Plant responses to low [CO2] of the past


Author for correspondence:
Joy K. Ward
Tel: +1 785 8645218



II.The case for low-[CO2] studies675
III.Experimental approaches for reducing [CO2]677
IV.Early low-[CO2] studies679
V.Low-[CO2] effects on the individual plant680
VI.Low [CO2] and plant evolution683
VII.Interactions of low [CO2] with other factors687
VIII.Low-[CO2] effects on community composition689
IX.Low-[CO2] effects on the ecosystem689
X.Low-[CO2] effects on early human societies690


During the Last Glacial Maximum (LGM; 18 000–20 000 yr ago) and previous glacial periods, atmospheric [CO2] dropped to 180–190 ppm, which is among the lowest concentrations that occurred during the evolution of land plants. Modern atmospheric CO2 concentrations ([CO2]) are more than twice those of the LGM and 45% higher than pre-industrial concentrations. Since CO2 is the carbon source for photosynthesis, lower carbon availability during glacial periods likely had a major impact on plant productivity and evolution. From the studies highlighted here, it is clear that the influence of low [CO2] transcends several scales, ranging from physiological effects on individual plants to changes in ecosystem functioning, and may have even influenced the development of early human cultures (via the timing of agriculture). Through low-[CO2] studies, we have determined a baseline for plant response to minimal [CO2] that occurred during the evolution of land plants. Moreover, an increased understanding of plant responses to low [CO2] contributes to our knowledge of how natural global change factors in the past may continue to influence plant responses to future anthropogenic changes. Future work, however, should focus more on the evolutionary responses of plants to changing [CO2] in order to account for the potentially large effects of genetic change.

I. Introduction

Rising atmospheric [CO2] (CO2 concentration) is expected to increase global temperatures and has been deemed a major threat to climate, economic development, and human health (IPCC, 2007a). It is perhaps ironic to be considering plant responses to low [CO2] during an era in which most research has focused on rising [CO2]. Here we will emphasize that a strong foundation in understanding plant responses to low [CO2] is critical for predicting the implications of rising [CO2], explaining plant evolutionary patterns over geologic time scales, and estimating past and future levels of net primary productivity (NPP). In addition, low-[CO2] studies provide a baseline for defining plant response to minimum [CO2] of the geologic past, such that changes in plant functioning in response to rising [CO2] can be assessed within a broader temporal context.

Atmospheric CO2 is the main source of carbon for photosynthesis and serves as a fundamental substrate for plant growth. CO2 assimilation during photosynthesis represents a critical exchange of carbon between the atmosphere and the biosphere within the global carbon cycle (Schlesinger, 1997). CO2 is unique in that plants assimilate this resource from the atmosphere, whereas other resources such as water and nutrients are acquired from the soil. Furthermore, CO2 is evenly distributed throughout the Earth’s atmosphere, and therefore CO2 availability to plants is similar across all terrestrial ecosystems (although there can be local gradients with altitude and within canopies). This is unlike other plant resources such as light, water, and nutrients that vary across ecosystems, as well as at much smaller spatial scales. Although similar within a time period, [CO2] has varied throughout geologic time, and during some periods may have been so low as to greatly limit plant growth and reproduction (Ward, 2005). Owing to the inability to substitute ‘space for time’, our knowledge of the evolutionary responses of plants to low [CO2] is rather limited compared with our understanding of evolutionary responses to other resource limitations, although advances have been made in this area and are reviewed here.

Beginning with the rise of vascular land plants through modern times, atmospheric [CO2] reached maximum values of 3300–3600 ppm during the early Devonian (Berner, 2006), possibly dropped as low as 150 ppm during the late Pliocene (Tripati et al., 2009), and consistently ranged between 170 and 190 ppm during glacial maxima of the past million yr (Petit et al., 1999; EPICA, 2004; Brook, 2005; Tripati et al., 2009; Fig. 1). Since CO2 is a major substrate for photosynthesis, such extreme changes in the availability of this resource likely had profound effects on plant productivity, community structure, and evolution through time. The LGM that occurred 18 000–20 000 yr ago represents a fascinating time when low [CO2] likely constrained the physiological functioning of C3 plants (Polley et al., 1993a; Dippery et al., 1995; Sage & Coleman, 2001; Ward et al., 2005; Lewis et al., 2010). During that period, [CO2] dropped to 180–190 ppm (Petit et al., 1999; EPICA, 2004), which is among the lowest concentrations predicted to have occurred during the evolution of land plants (Berner, 2003, 2006; Tripati et al., 2009). Following the LGM, [CO2] gradually increased to 270 ppm just before the Industrial Revolution, and has been rising rapidly in recent decades as a result of expanding industrialization (IPCC, 2007b). Currently, [CO2] is 392 ppm (recorded at Mauna Loa Observatory, Hawaii;, a value that may not have occurred since the mid-Miocene (14–16 million yr ago; Tripati et al., 2009). Thus, modern [CO2] values are more than twice the minimum concentrations that occurred during the LGM and c. 45% higher than pre-industrial values. [CO2] is expected to continue rising in the future, potentially reaching 1000 ppm by 2100, depending on the carbon emissions scenario that actually occurs (IPCC, 2007b).

Figure 1.

 Changes in atmospheric [CO2] throughout the evolution of vascular land plants. The upper right insert shows the past million yr expanded in order to show low [CO2] during glacial periods. The upper left insert is expanded to show low [CO2] periods over the last 10 million yr (data are from Petit et al., 1999; Monnin et al., 2001; Siegenthaler et al., 2005; Berner, 2006; Lüthi et al., 2008; Keeling et al., 2005; Tripati et al., 2009).

II. The case for low-[CO2] studies

To date, most CO2 studies have focused on elevated [CO2] that is predicted to occur within the next 50–100 yr as a result of anthropogenic carbon emissions (McLeod & Long, 1999; Ainsworth & Long, 2005; Körner, 2006; Miyagi et al., 2007; Springer & Ward, 2007; Lloyd & Farquhar, 2008). The motivation for many of these studies has been to determine the degree to which plants will serve as sinks to offset carbon emissions. Far fewer studies have focused on plant responses to low [CO2] of the past, even though this work is crucial for understanding long-term responses of plants to changing [CO2] over geologic and evolutionary time scales. From the studies that have been conducted, it is clear that modern C3 plant genotypes grown at low [CO2] (180–200 ppm) exhibit severe reductions in photosynthesis, survival, growth, and reproduction, suggesting that reduced [CO2] during glacial periods may have induced carbon limitations that would have been highly stressful on C3 plants (Polley et al., 1993a; Dippery et al., 1995; Sage, 1995; Tissue et al., 1995; Sage & Coleman, 2001; Ward & Kelly, 2004; Tonsor & Scheiner, 2007; see Fig. 2 for a photo of plants from Dippery et al., 1995). In addition, carbon limitations at low [CO2] may have altered plant tolerance ranges to other stressors such as drought, heat, and herbivory (Sage & Cowling, 1999). Furthermore, most studies focusing on the full range of plant responses to past through future [CO2] report much greater physiological and growth enhancements in response to increases in [CO2] below modern concentrations than to increases above modern concentrations (Sage & Reid, 1992; Polley et al., 1993b; Dippery et al., 1995; Ward & Strain, 1997). Thus, plants may have already exhausted much of their potential to respond to rising [CO2], unless, for example, major evolutionary changes occur in the future. From these findings, it is clear that assessing the full continuum of plant response to changes in atmospheric [CO2] through geologic time is essential for making accurate predictions regarding the functioning of both past and future ecosystems.

Figure 2.

 Representative plants of Abutilon theophrasti (C3) grown at glacial through future [CO2]. All plants were 14 d of age and were grown under similar water, light, and nutrient conditions. These plants were photographed during a study by Dippery et al. (1995). (Photograph is courtesy of Anne Hartley, Florida Gulf Coast University.)

Studies addressing the effects of low [CO2] on plants are also fundamental for understanding plant evolution in response to changes in resource availability through time – primarily since changing [CO2] has been shown to have major implications for plant fitness (Ward et al., 2000). Modern plants grown at low [CO2] (150–200 ppm) exhibit highly compromised survival (Ward & Kelly, 2004) and reproduction (Dippery et al., 1995) at conditions that occurred only 18 000–20 000 yr ago. Such findings beg the question of how glacial plants survived during low-[CO2] periods, especially considering the lack of evidence for plant extinctions during these times. Furthermore, past work has demonstrated that low [CO2] has the potential to act as a strong selective agent on plants, and therefore evolutionary responses may have ameliorated some of the negative effects of low [CO2] in the past (Ward et al., 2000). However, the full suite of mechanisms accounting for these adaptive responses is currently unknown, as well as how adaptive processes may have been influenced by other interactions with climate change (for a discussion of possibilities see Sage, 1994; Sage & Cowling, 1999; Ward et al., 2000; Beerling, 2005). Furthermore, it is also important to consider that any genetic changes that occurred in the recent geologic past as a result of low [CO2] may continue to affect the responses of plants to rising [CO2] throughout the next century (Strain, 1991; Sage & Cowling, 1999).

Overall, low-[CO2] studies are critical for understanding plant responses to past environments when carbon resources were most limiting, evaluating physiological and growth constraints for response to rising [CO2], determining the full continuum of plant responses to changes in [CO2] over evolutionary time scales, assessing the impacts of low [CO2] on plant community composition and ecosystem functioning, and understanding the influence that low [CO2] may have had on early human cultures via influences on the development of agriculture. Moreover, studying plant responses to low [CO2] provides information about past ecosystem functioning, such as estimates of glacial NPP (Prentice & Harrison, 2009), as well as insights into the availability of food resources for animals (Coltrain et al., 2004) and early humans (Sage, 1995; Richerson et al., 2001).

In this review, we mainly focus on the effects of low [CO2] that occurred during glacial periods, with an emphasis on plants possessing the C3 photosynthetic pathway (85–90% of all species), since they tend to be most responsive to low [CO2] (Ehleringer et al., 1991, 1997). For our purposes, ‘low’ or ‘glacial’ CO2 corresponds to values of 170–200 ppm, ‘pre-industrial’ values correspond to 250–300 ppm, ‘modern’ values range between 350 and 400 ppm, and ‘elevated’ values correspond to [CO2] at or > 500 ppm. Although numerous environmental factors have changed between glacial and interglacial periods (e.g. precipitation, temperature), few studies have focused on the interactive effects of low [CO2] with other variables. We incorporate interactive effects with low [CO2] when possible, recognizing that our understanding in this area is limited.

We emphasize that the majority of low-[CO2] studies involve the use of modern plants grown at simulated conditions of the past. Such studies allow us to characterize the effects of limiting [CO2] on physiological, growth, and reproductive processes. Unfortunately, however, because these studies generally report on only a single generation of growth at low [CO2], they limit the scope of our understanding of evolutionary processes. Therefore, we also discuss the results of studies that incorporate fossil material from glacial plants that were more fully adapted to low [CO2], as well as studies focusing on the evolutionary responses of plants to low [CO2] over multiple generations of artificial selection. We also fit low-[CO2] research into the larger context of global change studies, discuss technologies for reducing [CO2] during experimentation, and provide recommendations for future research directions in the field.

III. Experimental approaches for reducing [CO2]

Methods for elevating [CO2] above ambient concentrations are often achieved via external sources of CO2 and have been applied to growth chambers, open-top chambers, and free-air CO2 enrichment (FACE) sites, providing multiple scales for elevated [CO2] research. Reducing [CO2] below the modern value is much more difficult, since CO2 gas must be scrubbed from the atmosphere rather than added to it. This presents a major challenge when there is a large volume of air that must be scrubbed of CO2 in a controlled and rapid manner. Several approaches have been developed to address this issue, and the benefits and limitations of each are discussed in the following.

The earliest methods for reducing [CO2] took advantage of the finding that C4 plants can outcompete C3 plants under limiting [CO2]. In order to reduce [CO2], researchers enclosed C4 plants side by side with experimental C3 plants in small chambers, effectively lowering [CO2] below the C3 compensation point ([CO2] where carbon gain through photosynthesis equals carbon loss from respiration; e.g. Sharma et al., 1979; Fig. 3a). This approach did not allow for tight control of [CO2], and resulted in rapid onset of carbon starvation in plants, but was a clever way to lower [CO2] before the introduction of more sophisticated methods.

Figure 3.

 Experimental approaches for achieving low-[CO2] treatments. (a) Small growth chamber with C4 plants grown side by side with experimental C3 plants, used by Sharma et al. (1979, Fig. 1); (b) tunnel chambers for reducing [CO2] along a continuum, described by Mayeux et al. (1993) (photograph courtesy of Wayne Polley, Grassland, Soil, and Water Research Laboratory); (c) diagram of the tunnel chambers indicating where [CO2] (CO2), dew point (DP), and air temperature (T) are sampled. Wavy lines represent chilled water cooling coils and cross-hatched areas represent electrical resistance heaters (from Mayeux et al., 1993, Fig. 2); (d) experimental growth chamber with soda lime box underneath plants to reduce [CO2] within the chamber (photograph courtesy of Joy K. Ward, University of Kansas); (e) reach-in experimental growth chamber with low-[CO2] control (Conviron BDR-16; photograph courtesy of Joy K. Ward, University of Kansas).

In the mid-1990s, a later phase of technology was introduced at the Grassland, Soil, and Water Research Laboratory (USDA-Agricultural Research Service) that involves an outdoor tunnel system, whereby plants are grown across a continuum of modern to low [CO2] (Mayeux et al., 1993; Fig. 3b,c). During the day, air of known [CO2] is pumped into one end of the tunnel where plants experience relatively high [CO2]. Plants near the end of the tunnel, on the other hand, experience low [CO2] (similar to glacial values) as a result of photosynthetic removal of CO2 from air as it moves progressively through the tunnel. At night, airflow is reversed while plants are solely respiring, and this serves to equalize [CO2] throughout the tunnel. The desired [CO2] is maintained by varying the rate at which air flows through the chamber in response to changes in photosynthetic and respiration rates. A major strength of this system is that these chambers are housed outdoors and are exposed to full, natural lighting conditions with tight temperature control. In addition, plant data acquired from these experiments can be analyzed across a [CO2] gradient, as opposed to using discrete CO2 treatments. A disadvantage, however, is that the [CO2] gradient cannot be maintained during dark hours, and sample size within a specific, targeted [CO2] may be limited.

A different approach to lowering [CO2] is through the use of controlled growth chambers, which can range in size from small reach-in varieties (Ward et al., 2000; Mohan et al., 2004; Sage & Reid, 1992; Fig. 3e) to large walk-in chambers commonly housed in experimental phytotrons (Ward et al., 2008). In these chambers, plants can be grown with a variety of options, such as computer-controlled temperature, humidity, light and CO2 conditions. Most commonly, low CO2 treatments are obtained by scrubbing CO2 from the atmosphere, either by forcing air through ‘scrub boxes’ containing soda lime (contents may include calcium hydroxide, sodium hydroxide, and/or potassium hydroxide depending on the manufacturer; Fig. 3d) or by passing compressed air across soda lime filters. Even in these small and highly controlled systems, the maintenance of [CO2] can be inherently difficult. We have found that the best results occur when the scrubbing of CO2 within scrub boxes (Fig. 3d) is coupled with CO2 additions from an external tank with a computer-generated switch, allowing for more constant [CO2] control within the chamber (i.e. competing additions and scrubbing of CO2 gas tend to stabilize [CO2]; JK Ward unpublished). In addition, other researchers have utilized a system where CO2-free air is emitted into enclosed chambers in order to lower [CO2]. These units at the University of Florida (Gainesville, FL, USA) were made of glass and were housed outdoors, allowing for natural, full sun conditions (Baker et al., 1990; Allen et al., 1998). Overall, the tight control offered by the experimental systems described earlier is accompanied by the loss of realistic field conditions, and therefore these approaches may be less informative for strict ecological questions. These systems, however, can be highly useful for uncovering basic genetic, physiological, and growth mechanisms that may be driving individual plant responses to low [CO2].

In order to advance low-[CO2] research, it is hoped that future approaches will combine technologies for reducing [CO2], maintenance of environmental control, as well as employing conditions that more closely simulate natural, field conditions. Such plans are currently under consideration and are being discussed among the scientific community at large.

IV. Early low-[CO2] studies

The earliest studies focusing on plant responses to low [CO2] began in the early 1960s and were prompted by the development of new technologies for measuring plant gas exchange. These studies focused on determining the CO2 compensation point of plants, as well as comparing respiration and photosynthetic rates among species and genotypes. For example, Moss (1962) grew a variety of economically important crops (corn, sugar cane, orchard grass, tobacco, geranium, tomato, and Norway maple) in a closed system and allowed plants to draw down CO2 over time. The author found that corn and sugar cane (now known to be C4 plants) could draw down [CO2] below 10 ppm, whereas the other species (now known to be C3 plants) could only draw down [CO2] between 60 and 145 ppm. Interestingly, this low-[CO2] study distinguished plants with the C4 vs C3 photosynthetic pathways several years before the formal discovery of C4 photosynthesis. At the time, the author was unable to provide a specific mechanism to account for these differences, but realized he was working with two unique physiological systems. It was later shown that C4 plants have a CO2 compensation point that is close to 0 ppm, whereas that of C3 plants is in the vicinity of 50–60 ppm between 25 and 30°C.

During the early 1960s, Billings et al. (1961) also dealt with plant responses to low [CO2], but from a more ecological perspective. The authors compared gas exchange rates of two populations of Oxyria digyna– one from high altitude (Logan Pass, MT, USA) and the other from sea level (mouth of the Pitmegea River, AL, USA). Seeds from these populations were grown under common conditions and photosynthetic rates were measured across a wide range of [CO2]. The authors found that photosynthetic rates were higher and CO2 compensation points were lower in offspring that originated from alpine populations, and suggested that this may reflect adaptations to lower CO2 partial pressure at high altitude. The authors also concluded that low CO2 partial pressure, and not temperature, limited the upward distribution of the Oxyria populations in this case. A subset of more recent studies have supported these findings, while others have pointed out that there may be alternative explanations for these results. Similar to the findings of Billings et al. (1961), Körner & Diemer (1994) showed that species of the genera Ranunculus and Geum exhibited higher carbon gain when grown at elevated [CO2] compared with lowland species of the same genera. In addition, Ward & Strain (1997) found that Arabidopsis genotypes from high elevations produced greater seed numbers (indicating higher fitness) when grown at low [CO2] (200 ppm) compared with genotypes from lower elevations (where CO2 partial pressure was higher). Other studies, however, have noted additional factors with increasing altitude that would serve to enhance photosynthesis, thus overcoming some, if not all, of the negative effects of low CO2 partial pressure (Gale, 1972; Sage & Reid, 1992; Terashima et al., 1995). These include increased diffusivity of CO2 at high elevation that facilitates the movement of CO2 from air to intercellular spaces, as well as reductions in O2 partial pressure with elevation (proportionally similar to CO2) that would reduce photorespiration. These factors can potentially offset the drop in CO2 partial pressure up to 2500 m, after which stromal [CO2] may be reduced by 20% (although here low temperatures may begin to limit carbon gain over CO2; Sage & Reid, 1992). Taken together, these considerations suggest that high-altitude plants may not serve as a viable model for understanding adaptive responses of plants to global reductions in [CO2] of the past (Terashima et al., 1995).

A new era of low-[CO2] studies began during the late 1960s, with the objective of identifying plants with high photosynthetic efficiency and low photorespiration rates, with the overall goal of improving crop productivity. This research occurred in conjunction with the green revolution that emphasized initiatives for increasing food production. In several studies (Cannell et al., 1969; Nelson et al., 1975; Sharma et al., 1979), both crop and model plants were grown at extremely low [CO2] that was near or below the CO2 compensation point of C3 plants (achieved by growing newly identified C4 plants alongside experimental C3 plants). Subsequently, genotypes were screened for survival at these extremely limiting carbon conditions with the assumption that surviving genotypes would exhibit superior photosynthetic efficiency. While mostly unsuccessful with respect to the original goal, these studies provided valuable insights into the degree of physiological stress induced by low [CO2]. For example, Cannell et al. (1969) screened 2458 genotypes of soybean (Glycine max, C3) and found that none could survive at low [CO2] near the compensation point (ranging between 10 and 50 ppm CO2). A decade later, Sharma et al. (1979) conducted studies with 33 Arabidopsis thaliana (C3) genotypes grown below the CO2 compensation point. They found considerable variation, whereby some genotypes survived only 1 wk at these conditions, whereas others could survive longer than 2 wk (relatively large variation for such a short-lived species). They also noted that the crossing of parents with extreme differences in survival time near the CO2 compensation point resulted in a highly variable F2 population (with some genotypes being superior to both parents), suggesting that there is considerable natural genetic variation for survival at limiting [CO2]. In more recent advances (described in Delgado et al., 1994; Medrano et al., 1995), haploid tobacco (Nicotiana tabacum, C3) lines derived from mutagen-treated anthers were selected for survival near the CO2 compensation point (60–70 ppm). Approximately 5% of the lines survived, and, of those, plants had greater total leaf area, maintained higher photosynthetic rates specifically in mature and older leaves, and had lower respiration rates (on a leaf mass basis, but not an area basis) relative to parental plants. This study, however, did not find alterations in photorespiration rates and properties of Rubisco in selected plants, as originally hoped.

In the 1980s, the publication of ice core data characterized the CO2 composition of the atmosphere during the late Pleistocene, Holocene, and pre-industrial periods. Consequently, studies began focusing on the effects of low [CO2] on plants from a geological and historical perspective. One of the first of these studies was conducted by Overdieck et al. (1988), who found that C3 herbaceous annuals (Vigna unguiculata L., cowpea; Abelmoschus esculentus (L.) Moench, okra; Raphanus sativus L., radish) exhibited an average 8% reduction in growth at the pre-industrial value of 270 ppm CO2 relative to the modern value that was 350 ppm at the time. These results suggested that anthropogenic increases in [CO2] over contemporary time scales may have already modified plant functioning and productivity within modern ecosystems. Using Lyonia mariana, Overdieck (1989) was among the first to show that plants grown at pre-industrial [CO2] exhibited greater water loss as a result of higher stomatal conductance (g) (by as much as 16%) compared with plants grown at modern [CO2]. More recent studies have generally confirmed this result with instantaneous gas exchange measurements, but few studies have allowed for full physiological acclimation to pre-industrial [CO2] over a full generation as in the Overdieck (1989) study. In a more recent study, Dippery et al. (1995) found that the biomass production of Abutilon theophrasti was reduced by 24% when grown at pre-industrial (270 ppm) vs modern (350 ppm) CO2 (see Fig. 2 for a photograph), although specific effects of [CO2] on physiology and other leaf properties could not be identified between these treatments (Tissue et al., 1995).

In summary, these early studies laid the groundwork for more recent work by showing that plants with different photosynthetic pathways exhibit differential responses to low [CO2], modern C3 plants can become highly stressed when grown at low [CO2], and plants exhibit genetic variation in response to low [CO2] (whether natural or induced) that is often explained by shifts in biomass allocation or developmental timing rather than through direct alterations in photosynthesis/photorespiration rates or photosynthetic enzyme characteristics.

In the remaining sections, we describe the results of more recent low-[CO2] studies that investigate a variety of scales ranging from physiological to ecosystem-scale processes. We also address the potential for plant evolutionary responses to low [CO2] by reviewing studies that focus on plant fossils that occurred during glacial periods, as well as modern plants that were selected for high fitness over multiple generations at low [CO2]. We also discuss the current status of a hypothesis by Sage (1995), suggesting that low [CO2] may have influenced the timing of agriculture among early humans. Although we are unable to describe all low-[CO2] studies in detail, we highlight a wide range of work that reflects our current understanding of plant responses to low [CO2].

V. Low-[CO2] effects on the individual plant

1. Physiological responses

In C3 plants, low [CO2] affects net photosynthetic rates by reducing the rate of carboxylation of Rubisco resulting from substrate limitations and through higher photorespiration rates. Photorespiration is increased at low [CO2] because both CO2 and O2 compete for the same active site of Rubisco. A reduction in [CO2]/[O2] enhances oxygenation, resulting in carbon loss to the plant. Note that unlike [CO2], [O2] has remained unchanged in the atmosphere for at least the last several million yr (Berner et al., 2007), and thus [O2] changes are not relevant to the time periods covered in this review.

Studies that measure plant responses to low [CO2] most commonly involve growing plants at modern [CO2], followed by instantaneous gas exchange measurements on a small portion of leaf area across a large range of [CO2] (typically ranging between 0 and 1000 ppm). Referred to as Aci curves (photosynthetic assimilation vs leaf intercellular [CO2]), these measurements encompass low [CO2] that was characteristic of glacial periods (Fig. 4). Aci curves generally show steep linear increases in net photosynthetic rates (A) between ci values of 0 and 250 ppm CO2, with continued positive slopes through ci values as high as 1000 ppm (Sage & Coleman, 2001). These curves indicate that reductions in [CO2] during glacial periods produced Rubisco-limited photosynthetic rates (Sage & Coleman, 2001). A survey of Aci curves from a variety of C3 species indicate that net photosynthetic rates are reduced by 40–60% at low (180 ppm) vs modern (350–380 ppm) CO2 (ca values), with c. 30% of this reduction being induced by higher photorespiration rates when temperatures are optimal (Sage & Cowling, 1999).

Figure 4.

 Theoretical Aci curve showing the relationship between photosynthetic assimilation of CO2 (A) and leaf intercellular [CO2] (ci). The region within the vertical lines represents approximate ci values experienced by glacial plants (at corresponding atmospheric [CO2] of 170–200 ppm and based on ci values from glacial plants; Van de Water et al., 1994; Ward et al., 2005).

Aci curves are highly informative for understanding the biochemical properties of Rubisco in response to changing [CO2]. Because this enzyme has been highly conserved across evolutionary time scales (Griffiths, 2006), such measurements may have the potential to be extrapolated to glacial plants. It should be noted, however, that the extent of physiological acclimation at each [CO2] measurement is short-term in nature, and thus full acclimation responses are rarely assessed with Aci curves. To further highlight this point, changes in Aci curves have been observed in C3 plants grown at modern [CO2] and then transferred to low [CO2] (200 ppm) for several wk, and the direction of change was generally unpredictable (Sage, 1994; also see Overdieck, 1989; Sage & Reid, 1992). In addition, such measurements do not provide insights into whole-plant responses to low [CO2] over the full life cycle, such as modifications in both relative and total leaf area that can influence whole-plant carbon uptake. Another point worth noting is that the Aci curves of plants that are fully adapted to low [CO2] (over many generations of genetic change) may potentially deviate from that of modern plants, thus limiting the potential for physiological extrapolation to the past. Because we know essentially nothing about this issue, future work should focus on the extent to which modern physiological responses can be extrapolated to the past, with the inclusion of full acclimatory and adaptive responses.

In addition to having direct effects on photosynthetic rates and carbon gain, changes in [CO2] can also have large effects on g that can influence transpiration rates and water-use efficiency (WUE; carbon assimilation/stomatal conductance). In the vast majority of C3 plants, g increases with reductions in [CO2] by 35–50% (depending on the species) between modern (350–380 ppm) and glacial values (180–200 ppm). This response is observed during short-term gas exchange measurements (Lloyd et al., 1992; Flexas et al., 2007), as well as long-term exposure to low [CO2] over a full generation (Polley et al., 1993b; Ward et al., 1999; Tonsor & Scheiner, 2007). Higher g serves to enhance CO2 uptake by reducing the diffusional resistance of CO2 into the leaf interior, but with the cost of higher water loss. Furthermore, Polley et al. (2002) found that higher g within a C3/C4 grass assemblage grown at low [CO2] translated into greater depletion of soil water than seen in similar assemblages grown at modern [CO2] (Polley et al., 2002), suggesting that stomatal responses to low [CO2] can have implications for water availability at the ecosystem level.

Using carbon isotopes ratios, Polley et al. (1995) estimated WUE in a variety of C3 plants (Triticum aestivum, Bromus tectorum, Prosopis glandulosa) and found reduced values at low vs modern [CO2]. Interestingly, reductions in WUE were directly proportional to reductions in [CO2] in all species, suggesting that the ability to scale this response may be relatively straightforward. In the same study, nitrogen-use efficiency (NUE, biomass produced/plant N) decreased in B. tectorum and P. glandulosa at low [CO2], but this response was not as pronounced or consistent as changes in WUE, suggesting that low [CO2] may have increased the requirements for water to a greater extent than N in C3 species.

2. Biomass production

In order to determine how physiological responses to low [CO2] manifest themselves at the whole-plant level, plants must be grown for a full generation at low [CO2]. Such studies have shown that the average biomass production of modern C3 plants is reduced by c. 50% when grown at low (180–220 ppm) vs modern (350–380 ppm) [CO2], when other conditions are optimal (Sage & Coleman, 2001; Fig. 5). There is, however, variation in this response among C3 species (Fig. 5), as well as within C3 species, whereby reductions in biomass may vary by 40–70% among genotypes (Ward & Strain, 1997; Hovenden & Schimanski, 2000; Mohan et al., 2004). In addition, as [CO2] declines to 150 ppm, biomass production may be reduced by as much as 92%, as was observed in A. theophrasti (Dippery et al., 1995; Figs 2, 5).

Figure 5.

 Relative net photosynthesis (a) and biomass production (or seed yield) (b) for a variety of modern plants grown at [CO2] spanning from glacial to modern times. Data in the y-axis are responses at a given [CO2] relative to modern [CO2] (350–380 ppm), with the line representing responses for plants grown at the glacial [CO2] minimum (adapted from Sage & Coleman, 2001, Fig. 3, with updated data from Cowling & Sage, 1998; Ward et al., 1999; Hovenden & Schimanski, 2000; Bunce, 2001).

It is also generally found that the same absolute change in [CO2] below the modern value has a much greater effect on biomass production (and A) than the same absolute change in [CO2] above the modern value (Allen et al., 1991; Polley et al., 1992; Dippery et al., 1995; Tissue et al., 1995; Ward & Strain, 1997; Ward, 2005; Fig. 5). Thus, even small changes in [CO2] during glacial periods may have had large effects on plant productivity. It is possible, however, that modern plant responses to past [CO2] may be accentuated relative to glacial plants that were more fully adapted to low [CO2]. Moreover, the lack of evidence for plant extinctions during glacial periods indicates the likelihood that plants underwent adaptive changes in response to low [CO2] that may have altered biomass production, and these responses may no longer be evident in some modern genotypes (Ward et al., 2000; Ward & Kelly, 2004; Tonsor & Scheiner, 2007).

When comparing biomass production and net photosynthetic rates (A) of C3 plants, there is often a stronger correlation between these measurements at low [CO2] than at modern or elevated [CO2] (Tissue et al., 1995; Sage & Coleman, 2001; but also see Cowling & Sage, 1998 where this correlation becomes decoupled at very high temperature). This finding indicates that photosynthate is more directly converted into biomass at low [CO2], and this is often accompanied by lower starch and sugar accumulation within leaves (Sage, 1995; Allen et al., 1998). Furthermore, lower photosynthetic rates at low [CO2] often persist throughout a full generation, contributing to sustained reductions in biomass production. This is the case even when plants undergo acclimation responses to low [CO2] (reviewed in Sage & Coleman, 2001), such as higher concentrations of Calvin–Benson cycle enzymes (mainly Rubisco; Tissue et al., 1995; but also see Gesch et al., 2000 for an example of partial recovery), higher g (Polley et al., 1993b; Ward et al., 1999), increased activity of carbonic anhydrase that facilitates CO2 diffusion into the chloroplasts (Coleman, 2000), and greater leaf area ratio (leaf area/total plant mass, LAR) that may enhance whole-plant carbon uptake (Dippery et al., 1995).

3. Biomass partitioning

It has been hypothesized that plants grown at low [CO2] would partition a higher proportion of biomass to above-ground than to below-ground structures; this response would increase LAR and enhance overall investment in carbon assimilation under limiting [CO2] (Sage & Coleman, 2001). A variety of studies have found support for this idea, including Dippery et al. (1995) who showed that A. theophrasti partitioned a higher proportion of biomass to shoots relative to roots at 150 than at 350 ppm CO2 (root : shoot mass = 0.17 vs 0.34, respectively); this finding was additionally confirmed by allometric analysis that accounted for ontogenetic shifts (J. K. Ward, unpublished). Surprisingly, this response ended up being counterproductive, since reduced investment in roots resulted in reduced N uptake and lower Rubisco production, further compounding the negative effects of low [CO2] on carbon uptake (Tissue et al., 1995). Interestingly, Cowling & Sage (1998) observed that Phaseolus vulgaris also had reduced partitioning of biomass to roots at low [CO2], but this response was associated with increased stem biomass as opposed to leaf biomass. Such a finding may be associated with direct effects of [CO2] on biomass partitioning, or may be the result of the indirect effects of shifting plant ontogeny in response to differences in [CO2]. With respect to contemporary time scales, neither Dippery et al. (1995) nor Bunce (2001) found differences in LAR in A. theophrasti and other C3 annuals grown at pre-industrial (270 ppm) and modern (350–370 ppm) [CO2]. Also, Ghannoum et al. (2010) found no differences in root allocation in two Eucalyptus species grown at pre-industrial (290 ppm) and current (400 ppm) CO2 at ambient temperature. These findings suggest that increases in [CO2] over the past 100 yr may not have had a large influence on plant biomass partitioning (even though absolute biomass is often affected).

4. Developmental timing

Elevated [CO2] is known to affect plant developmental timing and such responses have been shown to influence plant fitness (Ward & Kelly, 2004; Springer & Ward, 2007; Springer et al., 2008). Unfortunately, very little is known about the effects of low [CO2] on the developmental timing of C3 plants. Sage & Coleman (2001) hypothesized that increasing carbon storage within roots before reproduction would enhance fitness at low [CO2] (assuming there was sufficient time remaining in the growing season to reproduce), and this could be achieved by delaying developmental milestones such as flowering. In support of this idea, Ward & Strain (1997) found that field-collected genotypes of A. thaliana required on average 9 d more to initiate flowering when grown at 200 vs 350 ppm CO2. This represented a proportionally large shift in developmental timing considering that this species only has a 40–60 d life cycle. Unlike this finding, Polley et al. (1993b) did not find modifications in developmental timing in two cultivars of wheat grown from 200 to 350 ppm CO2. These studies suggest that at this point, there is no predictable response pattern to describe the effects of low [CO2] on the developmental timing of C3 annuals and indicate that additional work is needed in this area. Furthermore, little is known about the effects of low [CO2] on the developmental timing of perennials. It is predicted that they would have had an advantage over annuals at low [CO2] in that reproduction could be delayed until adequate resources were acquired for successful reproduction (Cowling, 2001; Ward, 2005).

5. Reproduction and survival (fitness components)

Of the few studies measuring the survival and reproduction of C3 plants grown at low [CO2], all have reported large effects on these measurements. Dippery et al. (1995) found the most extreme response, where low [CO2] (150 ppm) prevented reproduction in the modern C3 annual, A. theophrasti, as a result of the abortion of all flower buds that drove the fitness response to zero. This finding suggested that 150 ppm CO2 may be near the threshold for successful completion of the life cycle in some C3 species. Campbell et al. (2005) found that tobacco was able to successfully reproduce at both 100 and 150 ppm CO2, although after a very large amount of time (16 wk), and the germination percentage of offspring was compromised at 100 relative to 150 ppm CO2. Both the Dippery et al. (1995) and Campbell et al. (2005) studies provide a strong reminder that the whole-plant CO2 compensation point allowing for full completion of the plant life cycle may be substantially higher than the leaf CO2 compensation point. In addition, much more time may be required for successful reproduction at low [CO2] since sufficient carbon must be accumulated and stored in order for reproduction to be successful. Such considerations must be kept in mind when predicting the full effects of low [CO2] on plant reproduction and fitness.

Tonsor & Scheiner (2007) investigated the effects of CO2 availability (ranging between 250 and 710 ppm) on patterns of trait integration among 35 genotypes of A. thaliana. They found that at low [CO2], whole-plant carbon assimilation was the main determinant of reproductive mass, although this relationship was not maintained at higher [CO2]. This study echoes the findings with A and biomass production (see Section V.2), but now additionally shows that A and fitness can also be closely correlated at low [CO2]. In addition, Ward & Kelly (2004) also worked with A. thaliana and found that six field-collected genotypes exhibited reduced survival, ranging between 20 and 49%, and reduced seed production ranging between 38 and 81% when grown at 200 ppm vs 350 ppm CO2. As a result, reductions in estimated fitness (percentage survival × total seed production) ranged between 59 and 87% among the genotypes, indicating substantial genetic variation for low-[CO2] response (significant CO2 × genotype interaction; Fig. 6). Furthermore, the reductions in reproductive output among the Arabidopsis genotypes were a result of overall reduced plant size, as opposed to changes in the partitioning of biomass to reproduction (Ward & Kelly, 2004). Taken together, these studies, as well as others (also see Mohan et al., 2004 with maple), indicate that there is ample genetic variation to account for potential evolutionary responses to low [CO2].

Figure 6.

 Responses of six field-collected genotypes of Arabidopsis thaliana grown at 200 and 350 ppm CO2 and measured for estimated total fitness (a), proportion of survival (b), and total seed number (c). Genotype numbers are from The Arabidopsis Biological Resource Center, Ohio State University (adapted from Ward & Kelly, 2004, Fig. 6).

In a different type of study using 13C as a label, Lehmeier et al. (2005) found that low [CO2] (200 ppm) did not alter the mobilization rate of seed-derived carbon within newly developing sunflower seedlings; such a response may have influenced survival (not measured in this study), since rapid movement of carbon reserves would be essential for maintaining a positive carbon balance in seedlings grown at low [CO2]. In addition, seed size may play a major role in determining survival at low [CO2], since greater carbon reserves would enhance the production of leaf area within seedlings, reducing the chances of destructive negative carbon budgets during rapid and early growth stages (Ward & Kelly, 2004; Ward, 2005; also see Metz et al., 2010 for experimental and theoretical considerations of seed size and survival). Likewise, the small seed size of A. thaliana may have contributed to the high mortality rates that were observed at low [CO2] in the Ward & Kelly (2004) study discussed earlier (Fig. 6).

VI. Low [CO2] and plant evolution

In past work, researchers have speculated about possible evolutionary responses of plants to low [CO2] during different geologic time periods (Beerling, 2005). For instance, Beerling et al. (2001) pointed out an association between decreasing [CO2] during the late Devonian and the evolution of megaphyll leaves, which had higher stomatal densities, higher transpiration rates, and greater capacity for cooling (based on modeling) compared with more primitive leaves. Furthermore, McElwain et al. (2005) suggested linkages between declining [CO2] during the Cretaceous and an increase in the relative abundance and radiation of the angiosperms.

The rise of C4 photosynthesis is also a major evolutionary event that has been attributed to low [CO2] (initially by Ehleringer et al., 1991; also in Ehleringer et al., 1997; Cerling et al., 1998; reviewed by Sage, 2004). Early work suggested that [CO2] decline during the late Miocene contributed to the evolution of C4 species, since these plants can concentrate CO2 in bundle sheath cells, even when atmospheric [CO2] is low. More recent work, however, has postulated that C4 photosynthesis originated much earlier, most likely during the Oligocene (20 million yr earlier), and this idea has yielded phylogenetic support (Sage, 2004). Similar to the initial idea, however, C4 evolution has still been linked with reductions in [CO2], although much earlier in time, which became possible as the geologic record of atmospheric [CO2] improved. In more recent geologic periods, it is now recognized that there was a major expansion of C4-dominated grasslands during the late Miocene-early Pliocene that would have resulted in dramatic shifts in plant productivity and ecosystem functioning (Sage, 2004; Tipple & Pagani, 2007; Edwards et al., 2010). This transition occurred during an abrupt [CO2] decline, as recently shown by Tripati et al. (2009) using boron : calcium ratios in foraminifera. Their results showed that atmospheric [CO2] dropped as low as 200–300 ppm between 5 and 10 million yr ago, suggesting a likely role for low [CO2] in this C4 expansion. However, Tipple & Pagani (2007), and more recently Edwards et al. (2010), noted other factors that may have played a role in C4 expansion, including aridity, a more seasonal climate, fire disturbance, and monsoon. These papers emphasized that this C4 expansion should be evaluated on a more regional scale, and with attention to multiple C4 adaptations (in addition to low [CO2] tolerance), since C4 expansion does not appear to be globally synchronous. In more recent geologic time, there is strong evidence that C4 plants continued to expand their range in response to low [CO2] during glacial periods of the past million yr, particularly in the tropics where warmer temperatures would have additionally favored C4 photosynthesis (see Section VIII for more details). In addition, this has also been deemed a period of recent C4 evolution, mainly through the rise of a variety of C4 dicot lineages (Ehleringer et al., 1997; Sage, 2004).

Even with this recent geologic expansion of C4 species, C3 plants still persisted in many ecosystems during glacial periods (e.g. Coltrain et al., 2004; southern California), and it is not yet fully understood how they functioned and adapted to low [CO2] during that time. It is clear, however, that modern C3 plants are products of an ancestry that had undergone at least 2 million yr of glacial–interglacial cycles, with corresponding changes in climate and [CO2] (Ward & Kelly, 2004). Thus, glacial plants must have had the genetic capacity to adapt to changing [CO2], and mechanisms accounting for this process may still be detectable in some modern species/genotypes. Furthermore, a better understanding of these responses will provide important insights into the capacity for plants to evolve in response to rising [CO2] in the future, which is a critical issue at this juncture.

While studies conducted for an entire generation at low [CO2] provide critical information on whole-plant responses with full acclimation (Section V), these studies do not incorporate adaptive changes at low [CO2] that occur over multiple generations with genetic change. The process of understanding plant evolution at low [CO2] has been addressed through two main approaches. In the first, modern plants are selected at low [CO2] (180–200 ppm) for high growth or reproduction, allowing for an understanding of how genetic change may drive adaptive processes in living plants. These studies are conducted over multiple generations, and differ from the screening methods described in earlier studies, where genotypes were selected for survival during only one generation of exposure to extremely low [CO2] (e.g. Sharma et al., 1979). A second approach involves studying fossilized plant material that actually occurred during glacial periods; such fossils are highly useful for determining evolutionary responses to low [CO2] since these plants had thousands of yr to adapt to these conditions over multiple generations. A relatively large number of plant specimens dating to the LGM have been naturally preserved in tar pits, bogs, and caves, and these may serve as critical resources for further addressing this issue (Wells & Jorgensen, 1964; Stock & Harris, 2001). Unfortunately, however, measurements on these fossils are generally restricted to stable isotopes, morphology, DNA sequencing (in some cases), and modeling approaches (because they are nonliving). By studying these specimens, along with modern plants that have been adapted to low [CO2] over multiple generations, we may better understand how plants functioned across a wider range of [CO2], and we may be able to pinpoint the mechanisms that enabled C3 plants to survive during the most limiting [CO2] periods of the past.

To investigate evolutionary processes, Ward et al. (2000) conducted a selection experiment with A. thaliana (derived from out-crossings) for high seed number over five generations at low [CO2] (200 ppm; control plants were also grown side by side with selected plants, but were randomly selected). At the fifth and final generation of selection, plants from selected populations produced 25% more seeds on average (with no changes in quality) than nonselected control plants when both were grown at low [CO2] (Fig. 7). In conjunction with this, selection plants delayed flowering by c. 4 d, had an extended life cycle, and produced 34% more total biomass than control plants (Fig. 7). Initially, no differences in physiology were detected between selection and control plants (Ward et al., 2000). However, in a more recent study with the same populations (and improved methods), Gonzàlez-Meler et al. (2009) found that respiration rates were lower in Arabidopsis plants selected at 200 ppm CO2 than in control plants (without reductions in N tissue content ), with a large portion of this response occurring from reduced activity of the alternative pathway. Reduced respiration rates improved the carbon budgets of the low-[CO2]-selected plants, representing a potentially important adaptive response to low [CO2]. Also with respect to the Ward et al. (2000) study, Bone & Farres (2001) surveyed the literature and found that the rate of selection at low [CO2] was among the fastest rates reported for a novel global change factor, based on both rates of change over time (yr) and generation number. Taken together, these studies show that low [CO2] can act as a strong selective agent on C3 plants, and suggests that single-generation studies may not be fully representative of the responses of glacial plants to low [CO2].

Figure 7.

 Mean values for seed number (a), time (d) to first flower (b), and total biomass (c) for selected and control (nonselected) plants of Arabidopsis thaliana at the fifth and final generation of selection for high seed number at 200 ppm. All plants were grown under the same conditions at 200 ppm CO2. Lines connect mean values for replicate control and selection populations, and open or closed circles designate different growth chambers. Note that low [CO2] was found to be the selective agent acting on these plants via subsequent reciprocal transplant experiments (see Ward et al., 2000) (adapted from Ward & Kelly, 2004, Fig. 3(a); Ward et al., 2000).

Using fossil plants from different regions, a variety of studies have shown that ancient plants exhibited higher stomatal density/index during low-[CO2] periods of the past (reviewed in Royer, 2001; also see Beerling & Chaloner, 1993; Beerling & Woodward, 1993; Wooller & Agnew, 2002; Roth-Nebelsick, 2005). A number of groups have stated that this may be an adaptive response to low [CO2] that reduced the resistance for entry of CO2 into the interior of leaves when [CO2] was most limiting, but potentially at the cost of higher water loss. Other groups, however, have not found a correlation between stomatal density/index and [CO2] and have argued against this possibility (see Körner, 1988; Bettarini et al., 1998; Maherali et al., 2002; and mixed responses reported by Knapp et al., 1994). We will briefly describe the data supporting both sides of this issue, and make some suggestions for possibly resolving this debate.

In support of this idea, Beerling et al. (1993a) reported higher stomatal density and index of European Salix herbacea L. that occurred during low [CO2] of the LGM relative to modern plants (Fig. 8). In addition, using specimens from packrat middens collected from the Great Basin, Van de Water et al. (1994) found that stomatal density was 17% higher in Pinus flexilis needles dating to the LGM relative to the Holocene; it is important to note, however, that in this study, higher stomatal density did not translate into higher leaf ci : ca (lower, in fact), possibly because of reduced g or changes in photosynthetic capacity. In addition to these findings, the relationship between stomatal density/index and [CO2] has been supported by reductions in stomatal density/index during the contemporary rise in [CO2], and such responses may serve to conserve water as CO2 becomes less limiting (Woodward, 1987; Beerling et al., 1993a). More specifically, Beerling (2005) found that UK populations of Selaginella selagenoides and Selaginella kraussiana exhibited a c. 30% decrease in stomatal index between the period when [CO2] rose from 280 to 360 ppm. Interestingly, Wagner et al. (1996) looked at this issue from an acclimation perspective, and showed that stomatal density decreased within the lifetime of a single birch tree (Betula pendula) during the contemporary rise in atmospheric [CO2] (stomatal density declined −0.6% for every 1 ppm increase in [CO2]). This group considered this a reflection of evolved plasticity that may have emerged from changing [CO2] over geologic cycles.

Figure 8.

 Stomatal density (a) and stomatal index (b) of fossil Salix herbacea leaves in response to changes in atmospheric [CO2] from the Last Glacial Maximum to the present (taken from Beerling et al., 1993a, Fig. 2).

Although many studies find evidence for higher stomatal density (or index) at low [CO2], c. 12% of studies find the completely opposite trend (Royer, 2001), and this may be an underestimate since negative results are usually more difficult to publish. For example, in a study of modern C3 grasses (Solanum dimidiatum, Bromus japonicus) grown across a [CO2] gradient (200–550 ppm), plants exhibited lower stomatal density at low [CO2] (Maherali et al., 2002), and instead had larger stomatal pore size, suggesting an alternative response to limiting [CO2]. In addition, Körner (1988) found no relationship between stomatal density and [CO2] for 200 plant species from central Europe that grew during the pre-industrial period (270 ppm CO2) vs modern times (340 ppm CO2), including both alpine and lowland species. Royer (2001) also pointed out that most studies supporting the notion of higher stomatal density or index at low [CO2] occur with fossil plants (88 and 94% of studies, respectively), with fewer studies showing this response in modern plants grown for short periods at low [CO2] (36 and 40% of studies, respectively). This may be indicative of an evolved response that occurs only after many generations at low [CO2], as suggested by Royer (2001), and/or may reflect the fact that other growth conditions (water, light, temperature) are less controlled in fossil studies than in modern experimental work. In opposition to Royer’s idea, Bettarini et al. (1998) studied 17 plant species growing near high [CO2] springs in central Italy where [CO2] was twice the modern value for at least two centuries, and found that stomatal density was unaffected by elevated [CO2] in the majority of species.

Despite these inconsistencies, the issue of stomatal index/density response to [CO2] of the past has increased interest and investment in the understanding of plant response to low [CO2]. Roth-Nebelsick (2005) laid out the challenges to this field, including high variance in stomatal measurements, especially in fossil samples, large inter-specific differences, nonlinear response to [CO2] increase, and differences between long-term and short-term studies. To date, a specific developmental mechanism linking stomatal index/density to [CO2] has not been determined, although the Arabidopsis HIC (high carbon dioxide) gene may play a role in this response (Gray et al., 2000). The authors suggest that these challenges may be addressed by striving to identify fossil material that is growing under more tightly controlled conditions (i.e. limiting differences in factors other than [CO2]), as well as advancing the understanding of molecular mechanisms tied to this response (Bergmann & Sack, 2007). One approach may be to conduct a thorough quantitative trait locus (QTL) analysis with model plants to determine the chromosomal regions (and eventually genes) that influence stomatal density/index across a range of [CO2] growth conditions. A more thorough understanding of the primary mechanisms driving stomatal density/index responses to [CO2] may eventually allow us to resolve the reasons for the inconsistencies reported in this field.

In addition to studying leaf morphology, researchers have also analyzed the carbon isotope ratios of fossil material to determine if there were evolutionary shifts in leaf physiology during low-[CO2] periods (Beerling et al., 1993b; Van de Water et al., 1994; Beerling, 1996; Pedicino et al., 2002; Ward et al., 2005). For example, Van de Water et al. (1994) calculated ci : ca from carbon discrimination values (that account for changes in carbon isotope ratios of source air through time) of P. flexilis leaves that occurred between the LGM and the Holocene. They found lower ci : ca during the last glacial period (0.55) relative to the Holocene (0.64), which was the result of either lower g or higher photosynthetic capacity. Most interesting here was the finding that neither ci : ca nor caci (representing the [CO2] gradient) were maintained through time, and ci values dropped as low as 109.7 during the last glacial period. Ward et al. (2005) also calculated ci : ca from carbon discrimination measurements in Juniperus trees from the Rancho La Brea tar pits in southern California. The authors found that ci : ca ratios were actively maintained over 60 000 yr (as hypothesized by Ehleringer & Cerling, 1993), and, as a result, glacial trees experienced ci values as low as 113 ppm CO2. If such reductions in ci were to occur in modern vegetation, plant growth would be reduced by half, according to studies with modern plants (Polley et al., 1993a; Van de Water et al., 1994; Ward et al., 2005). As a whole, these studies support the notion that trees were potentially carbon-starved during low-[CO2] periods because of glacial ci values that are, for the most part, unprecedented in modern vegetation.

VII. Interactions of low [CO2] with other factors

A variety of studies have examined the interactive effects of low [CO2] with other environmental factors, such as temperature, water, and nutrients. It is expected that low [CO2] would accentuate the effects of other stressors, mainly because plants would already be compromised by low carbon availability (Cowling & Sage, 1998; Cowling & Sykes, 1999; Ward, 2005). Certainly the interactions of low [CO2] with other environmental factors may have had synergistic effects on physiology, plant productivity, and evolution. To address this issue, recent studies incorporating the interactive effects of low [CO2] with temperature, water, and nutrients are described in the following sections.

1. Temperature

Sage & Cowling (1999) studied the interactive effects of low [CO2] and high temperature on several C3 species (P. vulgaris, wheat, and tobacco). They found that at high temperatures (day : night, 35 : 29°C), biomass production was reduced by 75–95% at 200 ppm CO2 relative to 380 ppm CO2, whereas at low temperatures (day : night, 25 : 20°C) biomass production was only reduced by 40–60%. Interestingly, the effects of high temperature on the low-[CO2] response could not be attributed to lower A (or, indirectly, higher photorespiration), but rather to reduced leaf area (relative to total mass) that resulted from lower leaf expansion and less leaf initiation (Sage & Cowling, 1999). From this and other similar studies, altered leaf development has been implicated as the primary response to high-temperature stress at low [CO2]. Moreover, these results suggest that C3 plants that occurred in warm regions, such as in the tropics, may have been most negatively affected by low [CO2] during glacial periods.

In a more recent study, Ward et al. (2008) grew A. theophrasti at low [CO2] (200 ppm), with both modern (30/24°C) and glacial temperatures (22/16°C). The authors discovered that any beneficial effects of low temperature on the C3 species, such as reduced photorespiration, did not overcome the very negative effects of low [CO2] on growth. Also, the authors could not detect an effect of temperature on physiology (A, respiration, g), biomass production, or total leaf area at low [CO2]. This may be explained by earlier empirical work (using Nerium oleander) showing that the sensitivity of C3 photosynthesis to increasing leaf temperature is diminished at low [CO2] (Berry & Björkman, 1980; see Fig. 4 therein). In the same study, Ward et al. (2008) found that C4 plants (Amaranthus retroflexus) were greatly inhibited by the lower glacial temperatures, but still maintained a large absolute growth advantage over the C3 species in response to low [CO2].

2. Water

It has been predicted that low [CO2] during glacial periods would have increased the water consumption of C3 plants, and this has been supported experimentally, mainly through single-generation studies (Baker et al., 1990; Polley et al., 1992, 1993a,b; Beerling & Woodward, 1993; Sage, 1995; Ward et al., 1999; Tonsor & Scheiner, 2007). Higher water loss is often a result of greater leaf biomass and/or area relative to total plant mass (proposed by Sage & Coleman, 2001; shown in Dippery et al., 1995; Ward et al., 1999 in well-watered plants; also see Medrano et al., 1995 where this response occurred during selection at low [CO2]), higher g, and greater stomatal density that enhance CO2 uptake on a whole-plant basis, but produce greater water loss. It is also worth noting that despite evidence indicating greater water demand of plants at low [CO2], precipitation was much lower during glacial periods on a global basis (Yung et al., 1996; Lambert et al., 2008). Such a combination of low water availability and low [CO2] may have been extremely stressful on plants and may have interacted to greatly reduce productivity during glacial periods.

In one example of work investigating this issue, Ward et al. (1999) grew both a C3 (A. theophrasti) and C4 (A. retroflexus) species at 180 (glacial), 270 (pre-industrial), 350 (modern), and 700 (elevated) ppm CO2 with severe drought treatments. In this case, the authors found that the C3 species responded to drought by dropping a large number of leaves, and retaining high water potential in remaining leaves at all CO2 treatments. At 180 ppm CO2, however, C3 plants retained relatively greater leaf area (by dropping fewer leaves) and delayed the lowering of g following the induction of drought relative to plants grown at 350 ppm CO2. The combined effects of these responses contributed to the maintenance of a positive carbon budget in the C3 plants grown at 180 ppm CO2. Surprisingly, the response of the C3 species produced similar degrees of relative recovery from drought at low [CO2] (assessed as biomass before and after drought) as in the C4 species. Thus, developmental and growth responses to low [CO2] and drought superseded the effects of physiological responses in this case, producing similar relative recovery of C3 and C4 plants that would not have been predicted from theoretical expectations. Note, however, that in this study (Ward et al., 1999) the C4 species still had much higher biomass on an absolute basis at low [CO2] compared with the C3 species.

In the same study (Ward et al., 1999), the C4 species (A. retroflexus) grown under well-watered conditions responded physiologically to changes in [CO2] with respect to water use. With increasing [CO2] between 180 and 700 ppm CO2, the C4 species showed gradual increases in A and decreases in g and transpiration that somewhat mimicked the responses of the C3 species. These physiological responses, however, did not result in enhanced leaf area and biomass production, but did serve to conserve water. Amaranthus was previously shown to reach CO2 saturation at a ci value of 200 ppm CO2 (Tissue et al., 1995), indicating that this species may not have been CO2-saturated at the low-CO2 treatment (180 ppm), explaining why it may have responded physiologically to increasing [CO2]. Similar to these findings, Polley et al. (1994) observed that Schizachyrium scoparium (C4) exhibited decreased g and had higher A between 200 and 340 ppm CO2, but plant growth remained unaffected. In addition, lower g and higher A have been observed in A. retroflexus (C4) and Setaria faberii (C4) in response to CO2 enrichment (Garbutt et al., 1990). Also, when grown in competition, elevated [CO2] stimulated the biomass production of A. retroflexus (C4) to a greater extent than A. theophrasti (C3) (Bazzaz et al., 1989). These results, as well as others (Cunniff et al., 2008), demonstrate that some C4 species can be quite physiologically responsive to changes in [CO2], suggesting that, like C3 plants, C4 species may have also had higher water demands during the LGM than in modern times.

3. Nutrients

Very little work has focused on the interactive effects of low [CO2] with nutrients. As mentioned earlier (Section V.1.), Polley et al. (1995) found that across a [CO2] gradient (200–360 ppm; without N treatments), WUE was decreased much more consistently at low [CO2] (200 ppm) than was NUE among C3 annual grasses and a woody perennial. This indicated that the requirements for water and nitrogen may have shifted between the LGM and modern times, with higher water demand appearing to be the more dominant and consistent factor. In a newer study with Populus deltoids (cottonwood), Lewis et al. (2010) showed that low [CO2] (200 ppm) inhibited the responses of A and biomass production to increases in phosphorus (P), whereas positive responses to P additions were more prominent at modern (350 ppm) and elevated [CO2] (700 ppm). Even more interesting was the finding that at high P concentrations (0.5 mM), biomass was less enhanced with increases in [CO2] below the modern value relative to increases above the modern value, reversing the trend for previous studies that did not include P treatments (see Section V.2.). In limiting P scenarios, Campbell & Sage (2006) found that white lupin (Lupinus albus L.) was restricted in its ability to respond to increasing [CO2] above 200 ppm with respect to A. As a result of these findings, the authors emphasized that nutrient availability must be factored into models predicting responses of ancient plants to low [CO2], as well as plant responses to increases in [CO2] between pre-industrial and modern times.

Sage (1995) predicted that symbiotic relationships (mychorrizal fungi, N-fixing bacteria) may have been less prevalent during glacial periods in response to low [CO2]. The high investment of photosynthate required to support symbionts may not have been as beneficial during a period when carbon was potentially more limiting than nutrients. In support of this idea, Polley et al. (1994) found that N fixation was reduced for plants (P. glandulosa) grown at low [CO2] compared with modern [CO2] with the same availability of soil N. Again, this speaks to the potential for major changes in ecosystem functioning as a result of shifts in limiting resources during low-[CO2] periods.

VIII. Low-[CO2] effects on community composition

Past work has incorporated low-[CO2] effects on community composition, with a focus on shifts in plants possessing the C3 vs C4 photosynthetic pathways. From stable carbon isotope ratios, it is clear that low-[CO2] periods were characterized by major shifts from C3 to C4 dominance in some regions (Ehleringer et al., 1991; Cerling et al., 1997, 1998; Street-Perrott et al., 1997; Flores et al., 2009). Other studies, however, have concluded that climate (temperature, precipitation) may have been a stronger force behind C3–C4 shifts relative to low [CO2] (Morgan et al., 1994; Latorre et al., 1997; Pagani et al., 1999; Huang et al., 2001, 2006; Keeley & Rundel, 2005; Flores et al., 2009; Edwards et al., 2010). Many of these discrepancies may be explained by the regional conditions of each study, and whether or not water availability and temperature further enhanced or negated the transition to C4 dominance.

Past studies have also investigated ancient shifts in plant communities in response to low [CO2] via empirical studies with pollen records, as well as modeling efforts. Using a process-based vegetation model, Jolly & Haxeltine (1997) predicted a dominance of xerophytic scrub in East Africa during the LGM compared with the current montane forest. In this case, modeled results showed that lowering [CO2] to the LGM value of 190 ppm produced a shift to the LGM scrub community, even under modern temperature and precipitation values (and may explain the enhanced African scrub belt from LGM pollen records). For this reason, the authors concluded that paleoclimate reconstructions based on present climate analogs may be unreliable if they do not incorporate the direct effects of low [CO2] on vegetation.

In a different study, Levis et al. (1999) predicted that low [CO2] of the past 21 000 yr favored grasslands and tundra over the forests currently located at temperate and boreal latitudes. These changes would have greatly affected evapotranspiration and possibly the entire water cycle of the region. This result suggested that large-scale changes in the characteristics of canopy cover (e.g. leaf area index, LAI) in response to low [CO2] may have had a more overwhelming effect on ecosystem functioning than smaller-scale physiological changes within individual plants (Levis et al., 1999). Furthermore, Cowling et al. (2008) modeled vegetation distributions in central Africa during the LGM and predicted that even though tropical broadleaf forests were not displaced during the LGM (with the exception of grassland encroachment on the edges), structural changes in forest canopy characteristics (e.g. lower LAI, tree height, and density) may have greatly altered ecosystem functioning. This change may have further impacted the migration patterns of numerous plant and animal species, including Homo sapiens. Taken together, these studies show that low [CO2] can influence the characteristics of community structure, ranging from alterations in canopy characteristics, shifts in C3 vs C4 dominance, to changes in biome structure.

IX. Low-[CO2] effects on the ecosystem

In order to better understand the effects of low [CO2] on whole ecosystems, global climate and biosphere models have been used to predict glacial values of NPP, terrestrial carbon storage, and isoprene emissions. The following section summarizes the results of these modeling efforts and discusses the benefits and limitations of scaling low-[CO2] effects on individual plants to the ecosystem level.

Using a mass-balance carbon isotope approach, Bird et al. (1994) constrained the increase in terrestrial carbon storage between the LGM and pre-industrial period to 310–355 Gt (or 270–720 Gt when oceanic δ13C uncertainties were incorporated), which supported previous estimates made by others using different approaches (e.g. Friedlingstein et al., 1992; Prentice et al., 1993; Van Campo et al., 1993). Later, François et al. (1998) coupled the CARAIB vegetation model to a general circulation model (ECHAM) in order to understand changes in terrestrial NPP and carbon storage across glacial–interglacial boundaries. Their modeling efforts predicted that NPP values increased from 38 Gt C yr−1 during the LGM to 53 Gt C yr−1 during the pre-industrial period; in addition, their estimates of changes in carbon storage overlapped that of Bird et al. (1994). François et al. (1998) attributed the possible low NPP during the LGM to changes in land area, cooler temperatures, and effects of low [CO2] on vegetation. In a subsequent analysis, they predicted that low-[CO2] effects on vegetation were the dominant factor in reducing NPP during the LGM, and a later sensitivity analysis supported this finding (Otto et al., 2002; François et al., 2006). In addition, Turcq et al. (2002) predicted from modeling efforts that carbon storage in Amazonia ranged between 44 and 94% of modern amounts, and this reduction was attributed to the effects of low [CO2] on vegetation, as well as cooler and drier conditions.

Harrison & Prentice (2003; using BIOME4) predicted that the effects of low [CO2] on plants may have reduced tropical forest cover by 44–69% during the LGM. Before this, Levis et al. (1999) also predicted that increases in C4 grasslands during the LGM, which occurred at the expense of forests, produced positive feedbacks associated with albedo that resulted in warmer and drier ecosystems, particularly in the tropics. Using a global-vegetation model, Crucifix et al. (2004) also found indications of major changes in vegetation distribution between the LGM and the pre-industrial period, with specific disappearance of the Siberian boreal forest, greater shrub cover in Europe, and an expansion of subtropical desert land area. Much of this change was attributed to differential effects of low [CO2] on plant physiology and water use, where grasses and shrubs had lower water requirements during glacial conditions than trees. By comparing modeled estimates to empirical continent-scale palaeoclimate data, Jolly & Haxeltine (1997) stated that efforts to determine paleo-temperatures based on tree-line depression may be faulty, and may overestimate reductions in temperature if low-[CO2] effects on vegetation are ignored. Taken together, these studies support the recent plea by Prentice & Harrison (2009) to incorporate direct physiological effects of low [CO2] on plants more widely when modeling past climates and ecosystem functioning, because failure to do so may result in erroneous conclusions.

Global vegetation models have also been used to better understand how changes in [CO2] may have affected post-fire recovery rates of trees and grasses from the LGM through modern times. In South African grasslands, Bond et al. (2003) predicted from modeling efforts that low [CO2] of the LGM reduced sapling growth to the point where young trees could not reach a fire-proof size between fire events, resulting in the elimination of trees in favor of C4 grasses. Furthermore, increased [CO2] during the Holocene may have allowed for the reappearance of savanna trees (Bond et al., 2003). Moreover, this study predicts that the interactive effects of low [CO2] and disturbance factors may have altered plant establishment patterns, which is an issue that had not been previously appreciated.

Plant emissions of trace gases can be influenced by changes in [CO2], climate, and other environmental factors, and can also feed back to influence climate change through the production of organic aerosols (Possell et al., 2005). In particular, isoprene emissions constitute approximately half of the biogenic volatile organic compounds released by the terrestrial biosphere, and therefore should be incorporated into the glacial–interglacial transition. To address this issue, Possell et al. (2005) grew the known isoprene-emitters Mucuna pruriens and Arundo donax in growth chambers at glacial (180 ppm), pre-industrial (280 ppm), and modern (366 ppm) [CO2]. On a leaf area basis, the lowest CO2 treatment produced a threefold increase in isoprene production relative to modern [CO2]. However, low [CO2]-induced reductions in above-ground biomass and LAI offset this increase, such that total canopy isoprene emissions showed no significant change between low and modern [CO2]. When cooler conditions of the LGM were incorporated, isoprene emissions were significantly reduced at the canopy level and on a leaf mass basis, but remained similar across all CO2 treatments. In a more recent study, Wilkinson et al. (2009) showed that Eucalyptus trees had significantly higher isoprene emissions at 240 ppm than at 380 ppm, while sweetgum showed no significant change across this gradient. It is still not clear why plants exhibit differential isoprene emissions in response to low [CO2] or even how best to represent these changes in modeling efforts. This emphasizes the need to better understand the interactive effects of climate and [CO2] on trace gas emissions of the past.

On a different note, little is known about how reduced [CO2] may have influenced the availability and quality (e.g. C : N) of food sources to herbivores and how shifts in vegetation may have affected animal diets during glacial periods (Ehleringer et al., 2002). The majority of studies dealing with these issues are mainly focused on other time periods such as the Cenozoic (MacFadden, 2005) and late Miocene–Pliocene boundary (Cerling et al., 1997). Hopefully in the future, more attention can be focused on the effects of low [CO2] on glacial animals.

In summary, a variety of modeling studies show strong evidence for reductions in NPP and carbon storage during glacial periods, with low [CO2] playing a major role in these changes. Much of this modeling work relies on empirical and theoretical changes in A and g with [CO2], mainly from Farquhar et al. (1980). As mentioned earlier, there tends to be a close correlation between A and growth (and possibly reproductive output) at low [CO2]. Therefore, the sole use of physiological data for estimating glacial NPP and carbon storage has validity, and is probably accurately representing our understanding of single-generation responses of modern plants to past [CO2]. It is quite possible, however, that as we gain further insight into our understanding of plant evolutionary responses to low [CO2], these large-scale estimates may need to be revised. As illustrated earlier (Section VI), Ward et al. (2000) found a 34% increase in plant biomass at 200 ppm CO2 following only five generations of selection for high fitness. Unfortunately, very few estimates of this kind exist. Thus, as we learn more about potential adaptive responses to low [CO2], these should be incorporated into future modeling efforts.

X. Low-[CO2] effects on early human societies

The emergence of agriculture among early human societies appeared throughout the world between 5000 and 10 000 yr ago, and this represents a rather short time span with respect to evolutionary change, particularly for crop plants. Sage (1995) proposed that such synchronous origins may have been the result of a common global factor, more specifically the rise in [CO2] from 200 to 270 ppm that occurred between 15 000 and 10 000 yr ago (Fig. 9). In this section, we discuss the role that low [CO2] may have played in the emergence of agriculture.

Figure 9.

 The timing for the emergence of agriculture among different human societies and changes in atmospheric [CO2] through time. Geographic names indicate the period when domesticated plants first appear in the archeological record for that region (from Sage, 1995, Fig. 1).

In the past, anthropological literature had mostly attributed the onset of agriculture to changes in human social and cultural structure and complexity (Bar-Yosef, 1998; Mannion, 1999). Although this has been a long-standing idea, cultural mechanisms were unlikely to have occurred simultaneously throughout all societies, and therefore this idea does not fully explain the synchronous emergence of agriculture (Cunniff et al., 2008). Other ideas have incorporated the effects of changing palaeoclimates, such as increasing precipitation and temperature (Bar-Yosef, 1998), although often, these are in the context of how climate impacted human society and not the crops themselves (Bar-Yosef, 1998). Although changes in climate are influential in their impact on human culture, a sole focus on this aspect of environmental change ignores the potentially large and direct effects of [CO2] on crops.

A plethora of past work strongly suggests that the increase in [CO2] that occurred between 15 000 and 10 000 yr ago may have been large enough to have had a profound impact on crop productivity, and hence on human subsistence patterns (Sage, 1995; Fig. 9). In general, glacial conditions would have been a hostile environment for C3 crops because of low [CO2], as well as drier soils and higher seasonal variation (Richerson et al., 2001). As the interglacial period commenced, the onset of rising [CO2] and other climatic changes would have removed an environmental limitation to the development of agriculture (Sage, 1995). Increasing [CO2] during the interglacial period may have directly enhanced plant productivity and may have reduced the effects of interactive stressors, such that crop production could be sustained year after year within human societies (Sage, 1995).

Anthropologists are beginning to incorporate the importance of CO2 into their ideas on agricultural development. For example, Bettinger et al. (2009) attributed the development of agriculture to the combined effects of climatic and cultural changes, including increasing [CO2]. He commented that once the external environmental constraints were removed, social innovation and more organized human communities became the predominant driver for the development of agriculture (Bettinger et al., 2009).

The importance of [CO2] in agricultural origins has met with some debate, mainly since some early crops were C4 (although these were in the minority), and it was initially thought that they would not have responded to increasing [CO2] (Cunniff et al., 2008). By contrast, however, a number of studies have shown that C4 species experience up to a 40% increase in biomass production between glacial and interglacial [CO2], as well as a 35% reduction in g, indicating that C4 species may have been strong responders to this rise in [CO2] (Cunniff et al., 2008). In addition, increasing WUE and A with rising [CO2] would have been most pronounced in the hot climates where development of C4 crops first occurred (e.g. millets in north China, sugarcane in New Guinea). Future research may be able to tease apart the impacts of [CO2], climate change, and cultural development on the emergence of agriculture among different societies. Such work will be enhanced by stronger collaboration among anthropologists, plant physiologists, and ecologists.

XI. Conclusions

From the studies highlighted in this review, as well as others, we now know that the influence of low [CO2] during glacial periods transcends many levels, ranging from physiological effects on individual plants to changes in ecosystem functioning, and may have even influenced early agricultural development. Results of these studies provide a baseline for plant response to minimal [CO2] that occurred during the evolution of land plants. Within a single generation of exposure to low [CO2], modern C3 plants show an average reduction in photosynthesis and biomass production of 50% (although this can range from 40 to 70% among species/genotypes). When scaled up in space and time, these physiological responses imply large reductions in NPP and carbon storage during glacial periods relative to the present, as well as changes in vegetation distributions, post-fire recovery, and trace gas emissions.

Despite these advances, we still have much to learn with respect to plant evolution at low [CO2]. Although numerous environmental factors were changing between glacial and interglacial periods, it is clear that CO2 alone can act as a strong selective agent on plants. For example, productivity increased by 34% within a C3 model system (Arabidopsis) after only five generations of selection for high fitness at low [CO2] (Ward et al., 2000). The potential for rapid genetic change in response to low [CO2] is great and carries implications for our understanding of ancient plant productivity and ecosystem functioning. For these reasons, an improved understanding of plant adaptation to changes in [CO2] is necessary to accurately predict plant function under both past and future conditions, and such information must eventually be incorporated into modeling efforts.

Overall, a better understanding of plant responses to low [CO2] will help us to understand how plants acclimated and adapted to changing carbon resources over geologic time scales. In order to advance our understanding in this area, it will be necessary to develop larger and more realistic growth environments, include more disciplines in order to move this field forward, extend experimental durations to allow for full acclimation and adaptation of plants, and continue studying the interactions between low [CO2] other environmental factors. Moreover, this will allow us to better predict future plant responses to anthropogenic increases in [CO2], since changes in plants that were a result of low-[CO2] periods in the past may still influence the potential for plants to fully utilize increasing carbon resources in the future.


The authors would like to thank the National Science foundation (CAREER award: 0746822), as well as the Wohlgemuth Faculty Scholar Award (with an endowment made possible by Ms Dorothy Lynch), and the Department of Ecology & Evolutionary Biology at the University of Kansas (KU) for support during the writing of this manuscript. In addition, LMG was supported by the C-CHANGE IGERT fellowship from the National Science Foundation (0801522) and from the Madison and Lila Self Graduate Fellowship at KU.