Atmospheric CO2 decline and the timing of CAM plant evolution

Abstract Background and Aims CAM photosynthesis is hypothesized to have evolved in atmospheres of low CO2 concentration in recent geological time because of its ability to concentrate CO2 around Rubisco and boost water use efficiency relative to C3 photosynthesis. We assess this hypothesis by compiling estimates of when CAM clades arose using phylogenetic chronograms for 73 CAM clades. We further consider evidence of how atmospheric CO2 affects CAM relative to C3 photosynthesis. Results Where CAM origins can be inferred, strong CAM is estimated to have appeared in the past 30 million years in 46 of 48 examined clades, after atmospheric CO2 had declined from high (near 800 ppm) to lower (<450 ppm) values. In turn, 21 of 25 clades containing CAM species (but where CAM origins are less certain) also arose in the past 30 million years. In these clades, CAM is probably younger than the clade origin. We found evidence for repeated weak CAM evolution during the higher CO2 conditions before 30 million years ago, and possible strong CAM origins in the Crassulaceae during the Cretaceous period prior to atmospheric CO2 decline. Most CAM-specific clades arose in the past 15 million years, in a similar pattern observed for origins of C4 clades. Conclusions The evidence indicates strong CAM repeatedly evolved in reduced CO2 conditions of the past 30 million years. Weaker CAM can pre-date low CO2 and, in the Crassulaceae, strong CAM may also have arisen in water-limited microsites under relatively high CO2. Experimental evidence from extant CAM species demonstrates that elevated CO2 reduces the importance of nocturnal CO2 fixation by increasing the contribution of C3 photosynthesis to daily carbon gain. Thus, the advantage of strong CAM would be reduced in high CO2, such that its evolution appears less likely and restricted to more extreme environments than possible in low CO2.


ONLINE SUPPLEMENTAL APPENDIX S1
Sources and Assumptions for Dating the CAM Lineages Presented in Figure 1B and 1C.
To generate Figure 1 panel 1B, we report dates from phylogenetic chronograms for stem and crown nodes of CAM-specific clades.The identification of CAM specific clades was based on densely-sampled δ 13 C surveys, gas exchange analyses, or diurnal acid accumulation, which were then mapped onto phylogenies reported below.For lineages where the CAM status of sister clades are not known, we assumed they were C3 if not listed as CAM by Gilman et al. (2023), with the notable exception of the Crassulaceae where uncertainty in CAM distribution led us to not have confidence in this assumption.As Gilman et al. note, follow-up surveys of sister clades should be conducted as prior work did not always report non-CAM determinations, or did not survey sister clades if they lacked strong succulence.
For Fig. 1C, we used Gilman et al. (2023) and Smith and Winter (1996) to identify clades containing some CAM species, and use dated phylogenies for divergence dates of these clades.Generally, CAM determinations for species within these clades are limited to a handful of species, so it is not possible to identify CAM specific clades, nor even when CAM may have arisen.Therefore, the date of the divergence given are for the clade in which the limited number of CAM species occur.This approach generally will not identify CAM age in the clade, but does set the oldest boundary for CAM.Generally, CAM will be much younger than the clade age.
What follows is a listing of clades examined and the source of the phylogenetic dates.For each clade, see the Fig. 1 legend for the clade names corresponding to each histogram as given in the sources listed below.The order of presentation matches that shown in Fig. 1B and 1C.
We note that our survey identified 73 CAM lineages, in contrast to the minimum of 66 CAM lineages noted by Gilman et al., (2023).This discrepancy reflects different assumptions and authorities in the two studies, and the preliminary nature of the respective efforts given the incomplete knowledge of CAM diversity and phylogenetics within many CAM families.Rather than harmonize the two estimates, we chose to stay with our estimate of 73 lineages to recognize that identification of CAM lineages is an ongoing area of research.Li et al. (2019).et al., 2015).
Agave clades are Hesperaloe, the crown Yucca, and Agave clades as delineated by Heyduk et al. and dated by McCain et al. (McKain et al., 2016;Heyduk et al., 2022).Horn et al. (2014), where the histograms of Fig. 1B represent 15 of their 17 postulated strong CAM clades, as listed in the Fig. 1 legend.Anthacanthae + Balsamis was assumed here to represent one clade, Anthacanthacae clade 8.

Aizoaceae:
The three CAM clades from bottom to top in Fig. 1B are Tetragonia, Mesembyranthoideae, and core Rushoideae.Divergence dates for these clades follow Klak et al. for the Tetragonia clade, and Liede-Schumann et al. for the other two clades (Klak et al., 2017;Liede-Schumann et al. 2020).To estimate the divergence of these strong CAM clades we assumed the sister clades were either C3+CAM or C3, following their absence from the CAM genera listed by Gilman et al. (2023).Rubiaceae: Two lineages of epiphytic ant-plants in the tribe Psychotrieae of Rubiaceae are dated from Fig. 1 in Chomicki and Renner (2016).Myrmecodia beccarii, which occurs in northern Queensland, Australia, is maximum 3.6 Ma old.Three CAM species in Squamellaria endemic to Fiji are estimated at 1.6 Ma.
Zygophyllaceae: Bulnesia retama in the subgenus Larreiodeae is recently shown to be the only C3+ CAM species in the family (Mok et al., 2023).It splits from its sister species at 1 (crown) to 4 (stem) Ma (Böhnert et al., 2020).(2015).The ancestral condition of Peperomia appears to be C3, with CAM arising multiple yet uncertain times at more distal nodes in the Peperomia tree.

Asphodelaceae (Aloid clades):
The origin of the strongly succulent and largely CAM clade Aloe and relatives (=genera in the Aloiod clade which includes Aloidendron and Haworthia) is based on Additional Files 2 and 3 in Grace et al. (2015).We presumed the crown of Bulbine + Asphodeloideae indicates when CAM evolved in this clade.
Ottelia in the Hydrocharitaceae follows divergence dates presented in Li et al. (2019).
Pelargonium divergence is estimated by van der Kerke et al. (2019).Because the CAM clade is subtended within the genus, we used their crown node date.
Coleus dates are for the succulent, CAM containing clade B (Paton et al., 2018).
Apocynaceae: Four independent CAM clades are apparent in the Apocynaceae (Ceropegia, Cynachrum, Dischidia-Hoya, and Pachypodium; Gilman et al., 2023).Dates for divergence of the Ceropegia clade follows Bruyns et al. and Liede-Schumann et al. for the Dischidia-Hoya clade (Bruyns et al., 2015;Liede-Schumann et al., 2022).Date estimates for the Cynachrum divergence, and the Pachypodium divergence were estimated by taking their respective crown node length presented in Wang et al., dividing by the stem node length of the Dischidia-Hoya clade, and then multiplying these relative lengths times the stem node ages of the Dischidia-Hoya clade.This method was able to reproduce the stem node age of the Ceropegia clade in Bruyns et al. (Bruyns et al., 2015;Wang et al., 2023).et al., 2015;Roalson and Roberts, 2016).However, the known CAM clade of Codonanthe appears younger, diverging 3 to 6 Ma.This clade is shown to have weak CAM (Guralnick et al., 1986).

Asteraceae tribe
Crassulaceae: Dates are based on Messerschmid et al. (2020) for Crassulaceae and Crassula divergence, as discussed in the text.Uncertainty in phylogenetic dating and CAM occurrence lead us to show the bar for Crassulaceae fading in color towards the left of the plot.

Portulacineae:
The crown node of the Portulacineae is thought to be C3+CAM, given the distribution of C3+CAM character states in the clades that branch at distal nodes in the phylogeny.Strong CAM evolved multiple times in the individual clades of Portulacinae shown in Fig. 1B.Stem and crown node dates follow Wang et al. and Arakaki et al. for the Portulacineae and its CAM-specific clades Alluadia + Allauadiopsis, core Cactoideae, Grusonia, and a clade of Tephrocactus and Opuntia (Arakaki et al. 2011; Wang et al., 2019).Stem and crown node dates for strong CAM in the Anacampseros are from Ocampo and Columbus (2010) following CAM determinations by Guralnik et al. (2008).

Figure 1C :
Figure 1C: Clades containing CAM species, but CAM position uncertain.Pyrrosia ferns follow Wei et al. (2017) who conclude CAM likely evolved in a distinct Austral-Asian clade with deeply sunken stomata at 17.98 Ma.Peperomia dates are from the average of crown dates in Nauman et al. (2013) and Massoni et al.(2015).The ancestral condition of Peperomia appears to be C3, with CAM arising multiple yet uncertain times at more distal nodes in the Peperomia tree.
Senescioneae: Three succulent clades with confirmed CAM are present in the Asteraceae -the Gynuroid clade with multiple CAM genera (for example, Kleinia, Curio and Senecio meuselii), the Caputia clade, and the Othonna/Crossothonna clade.The divergence dates of these clades are based on the ITS/ETS chronogram presented in Fig. 5 of Pelser et al. (2010).Other succulent Senecioneae not shown (e.g Pictocaulon; Cicuzza et al., 2017) branch at nodes that are less than 5 Ma but the presence of CAM requires confirmation.Clusia: Late-Miocene Clusia divergence dates are based on Ruhfel et al. (2016).Lujan et al. (2022) show strengthening of CAM across the Clusia phylogeny following its divergence.Curcurbitaceae: CAM divergence of two separate CAM-containing clades, Seyrigia and Xerosicyos, are dated in the Fig. 2 and Fig. A1 chronograms of Guo et al. (2020).Gesneriaceae: Two CAM lineages are identified in the African violet family, one in the Ramondinae (Haberlea/Ramonda) clade, and a second in the Cadonanthe clade.Petrova et al. date the Ramondinae spilt from a non-CAM sister clade at a crown age of 24.5 Ma, and a stem of 30.5 Ma while Roalson and Roberts date the CAM-inclusive Codonanthe-Nematanthus clade to diverge at 14 to 15 Ma (Petrova