50% of angiosperms in the pre-Albian Early Cretaceous, 7% in the Albian, and no more than 4% in the Late Cretaceous and Paleocene. This provides additional evidence that the patterns seen among nonangiosperm leaf fossils cannot be dismissed as an artifact of the perpetual existence of plants environmentally limited to low productivity. Any such ecological caveat should have applied equally to the angiosperm record, particularly because obligate shade plants are both ancestral and continuously abundant for angiosperms (46).

Although modeled in terms of limitations on the maximum rate of carboxylation, actual lineage-specific limitations on maximum photosynthetic capacity could take any number of forms. Rubisco activity is well known as a CO₂-dependent limit on photosynthesis that suggests highly elevated productivity as CO₂ increases (5, 15, 47). However, other limitations provide more opportunity for lineage-specific differences that are much less CO₂-dependent, such as investment in photoreceptors or the Calvin cycle regeneration of RuBP (47, 48), or may actually be exacerbated by increasing CO₂, such as the nitrate assimilation (49) and inorganic phosphate recycling that are also necessary for primary production (48, 50). Without constraining which factors may be involved, the stability of vein density distributions throughout the fossil record outside of the angiosperms suggests substantially lower photosynthetic capacities in most nonangiosperms regardless of CO₂ concentrations. Rather than productivity potentially ranging from 10 to more than 40 μmol⋅m⁻²s⁻¹ through time (13, 15), no more than the bottom half of that range may have been available before angiosperm evolution. Previous workers have been careful to acknowledge that those values are maxima that may not be reached because of other limitations, such as water availability. However, those elevated levels of productivity are reached by flowering plants in many extant environments, including rainforests and more temperate environments where moisture is adequate (14), and we argue no plant could achieve those levels seen in the modern world before the advent of angiosperms regardless of CO₂ concentrations.

Plant Productivity Through Time and Environmental Implications.

If most nonangiosperms do indeed have limited capacity for assimilation much greater than their modern levels, the large swings in productivity through time predicted by modeling vegetation responses to CO₂ are unlikely to have happened, except for a potential increase over the Cretaceous and Cenozoic with the spread of angiosperms. This would be more consistent with carbon cycle modeling than the doubling or tripling of assimilation rates that has been predicted with elevated CO₂ and would have been likely to prevent elevated CO₂ in the first place through silicate weathering feedbacks. This hypothesis of relative stability for plant productivity through vascular plant history at least up to the Cretaceous would be an important grounding for consideration of the evolution of terrestrial animals and ecosystems. Contradictory views of how productivity has changed since the Cretaceous exist in the literature (8, 13–16, 19–22, 25). Have flowering plants brought flourishing abundance that also benefited animal life (perhaps even in the oceans), or are angiosperms simply the best at enduring the deprivations of a diminished, carbon-starved world? Consideration of photosynthetic models in light of the fossil record suggests that productivity has not declined with angiosperm evolution and is likely to have increased.

The potential of increased productivity with angiosperm evolution is bolstered by a final discrepancy between model output and plant morphology: Although flowering plant leaf vein densities average around 10 mm mm⁻² and can surpass 25 mm mm⁻², photosynthetic models suggest there is never an advantage to densities greater than about 8 mm mm⁻² regardless of CO₂ concentrations. This expectation is based on the modeling of maximum instantaneous assimilation rates dependent on the maximum flux of CO₂ through the stomata when they are open. Along with other aspects of angiosperm hydraulic physiology, such as vessels and reduced safety margins (51), high vein densities greater than 10 mm mm⁻² may instead be important for maximizing the proportion of stomata that are not closed. Thus, the expectation is further bolstered that angiosperms have productivity advantages over other plants whether CO₂ is high or low.

Future tests regarding changes in productivity through time with CO₂ fluctuations and angiosperm evolution may be challenging but could come from a variety of directions. Increased photosynthetic capacities in angiosperms may not have had much impact on the carbon isotopic values of plants: For a given stomatal conductivity, the lower internal CO₂ concentrations that might be expected to accompany higher assimilation rates would decrease isotopic discrimination, but internal CO₂ concentrations might not have been lower if angiosperm hydraulic modifications allowed stomata to be open a greater proportion of the time. Increased Cretaceous abundance of charcoal has been argued to reflect increased angiosperm assimilation rates and fuel production (20), although over longer time periods, charcoal abundance has been considered more as an indicator of atmospheric oxygen levels than of plant productivity (20, 52, 53). Tests resulting from plant paleoecology may be the most informative. Of all plant growth forms, annuals are the most dependent on high growth rates. If productivity was tightly correlated with CO₂ concentrations, this might be expected to carry over to the abundance and diversity of annual plants through time as well. Annuals are effectively absent outside of the angiosperms throughout earth history (54), regardless of CO₂ concentrations.

Materials and Methods

All vein density measurements were performed with ImageJ (National Institutes of Health). Extant leaf vein density measurements were performed on a minimum of five leaves for each taxon. Fossil leaf vein density was measured from individual figured specimens or compiled from previously published studies (details are provided in Table S1). Angiosperm vein density measurements were based on three subsamples of each leaf at a magnification of 20× to 40×; however, this is not necessary for the thick veins and lower vein densities of nonflowering plants, for which the vein density of the entire lamina was measured. The photosynthetic model was derived from a previous source (14) with some corrections of parameter values as explained in SI Materials and Methods. Empirical measurements of photosynthetic rates in living plants were collected using a photosynthesis measuring system (LiCor 6400; with photosynthetically active radiation = 2,000 μE and CO₂ concentration = 1,000 ppm; leaves were allowed 20 min of acclimation time in the chamber before measurements).