Grass species with smoke‐released seed dormancy: A response to climate and fire regime but not photosynthetic pathway

Abstract Both C3 and C4 photosynthetic pathways and smoke‐released seed dormancy occur among grasses. C4 species evolved from C3 species as seasonality and fire frequency increased and might therefore imply that their smoke sensitivity increased. I searched the worldwide literature for reports on germination responses among grasses, whose photosynthetic pathway was known, to treatment by smoke. Data were obtained for 217 species and 126 genera. While subfamilies tended to be C3 (Pooideae), C4 (Chloridoideae) or a mixture (Panicoideae), a beneficial smoke response was independent of their photosynthetic pathway. The only exceptions were Danthonioideae (C3, non‐smoke responsive) and Triodia (C4, smoke responsive). One third of both C3 and C4 genera were smoke responsive. Even within genera, 90% of species showed contrasting smoke responses, confirming that smoke sensitivity is rarely taxonomically constrained. Data on photosynthetic pathway, climate, fire regime and vegetation were compiled for 15 regions that formed four distinct groups: 1) In warm climates with aseasonal rainfall, C4 grasses are moderately better represented, with crown fires and limited smoke responses. 2) In cool regions, most species are C3, with surface‐crown fires and lack smoke responses. 3) In warm regions with summer rain (savannas), most species are C4, with surface fires and lack smoke responses. 4) In Mediterranean‐climate regions with summer drought, most species are C3, with crown fires and smoke‐released dormancy. Thus, even though C3 and C4 grasses are equally capable of expressing smoke sensitivity, their response depends on the region’s climate and fire regime that also dictate which photosynthetic pathway dominates.


INTRODUCTION
Grasslands with C4 began to replace the C3 grasslands in Africa 14-16 million years ago (Ma), with a marked expansion at 4-9 Ma, which has been attributed to the onset of many environmental changes, especially the onset of summer rain and more frequent fires (Retallack 1992;Jacobs et al. 1999;K€ urschner et al. 2008;Osborne 2008;Scheiter et al. 2012). C4type photosynthesis has arisen at least 20 times among tropical grasses (family Poaceae) throughout the world (Linder et al. 2018), and it is now widespread at both the subfamily and generic levels (Edwards & Smith 2010). Subsequent reviews either make no reference to fire in the conversion of C3 to C4 grasses (Linder et al. 2018) or only mention fire in relation to resprouting (Simpson et al. 2021). None mentions fire effects on germination, although it is well-established that dormancy release among many grasses is promoted by smoke (Ghebrehiwot et al. 2012;Schwilk & Zavala 2012;Carthey et al. 2018).
Seed dormancy is expected to be most adaptive in highly seasonal climates, where seedling recruitment is only likely to succeed once the soil is moist and moderately warm. In fire-prone habitats, optimal recruitment conditions are favoured by the post-fire environment when large bare patches, with increased levels of light, moisture and nutrients, are created. Thus soil-stored seeds will have an adaptive advantage if their inherent dormancy is broken by smoke released from fire that heralds optimal recruitment conditions (Pausas & Lamont 2022). Environmentally induced dormancy will then be maintained until the onset of the first substantial rains at mild temperatures. This imposed dormancy period may be short-circuited if the smoke chemicals are not absorbed until dissolved in soil water associated with the wet season. Once absorbed, the karrikins link with dedicated proteins to produce hydrolytic enzymes that degrade stored carbohydrates and proteins, which enable germination to commence once hydrothermal conditions are suitable (Lamont et al. 2019). There is some evidence that smoke chemicals may also hasten the rate of germination, as distinct from their effect on dormancy release (Hodges et al. 2021), which might be especially adaptive under summerrainfall climates with their less effective rain.
This raises the question as to whether the germination of C4 grasses is therefore more likely to be promoted by smoke chemicals than that of C3 grasses. Is this another trait, like resprouting (Paula & Pausas 2011), that gives C4 grasses a competitive edge in frequently burnt savannas but that has been ignored? However, current evidence actually shows smoke-released dormancy is poorly represented among some C4 grasses, and the reverse may be true. The possible explanations are numerous, including fires are so frequent that seed dormancy is no longer adaptive (Pausas & Lamont 2022). C3 grasses are widespread in uniform rainfall, cool to temperate climates that may or may not be highly fire-prone (Tsuyuzaki & Miyoshi 2009). In particular, they are also characteristic of highly seasonal, fire-prone mediterranean climate regions, where smoke-released dormancy is also expected to be well represented (Roche et al. 1997;Reyes & Trabaud 2009). To resolve these conflicting expectations on the relationship between photosynthesis type and smoke responsiveness, I collated data from the literature on the effect of smoke on germination of C3 and C4 species located throughout the world and show that the answer is almost entirely context-driven (climate plus the associated fire regime), rather than dependent on their photosynthetic pathway or taxonomic affinity.

METHODS
I fed the keywords 'grass', 'smoke', 'seed' and 'dormancy' into Google â and Google Scholar â . Papers were selected that met the criteria of: (i) >3 grass species tested for the effect of smoke on breaking dormancy versus controls including their statistical significance; (ii) covered most fireprone regions of the world and included data on climate, fire regime and vegetation type; and (iii) could be downloaded from the web or were available through the Curtin University library. Since most papers used smoke water, there was no reason to suspect a bias in the type of smoke application between C3 and C4 species, and the focus was on presence/absence of a smoke response rather than its quantitative effect. No distinction was made between tests using dry smoke, smoke water or karrikins. This yielded 40 suitable papers that are cited in the reference list. Results for each species were placed in the categories 'plus' (smoke beneficial), 'nil' or 'minus' (smoke inhibitory) based on statistical tests in those papers. I then searched the web for whether the photosynthesis type of each species was known. Suitable data were available for 217 species and 126 genera. The subfamily for each species was noted. I also tabulated the percentage germination of controls and smoke treatment, if provided, and whether considered statistically significant. Suitable data were available for 145 species in 61 genera. The data were collated into 14 regions based on climate, fire-proneness and vegetation type. The raw data are given in Table S1.
To test whether smoke responses were not constrained by taxonomic affinity or photosynthesis type, I undertook four tasks. Species were allocated to their smoke response (positive versus nil or negative response) within their subfamily that was either C3 (Pooideae), C4 (Chloridoideae) or mixed (Panicoideae), with the 32 remaining species placed in a miscellaneous group of five families, to see if smoke responses were unaffected by subfamily/photosynthesis type. Since there were seven C3 and 81 C4 species in Panicoideae, this provided an opportunity to test if they were allocated equally between the two smoke-response types as a mark of their independence within a subfamily. To test for independence at the withingenus level, whether the smoke response was the same or mixed for those with two or three or more species was collated. To ensure possible bias introduced by varying number of species per genus was removed, each genus was represented by a value of one if all its species had the same smoke response, and 0.5 in each category if they differed. Where actual germination data were compared, the means of the smoke response and control for all species in the genus were calculated so that each genus contributed only one value to each category in the analysis.
The categorical data were submitted to Yates-corrected Chisquare or Fisher's Exact tests, depending on class sizes (Vas-sarStats, 2012-21©). Tests were two-tailed when there was no a priori expectation of a difference between the smoke-response types, and one-tailed on the expectation that smoke would promote germination over the controls. Since these analyses showed that smoke-released seed dormancy is spread widely and equally across the two photosynthesis types at taxonomic ranks from subfamilies to subgenera, this indicated that, even when there were taxonomic constraints on photosynthesis type, there were no photosynthesis-type constraints on the smoke response. The smoke response of each species was therefore treated as an independent data point in subsequent analyses. This approach was also supported by the species occurring in widely disparate geographic locations, minimizing opportunities for spatial autocorrelation of taxonomically related species (except perhaps for Triodia). The 2 9 2 quantitative (germination) data could not be analysed by conventional two-way ANOVA as the smoke-control pairings created a third variable, a species effect, so were not independent. Thus, a non-paired ttest was conducted on the C3 versus C4 data per genus, and a paired t-test was conducted on the smoke versus control data per genus.
When relative number of species in the region with a smoke response was regressed against relative number with C4-type photosynthesis, the data formed four highly disparate groups: groups 1/3 were separated from 2/4 by at least a 25% difference in percentage C4, and group 1 was separated from groups 2-4 by at least 25% smoke-responsiveness, making more formal ordination approaches to identify groups unnecessary. Data for the Restionaceae in the Cape province of South Africa were added as one dataset among the 15 collated by region at this stage as: (i) no grass data were available from this region, (ii) it is sister to the Poaceae and its growth form and habitats are Highlights • On a world scale, smoke-released seed dormancy exists equally among 40% of C3 and C4 grass species and one-third of genera and is not taxonomically constrained. • Some floras have only C3 grasses (summer dry) and some only C4 (summer wet) and most are dominated by one or the other. • Smoke sensitivity and photosynthetic pathway are readily divided into four regional types based on climate and fire regime. • Smoke sensitivity is poorly represented in frequentlyburnt C4 grassland/savannas but is almost universal in intensely-burnt C3 mediterranean shrubland/forests. • Moderate levels of smoke sensitivity among C3/C4 grasses under intermediate climates confirm that the presence/absence of smoke sensitivity elsewhere is unrelated to photosynthetic pathway.
matched, and (iii) this clade is well studied for its smoke requirements (He et al. 2016). The means AE confidence intervals at 95% of these groups were then calculated for comparative purposes. The common climate, fire regime and vegetation types for each group were obtained from the relevant papers and placed in categories for comparative purposes.

RESULTS
Ignoring their geographic distribution, the germination of 39.2% of 217 worldwide grass species benefitted from smoke (Table 1A). While the best-represented subfamilies were either C3 (Pooideae), C4 (Chloridoideae) or mixed (Panicoideae), their smoke responses were matched (41-45% of species). The only exception appears to be Danthonioideae, whose ten species were all C3 and non-smoke-responsive (Table S1). Within the mixed subfamily (Panicoideae), there was no difference in smoke response between C3 (43%) and C4 (41%) species (Table 1B). Within genera, for half of those represented by two species, half had one smoke-responsive and one non-smokeresponsive species; of those with three or more species, almost 90% were mixed smoke and non-smoke responsive (Table 1C).
The only exception appears to be Triodia whose seven species were C4 (or CAM) and smoke-responsive (Table S1). Removing any bias caused by varying number of species per genus on the results by representing each genus by one type or another (or 0.5 to each if mixed) resulted in identical smoke responses between C3 (34%) and C4 (33%) (Table 1D). Overall, on a per genus basis, C3 species germinated~13% better than C4 species, for both controls and smoke-treated (P = 0.014, twotailed) (Table 1E). Smoke improved germination of both C3 and C4 genera by~8% (P < 0.001, one-tailed). The smoketreated C3 species germinated 21% more than the control C4 species. That global analyses as described above obscure patterns revealed at other scales become clear when the datasets per region are compared (Table 2). Five regions have no C4 grasses, and three regions have only C4 grasses, with an average of 46.7% of grasses possessing C4 across all regions. One region has only smoke-responsive grasses (SW Australia)although the sample is small -, and one region has no smoke-responsive grasses (Cerrado in Brazil), with an average of 33.9% positively smoke-responsive on a regional basis. When inhibition cases are considered, this drops to a mean of 27.0%. Overall, the data are readily divisible into four groups based on climate, vegetation type, fire regime, the relative incidence of C4 and smokeresponsive grasses.
The first group is composed solely of C3 grasses (or more generally graminoids), with 75-100% smoke-responsive in mediterranean (summer-dry) shrublands to forests with moderately frequent crown fires (Fig. 1). The second group are essentially C3 with 0% (wetter)-25% (drier) smoke-responsive in grasslands/shrublands in cool-cold, summer-wet to aseasonal climates with surface fires, sometimes almost fire-free. The third group are essentially C4 grasses with 0% (wetter)-40% (drier) smoke-responsive in grasslands, summer-wet often frequent surface fires except in semiarid regions. The fourth group is intermediate, with a mixture of C3/C4 species, grassy woodland to forests with uniform rain and moderately frequent crown fires. In groups 3 and 4, there was a gradient from relatively wet to dry in terms of increasing abundance of smoke-responsive species. This was most notable in group 3 with the high-rainfall, Cerrado grasslands lacking any smokeresponsive species (90% C4) to semiarid NW Australia with 40% of grasses being smoke-responsive species (100% C4). This pattern was not evident among the three mediterranean regions.
The trends become clearer when the data for the various regions in each group are collated (Fig. 2). When they are arranged to give an increase in C4 species representation across the four groups, from 0% to almost 100%, equivalent gradient patterns for the other three attributes only arise when the cool-region group, with lowest incidence of smoke-responsive species, is omitted (as in Fig. 1). Here, smoke-responsive species in the mediterranean group (essentially C3) is 2.5 times that in the uniform-rainfall group, with similar total rainfall (C3 and C4 equally abundant) and 4.5 times that in the summer-rainfall group (essentially C4). All smoke-responsive species in the mediterranean regions were C3, and this was four times the level for C3 species in the uniform-rainfall group. Half of the C4 species in the uniform-rainfall group were smoke-responsive, and this was 2.5 times that in the summer-rainfall group which was essentially C4. Thus, the percentage of C3 species that was smoke-responsive in the mediterranean group was much higher than in the cool and uniformly wet groups, but the percentage of C4 species that was smoke-responsive in the groups was much higher than in the summer-wet group.

DISCUSSION
This meta-analysis of C3/C4 grasses versus smoke-released seed dormancy at a global scale shows that there are almost no taxonomic or photosynthetic pathway constraints to the presence of smoke responsiveness by a species. At any taxonomic rank, a given species is~40% likely to be smoke-responsive, whether its photosynthesis type is C3 or C4 (Table 1). Thus, each species can be treated as an independent unit for statistical purposes. The major constraints instead are centred on climate and fire regime to which the incumbent species have adapted over time. Clearly, there is something about the mediterranean environment that promotes both the presence of C3 grasses and smoke-responsiveness, and the savanna environment that promotes both the presence of C4 grasses and dearth of smokeresponsiveness, even though both their vegetation types are highly fire-prone. To this we can add: there is something about intermediate and cool climates that promotes intermediate levels of both types.
The stimulus for germination of most species, here C4 grasses, proved not usually to be caused by smoke under grassy savanna climates (with frequent surface fires). For example, the seeds (caryopses) of many grass species in the Cerrado have a large non-dormant component (Fontenele et al. 2020). However, it would be misleading to interpret this to mean that the seeds of C4 grasses are not fire-adapted. Thus, seeds of three of five C4 grasses from Brazilian grasslands germinated as well as the controls after 2 min at 90°C (Overbeck et al. 2005), showing substantial heat tolerance. Cuello et al. (2020) recorded 21 grass seedlings in a burnt Uruguay C4 grassland, compared with just four in the unburnt sites. And non-dormancy coupled with (a) high heat tolerance and/or (b) fire-stimulated flowering (Fontenele et al. 2020) enables seeds to take advantage of exceptionally wet seasons either (a) following or (b) between fires, which stored, smoke-responsive seeds cannot do, as they must germinate whenever fires occur, even if followed by a poor growing season.
On the other hand, most species, here C3 grasses, under mediterranean climates (with moderately frequent crown fires), store their seeds in the soil and dormancy breakage requires a fire-related stimulus, here smoke (Lamont et al. , 2019. Again, species in cool-cold climates, here predominantly C3 grasses (as noted by Simpson et al. 2021), with a wide range of fire frequencies, may rely on factors other than smoke, such as cold stratification, to promote dormancy release (Montalvo et al. 2002). Fire-prone woodlands and forests have species, here a mixture of C3 and C4 grasses, with stored and nonstored seeds that may or may not be smoke-responsive as recruitment may also occur inter-fire (Carthey et al. 2018).
Thus, the smoke response is driven solely by the fire regime (seasonality and frequency) as controlled by the climate and its associated vegetation. No physiological interaction with the photosynthetic pathway is required as both C3 and C4 grasses can be equally smoke-sensitive. The gradual evolution of C4 grasses out of a C3 heritage has been attributed to a plethora of factors (see Introduction), including a strong correlation between resprouting in response to frequent fire and the adoption of the C4 pathway (Simpson et al. 2021). But the concurrent adoption of smoke-released seed dormancy, as another potential trait to cope with frequent fire, is not one of them. In fact, the dearth of seed storage among summer-wet grasslands Table 2. Number of Poaceae species with germination stimulated (plus), unaffected (zero) or inhibited (minus) by smoke treatment. Number of species with C4-type photosynthesis (rather than C3) used to rank sites from 100% to 0%. Note that the Restionaceae ('grass') was used for the Cape province, South Africa, as no data were available for species in its sister family, Poaceae, in this mediterranean region. The four shaded bands represent the four response groups recognized in Figs  is notable Pausas & Lamont 2022). Meanwhile, seed dormancy and its breakage by fire-related stimuli became almost mandatory among the C3 grasses in the developing mediterranean climate regions over the last 10 million years as the increasingly severe summer-drought, fire-prone environments deterred inter-fire recruitment Rundel et al. 2018). Note that, since the binary data required were only available for a fraction of grass taxa, the Fig. 1. Relative number of grass species with smoke-assisted seed dormancy breakage in 15 regions, as described in Table 1, versus relative number of grass species in those studies with C4-type photosynthesis. Four types of response can be recognized, based on percentage C4 species, vegetation type, rainfall seasonality and type and frequency of fire. An exponential-type curve has been fitted to the 11 sets of studies in warm regions (circles) omitting the four C3-dominated floras in the cool-cold regions (diamonds). The regions dominated by C3, C4 or mixed grasses are indicated by the coloured columns. Fig. 2. Allocation of C3, C4 and smoke-responsive species to the four groups recognized in Table 2 and Fig. 1, except that Restionaceae data were omitted from group 1. Contingency table analyses given are based on raw data for presence or absence in each category. Temperature = mean annual temperature class (see Table 2 notes). Note the rise in C4 species representation (first bar) across the four groups but no gradient pattern for the other three attributes assessed.
results of this meta-analysis must be regarded as indicative rather than absolute.

ACKNOWLEDGEMENTS
This paper was prompted by work for a major review on firereleased seed dormancy (Pausas & Lamont 2022) but this topic was too complex to be pursued there. Thanks to colleagues Hermann Heilmeier and Fernando Silveira, and a reviewer, for their useful comments on drafts of this manuscript. My retirement is funded by a university pension that allows me to pursue my research interests at my leisure. I declare no conflict of interest. Open access publishing facilitated by Curtin University, as part of the Wiley -Curtin University agreement via the Council of Australian University Librarians.

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of the article.