Reconstructing relative genome size of vascular plants through geological time

Early Devonian plants have exceptionally large stomata

The range of GCL among the earliest Devonian (c. 410 Ma) plants (22–86 μm) encompasses almost the full range of GCL observed in extant plants (Beaulieu et al., 2008) and the present work (15–100 μm). Given the strong correlation between GCL and GS, these data suggest that a wide range in GS emerged during the primary diversification phase of vascular land plants. Several species from the Rhynie Chert (c. 410 Ma) have a large GCL (Horneophyton lignieri, 86 μm, Aglaophyton major, 82 μm, Nothia aphylla, 67 μm, Asteroxylon mackiei, 73 μm and Rhynia gwynne-vaughanii, 63 μm). The architecturally similar extant species Psilotum nudum has very large GCL (99 μm: Pant & Khare, 1971) and a correspondingly large GS (2C = 142 152 Mbp: Obermayer et al., 2002). The large GCL of Psilotum nudum (and of the closely related extant ‘fern’ Ophioglossum petiolatum, measured at 61 μm: Peterson & Hambleton, 1978; 2C 128 216 Mbp: Obermayer et al., 2002) compares well with the Early Devonian species, although the Devonian plants are only distantly related to Psilotum and Ophioglossum, and they occupied radically different, hot-spring habitats. When combined with our understanding of extant plants, these fossil data imply that many Early Devonian vascular plants were characterised by large genomes. However, given the early-divergent position that these Devonian plants occupy in embryophyte phylogeny, it is likely that at least some of the mechanisms that link GCL with GS were less constrained by canalisation and/or followed contrasting pathways of epidermal development (Rudall et al., 2013).

Extant plants with large genomes are absent from extreme environments and under-represented in disturbed environments (Knight et al., 2005). However, this does not seem to be the case for some of the Early Devonian plants. Horneophyton lignieri and Rhynia gwynne-vaughanii are characterised as primary colonisers of ever-wet and drier ground, respectively. Aglaophyton major is interpreted as being adapted to periodic flooding and waterlogging, whereas Nothia aphylla and Asteroxylon mackiei are regarded as successional plants rather than primary colonisers (Powell et al., 2000). This indicates that Early Devonian plants occupied a variety of niches associated with hot springs. This environment may have been equable in the short-term (seasonal) but in the longer-term (generational) it undoubtedly represents an extreme and dynamic, disturbance-prone environment. The relationship between ecology and genome size suggests that, in angiosperms at least, this environment would in theory be dominated by species possessing small genomes (Knight et al., 2005). However, the fossil data imply that this relationship does not hold through geological time or across all phylogenetic groups. Unfortunately, comparison with modern hot-spring floras is weakened by radical taxonomic differences.

Gaps and mass extinctions in the fossil record

There are two notable stratigraphic gaps within our database (Fig. 2a). These span the Late Devonian to Mississippian (c. 380–330 Ma), and a second shorter gap in the earliest Triassic (250–236 Ma). The paucity of Late Devonian GCL data can be ascribed to a lack of wetland biotas, whereas Mississippian data are scarce due to an ongoing lack of systematic investigations (Scott & Galtier, 1996). The Early Triassic gap is driven by climate, widespread aridity favouring oxidation and the loss of plant material before it can be preserved. Both of these gaps in our database are problematic as they coincide with major evolutionary and climatic events. The Late Devonian gap is associated with a large-scale draw-down in CO₂ (Berner & Kothavala, 2001; Berner, 2006), the end-Devonian extinction event (c. 360 Ma; primarily a feature of the marine fossil record: Bond & Wignall, 2008), and the origin of true megaphyllous leaves (Kenrick & Crane, 1997; Osborne et al., 2004; Galtier, 2010). The Early Triassic gap represents the aftermath of the end-Permian (c. 250 Ma) mass extinction event (e.g. Hallam & Wignall, 1997). Consequently, plant genome-size responses to these important global events remain elusive.

In order to determine whether patterns of change in GCL, and therefore in inferred GS, could be associated with any of the other mass extinction events (asterisked in Fig. 2b), we calculated the running average in GCL through 25 Myr time-bins (Fig. 2b). Since the earliest of these extinction events, the end-Ordovician (443 Ma), occurred before the earliest well-documented diversification of vascular plants, this event was not analysed further. As for the Triassic/Jurassic (200 Ma) or the Cretaceous/Tertiary (K/T: 65 Ma) extinction events, no long-term impacts in GS could be identified from the available data, despite the fact that both events left excellent plant fossil records. For example, cuticle from macrofossils is readily recoverable from the Triassic/Jurassic boundary sections of Greenland (McElwain et al., 1999), and suites of dispersed cuticle assemblages are available from the Cretaceous/Tertiary boundary of the western continental interior of the USA (Wolfe & Upchurch, 1987a; Upchurch, 1995).

The lack of any clear change in GCL (and hence GS) associated with the K/T extinction event is perhaps surprising, given that a recent phylogenetic investigation by Fawcett et al. (2009) highlighted a statistically significant increase in the frequency of angiosperm polyploids immediately following the K/T boundary that would be predicted to lead to larger genomes. Clearly, denser sampling of the well-preserved fossil record on either side of the K/T boundary would provide a more rigorous test of any perceived association.

Stomatal density (SD) and stomatal index (SI) analysis through both boundary sequences has determined that both mass-extinction events are accompanied by a large increase in pCO₂ (Triassic/Jurassic boundary, McElwain et al., 1999; Steinthorsdottir et al., 2011; K/T boundary, Beerling et al., 2002). Increases in atmospheric pCO₂ typically result in a decrease in SD and SI (e.g. Woodward, 1987). Also, GCL increases as stomatal density declines, resulting in the plant having fewer but larger stomata (Hetherington & Woodward, 2003). This negative correlation implies that plants with larger GCL, and hence larger GS values, might be expected to become more frequent during angiosperm re-colonisation and recovery, given the high CO₂ of the post-impact environment of the early Paleogene (c. 65 Ma; Lomax et al., 2000; Beerling et al., 2002). Thus, the relationships between stomatal density, GCL and CO₂ may explain why some plants with large genome sizes, including perhaps early Paleogene polyploids, may have preferentially prospered in the early Paleogene high-CO₂ world (Fig. 2a).

Extrinsic influences: changes through time in atmospheric CO₂

Two key palaeopolyploidy events have recently been placed within, and dated using, broad-brush molecular phylogenies of land plants (Jiao et al., 2011) (summarised in Fig. 2c). The earlier event (c. 350 Ma) coincides approximately with the diversification of major groups of gymnosperms, and the later event (c. 200 Ma) is purportedly contemporaneous with the initial diversification of extant angiosperm groups that preceded the separation of the monocots and eudicots. Both palaeopolyploidy events appear to coincide with relatively high, yet declining, atmospheric CO₂ concentration (Fig. 2c,d). However, using the fossil record rather than molecular dating techniques, the earlier event actually corresponds with the initial radiation of seed plants in the Late Devonian and Mississippian, while the latter is contemporaneous with the Triassic–Jurassic boundary and hence significantly pre-dates the earliest morphologically recognisable angiosperms (e.g. Bateman et al., 2006). To explore the relationship between GCL/GS and CO₂ we compared our GCL trend with the GEOCARB III model (Berner & Kothavala, 2001; Fig. 2d). Like the taxonomically broad analysis of Franks et al. (2012a), we see a positive trend between pCO₂ and GCL/GS within our species-level investigation. The broad correlation between CO₂ and climate is well illustrated by the Permo-Carboniferous glaciation (c. 330–255 Ma), which corresponds with a period of low CO₂ (Mora et al., 1996), and the Cretaceous period, when warm poles and low latitudinal temperature gradients (e.g. Wolfe & Upchurch, 1987b; Spicer & Hermann, 2010) coincided with high CO₂. Evidently, additional broad-scale climatic variables may influence plant GS through physiological and ecological processes.

Many tracheophytes have recorded episodes of palaeopolyploidy within their genome (e.g. Jiao et al., 2011). However, analyses of neopolyploid plants have also shown that the average speciation rate of neopolyploids may be significantly lower than that of diploids whereas their extinction rates are significantly higher (Mayrose et al., 2011). These results may seem counter-intuitive, but Mayrose et al. (2011) suggested that this dichotomy is a result of successful palaeopolyploids having an expanded genomic potential that drives their longer-term evolutionary success. Plants with fewer, larger stomata (polyploids) are typically less efficient in terms of stomatal responses to environmental stimuli (e.g. Hetherington & Woodward, 2003), which may then result in concomitant changes in water-use efficiency and photosynthetic performance – these factors may in turn eventually impact on overall plant fitness. Most studies investigating the relationship between GCL and perceived plant fitness have focused on comparisons between diploid and polyploid angiosperms. However, the precise nature of these relationships remains contentious, given that results differ between treatment and species. For example, polyploids can be more (Garbutt & Bazzaz, 1983; Watanabe, 1986; Li et al., 1996) or less (Baldwin, 1941) tolerant of water stress when compared with their diploid progenitors. Similarly dichotomous results are observed for other abiotic factors that are recorded in the literature (reviewed by Maherali et al., 2009). Such studies highlight the importance of understanding the role played by polyploidy in the crucial relationship between speciation and extinction (Arrigo & Barker, 2012).

Given elevated pCO₂, there is typically a reduction in stomatal conductance as a result of increasing stomatal closure (Drake et al., 1997; Ainsworth & Long, 2005; and references therein). Elevated atmospheric CO₂ may therefore provide an alternative explanation for the apparent dichotomy of selective extinction of neopolyploids noted above (Mayrose et al., 2011). In the modern atmosphere relatively deprived of CO₂, increases in GCL associated with increases in ploidy may select against neopolyploids. By contrast, events responsible for palaeopolyploidy observed today in extant lineages would have occurred in the CO₂-enriched atmospheres of the past, when stomatal closure via elevated CO₂ may have reduced this potential selective pressure. However, this intriguing hypothesis requires testing in experimental conditions, where polyploid plants can be compared with their diploid progenitors over a range of CO₂ concentrations relevant to the evolutionary history of vascular land plants.

Effect of CO₂ on stomatal size

Allium cepa showed a statistically significant 11% decrease in GCL whereas A. tuberosum showed no statistical significant difference in GCL (ambient vs elevated CO₂ 800 ppm). These new experimental data partially support many previous studies that cover a range of taxa (angiosperms and gymnosperms) and genome sizes. For example, Lomax et al. (2009) demonstrated that GCL values were independent of CO₂ concentration in Arabidopsis thaliana (2C = 157 Mbp). Ryle & Stanley (1992), working on Lolium perenne (2C = 5390 Mbp), found no statistical difference in GCL with CO₂ (340 vs 680 ppm). Radoglou & Jarvis (1990) and Ceulemans et al. (1995) found no statistical difference in GCL with altered CO₂ in poplar clones. Knapp et al. (1994) examined Andropogon gerdarii (2C = 8839 Mbp) and Sorghastrum nutans and found no statistical difference in GCL with CO₂ treatment (ambient vs double ambient) in either species. Ogaya et al.'s (2011) study of ‘living fossils’ found that GCL did not change in Araucaria araucana (2C = 44 499 Mbp), Sequoia sempervirens (2C = 62 856 Mbp) or Metasequoia glyptostroboides (2C = 21 594 Mbp). But Taxodium distichum (2C = 19 462 Mbp) showed a statistically significant 8% decrease in GCL and Nothofagus cunninghamii showed a statistically significant 10% increase in GCL between the CO₂ treatments (400 vs 800 ppm). Haworth et al. (2013) examined the stomatal parameters of eight species with both passive (Osmunda regalis, 2C = 27 393 Mbp; Ginkgo biloba, 2C = 22 983 Mbp; Podocarpus macrophyllus, 2C = 18 973 Mbp; and Agathis australis, 2C = 30 905 Mbp) and active (Lepidozamia peroffskyana, 2C = 54 083 Mbp; Nageia nagi, 2C = 10 954 Mbp; Solanum lycopersicum; and Hordeum vulgare, 2C = 10 855 Mbp) stomatal control and found no statistical difference in stomatal pore length (equivalent to GCL) with pCO₂ (380 vs 1500 ppm).

However, Franks et al. (2012b) showed that in four phylogenetically disparate species (Selaginella uncinata, 2C = 176 Mbp; Osmunda regalis; Commelina communis; Vicia faba, 2C = 26 064 Mbp) stomatal size was significantly affected by CO₂; stomatal size increased with elevated pCO₂ (240, 450 and 1000 ppm). These authors also reported an apparent change in genome size, but this was attributed to a change in packaging of the DNA leading to altered stoichiometry in the binding of the fluorochrome to the DNA rather than to an actual increase in the amount of DNA (Franks et al., 2012b).

Plotting the range in GCL observed in Allium cepa and A. tuberosum due to growth environment onto the herbaceous dataset of Beaulieu et al. (2008) shows that changes in GCL for the two Allium species are greater than the changes observed in Arabidopsis (Fig. 3b), despite A. thaliana being subjected to a wider range of environmental perturbations (atmospheric CO₂ 400–3000 ppm, drought, relative humidity, irradiance, ultraviolet radiation and pathogen attack; Lomax et al., 2009). These data suggest that plants with a larger GS may have a higher degree of flexibility with respect to scaling GCL to GS when compared with plants possessing a smaller genome. Furthermore, these observations emphasise the complexity of the relationship between GCL and CO₂ and indicate the existence of both responders (both increasing and decreasing in GCL with elevation of pCO₂) and nonresponders, analogous to the stomatal density and index response to altered CO₂ concentrations (Woodward & Kelly, 1995; Royer, 2001). However, this variation in response does not appear to relate to any one variable, including GS, phylogenetic history or the mechanism of stomatal control. These observations imply not only that broad-scale palaeo-GS reconstructions based on limited datasets (e.g. Franks et al., 2012a) might be overly simplistic, but also that the relationship between CO₂, GCL and GS requires further investigation.

Intrinsic influences on genome size

Angiosperm-focused research shows that two processes are particularly common in genome evolution – polyploidy and the accumulation of noncoding repetitive DNA sequences, especially transposable elements, in the genome. Evidence of their widespread occurrence in angiosperm genomes suggests that the null hypothesis of genome size evolution should be a trend of increasing genome size through time (Bennetzen & Kellogg, 1997). Certainly, polyploidy can lead to an increase in cellular DNA content and a concomitant increase in GCL; for example, within the Coffea arabica polyploid series there is a 43% increase in GCL from diploid through tetraploid to the octoploid cytotype (Mishra, 1997). Similarly, diploid species that differ in genome size due to variations in the amount of repetitive DNA also show differences in GCL (e.g. Cypripedium; Khandawala, 2009). Analysis of our database, which is novel in being dominated by nonflowering seed-plants (83/130; 64%), enabled testing of whether GS (as inferred by changes in GCL) has increased within the gymnosperms (Fig. 4a–c).

Maximum GS has increased through gymnosperm evolution, driven primarily by an increase within some Coniferales of GCL/GS (Fig. 4a), but despite this trend, species with a small GS do occur in some extant gymnosperms (e.g. the gnetaleans Gnetum and Welwitschia, 2C = 4401–7785 Mbp and 14 083 Mbp, respectively). When the data are analysed at a finer taxonomic resolution, we detect increases in the GCL of Cordaitales and Bennettitales and hence infer an increase in GS (Fig. 4b), though admittedly data from these groups are based on a limited sample size. For groups yielding more numerous data points (n ≥ 7), such as the Medullosales, Glossopteridales and Ginkgoales, we identified no significant trends in GS (Fig. 4b, Table S2). In the case of Medullosales and Glossopteridales, the observed ranges of GS may include evidence of genome duplication events within the data or genome size expansion via amplification of repetitive DNA sequences at the diploid level. However, in the absence of rigorous phylogenetic frameworks in which to analyse such ‘localised’ GS distributions, this hypothesis cannot be adequately tested. The contemporaneous (in a geological sense) nature of the data for these groups (Fig. 4b) precludes confident identification of long-term trends. Analysing conifer GCL sizes at family level suggests that average GS increased through time in the Cheirolepidiaceae and Araucariaceae but decreased in Cupressaceae (Fig. 4c); however, these trends are not (yet) statistically significant (Table S2).

Directions of future research

We view the present contribution as an interim report on progress achieved in this relatively novel field of research, intended to stimulate more detailed investigations. Although additional data from any taxonomic group and time period should prove helpful, factors potentially confounding the positive correlation between stomatal guard-cell size and genome size will be fewer and/or of lesser impact on future studies if time periods are deliberately shortened and taxonomic spectra are narrowed to closer relatives. The comparatively large guard-cell size evident in Early Devonian land plants is probably best explored by more detailed investigation of ‘breathing pores’ in extant nonvascular land plants and the lycopsid descendants of Asteroxylon. In terms of interpreting both the effects of developmental canalisation and ecophenotypic modification of guard-cell size, controlled experiments conducted in geologically relevant conditions and on palaeobotanically relevant species will clearly continue to pay dividends. In addition, as noted above, some evidence exists that molecular responses to genome duplication events may differ between major groups of land plants (e.g. lycophytes, monilophytes, gymnosperms, angiosperms). Given that most research on polyploidy to date has focused on angiosperms, expanding these studies to include other land-plant groups will also be essential for understanding how polyploidy impacts on the evolution of guard-cell size and genome size.