Despite a wide range of variation, the taxa of the Gossypium share a number of common features. The pollen grains of all species examined are monad, radially symmetrical, spherical, echinate, panporate and operculate (Saensouk and Saensouk, 2021). Pollen shape is determined based on the ratio of polar diameter (P) to equatorial diameter (E) and can be classed into five categories, i.e. oblate (P/E < 0.50), suboblate (P/E = 0.5–0.88), spheroidal (P/E = 0.88–1.14), subprolate (P/E = 1.14-2.00), and prolate (P/E > 2.00) according to Erdtman (1987). We find that pollen shape is spheroidal in all investigated species, which is congruent with the results of previous studies (Jia et al., 1988; Lan and Xu, 1996). Significant variation in size and sculpture of pollen shows great taxonomic potential in identification and delimitation of species (Ullah et al., 2018) In this study, there is significant interspecific differences in pollen size, but little intraspecific variation of pollen size was observed in most cotton species, indicating that pollen size can be useful index for taxonomy at (or above) the species level.
It is widely accepted that the pollen size usually increases with chromosome number. It has been reported that the pollen grains size of the tetraploid species in the genus Skimmia tend to be larger in comparison with other diploid species, though it is not always true (Fukuda et al., 2008). In this study, we also observe that the allotetreaploids (4n = 52) with tetraploid had larger pollen size than the diploid cotton species (2n = 26) with diploid. The most widespread consequence of polyploidy in plants is the increase in cell size, caused by the larger number of gene copies. Consequently, polyploidy individuals may exhibit larger organs compared to their diploid counterparts, such as roots, leaves, tubercles, fruits, flowers and seeds (Sattler et al., 2016). Interestingly, it was also observed that the pollen size of diploid K genome species (G.rotundifolium) was larger than those of others diploid cotton species. That may be due to its larger genome sizes. It has been reported that the genome size of K genome species (~ 2570 Mb) are even slightly larger than the AD genome Allotetraploids (~ 2400 Mb) (Wendel et al., 2012). Therefore, there may be a positive correlation between pollen size and chromosome dosage and genome size.
It has been point out that exine ornamentation of pollen grains plays a significant role within the tribe, family and between species in the same genus of Brassicaceae (Khalik et al., 2002; Erik, 2012). Baser et al. (2016) also stated that the ornamentation is useful to distinguish between closely related species in the same genus such as Pelargonium endlicherianum and P. quercetorum. From the investigated species in present study, we find that the pollen exine of all cotton species is densely covered with echini except for G. armourianum, whose pollen surface is covered with a thick layer of wax. According to the shape of exine echini, cotton pollen grains are separated in two main types with conical or rod-like exine echini, respectively. The most common shape of exine echini in Gossypium is conical (nineteen species) and rarely rod-like (three species). From the SEM photographs, it is common phenomenon that the base of the exine echini bulges into the tuberculate form, but the size of tuberculate form is varied among the investigated cotton species. Although Gossypium species has usually echinate exine ornamentation, most of them have different densities and lengths of exine echini. In addition, the t test results showed that cotton pollen size is highly correlated to the length of exine echini (r2 = 0.69, p < 0.001) but low correlated to the number of exine echini (r2=-0.04, p > 0.05), which suggested that the cotton species with larger pollen size very possibly has longer exine echini. For example, in the diploid cotton species G. schwendimanii has larger pollen size, longer length but least number of exine echini. On the contrary, G. incanum has smaller pollen size, shorter length but most number of exnine echini. Due to it is hard to remove the thick wax from the surface of pollen grain of G. armourianum, we can't observed any echini on the pollen grain exine though we have repeated the preliminary experiment for three times using the pollen grain collected from different years. Nevertheless, we speculated that it should be have exine echini because having echinate sculpture is a conservative evolutionary feature for the family of Malvaceae and its sister species G. harknessii also has the normal exine echini. Thus, more materials are still needed for further study to observe the detailed pollen characteristics of G. armourianum.
Cluster analysis based on the six pollen traits show that all the allotetraploid species (AD genome group) are completely ranked into Cluster I, and the wild cotton G. tomentosum (AD3) showed a closer relationship with the cultivated cotton G. hirsutum (AD1) compared to the other allotetraploid cotton. However, there are also confused findings that the nine cotton species from D genome group are scattered into different cluster branches and G. incanum (E4) is separated from the other E genome species (E1, E2, and E3), which are out of our expectation. Although cluster analysis is not exactly coincident with the molecular phylogenetic tree proposed by Wendel and Cronn (2003), the results of cluster analysis would provide new information for the phylogenetic relationships of the Gossypium genus from the palynological perspective. Therefore, it is clear that pollen characteristic alone is insufficient to reconstruct taxonomic and systematic relationships within the genus Gossypium, but pollen variations are useful for further taxonomic revisions at the species level.