Physiological selectivity and plant–environment feedbacks during Middle and Late Pennsylvanian plant community transitions Open Access

Supplementary material: Information about fossils used in this study and supplementary quantitative analysis are available at https://doi.org/10.6084/m9.figshare.c.6392423

Land plants, which lose water from their tissues to fix carbon dioxide (CO₂) into organic carbon, releasing molecular oxygen (O₂) in the process, are key homeostatic agents in the interface between the biosphere and the geosphere. Terrestrial plants respond to environmental change (e.g. changing CO₂ concentrations, rainfall patterns, etc.) but also drive environmental change, because they participate in key geobiological cycles, including the carbon, oxygen and water cycles. Although nearly all land plants share the same core physiological toolkit, such as evapotranspiration, carbon dioxide assimilation and photosynthesis using the enzyme Rubisco, different plants conduct these physiological operations at different rates and magnitudes (Taiz and Zeiger 2006). Therefore, a change in composition of, or relative abundance within, terrestrial plant environments may drive environmental change, which may, in turn, feed back on the plant community itself. One result of this plant–environment feedback over geological timescales is the convergent evolution of stable solutions to environmental problems (Niklas 1994, 1997), such as the multiple origins of laminar leaves, the evolution of complex root systems and multiple independent origins of vascular adaptations. The resulting ‘physiological adaptive landscape’ means that rapid environmental changes have differential effects on plants within an ecosystem: some plants may survive whereas others may not.

The Late Paleozoic Era is a crucible of plant evolution and documents substantial change in plant community composition, biodiversity and anatomy. The Devonian Period closes with forests of spore-bearing plants, including Archaeopteris, the first modern tree (Beck 1962; Scheckler 1978; Meyer-Berthaud et al. 2000), but by the end of the Permian Period forests are dominated by seed plants, including relatives of the stem groups leading to modern groups of gymnosperms (e.g. conifers, cycads) and, perhaps, even the stem groups leading to the angiosperms (Doyle 2006). Throughout the Late Paleozoic, plants expand into different biomes, develop new modes of life, including the seed habit, and record significant changes in global environments.

One singular event during this time is the floral transition at the Middle–Late Pennsylvanian Boundary (MLPB), which has been well documented in the palynoflora and macroflora from the equatorial and tropical deposits in the Illinois Basin (Phillips et al. 1974; DiMichele and Philips 1996). The transition is notable because it records a transition from forests dominated by arborescent lycophytes (‘lycopsids’) and their replacements as dominants or subdominants by stem group marattialean tree ferns (e.g. the genus Psaronius). Other changes take place, including a decrease in the abundance of the stem group seed plant Medullosa (DiMichele and Philips 1996), and recent work has shown that the MLPB transition is one particularly notable transition in a time period notable for several intercalating floral changes whose causes remain elusive (Phillips et al. 1974; DiMichele and DeMaris 1987; DiMichele and Aronson 1992; DiMichele and Phillips 1994, 1996; DiMichele et al. 2008, 2010; Falcon-Lang et al. 2009; Falcon-Lang and DiMichele 2010; Tabor et al. 2013; van Hoof et al. 2013; DiMichele 2014; Montañez 2016).

Physiological selectivity has become a powerful tool for gaining new insights into marine mass extinctions and biodiversity turnover events (Bambach 2006; Bush et al. 2007; Knoll et al. 2007). In these cases, examining the physiological capabilities of organisms that were dramatically affected by mass extinction events, and contrasting those capabilities with organisms that did not suffer high biodiversity loss, can illuminate potential kill mechanisms during biotic turnover. This tool has been infrequently applied to terrestrial events, although plants’ conserved physiological capabilities make insights easier to obtain, and these capabilities can be measured in a quantitative way (Wilson et al. 2017).

Water transport is a particularly useful lens to examine the physiology of extinct plants. In vascular plants, structural and functional tradeoffs are a consequence of biophysical properties of plant anatomy (Tyree and Zimmermann 2002; Sperry 2003; Venturas et al. 2017), which can be measured in the fossil record (Niklas 1985; Cichan 1986a; Roth-Nebelsick et al. 2000; Roth-Nebelsick 2001; Roth-Nebelsick and Konrad 2003; Wilson et al. 2008, 2015, 2017; Pittermann 2010; Wilson and Knoll 2010; Wilson 2013). The most important tradeoff is ‘safety v. efficiency’ (Wheeler et al. 2005; Hacke et al. 2006, 2007; Pittermann et al. 2010, 2011): structural adaptations that increase hydraulic conductivity through the xylem conflict with those features that reduce the frequency of cavitation and embolism, water-stress-induced damage that can fill the xylem with air and halt water transport. The xylem cell morphological correlates of this tradeoff are clear. Wider and more porous cells allow more water volume to be transported with less hydraulic resistance, at the cost of a higher risk of nucleating cavitation and allowing an embolism to spread, whereas narrower tracheids yield lower hydraulic conductivity but provide greater resistance to cavitation and embolism, thus increasing the plant's safety margin (Tyree and Sperry 1988, 1989; Sperry and Tyree 1990; Tyree and Ewers 1991; Pittermann and Sperry 2006; Pittermann et al. 2006a, b, 2011). Key factors therefore include the xylem cell morphology, particularly its diameter, and the abundance of porous areas on the later walls of the xylem conduits, called pits, through which water passes from one xylem cell to another. Previous work has shown substantial variation in the relative balance between conductivity and safety among extinct plants (Cichan 1986a; Wilson et al. 2008, 2015, 2017; Wilson and Knoll 2010; Wilson and Fischer 2011b; Wilson 2013, 2016), but a systematic investigation of Pennsylvanian plants in the context of the documented floral transitions has not been performed. We close this gap here.

In this study, we provide new data on Pennsylvanian plant xylem, including more than 2000 measurements of tracheids and details of pit structure. We use quantitative analysis of fossil xylem, derived directly from scanning electron and light microscopy of fossilized anatomy, to test the hypothesis that the key driver of the MLPB floral transition was drought. Using fluid dynamics modelling and statistical analysis of xylem cell anatomy, we assess the stem hydraulic capacity and drought resistance of five key Carboniferous Period plants, as representative of the major vegetation groups present at the time: Lepidophloios, a lycophyte; Medullosa, a stem group seed plant; Sphenophyllum and Arthropitys, morphologically disparate representatives of the horsetail clade; Psaronius, a stem group marattialean tree fern; and Cordaites, a stem group coniferophyte. We combine this analysis with the same quantitative analysis of hydraulic capacity and drought resistance applied to extrabasinal and dryland plants of the Carboniferous Period, as representatives of the flora that succeeds the Carboniferous wetland forest.

The equatorial tropical forests of the Pennsylvanian Subperiod contained a diversity of land plants, including seed plants and spore-bearing plants. These well-known biomes expanded during part of the glacial periods during the Pennsylvanian, where they covered vast areas of the continents, and were repeatedly drowned by rising sea-levels. In these ecosystems, five key taxonomic groups form much of the vegetation in these landscapes. They are the arborescent lycopods (lycopsids), medullosans, calamitaleans, cordaitaleans and tree ferns. Each of these taxonomic groups contains morphological, anatomical and taxonomic diversity, and has been well studied for decades (Taylor et al. 2009): for example, several different species of medullosans vary in their stem anatomy, vascular arrangement (Andrews 1940, 1945; Andrews and Mamay 1953; Delevoryas 1955; Pfefferkorn et al. 1984) and leaf size (Laveine 1986, 1997). This study investigates representative forms of each of these five key taxonomic groups to establish a baseline for comparison with future investigations (Fig. 1).

Recent palaeobotanical work has established the presence of unusual plants outside of the wetland tropical basins at this time. Many of these plants are coniferophytes, some related to the cordaitaleans but many others related to the walchian coniferophytes, and it has been known for some time that these plants progressively came to dominate the lower-latitude basins throughout the later portion of the Pennsylvanian, rising to full dominance in the more arid landscapes of North America during the Permian Period (Hernandez-Castillo et al. 2001, 2003, 2009a, b; Falcon-Lang and Bashforth 2004, 2005; Falcon-Lang et al. 2009, 2014, 2016; Falcon-Lang and DiMichele 2010; Bashforth et al. 2014; DiMichele 2014; Looy and Hotton 2014; Montañez 2016). Therefore, these plants are appropriate to study as representatives of lineages that persist and diversify through, and after, the MLPB, into the Permian Period. In this study, they are collectively described as ‘extrabasinal’ plants. The following sections will outline key information about the plant groups that are the focus of this study.

The best-known group of Carboniferous plants, with a distinctive anatomy and morphology, are the arborescent representatives of the lycophyte clade, known as the arborescent or rhizomorphic lycopsids (Eggert 1961; DiMichele 1979a, b, 1981; DiMichele and Bateman 1996). Several distinct genera can be distinguished from the fossil record, including Lepidodendron, Sigillaria and Lepidophloios, and they form unique deposits throughout the Carboniferous Period (DiMichele and DeMaris 1987; DiMichele et al. 2009; DiMichele and Falcon-Lang 2011). The arborescent lycopsids produced a relatively small amount of secondary xylem in their stems, which were covered by single-veined leaves and leaf bases with stomata, atop large, dichotomizing root systems, assigned to the form genus Stigmaria (Rothwell and Pryor 1991; Hetherington et al. 2016). Arborescent lycopsids were the dominant plant in the wetland, tropical, coal swamps before the MLPB, according to the macrofossil and palynological record (Phillips et al. 1974; DiMichele and Philips 1996; DiMichele 2014). The reduction in diversity and abundance of arborescent lycopsids in the lowland tropical forests after the MLPB is a notable feature of the floral transitions in this period.

The medullosans are a morphologically distinct clade of stem group seed plants with unusual anatomy and morphology (Andrews 1940, 1945; Andrews and Kernen 1946; Stewart and Delevoryas 1952, 1956; Andrews and Mamay 1953; Stidd 1981). Medullosan stems contain anomalous development of their vascular cambium, which results in each stem containing anastomosing vascular bundles of varying number and size within a single stem (Delevoryas 1955; Basinger et al. 1974). These bundles also contain wide and long tracheids, in excess of 150 µm in diameter and more than 20 mm in length (Andrews 1940; Cichan 1986a; Wilson et al. 2008; Wilson and Knoll 2010). Plants from the genus Medullosa grew in a variety of habits, from self-supporting trees to leaning, or even climbing, forms (Pfefferkorn et al. 1984; Wnuk and Pfefferkorn 1984; Mosbrugger 1990; Wilson and Fischer 2011a). Branch and leaf systems of individual medullosan plants were immense (Pfefferkorn et al. 1984; Wnuk and Pfefferkorn 1984): fronds more than 7 m long have been reported (Laveine 1986). Medullosans survive into the Permian Period (Roberts and Barghoorn 1952), but their biodiversity decreases across the MLPB, and there is some turnover within the group, according to the stratigraphic record of their reproductive organs (DiMichele and Philips 1996).

Calamites are seed-free plants with diverse forms, all related to the extant horsetail genus Equisetum (Taylor et al. 2009). A diversity of calamites habits exist (Spatz et al. 1998; Pfefferkorn et al. 2001; DiMichele et al. 2009; Mencl et al. 2013). It is notable that there are arborescent and scandent/climbing forms within the group, or even within the same genus: for example, Arthropitys kansana is a self-supporting tree, but Arthropitys deltoides has been described as a climbing plant (Cichan and Taylor 1983). Two genera are investigated in this study, the climbing or scandent Sphenophyllum (Baxter 1948; Schabilion and Baxter 1971; Good and Taylor 1972; Eggert and Gaunt 1973; Good 1973; Batenburg 1982; Cichan and Taylor 1982; Cichan 1985) whose vasculature has received previous attention (Cichan 1985, 1986a; Wilson 2013) because of its extraordinary xylem cell size (Fig. 1b), and the arborescent Arthropitys kansana and Arthropitys communis, which are distinguished from one another on the basis of secondary xylem pitting (Andrews 1952). Arborescent calamites persist through the MLPB into the Permian Period, whereas Sphenophyllum plants decrease in abundance and, eventually, appear to become extirpated (locally extinct) from the lowland basins.

The cordaitaleans are a group of stem group coniferophyte seed plants with strap-shaped leaves and dense, pycnoxylic wood (Florin 1950, 1951; Costanza 1985; Cichan 1986c; Taylor et al. 2009). A variety of forms exist, from smaller, scrambling forms, to large trees (Rothwell 1993). Cordaitalean trees reached more than 30 m in certain extrabasinal zones and inhabited a diverse set of environments outside of the wetland tropics (Falcon-Lang and Scott 2000; Falcon-Lang and Bashforth 2004, 2005). All cordaitalean stems contain a eustele of secondary xylem containing tracheids with circular-bordered pits (Fig. 1a); other than this distinctive pit morphology, cordaitalean xylem cells are very similar to secondary xylem from extant conifers. Cordaitaleans were a conspicuous part of Pennsylvanian lowland tropical ecosystems and persisted into the Permian Period (Taylor et al. 2009).

Pennsylvanian tree ferns are best represented by Psaronius, a genus of stem group marattialean tree ferns (DiMichele and Phillips 2002). Psaronius morphology is distinctive among Carboniferous plants: the stem is surrounded by a large root mantle, and it appears that each frond emergence was paired with aerial root emergence, even far above ground (Dawson 1871). Tree ferns like Psaronius are abundant before the MLPB, but after the event, tree ferns rose to become the dominant plant of most tropical biomes (Phillips et al. 1974; DiMichele and Philips 1996; Baker and DiMichele 1997). A consequence of this unique morphology is that a large fraction of the cross-sectional area of a Psaronius trunk is root mantle, which surrounds a comparatively small stem, even far above the ground (Fig. 1c).

In addition to the extrabasinal cordaitaleans described above, a diverse set of stem group coniferophytes have been identified from these environments as well, including the genera Giblingodendron, Thucydia, Macdonaldodendron and Emporia (Hernandez-Castillo et al. 2001, 2003, 2009a, b; Falcon-Lang and Bashforth 2004, 2005; Falcon-Lang et al. 2009, 2014, 2016; Falcon-Lang and DiMichele 2010; Bashforth et al. 2014; DiMichele 2014; Looy and Hotton 2014). Although complete specimens of each of these plants are not known, their internal stem anatomy includes a eustele containing pycnoxylic wood, whose tracheids contain uniseriate and biseriate circular-bordered pits (Hernandez-Castillo et al. 2001, 2009a; Falcon-Lang et al. 2014, 2016). These plants are representative of the coniferophytes that increase in diversity and abundance through the latest Pennsylvanian and come to dominate many ecosystems during the Permian. Five species are considered here: Giblingodendron nudifolia, Giblingodendron aridus, Macdonaldodendron giganticus, Thucydia mahoningensis and Emporia royalii.

Fossils used in this study were obtained from the National Museum of Natural History collection of the Smithsonian Institution, the Harvard University Paleobotanical Collection and the Holden Arboretum (Fig. 1; see also Supplementary Table S1). All specimens are permineralized stems in coal balls (carbonate concretions) that are Late Pennsylvanian in age: Arthropitys, Psaronius, Lepidophloios and Sphenophyllum specimens are from West Mineral, Kansas; the two Cordaites samples are from Lewis Creek and Cawood, Kentucky; and the Medullosa specimen is from Waukee, Iowa (the Shuler Mine). This paper follows the logic of Baxter (1948) and Schabilion and Baxter (1971) that each of the Sphenophyllum samples may be assigned to Sphenophyllum plurifoliatum. All fossils were polished and scanned at high resolution using an Epson desktop scanner before maceration.

Fossil stems or roots were disaggregated from coal ball matrix using a rock saw, hammer and chisel. Separated stems were immersed in a macerating solution of 10% hydrochloric acid (HCl) for 48–96 hours. During the maceration process, material was kept as still as possible, to avoid mechanical damage to tracheids. Macerated material was rinsed with deionized water a minimum of five times and pipetted onto microscope slides for light microscopy investigation.

Living plant tissue material (i.e. Equisetum species) was macerated using the technique described by Ruzin (1999): immersion in a 1:4:5 solution of 30% hydrogen peroxide–deionized water–glacial acetic acid. Slivers of Equisetum stems were immersed in 50 ml of this solution and incubated in a water bath at 56°C for two days. After plant stems had been decolourized and broken down, material was rinsed in deionized water five times, sectioned using a razor blade and stained with a 1:5 dilution of the metachromatic dye toluidine blue O (TBO, 0.1% solution, Fisher). Stained segments were photographed using light microscopy and analysed using standard measurement techniques (see below).

Material for scanning electron microscopy was gold coated in a sputter coating chamber and imaged using an environmental scanning electron microscope, a JEOL 6510 LV, at Haverford College.

Images of tracheids that were captured through cameras attached to light microscopes, from images published in the literature or through scanning electron microscopes, were analysed using the open-source software package ImageJ (https://imagej.nih.gov/ij/; Schneider et al. 2012).

For rectangular tracheids, lumen diameter was calculated by converting lumen cross-sectional area to the diameter of an equivalent circle, as in previous work (e.g. Wilson et al. 2020). Stems exhibiting obvious evidence of post-depositional mechanical distortion, including crushed or torn vascular segments, were excluded from this analysis. Pit membrane area per tracheid was measured by measuring membrane area directly; for tracheids with mixtures of circular-bordered pits and larger, elliptical, nearly scalariform pits (only found within Arthropitys; Fig. 2b), pit area was expressed as a fraction of radial wall area.

Tracheids were modelled according to their pitting type: individual models were constructed for each genus of plant with circular-bordered pits (e.g. Medullosa, Cordaites, Arthropitys (CBP), Giblingodendron, etc.), and separate models were constructed for each genus of plant with scalariform pits (e.g. Lepidophloios, Arthropitys (scalariform), etc.). Parameters used for hydraulic conductivity calculations, including conduit diameter and pit area per millimetre tracheid length, are shown in Tables 1 to 4. Because tracheid length is the most difficult parameter to measure in fossils, coniferophytes (e.g. Cordaites and the extrabasinal plants) and Arthropitys were assigned a length of 4 mm, which accords within the range of conifer tracheids reported in the literature and reported for calamitaleans, respectively (Bannan 1965; Cichan and Taylor 1983, 1984; Cichan 1986a, b). Tracheid lengths for Psaronius and Lepidophloios were estimated to be 100× longer than their diameter, approximating length–width ratios found within many vascular plants and extinct plants, and values for Sphenophyllum and Medullosa reflect measurements reported in the literature (Bailey and Tupper 1918; Andrews 1940; Bannan 1965; Cichan and Taylor 1984; Cichan 1985; Wilson et al. 2008; Wilson and Knoll 2010; Wilson 2016).

Anatomical values used to calculate pit membrane area per tracheid are shown in Table 4.

Multiple factors influence the mean cavitation pressure in living plants, including conduit diameter and length, intertracheid pitting type, interconduit pit area, xylem strand connectivity and topology, wood density and other properties of the plant vascular system, which are often a consequence of the specific xylem cell anatomy within a particular plant lineage (Comstock and Sperry 2000; Hacke and Sperry 2001; Hacke et al. 2001, 2004, 2006, 2007; Lancashire and Ennos 2002; Sperry and Hacke 2004; Pittermann et al. 2005, 2006a, b, 2011, 2013; Sperry et al. 2005, 2006, 2007, 2008; Wheeler et al. 2005; Christman et al. 2009, 2012; Brodersen et al. 2014). One of the strongest correlations found is that between interconduit pit area and mean cavitation pressure, which has been observed in seed plants and seedless vascular plants, from vessel-bearing angiosperms to ferns that rely on primary xylem (Wheeler et al. 2005; Hacke et al. 2007; Christman et al. 2009; Brodersen et al. 2014). Because of the unusual tracheid morphological dimensions and pit structure found within Pennsylvanian plants – Sphenophyllum and Medullosa contain tracheids that are wider than most vessels, and the same tracheids are longer than nearly all tracheids from living plants; consequently, the interconduit pit area per tracheid is high – the only correlation that covers the range of pit areas observed in this study, and therefore has statistical resolution, is that observed within vessel-bearing eudicots. Applying this relationship to Pennsylvanian plants is a conservative modelling strategy: as others have noted, angiosperm vessels have less porous, and therefore more cavitation-resistant, pit membranes (e.g. Hacke et al. 2007). It is therefore possible that applying this correlation to the extreme morphologies found in Sphenophyllum and Medullosa may overestimate their cavitation resistance.

Imaging, and image analysis, of tracheids from Carboniferous plants shows substantial variability between the five major groups of wetland plants. Tracheid size, expressed as lumen diameter, varies by more than an order of magnitude from the largest group to the smallest group (Table 1). Four Sphenophyllum plurifoliatum stems contain tracheids with mean diameters exceeding 200 µm in diameter: 235 ± 38, 210 ± 49, 226 ± 37 and 232 ± 39 µm. Medullosa noei contains the next widest tracheids (156 ± 8.7 µm), followed by Lepidophloios (62 ± 34 µm), Psaronius stems (50 ± 16 µm) and Arthropitys communis (48 ± 13 µm). Cordaites, Psaronius roots and Arthropitys kansana contain substantially narrower tracheids. Notably, even the smallest Arthropitys tracheids are wider than those found within four extant Equisetum species, including the largest living horsetail species (Equisetum giganteum and Equisetum myriochaetum; see Supplementary Fig. S1).

Variability of pitting, even within a single specimen, is demonstrated, particularly within the arborescent sphenophytes Arthropitys kansana and Arthropitys communis (Fig. 2): these specimens contain tracheids with circular-bordered pitting (Fig. 2a–c, e), scalariform pitting (Fig. 2d, e) and tracheids that contain extremely elliptical bordered pits, resembling an intermediate form (Fig. 2b). More uniform pitting is seen in other plants. Psaronius tracheids contain scalariform pitting in most of their tracheids (Fig. 3), except in one root specimen that has nearly circular pits. Cordaites tracheids contain uniseriate, biseriate and multiseriate pitting (Fig. 4), Medullosa noei tracheids contain multiseriate circular-bordered pits (Fig. 5) and Lepidophloios tracheids contain scalariform pits (Fig. 6c–e). Tracheids of Sphenophyllum are regrettably fragmentary because of their wide diameter (Table 1), but some contain scalariform pitting (Fig. 6f, g): these may also be interpreted as circular-bordered pits that developed into an elliptical shape (see fig. 2 in Cichan 1985). Differences may be observed between stem and root tracheids in Psaronius: stem tracheids contain wider areas of pit membrane (Fig. 3), in contrast to tracheids isolated from Psaronius root mantle (Fig. 6a, b).

Image analysis of these tracheids yields substantial differences in pit area per tracheid between, and among, the five major groups of wetland plants (Table 2). Sphenophyllum plurifoliatum and Medullosa noei tracheids contain the most pit membrane area per millimetre tracheid length (0.2 and 0.7–0.16 mm², respectively), followed by Lepidophloios (0.04 mm²). Psaronius and Cordaites contain less pit membrane area per millimetre tracheid length, followed by Arthropitys kansana and Arthropitys communis.

When specimens of the five Carboniferous taxa are compared with extrabasinal plants, substantial differences between their tracheid size and pit membrane area can be noted (Table 3). The extrabasinal plant with the widest tracheids, Macdonaldodendron giganticus (44 ± 9 µm), contains tracheids that are narrower than Sphenophyllum, Medullosa noei, Lepidophloios, Psaronius stems and Arthropitys communis. The other four extrabasinal plants – Giblingodendron nudifolia, Giblingodendron aridus, Thucydia mahoningensis and Emporia royalii – contain even narrower tracheids, reaching a minimum of 9.5 ± 2.4 µm in Emporia royalii. Pit membrane area per millimetre tracheid is similarly low in these plants, with only Giblingodendron aridus falling within the range of the five wetland taxa; all three other extrabasinal plants whose pit frequency could be measured (Macdonaldodendron giganticus, Giblingodendron nudifolia and Thucydia mahoningensis) contain less pit membrane area than all wetland plants.

In summary, there are notable differences in conduit width, pitting type, pit dimensions, pit frequency and pit membrane area per tracheid between the wetland and extrabasinal plants. Sphenophyllum, Medullosa noei and Lepidophloios stand apart from the remaining plants in their conduits’ breadth, pit frequency and pit membrane area per tracheid.

The functional consequences of these plants’ tracheid dimensions, pitting type, pit dimensions, pit frequency and pit membrane area per tracheid are vast (Figs 7 & 8). Hydraulic conductivity modelling of Sphenophyllum plurifoliatum and Medullosa noei shows high hydraulic conductivity for single tracheids: these values are a consequence of the enormous width of their tracheids and the high volume of pit area per tracheid. The extreme tracheid lengths (>20 mm) observed in these plants also contribute to their high hydraulic conductivity: they are more than six times longer than typical tracheid lengths in conifers (Bannan 1965) and more than double the tracheid length found in ferns (Pittermann et al. 2011). Sphenophyllum and Medullosa appear to contain the highest-conductivity tracheids heretofore found and modelled in the fossil record. Lepidodendron and Psaronius stem tracheids also contain higher hydraulic conductivity than stem group gymnosperms, and both plants’ conductivity values are a consequence of their relatively broad cells and pit area. In contrast to Sphenophyllum, the tracheids within Arthropitys – whether they contain scalariform or circular-bordered pits – have a lower hydraulic conductivity, and their range marginally exceeds or overlaps the values from Cordaites and other extrabasinal coniferophytes, except for Thucydia mahoningensis and the narrow, less-pitted tracheids from Psaronius root mantle.

The high hydraulic conductivity of Sphenophyllum, Medullosa noei and Lepidophloios comes at a cost: they have reduced resistance to water-stress-induced cavitation (Fig. 8; Table 4). Sphenophyllum and Medullosa noei, in particular, have very low estimated values of mean cavitation pressure: −0.8 and −0.9 MPa, respectively. These barely negative values imply restriction to permanently drought-free environments, or frequent cavitation during drought events if those conditions were present in their habitat, and, if so, this suggests that these plants likely contained a physiological mechanism to repair embolisms. Lepidophloios tracheids, also, likely cavitated at moderate levels of water stress, with an estimated mean cavitation pressure of −2.7 MPa. Psaronius, Cordaites, Arthropitys and the four extrabasinal plants contain moderate to extremely high mean cavitation pressures, suggesting a strong ability to resist damage from water stress.

When plants are placed in an ecospace that combines the two dimensions of conductivity and cavitation resistance, three ecophysiological strategies emerge (Fig. 8). One cluster demarcates plants with high hydraulic conductivity and low cavitation resistance, denoted by Sphenophyllum, Medullosa noei and Lepidophloios. A cluster with moderate conductivity and moderate cavitation resistance is formed by Psaronius, Arthropitys communis with scalariform pitting, Macdonaldodendron giganticus and Cordaites. It is notable within this cluster that the stem tracheids of Psaronius are more highly conductive, and less cavitation resistant, than its root tracheids – a pattern that is distinct from most vascular plants (see below). The third cluster, consisting of tracheids with low hydraulic conductivity and high cavitation resistance, contains Arthropitys kansana (both circular-bordered pitting and scalariform pitting tracheids), Arthropitys communis with circular-bordered pits and the three extrabasinal plants Thucydia mahoningensis, Giblingodendron nudifolia and Giblingodendron aridus. Comparing the coniferophyte lineages in this ecospace is instructive: despite the similarity between the tracheid dimensions in Cordaites and the extrabasinal plants, the differences in pit size, pit frequency and overall pit membrane area result in substantially increased water-stress-induced cavitation resistance in the extrabasinal plants, compared with Cordaites. Although living coniferophyte xylem cell size and pitting variability has been studied extensively (Bannan 1965; Pittermann et al. 2006a, b, 2010, 2011), there has been little quantitative study of pitting variability among extinct plants, particularly Paleozoic plants, using multiple specimens and, therefore, the differences documented here should be studied further. Detailed investigation of extinct plant xylem is, therefore, critically important to distinguishing subtle differences in plant palaeoecophysiology and consequences for the environment as a whole (Matthaeus et al. 2022).

Within the Pennsylvanian terrestrial landscape, several lineages of plants developed xylem that could support high hydraulic conductivity, including Sphenophyllum, Medullosa and Lepidophloios. Stems of the tree fern Psaronius could also support higher hydraulic conductivity than values found within Psaronius roots and the cordaitaleans and coniferophytes. Within the latter two groups, however, tracheids with biseriate and multiseriate pitting (e.g. within Cordaites sp. sample 6034F) have higher hydraulic conductivity as a consequence of this increased pit membrane area – a pattern observed within other land plants (Hacke et al. 2006, 2007; Pittermann et al. 2006a, b, 2010, 2011; Wilson et al. 2008; Hacke and Jansen 2009; Wilson and Knoll 2010; Wilson and Fischer 2011b; Wilson 2013, 2016).

Increased hydraulic conductivity comes at a cost: the same plants with high hydraulic efficiency (Sphenophyllum, Medullosa and Lepidophloios) had the lowest thresholds for mean cavitation pressure. Indeed, estimated mean cavitation pressures for Sphenophyllum and Medullosa are comparable to those found in grapevine (Vitis vinifera) and kudzu (Pueraria montana), two vessel-bearing, climbing angiosperms (Hacke et al. 2006). The great length of the tracheids in Sphenophyllum and Medullosa may have increased their cavitation vulnerability further: longer conduits with abundant pit connections between them are highly efficient at spreading embolisms through stems (Loepfe et al. 2007), and these plants’ hydraulic architecture was therefore embolism-prone. The estimated mean cavitation pressure of Lepidophloios is comparable to sourwood (Oxydendron arboreum) and box elder (Acer negundo) based on the amount of pit area in each tracheid (Hacke et al. 2006). Sourwood and box elder are found in different environments today – ranging from well-drained slopes and ridges to other subxeric environments, to moist, riparian and palustrine habitats, respectively – and therefore pursue different ecophysiological strategies. Further study could illuminate whether Lepidophloios resembles sourwood or box elder, although the presence of Lepidophloios in mesic, palustrine environments suggests the latter. In contrast to Sphenophyllum, Medullosa and Lepidophloios, then, other plants were safer: coniferophytes with small pits in a uniseriate arrangement, such as Giblingodendron nudifolia, have extremely negative mean cavitation pressures, suggesting substantial ability to resist damage from water-stress-induced embolism. The extrabasinal plants tend to contain xylem that is safer from water-stress damage at the expense of reduced conductivity.

Vascular plants within the Middle and Late Pennsylvanian ecosystems exhibited a diversity of ecophysiological strategies and compromises, from high-conductivity stems with low safety margins to lower-conductivity stems with greater ability to resist damage. These tradeoffs have been observed in other vascular plants throughout geological time (Sperry and Tyree 1990; Sperry 2003; Pittermann et al. 2006a, b, 2010, 2011; Hacke et al. 2007; Pittermann 2010; Wilson and Knoll 2010; Wilson 2013, 2016). One unusual pattern, not heretofore observed, is within the tree fern Psaronius: in this plant, the tracheids within its root system are narrower and less pitted than those found within the stem; that is, Psaronius root tracheids have a higher safety margin than stem tracheids from the same sample. This is the opposite of the pattern usually seen in land plants, in which root tracheids are frequently broader and more porous than stem tracheids (Bannan 1965; Alder et al. 1996; Sperry and Ikeda 1997; Hacke et al. 2004; Sperry and Hacke 2004) and are consequently more vulnerable to cavitation and embolism. The ‘inverse’ hydraulic pattern of Psaronius is likely a consequence of its unusual morphology, specifically the aerial nature of its root mantle: tracheids in roots are more directly exposed to the atmosphere and therefore more vulnerable; thus, their narrow and less-porous nature serves as a protective function to reduce the introduction of embolism into the whole plant's vascular system. Closer examination of other extant ferns that produce root mantles (e.g. Dicksonia antarctica) may reveal that this pattern is not unique to Psaronius, and functional consequences of this pattern may contribute to patterns of ecophysiological variability observed in living seedless vascular plants (Watkins et al. 2010; Pittermann et al. 2011, 2013; Holmlund et al. 2016).

Examination of the pattern of extinction, extirpation and/or biodiversity reduction among the Pennsylvanian tropical ecosystems provides some insights into environmental mechanisms that shaped these diversity patterns. First, the loss of the arborescent lycopsids and reduction among medullosans can be interpreted as an extirpation of high-conductivity, low-safety-margin stem plants across the boundary (Phillips et al. 1974; DiMichele and Philips 1996); it is likely that this is, at least in part, a consequence of increased water stress, among other environmental and life history factors (e.g. the effect of drought on the lycopsid reproductive cycle, which required water for sexual reproduction; although this would also affect other spore-bearing plants). In contrast, plants with the capability to resist damage from extreme water stress (e.g. drought and/or frost), including those that evolved outside of the tropics, survived and expanded their ranges, including tree ferns, cordaitaleans and coniferophytes (DiMichele and Aronson 1992; DiMichele and Phillips 2002; Falcon-Lang et al. 2009; Falcon-Lang and DiMichele 2010; DiMichele and Falcon-Lang 2011; Stull et al. 2012; DiMichele 2014). Within this broader pattern – represented as end-members of the ecophysiological spectrum here – differences would be expected, even among phylogenetic groups. For example, arborescent calamitaleans, such as Arthropitys kansana, would be expected to survive increasing aridity, whereas the high-conductivity but vulnerable xylem of Sphenophyllum would put it at significant risk of lethal water stress, except in locations where precipitation and water availability remained high. As aridification intensified and expanded through the Late Pennsylvanian and into the Permian, these regional and biome-level changes to the hydrologic cycle would have contributed to the repeated shifts observed in the equatorial forests (DiMichele 2014).

Further tests of the extinction selectivity hypothesis are possible by examining the functional consequences of within-lineage anatomical variation. For example, the palaeoenvironmental distribution of the arborescent lycopsid Sigillaria implies that its xylem should be more drought-resistant than xylem from Lepidophloios or Lepidodendron. Furthermore, medullosan or Sphenophyllum species that persist into the more seasonal or arid environments of the Permian Period (Roberts and Barghoorn 1952; Stull et al. 2012; Luthardt et al. 2021) should exhibit xylem containing narrower tracheids, fewer pits or other adaptations to increase resistance to drought stress when compared with their wetland Pennsylvanian counterparts. The interactions between hydraulic conductivity, abiotic environmental parameters (e.g. rainfall seasonality), cavitation resistance and biome-scale transpiration have received recent examination (White et al. 2020; Matthaeus et al. 2021, 2022). These results have demonstrated that the composition of mixed ecosystems and the individual plant capacities and limits found within these ecosystems have a determinative effect on the impact of these biomes’ geobiological fluxes. For example, a biome with high-transpiration plants (e.g. medullosans, Sphenophyllum) may support high transpiration rates when water is abundant on the landscape, but have a significantly lower transpiration rate when rainfall decreases, because of cavitation and embolism among these vulnerable plants. Conversely, an ecosystem with lower-transpiration plants (e.g. Psaronius) may maintain low-to-moderate transpiration rates through dry periods (Matthaeus et al. 2022). Closer examination of previously collected plant stems throughout the Pennsylvanian, and a detailed census of palaeovegetation abundance and diversity on the landscape-scale, could complement the higher-resolution picture of a dynamic landscape that is emerging from a variety of interdisciplinary perspectives (DiMichele 2014; Montañez et al. 2016).

From an evolutionary perspective, several adaptations present in Late Pennsylvanian forests would shape the later course of plant evolution and have important consequences for coevolution of terrestrial ecosystems and the environment. The pycnoxylic wood observed in the cordaitalean and other coniferophyte lineages is a significant adaptation, despite its low hydraulic conductivity on a single-cell basis, because it confers increased resistance to water-stress-induced damage from low water availability (or high temperatures) and low temperatures (e.g. frost damage) at the same time (Pittermann et al. 2005, 2006a, b, 2011; Pittermann 2010; Wilson and Knoll 2010; Brodribb et al. 2012; Wilson 2013; Matthaeus et al. 2021). Plants with pycnoxylic wood would have resisted both ends of the climate extremes in the Pennsylvanian and been positioned to radiate as atmospheric CO₂ concentrations decreased and aridity expanded into the Permian (Matthaeus et al. 2021). At the same time, the consequences of the loss of high-conductivity and vulnerable taxa on environmental feedbacks could be large and could amplify the initial environmental change that caused the major vegetation shift. Extirpation of plants with high hydraulic conductivity, leaf area and evapotranspiration rates would have increased the rate at which rainfall entered surface hydrologic systems (rather than vegetation) and accelerated runoff rates, reducing overall soil moisture available for plants to draw upon and transpire, further aggravating the initial drought (Montañez et al. 2016; Wilson et al. 2017; White et al. 2020; Matthaeus et al. 2021). This positive feedback – loss of high-conductivity plants exacerbating the water availability or lack of humidity problem, which causes further stress to drought-sensitive plants – could have contributed to the breakdown of the Pennsylvanian vegetation-sustained tropical rainforest. In effect, this phenomenon would have been the reverse of the terrestrialization process observed during the Late Silurian and Early Devonian: at that time, plant colonization on unvegetated surfaces and transpiration from early plants would have changed surface albedo and the hydrologic cycle, respectively, enabling vegetation to expand on the landscape. Mirroring this process, the reduction and eventual extirpation of plants whose unique anatomy and physiology supported this ecosystem would have amplified the floral transitions observed beginning in the Pennsylvanian and continuing into the Permian, and shaped terrestrial ecosystems during this critical period in plant evolution (DiMichele and Aronson 1992; DiMichele et al. 2001, 2008; DiMichele and Gastaldo 2008).

The Late Paleozoic Era is a crucible of plant evolution: interaction and coevolution between plants and the environment shaped the course of subsequent terrestrial ecosystems. The floral transitions during the Late Pennsylvanian and early Permian, including the MLPB, are notable for their effect on plant communities. Examination of the xylem ecophysiology of plants whose biodiversity decreased during these events, when compared with the ecophysiology of those plants that survived or increased, indicates that plants in the first group (including Sphenophyllum, Medullosa and Lepidophloios) tended to contain high-conductivity and low-safety-margin xylem, making them vulnerable to water-stress-induced damage. Other plants in the same ecosystem – including the stem group marattialean tree fern Psaronius, stem group coniferophytes and the arborescent sphenopsid Arthropitys, a close relative of Sphenophyllum – contained lower-conductivity, higher-safety-margin vasculature, and survived these events. Taken together, these results suggest a role for different hydraulic adaptations contributing to terrestrial ecosystem extinction selectivity. Further investigation could elucidate additional within-lineage differential selections and illuminate the role of Paleozoic plant physiology as a response to, and potential amplifier of, the breakdown of Pennsylvanian tropical forests.

We are grateful for materials provided by Juliana Medeiros (Holden Arboretum), Bill DiMichele (Smithsonian Institution) and Suzanne Costanza (Harvard University) and thank them for their support. We thank Luke Troyon (Haverford College) for assistance with sample preparation and scanning electron microscopy and Remmy Chen and Deana Rauh (Haverford College) for related palaeobotanical research. We thank two anonymous reviewers for helpful and constructive reviews. We are particularly grateful to Joseph White, Will Matthaeus, Isabel Montañez, Jennifer McElwain, Bill DiMichele and Jarmila Pittermann for helpful discussions throughout the duration of this project.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

JPW: conceptualization (lead), data curation (lead), formal analysis (lead), investigation (lead), methodology (lead), project administration (lead), supervision (lead), visualization (lead), writing – original draft (lead), writing – review & editing (lead); GO: data curation (supporting), investigation (supporting), visualization (supporting); ER: data curation (supporting), investigation (supporting), visualization (supporting); JS: data curation (supporting), investigation (supporting), visualization (supporting); CM: data curation (supporting), investigation (supporting), visualization (supporting); BK: data curation (supporting), investigation (supporting), visualization (supporting).

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

The datasets generated during and/or analysed during the current study, and higher-resolution images shown in the figures, are available from the corresponding author on reasonable request.