ABSTRACT
Premise of the Study Male gametophytes of seed plants deliver sperm to eggs via a pollen tube. Pollen tube growth rate (PTGR) may evolve rapidly due to pollen competition and haploid selection, but many angiosperms are currently polyploid and all have polyploid histories. Polyploidy should initially accelerate PTGR via “genotypic effects” of increased gene dosage and heterozygosity on metabolic rates, but “nucleotypic effects” of genome size on cell size should reduce PTGR. How are such cell-level consequences of genome size change related to the evolution of orders-of-magnitude faster PTGRs in angiosperms versus gymnosperms?
Methods We assembled a phylogenetic tree of 451 species with known PTGRs and added ploidy and DNA content values from the literature. We then used comparative phylogenetic methods to detect phylogenetic signal, differences in selective optima, and correlated evolution of PTGR and genome size.
Key Results Gymnosperms had significantly higher C-value and slower PTGR optima than angiosperms. DNA content was negatively correlated with PTGR in gymnosperms, but non-significant in angiosperms. Among angiosperms, model-based analyses indicated a single PTGR selective optimum for diploids and polyploids. Sister-taxon and intraspecific-cytotype comparisons indicated either no difference or slower PTGRs in polyploids than diploids.
Conclusions Results from closely-related taxa indicate nucleotypic effects are initially equal to or outweigh genotypic effects, and none of the analyses found that genome duplication accelerated PTGR. Thus, our results suggest that gradual and sustained expansion of PTGRs in angiosperms has mostly occurred in diploid or diploidized species, where pollen tubes are most subject to haploid selection.
INTRODUCTION
In seed plants, the male gametophyte is a highly-reduced haploid organism, developing within the pollen grain until pollination, when it germinates and a pollen tube emerges to invade maternal tissues. The pollen tube has been recruited for the novel function of siphonogamy, the transport of non-motile sperm cells to the egg-bearing female gametophyte, in conifers and Gnetales (two related gymnosperm groups) and in angiosperms (Friedman, 1993). In siphonogamous groups, pollen germination speed and pollen tube growth rate (PTGR) together determine fertilization timing. Much of the haploid genome of the male gametophyte is expressed during pollen tube development and also after fertilization, in the diploid sporophyte generation (Tanksley et al., 1981; Twell et al., 2006; Otto et al. 2015). Pollen competition is thought to be much more intense in angiosperms, causing greater exposure to haploid selection and faster evolution of genes for pollen germination speed and tube growth rate (Mulcahy, 1979; Snow, 1990; Arunkumar et al. 2013; Otto et al. 2015). Indeed, angiosperms have evolved orders-of-magnitude faster pollen tube growth rates than gymnosperms (Williams 2009, 2012).
One complication with the idea that pollen performance genes have evolved rapidly under haploid selection is that whole genome duplications (WGDs) have been especially common in angiosperms relative to gymnosperms (Leitch & Leitch, 2012, 2013). WGDs elevate pollen tube gene expression above the haploid level, which has consequences for growth rates. First, WGDs mask haploid selection on pollen-expressed genes via the effects of heterosis (sheltering of deleterious alleles and/or new allelic interactions) (Lande and Schemske, 1985; Husband and Schemske, 1997; Comai, 2005; Birchler et al. 2010; Husband, 2016). Secondly, WGDs enable gene dosage effects that could increase the capacity for protein synthesis and hence increase metabolic rates (Stebbins, 1974; Comai, 2005; Conant and Wolfe, 2008). Thirdly, large changes in DNA content (whether by WGD or other processes) are known to have “nucleotypic” effects on nuclear size, cell size and the duration of the cell cycle, independent of “genotypic” effects at individual loci (Bennett, 1971, 1972; Cavalier-Smith, 1978; Price 1988; Cavalier-Smith, 2005).
We know surprisingly little about the mechanistic effects of WGDs on gametophyte performance in general, and how they might balance each other out in any one species, or if there are any universal effects of WGDs on pollen performance specifically. Almost all studies of the effects of increased DNA content on cell size and growth rate have been on relatively isotropic cells with isotropic growth (Commoner, 1964; Holm-Hansen, 1969; Price et al., 1973; Grime, 1983; Shuter et al., 1983; Beaulieu et al., 2008). In these, increases in DNA content, whether by WGDs or other mechanisms are known to have the following effects: 1) a longer S-phase is required to replicate more DNA (Van’t Hof, 1965; Bennett, 1972), and 2) a longer period of cell growth is needed to grow a larger cell, because more DNA requires a larger nucleus and a proportionally larger cytoplasm (Gregory, 2001; Cavalier-Smith, 2005). As a result, in cell populations, including multicellular tissues, the population (tissue) growth rate is slower because the cell cycles of cell lineages are longer. Importantly, the volumetric growth rate of individual cells is rarely of interest, and may in fact stay the same after WGD, if the longer duration of the cell cycle is proportional to the increase in cell size.
If increases in cell cycle duration and cell size are disproportionate after an increase in DNA content, then volumetric growth rate of an individual isotropic cell must also evolve. Most roughly isotropic plant cells grow by diffuse growth, in which cell wall construction occurs simultaneously across the entire expanding plasma membrane surface (Albersheim et al., 2011). In such cells, surface area increases as the 2/3 power of volume, and consequently ever more cell wall material must be made per unit time during growth. This means that wall and plasma membrane production rates must either keep increasing just to maintain a constant rate of volumetric growth or that volumetric growth rate slows down as surface area increases. Diffuse-growing plant cells can partially reduce this problem by delaying some of their cell wall production until after they reach mature size (Cosgrove, 2005; Albersheim et al., 2011). But a large increase in cell size, such as after WGD, is expected to cause a disproportionate amount of extra construction costs, making it likely that a slower cell growth (expansion) rate will be the consequence.
The problem is quite different for anisometric, tip-growing cells, such root hairs and pollen tubes. First, pollen tubes are terminally-differentiated cells, so their cell growth rate is entirely restricted to G1 of the cell cycle. Secondly, as shown in Fig. 1, new cell wall production in a pollen tube occurs only in a very small region near the elongating tube apex (Chebli et al., 2012), and growth rate is directly determined by the rate of wall production (Rounds et al., 2011). Thus, in a pollen tube, cell surface area and cell volume are directly proportional to each other throughout growth, because the area of active wall synthesis remains constant during growth (Fig. 1). Given a constant nutrient supply and temperature, PTGR is also constant (Brewbaker and Majumder, 1961). Tube diameter is determined as the tube tip is formed and after germination a tube of constant diameter is maintained by self-similar growth (Fayant et al., 2010; Geitmann, 2011; Nezhad et al., 2013). This means the same amount of work is done at all stages of elongation, and consequently wall production rate does not become limiting at later stages and larger sizes as it does in isotropic, diffuse-growing cells. Furthermore, since a pollen tube only functions during growth, rather than as a mature cell, wall production is not delayed until after mature size is attained. In sum, the volumetric growth rate of a pollen tube directly reflects the amount of wall material being produced per unit of time (Williams et al., 2016). For a neo-diploid pollen tube relative to its haploid ancestor, the nucleotypic effect on PTGR will not depend on cell cycle duration or mature cell size per se, but only on the degree to which the amount of wall materials has been modified to accommodate its new DNA content (Fig. 1).
How should the nucleotypic effect play out in a cell that only functions during growth? In isotropic, diffuse-growing cells, genome size scales with cell size because the nucleo-cytoplasmic ratio is thought to be conserved for optimal function of the mature cell (Bennett, 1972; Cavalier-Smith, 1978; Price, 1988; Cavalier-Smith, 2005; Barow, 2006). In a pollen tube (Fig. 1), most of the functional cytoplasm, including the sperm cells, tube nucleus, and organelles, are confined to a space of constant size between the leading edge of the enlarging tube vacuole and the elongating tip (Hepler and Winship, 2015). Hence, a reasonable expectation is that the volume per unit of tube length of a growing diploid pollen tube would be up to double that of its progenitor haploid tube. Doubling volume per unit length results in a 41% increase in tube diameter and circumference (Fig. 1). That value is consistent with in vivo observations of a 39% larger diameter tube in an allotetraploid Nicotiana relative to the mean of its presumed diploid progenitors (Kostoff & Prokofieva, 1935), or of 8-53% larger tube diameters in tetraploid versus diploid species of Gossypium (Iyengar, 1938).
Tube shank wall thickness is generally seen as being near its minimal value to function in resisting turgor pressure and compression stress (Parre & Geitmann, 2005). Wall thickness did not scale with tube diameter in several angiosperms (Williams et al. 2016), and turgor did not scale with either tube diameter in fungal hyphae (Harold et al., 1996) or with PTGR lily (Benkert et al., 1997). Therefore, all else being equal, nucleotypic effects of genome doubling are expected to cause up to a 41% increase in the amount of wall material needed to extend a pollen tube tip a unit of length, which in turn is predicted to slow PTGR by up to 41%,.
Genome doubling can also affect PTGR by its genotypic effects on pollen tube energetics, which ultimately affect the rate of wall synthesis. Pollen tube energetics are still not well understood, and there are redundant aerobic and anaerobic respiratory systems (Rounds et al. 2010, 2011; Colaço et al., 2012; Obermeyer et al., 2013; Selinski and Scheibe, 2014). Yet it is well accepted that most of the metabolic energy expended by a growing pollen tube is allocated in one way or another to the production of new wall material (Bove et al., 2008). Genome duplication might be expected to increase wall production rates because of the effect of increased gene dosage on metabolic processes. Although gene expression levels can become rapidly modified during and after polyploidization due to gene loss, gene silencing, and dosage compensation (Freeling et al., 2015; Schoenfelder and Fox, 2015; Dong et al., 2016; Panchy et al., 2016), there is evidence that dosage effects have played an important role in the evolution of metabolic rates. For example, in yeast cells, alcohol dehydrogenase activity per cell increased linearly with ploidy across four ploidy levels (Dilorio et al., 1987). In plant tissues, endopolyploidy is common in cells that have secretory or transfer functions and endopolyploid cells have higher metabolic rates (d’Amato, 1984; Galbraith et al., 1991; Scholes & Paige, 2015). Finally, WGD-generated paralogs involved in metabolic pathways, such as glycolysis, can be preferentially retained as genomes undergo gene losses after WGD (Conant and Wolfe 2007).
Another genetic consequence of WGD on PTGR occurs via the fitness effects of diploidy over haploidy at the gene level. Diploidy is expected to initiate or maintain faster PTGRs for two reasons. First, via the sheltering of deleterious performance alleles that were present in one or both haploid ancestors or that arose after WGD (Husband and Schemske, 1997; Husband, 2016). Secondly, via the effect of heterozygosity (overdominance), which is known to cause heterosis for growth rates of sporophytes in intra- and inter-specific hybrids (Lippman and Zamir, 2007; Feys et al., 2018). These potential effects of diploidy on PTGR are expected in all neo-polyploids (Lande and Schemske, 1985), but the heterotic effects of allelic interactions and sheltering become more evident as genetic variation in the polyploid increases. Still, in all but the special case where there is no genetic variation, diploidization of an ancestrally haploid pollen tube predicts genetic effects on performance speed to result in greater than or equal to ancestral haploid speed.
The predicted effects of WGDs on pollen performance speed are summarized in Fig. 2. Increases in DNA content, by WGD or any other process, are expected to reduce PTGR by nucleotypic effects, primarily that of increased tube cell size on wall volume. In contrast, increases in DNA content by WGD involves genome-wide gene duplication, and are expected to increase PTGR through the genotypic effects of gene dosage and heterosis. Given that haploid selection on pollen tubes has been seen as an important arbiter of the evolution of flowering plant sporophytes, how robust has pollen tube growth and development been to WGDs? Is there evidence that WGDs been involved in generating the pattern of accelerated PTGRs within angiosperms?
In this study, we test the hypothesis that changes in DNA content have affected PTGR by asking if there are consistent macroevolutionary patterns that might distinguish among the alternative predictions described above. Specifically, we tested for differences in PTGR between diploids and polyploids at the intraspecific, sister-taxon, and macroevolutionary scale; and for correlations between PTGR and DNA content at the macroevolutionary scale. We examined the macroevolutionary pattern at the level of all plants that have pollen tubes (seed plants), and within the two major lineages that have different pollen tube cell biology and vastly different levels of species diversity, gymnosperms and angiosperms.
MATERIALS AND METHODS
Data collection
Data on PTGRs were taken from Williams (2012) and more recent literature (cited in Supplemental Information). Consistent with other comparative analyses of physiological traits, and with the way researchers measure PTGR from the longest pollen tubes, PTGR values for each species represent an average of maximum in vivo growth rates, or if there was more than one report for a species the average of those values (see Williams 2012 for details). PTGRs were taken from within-diploid or within-polyploid (i.e., never interploidy) crosses, in keeping with our overall goal of finding mechanisms underlying the pattern of PTGR evolution in stabilized polyploids. DNA content was analyzed using C-value: the amount of nuclear DNA in the unreplicated gametic nucleus, irrespective of ploidy level (Swift, 1950; Bennett and Leitch, 2012). While DNA content is equivalent to genome size in diploid organisms, it is a multiple of genome size in polyploids, thus we use the term DNA content throughout. C-value data was collected from the Kew Royal Botanic Gardens Plant C-Value Database (Bennett and Leitch, 2012). Ploidy level data was extracted from the same source as PTGR data where possible, and otherwise from the Kew Plant C-Value Database, TRY databases, or the Index to Plant Chromosome Numbers. As ploidy level is determined either by chromosome count or segregation pattern in these data, paleopolyploids and ancient duplication events are not included.
Tree Construction and Dating
GenBank accessions for 16 gene regions (rbcL, matK, trnL-F, 18s_rDNA, atpB, ndhF, adh, trnL, rpl32, trnT-L, psbA-trnH, rpl32-trnL, ITS, 5.8s_rRNA, rps16, and 26s_rDNA) for the 451 seed plant species with pollen tube growth rate data were retrieved, cleaned, and assembled into multiple gene alignments using PHLAWD and phyutility (Smith and Donoghue, 2008; Smith and Dunn, 2008). Tree inference was performed using maximum likelihood in RAxML version 8 (Stamatakis, 2014) on CIPRES. A pruned version of the seed plant tree from Magallón et al. (2015) was used as a guide tree to enforce topology of major clades. The resulting maximum likelihood estimate of the tree was rooted and ultrametricized using the ape (Paradis et al., 2004) and geiger packages in R (Harmon et al., 2008). Time-calibration was performed using the Congruification method (Eastman et al., 2013).
Character scoring
The PTGR value used for each species represents an estimate of maximum sustained growth rate. Since each PTGR value represents a species mean obtained from multiple measurements, we attempted to incorporate error into phylogenetic comparative analyses. OUwie allows for incorporating estimates of standard error (SE), and since species means were log-transformed for analysis, log-transformed SEs are required. As there is no reliable way to calculate the log-transformed SE from the literature without the original data for each species, we used the following conservative approximation. First, we assumed all species had similar SEs in PTGR, and we applied an empirically-determined SE from an exemplar species to all. Magnolia grandiflora has an average PTGR of 828 (±141) μm h-1 (N = 25 outcrosses), close to the angiosperm median of 587 μm h-1 (Williams, 2012 and this study) (Table S1). The standard deviation (SD) of log-transformed data was calculated and divided by the mean of the log-transformed data to acquire a coefficient of variation (CV) of 0.0237. We then multiplied the log-transformed mean PTGR of each species by 0.0237 to provide an estimate of taxon-specific standard deviation. The standard deviation was used as a conservative estimate of error because sample sizes were generally not available for calculating SE. We also performed a sensitivity analysis by running the OUwie analyses with hypothetical error values obtained set as 0, 0.05, 0.1, 0.25, and 0.5 (Table S2). Unless otherwise noted, all measures of uncertainty around parameter estimates are in standard error.
We analyzed two different sets of ploidy data. In the first, ‘restricted dataset’ (N = 273 taxa), we recognized only ploidy levels as stated in the literature. In the second ‘expanded dataset’ (N = 335 taxa), we inferred ploidy levels of 62 additional species by comparing their chromosome count to their generic base count as reported in Wood et al. (2009). Taxa with a chromosome number equal to or greater than twice the generic base count were considered polyploid.
Phylogenetic Comparative Analyses
Phylogenetic signal was assessed for both PTGR and gametophytic DNA content using Blomberg’s K in the picante package in R (Blomberg et al., 2003; Kembel et al., 2010). C-values and PTGRs were log10 transformed in this and all other phylogentic comparative analyses below. To visualize changes in DNA content and PTGR along tree branches and to generate estimates of node states, ancestral state reconstructions were performed and plotted using the contMap function in phytools (Felsenstein, 1985; Revell, 2012). Given many known biological differences between gymnosperms and angiosperms for both pollen tube growth (Friedman, 1993; Williams, 2008) and DNA content /polyploidy (see Discussion) (Ohri and Khoshoo, 1986; Leitch et al., 1998), all analyses were performed on gymnosperms only, angiosperms only, and the full dataset (all spermatophytes). Whether PTGR evolves under different selective regimes between the two seed plant groups was also evaluated using the OUwie function (OUwie package) in R (Beaulieu and O’Meara, 2014) to compare different models of evolution (detailed below). In multiple-regime models, the regimes were either gymnosperms (1) or angiosperms (2). The same analysis was used to test for differences in DNA content evolution between gymnosperms and angiosperms as well.
To detect whether pollen tube growth rates of polyploid species evolve under a different selective regime than diploid species, OUwie was used to compare single-regime and two-regime models. The following models were tested: single-rate Brownian motion (BM1), multi-rate Brownian Motion (BMS), single-regime Ornstein-Uhlenbeck (OU1), multi-regime OU with single alpha and sigma^2 estimates (OUM), multi-regime OU with single alpha and multiple sigma^2 (OUMV), and multi-regime OU with multiple alpha and single sigma^2 (OUMA).
The association between polyploidy and PTGR was also assessed among 20 diploid-polyploid sister taxa on the tree (within-genus or within-family level). Only polyploid taxa with a single diploid sister were used. The PTGRs of 11 intraspecific diploid-polyploid pairs from the literature were also compared. A two-tailed binomial (sign) test was used to test significance.
The relationship between the continuous traits, pollen tube growth rate and gametophytic DNA content, was assessed with phylogenetic generalized least squares (PGLS) regression using the phylolm package in R. Gametophytic DNA content was used as the predictor variable and PTGR the response variable. BM (Grafen, 1989) and OU (Martins and Hansen, 1997) models were both used, in addition to Pagel’s lambda, kappa, and delta models (Pagel, 1997, 1999).
To explore the level of convergent evolution within each trait and determine coincidence of selective regime shifts in PTGR and DNA content, the SURFACE package in R was used (Ingram and Mahler, 2013). SURFACE adds selective regime shifts to a single regime model in a stepwise fashion until adding another regime shift decreases the model likelihood. Regimes are then collapsed in a pairwise fashion until further collapses decrease the likelihood. Nodes that included a regime shift in both traits were identified.
RESULTS
PTGR evolution and C-value evolution in angiosperms versus gymnosperms
Blomberg’s K indicated significant phylogenetic signal across seed plants and within angiosperms, for both the PTGR and C-value datasets (Table 1). OUwie analyses of seed plants rejected both the Brownian Motion (BM) and Ornstein-Uhlenbeck (OU) single-optimum models for the evolution of PTGR and C-value. OU models that allowed separate selective regimes for angiosperms and gymnosperms accounted for > 99.9 % of the model weight in both PTGR and C-value (Tables S3, S4). Log10 PTGR selective optima were more than a magnitude of order higher in angiosperms (2.69 ± 0.048 μm h-1) than in gymnosperms (0.187 ± 0.123 μm h-1). Log10 DNA content selective optima were more than a magnitude of order smaller in angiosperms (0.184 ± 0.051 pg) than in gymnosperms (1.231 ± 0.041 pg). The maximum likelihood (ML) reconstructions of angiosperm and gymnosperm common ancestors (CAs) showed a similar pattern. Ancestral Log10 PTGR for angiosperms was higher than that of gymnosperms, 2.44 μm h-1 (95% CI: 1.09-3.69) versus 0.215 μm h-1 (95% CI: −1.48-1.92), respectively (Fig. 3), and ancestral Log10 DNA content was 0.29 pg (95% CI: −0.45-1.04) for angiosperms and 1.10 pg (95% CI: −0.28-2.47) for gymnosperms.
Joint evolution of PTGR and ploidy
- In the restricted ploidy dataset, the best fitting OUwie model for angiosperms using the Magnolia error estimate of 0.024 was the single-rate OU model which had 50.0% of the model weight (Table 2a); whereas the three two-regime OU models accounted for the remaining 50.0% of the model weight (both BM models were rejected). This pattern held up for a wide range of error values in the sensitivity analysis (Table S2), except that when error reached 0.5, a value > 20 times that seen in Magnolia. In the expanded ploidy dataset (Table 2b), the weight of the single-regime OU model was 57.2% compared to a combined 42.8% for the three two-regime models. However, the estimates for the polyploid selective optima are different from those of the restricted dataset; diploid optima in both datasets were 2.7 ± 0.07 log10 μm h-1, whereas polyploid optima were 1.4 ± 1.5 log10 μm h-1 in the restricted dataset versus 2.8 ± 1.00 l og10 μm h-1 in the expanded dataset.
A survey of intraspecific cytotypes found autopolyploids had slower PTGR than diploids in 9 of 11 pairs and no difference in the remaining two (Binomial test, P = 0.002; Table S4b). In the within-genus sister-taxon comparison, polyploids had slower PTGR than diploids in 9 pairs, faster PTGR in 2, and no difference in two (Two-tailed binomial test, P = 0.065). If expanded to within-family sisters, there were twelve, six, and two pairs, respectively (P = 0.238; Table S4a).
Joint evolution of PTGR and DNA content
For seed plants, simple linear regression (SLR) showed a significant negative correlation between DNA content and PTGR (P<0.0001), but that result was clearly driven by the large DNA contents and slow PTGRs of gymnosperms relative to angiosperms (Fig. 5), because the PGLS regression was non-significant (P=0.463; Table 3). Taking these two clades separately, DNA content was negatively correlated with PTGR in gymnosperms in the PGLS regression (SLR: P = 0.200; PGLS: P < 0.02, OU model; Table 3), but not SLR (phylogenetic signal was not significant for either trait in gymnosperms; Table 1). For angiosperms, it was positively correlated using SLR (P=0.0005), but non-significant using PGLS (P = 0.284; Table 3). In angiosperms, the PGLS regression was non-significant when polyploids were excluded (P = 0.457, kappa model).
Detecting coincident regime shifts in PTGR and DNA content within seed plants
- The SURFACE analysis of PTGRs found 13 distinct selective regimes (N = 451 taxon tree), with 51 selective regime shifts (22 to faster and 29 to slower regimes). For C-value, there were 9 distinct selective regimes (N = 184 taxon tree), with 4 shifts to larger and 7 shifts to smaller regimes. Regime shifts in PTGR and DNA content were coincident at only two nodes, a PTGR acceleration and genome downsizing in the CA of extant angiosperms and a PTGR slowdown and genome size decrease in the CA of rosids and (Fig. 6). Excluding terminal, branches, all other regime shifts were at least 3 nodes away from each other. PTGR shifts on the smaller tree (Fig. 6) were in the same locations as on the larger PTGR tree (not shown).
DISCUSSION
DNA content variation and the evolution of PTGR in angiosperms versus gymnosperms
We found that PTGR has evolved around a significantly faster selective optimum in angiosperms than in gymnosperms. Though it has long been known that most angiosperm PTGRs far exceed those of gymnosperms, this result provides quantitative support to the near-universal assumption that shifts to faster angiosperm-like PTGRs began prior to the origin of extant angiosperms. However, the slow angiosperm-wide selective optimum and reconstructed ancestral PTGRs (both < 490 μm/h), are consistent with gradual evolution of faster PTGRs in angiosperms.
There are several hypotheses for how and why angiosperms evolved such fast PTGRs. First, Mulcahy (1979) invoked a shift to much higher intensity of pollen competition in angiosperms as a driver of the origin and continued evolution of faster growth rates. Notably, no other type of tip-growing cell (haploid or diploid) in land plants has evolved comparably fast tip-growth rates and none of those cell types, including gymnosperm pollen tubes, experience intense competition for resources (Williams et al., 2016). Secondly, rapid PTGR may have been advantageous as angiosperms transitioned to a much faster reproductive cycle (Stebbins, 1974; Williams, 2012; Williams and Reese, in review). Thirdly, biophysical or physiological attributes of pollen tubes may have either constrained PTGRs of gymnosperms to be slow or enabled faster PTGRs in angiosperms, or both (Hoekstra, 1983; Derksen et al., 1999; Fernando et al., 2005; Williams, 2008, 2009). A fourth possibility, tested here, is that strong differences in genome-level processes between angiosperms and gymnosperms have impacted the evolution of angiosperm PTGRs relative to their living and extinct seed plant relatives.
We found that DNA content has evolved around a significantly lower selective optimum in angiosperms than in gymnosperms, even though angiosperms have a broad range of DNA C-values that encompass the entire range of seed plant genome sizes (Fig. 5; see Leitch and Leitch, 2013 for a larger survey). Paradoxically, angiosperms have great variation in ploidy level, a history of speciation by polyploidy, and much evidence of past genome duplication in both diploids and polyploids, including a WGD in a common ancestor of extant angiosperms (Wood et al., 2009; Husband et al., 2013; Van de Peer et al., 2017; Landis et al., 2018). Evidence for more recent WGDs is found in most modern diploids, which is consistent with studies of experimental WGDs that show that genome duplication is commonly accompanied by rapid loss of DNA sequences, gene fractionation by large-scale deletions, biased retention of genes with beneficial dosage effects, and ultimately a return to the diploid state (Conant and Wolfe, 2008; Conant et al., 2014; Freeling et al., 2015; Dodsworth et al., 2016; Wendel et al., 2018).
In contrast, WGDs have been very rare in gymnosperms (Leitch et al., 2005; Wood et al., 2009; Soltis et al., 2009; Husband et al., 2013; Leitch and Leitch, 2013), and their high DNA contents are thought to be due mainly to high transposon activity without repeated rounds of genome duplication (Leitch & Leitch, 2013; Lee and Kim, 2014). Hence, gymnosperms have experienced the effects of higher DNA content on pollen tube dimensions, which is predicted to reduce PTGR, without the potential counter-balancing effects of gene dosage and heterosis incurred by repeated rounds of genome duplications (Fig. 2), gene sorting, and diploidization.
Angiosperm pollen (Hoekstra, 1983) and sporophytes (Bond, 1989; Feild and Arens, 2005; Beaulieu et al., 2007) are known to have evolved much higher respiration rates than the pollen and sporophytes of gymnosperms. Respiration rates of angiosperm pollen are ~10 times higher than those of sporophytic tissues (Tadege and Kuhlemeier, 1997), and other energetic pathways also function (Rounds et al., 2011; Obermeyer et al., 2013). A major consequence of WGDs over single-gene duplication events is that WGDs can increase gene dosage of whole metabolic pathways at once (Conant and Wolfe, 2007). When particular metabolic pathways are limiting to growth rate, duplication can upregulate the pathway via gene dosage effects, and positive selection on dosage can shelter gene duplicates, allowing time for sub- or neo-functionalization, and hence much faster evolution of those genes (Kondrashov and Kondrashov, 2006; Conant and Wolfe, 2007). Pollen tube growth rate genes are widespread in the genome, for example, occurring on every chromosome arm of corn (Sari-Gorla et al., 1992). Thus, PTGR is a trait governed by many interacting genes that are widespread in the genome, which is exactly the kind of trait in which WGDs are favored over individual gene duplications or segmental duplications of chromosomes, due to the need to conserve gene dosage balance (Birchler et al., 2012; Conant et al., 2014; Freeling et al., 2015).
Growth rate evolution in sporophytic systems is consequent on growth of tissues, populations of cells, and rates and durations of many individual cell cycles. In contrast, PTGR genes with gametophytic expression govern a very restricted aspect of individual cell growth – mostly the intracellular transport system and cell wall synthesis machinery operating within a single, terminal cell cycle stage (G1, or G0) (Russell et al., 2012; Geitmann and Nebenführ, 2015). This means that the evolution of faster PTGR in angiosperms is likely to have imposed selection on a restricted set of gametophytically-expressed genes in metabolic and intracellular transport pathways (Arunkumar et al., 2015; Gossmann et al., 2016). Genes within such biochemical pathways might be preferentially retained after WGD due to the need to maintain balanced gene dosage effects (Birchler and Veita, 2012; Conant et al., 2014). Starting with the origin of flowering plants, WGDs have provided repeated opportunities for genome-wide, gene-level effects (gene sorting via positive and purifying selection, sheltering of deleterious alleles, and heterosis) to affect PTGR evolution. At the same time, given the small genome sizes of so many plants with a history of WGDs, the nucleotypic effect of doubled DNA content on cell size must often have been tempered by rapid genome downsizing following WGD (Freeling et al., 2015; Dodsworth et al., 2016; Wendel et al., 2018).
In summary, increases in DNA content by any mechanism are expected to initially decrease PTGR via the nucleotypic effects of larger genome size on cell size and growth, whereas increases in DNA content by WGD are expected to initially increase PTGR via the genotypic effects of increased gene/allele numbers (Fig. 2). PTGR evolution of gymnosperms may have been relatively stagnant for many reasons, such as lack of pollen competition, relaxed selection on PTGR due to slow reproductive cycles, and developmental constraints on their pollen tube structure and physiology. However, it may also be that their high degree of heterochromatin and rarity of WGDs have given them all the limitations of high DNA content but none of the evolutionary opportunities provided by large-scale redundancy due to WGD and fractionation cycles. Angiosperm PTGRs have evolved via repeated cycles WGDs, followed by fractionation, gene retentions due to advantageous dosage effects, sub- or neo-functionalization of paralogs, and eventually diploidization. Consequently, growth rate genes have alternately experienced periods of relaxed selection due to WGD and haploid selection on performance. Our findings that angiosperms have evolved around a higher optimum PTGR but a lower optimum C-value than gymnosperms, no coincident regime shifts to higher PTGR and higher C-value, and the lack of evidence for faster PTGRs in polyploids than diploids, suggests that the rapid PTGRs of angiosperms have largely evolved at the haploid level, in diploid or diploidized species.
Polyploidy and PTGR evolution within angiosperms
Diploid and polyploid angiosperms evolved under a single selective regime, irrespective of how ploidy level was scored. However, in both analyses the three two-regime OU models also accounted for substantial model weight (43% and 50%), and in these, the polyploid PTGR optimum was similar or slightly slower than the diploid optimum. Thus, we find no evidence that WGDs result in higher PTGRs. Two caveats are first, only 13-14% of the taxa in our two analyses were scored as polyploids versus 25-35% estimated for angiosperms (Wood et al., 2009; Landis et al., 2018); and secondly, on our tree polyploid taxa were generally recently-derived with relatively short branches, comprising a very small proportion of the total branch length of the tree. Hence, we may have had lower power to estimate parameters for polyploids relative to diploids. Still, the sister-taxon comparisons avoided these weaknesses and were consistent with the model-based results – polyploids had similar or slightly slower PTGR than diploids in most sister taxa on our tree.
Neo-polyploid PTGR is affected by the balance between genotypic and nucleotypic effects (Fig. 2). Since polyploid genetic variation can vary but affects only the potential for genotypic effects, our species might not have been a random sample of polyploid variation. Mating systems, modes of polyploid origins, and patterns of chromosomal segregation have effects on the degree of genetic variation and the nature of recombination in neo-polyploids. Yet, among 16 species scored as polyploid in both our analyses, eight were fully outcrossing, seven were self-compatible (two autogamous, two mixed mating, and four unknown), and 1 was apomictic – a not unusual distribution (Goodwillie et al., 2005; Gibbs, 2014; Ashman et al., 2014). Whitney et al. (2010) found only a weak correlation between DNA content and mating system, and this may reflect the fact that genetic variation in polyploids cannot be predicted easily. For example, autotetraploids originate with a subset of the genetic variation in the diploid progenitor population but they often outcross and hybridize, whereas allopolyploids can be highly heterozygous when they originate, but often are highly selfing (Stebbins, 1974; Soltis and Soltis, 1999; Whitney et al., 2010). In the special case in which there is no genetic variation in a diploid progenitor and its neo-autopolyploid descendant, PTGR would be solely determined by the balance between nucleotypic effects and gene dosage effects (which also depend on the degree of dosage compensation; Guo et al., 1996) (Fig. 2).
The closest approximation of the initial effect of genome duplication on PTGR, independent of levels of genetic variation, is the comparison of diploids with their intraspecific, autopolyploid cytotypes. In all 11 pairs, PTGRs of autopolyploid cytotypes were slower than or equal to those of their intraspecific diploid progenitors. We should re-emphasize that all studies involved in vivo diploid crosses (1x pollen on 2x pistils) compared to tetraploid crosses (2x pollen on 4x pistils). The lack of any examples of faster PTGR in neo-autotetraploid cytotypes than in their diploid progenitors (in which the effects of heterosis are minimized) seems to suggest that increased gene dosage generally cannot fully offset nucleotypic effects, causing slower or at best similar PTGR upon autotetraploid formation.
Disentangling nucleotypic from genotypic effects on PTGR evolution
Genome size is positively correlated with cell size and often negatively correlated with organismal growth rate across all kingdoms of life (Cavalier-Smith, 1978; Gregory, 2001). One major proposed explanation for the correlation between DNA content (which strongly influences nuclear size) and cell size in isotropic cells is that there is an optimal ratio of nuclear to cytoplasmic volume for metabolic efficiency (nucleoskeletal theory) (Cavalier-Smith, 1978, 2005). “Growth rate” in most studies refers to the doubling time of lineages of cells, and hence to the average duration of cell cycles, not to individual cell growth rates per se. Cell cycles are longer after genome duplication due to a longer DNA replication phase and also to the longer growth period needed to reach a larger size. Thus, the speed of cell enlargement in such studies is conflated with the duration of S-phase. This is not the case for a pollen tube, which is a terminally-differentiated cell that only grows and functions in the G1-phase of its cell cycle. Therefore, in pollen tubes, unlike in diffuse-growing cells of multicellular tissues, volumetric cell growth rate is the primary target of natural or sexual selection.
Large-scale genome size increases, whether by WGD or any other method, are expected to increase cell size to maintain the nucleo-cytoplasmic ratio. PTGR is directly determined by the amount and rate of cell wall production, which depends on tube size (Winship et al. 2010; Williams et al. 2016). A neo-diploid pollen tube with doubled tube volume must produce 41% more cell wall material per unit time to maintain its ancestral haploid PTGR (Fig. 1). Size increases of that order have been observed in diploid over haploid pollen tubes (Kostoff & Prokofieva, 1935; Iyengar, 1938). Hence, without increases in metabolic rates that could compensate for that extra work, PTGR and genome size are expected to be negatively correlated.
Importantly, we found such a negative correlation in gymnosperms. Nucleotypic effects should outweigh genotypic effects in gymnosperms, since the group has experienced large genome expansions without widespread gene duplication by WGD after their origin (Leitch & Leitch, 2013; Lee and Kim, 2014). Similar support for the existence of nucleotypic effects is that PTGR was generally slower in the youngest angiosperm polyploids (eg. intraspecific cytotypes or within-genus sisters). Over time, nucleotypic effects remain more or less constant in polyploids, but positive genotypic effects can accumulate via all the usual population genetic processes, counterbalancing or outweighing nucleotypic effects. Hence, the lack of correlation between DNA content and PTGR in angiosperms as a whole should be interpreted as a consequence of variation in the degree of compensating genotypic effects on PTGR, due to time since origin and genetic variation.
A weakness of comparative analyses of the question of DNA content and pollen performance is that all angiosperms, including “diploids” as determined by diploid segregation of chromosomes or by diploid base chromosome number, have at least one WGD in their ancestry (Landis, 2018, and reference therein). Thus, the most relevant measures of ploidy level are probably cryptic. For example, is pollen tube gene expression in a particular taxon haploid or polyploid? Perhaps a more relevant measure for comparisons is the number of WGDs among different lineages, traced back from each tip. Such data are already becoming available as more genomic data is accumulated.
Conclusions
Studies across the tree of life have consistently shown that ploidy level and DNA content are correlated with cell size and metabolic rate. Pollen tube dimensions and energetics affect the amount of cell wall material produced per unit of growth and the rate at which cell wall is produced, which together determine PTGR. In gymnosperms, PTGR was negatively correlated with genome size, but in angiosperms, where the effects of WGDs are more prevalent, PTGR seems to be somewhat robust to genome duplication. Neo-autopolyploids inherited similar or slower PTGRs than their diploid ancestors, and polyploids compared to their diploid near-relatives seem to follow the same pattern. With changes in genome size, nucleotypic effects act as a brake on growth rate and are always present, but the degree to which genotypic effects counterbalance these depends on the historical nature and time since genome size increase in any particular lineage. Understanding causal relationships between genome size, ploidy and PTGR will involve mechanistic studies of tube cell dimensions and wall synthesis rates in haploid and polyploid gametophytes. On the other hand, there appears to be great variation in the tug of war between genotypic and nucleotypic effects, and there are likely to be deeper evolutionary patterns underlying that variation.
Authors Contributions
J.B.R. and J.H.W. jointly conceived of the study and wrote the paper; J.B.R. collected data on genome sizes and ploidy levels, constructed the phylogenetic tree and performed all comparative analyses; J.H.W. collected data on PTGRs and diploid-autopolyploid PTGRs.
ACKNOWLEDGEMENTS
We thank B. O’Meara and J. Beaulieu for advice on phylogenetic analyses, I. Leitch for data on DNA content, and J. Edwards and M. Rankin for assistance in the lab. Several anonymous reviewers also provided useful advice. Partial support to J.B.R. was provided by National Science Foundation award IOS 1052291 to J.H.W.