Sugarcane is an ancient crop. The early canes were domesticated around 7,000 BCE and the simplest scenario is that Saccharum officinarum was domesticated from Saccharum robustum in the New Guinea region. Humans then spread these cultigens over large distances. In Southeast Asia, S. officinarum hybridized with local Saccharum spontaneum giving rise to Indian and Chinese cultivars (reviewed in Grivet et al., 2004). Cultivation and sugar processing had become established in Persia by the 600 CE and within a century had reached the Mediterranean and North Africa. The Spanish and Portuguese conquistadors carried sugar to the southwest of Iberia (Madeira, Canary Islands) and brought sugarcane to the New World.

Despite this long ongoing period of cultivation, the early stages of plant selection occurred fairly recently. From the 6th to the 18th century, all cane grown in the Western world was of a single variety, “Creole” cane from Java. Subsequently, other canes dominated their adopted habitats for lengthy periods in Hawaii, Mauritius (from Java), Australia (from New Guinea), South Africa, and Brazil. However, traditional cultivars proved highly susceptible to diseases, and this led breeders to focus on the hybridization of S. officinarum clones in the early 1900s, but they soon progressed to interspecific crosses incorporating the wild S. spontaneum resulting in disease resistance. A backcrossing program to S. officinarum, called “nobilization”, was then required to boost sucrose content (reviewed in Dillon et al., 2007). Clonal selection across generations and incorporation of other germplasm into cultivated backgrounds established the (narrow) genetic base of modern cultivars, which have a man-made genome (Grivet et al., 2004; Jannoo et al., 2004).

Due to female meiotic restitution, the F₁ hybrid conserved all S. officinarum chromosomes and half the S. spontaneum chromosomes (2n+n), and then a few backcrosses later, this hybrid broke down to n+n, establishing the constitution of sugarcane (Bremer, 1961a). Current cultivars (Saccharum spp.) derived from crosses between varieties or clones, are highly polyploidal and tolerant of aneuploid constitution, which makes the chromosome combination in each offspring unique and unpredictable (Grivet and Arruda, 2002). The crop is grown primarily for sugar production in the warm climates of South America, North America, Asia, and Australia. In recent years, especially in Brazil, sugarcane has become of enormous economic importance with multifaceted end-uses, such as the production of ethanol and bioelectricity (Furlan et al., 2013 and references therein).

Taxonomically speaking, the “Saccharum complex” is a closely related group, containing five genera: Erianthus (section Ripidium), Miscanthus (section Diandra), Narenga, Sclerostachya, and Saccharum, encompassing six species forming a wild (S. spontaneum and S. robustum) and a cultivated group. This cultivated group, in turn, contains Saccharum edule, eaten as a vegetable in Pacific islands and Papua New Guinea; S. officinarum, the most widely grown species; and Saccharum barberi and Saccharum sinense. It has been suggested that these last two are interspecific hybrids of S. spontaneum and S. officinarum rather than true species (D’Hont et al., 2002).

In addition to the breeding techniques that led to the formation of Saccharum spp. discussed above, a few natural and resynthesized polyploids are also worthy of consideration. Brassica napus is a good example that has been extensively studied. B. napus originated from interspecific hybridization between two diploid ancestors, followed by polyploidization. The occurrence of homoeologous exchanges, duplications, and deletions was detected in experimental B. napus at the onset of genome merging or within a few generations after the confrontation of the two genomes (see Nicolas et al., 2007; Szadkowski et al., 2010). Moreover, by using fluorescent in situ hybridization (FISH) to distinguish all chromosomes present in B. napus, Xiong et al. (2011) reported evidence for homoeolog pairing and chromosome rearrangements, aneuploidy, and homoeologous chromosome compensation in the resynthesized plants.

In natural populations as well as in synthetic lines of Tragopogon mirus and Tragopogon miscellus, two young allopolyploid species (<200 years old), normal bivalent formation and an array of meiotic abnormalities (including multivalent formation, lagging chromosomes, and the formation of anaphase I bridges) have been observed (Lim et al., 2008; Tate et al., 2009). In particular, intergenomic translocations and/or aneuploidy were identified, but variation typically followed a compensated pattern (Chester et al., 2013, 2015). These key reports of such extensive chromosome rearrangements in nature are similar to those observed in experimental B. napus allopolyploids (Xiong et al., 2011).

However, the regularity of meiosis needs to be in place soon after polyploid formation and was found to increase rapidly in experimental polyploids that were initially chromosomally unstable (reviewed in Cifuentes et al., 2010). In wheat, for instance, the Ph1 gene is the major component of a multigenic system inhibiting recombination between homoeologous chromosomes favoring cytological diploidization. Similarly, B. napus shows complete diploid-like meiotic behavior, with only bivalents and disomic inheritance; thus, in all euploid genotypes, crossovers mostly occur between homologous chromosomes. The so-called pairing regulator gene PrBn is the most influential locus reducing homoeologous pairing in B. napus (Jenczewski et al., 2003; Suay et al., 2014). The BYS locus has also been identified as a dominant locus contributing to homologous pairing during meiosis in the allotetraploid Arabidopsis suecia (Henry et al., 2014).

The unusual genome complexity and inter-relationships within Saccharum (with ploidy levels ranging from 5× to 16×) has inspired several molecular studies that have elucidated aspects of sugarcane genome constitution (Souza et al., 2011; Setta et al., 2014), architecture (Garcia et al., 2013), and cytogenetics, studied mainly by D’Hont et al. (1993; 1996; 1998), D’Hont (2005) and Piperidis et al. (2010). However, there is a critical shortage of information on chromosome behavior throughout meiosis. In this study, we examined the microsporogenesis of the IACSP93-3046 variety, providing a detailed analysis of meiotic process and chromosome association at diakinesis, using FISH with centromeric probes. Possible reasons for meiotic abnormalities and the prevalence of bivalent formation are discussed on the basis of information in the literature and data generated here.

In brief, our hypothesis is that Saccharum sp. experienced intergenomic exchanges during meiosis of the interspecific hybrids and nobilized clones. However, during the nobilization process sugarcane acquired the ability to suppress multivalent formation by restricting meiotic crossovers to homologous chromosomes or homologous chromosome regions. We suggest that a genetically controlled strategy that enforces bivalent formation was possibly inherited from S. officinarum (an autopolyploid with diploid-like meiotic behavior) and this became evident once the hybrid-derived clones broke down to n+n, establishing the constitution of sugarcane.

Pioneering studies (Moriya, 1940; Bremer, 1961a,b) determined the chromosome numbers of POJ cultivars, including POJ2878 with 2n = 119. In India, at the beginning of the 20th century, nobilization of S. barberi and S. spontaneum with S. officinarum produced early nobilized tri-species hybrids of Co seedlings (from Coimbatore), which also gained acceptance in South Africa, Australia, Louisiana (United States), Argentina and Brazil (see Paterson, 2013). Crossing POJ 2878 with Co290 produced Co419 (2n = 113). Both POJ2878 and Co419 are ancestors of the variety herein analyzed (Figure 1), which we determined to have 2n = 112.

To clarify, the wild S. spontaneum (2n = 40–128) evolved via polyploidy and aneuploidy. It is widely distributed, a fact ascribed to its adaptability, and five major cytotypes were described with 2n = 64, 80, 96, 112 and 128 (x = 8). In contrast, S. officinarum (2n = 80) is an autopolyploid, with 8 sets of chromosomes (x = 10); gametes have n = 40. Molecular evidence has confirmed the direct descent of S. officinarum from the wild species, S. robustum (2n = 60 to 220), and two major cytotypes were identified with 2n = 64 and 80 (x = 10) (Panje and Babu, 1960; D’Hont et al., 1993; Lu et al., 1994; Amalraj and Balasundaram, 2006).

In Brazil, chromosome numbers have been determined for NA56-79 (2n = 114) and Co419 (2n = 113; Silvarolla and Aguiar-Perecin, 1994), some RB cultivars (2n = 110–112; Ferrari, 2010), and IAC91-1099 (2n = 112; Melloni et al., 2016). There are indeed some inconsistencies regarding the mitotic number and this can be ascribed to incorrect counting and the complexity of sugarcane’s origin. In fact, the majority of canes currently cultivated have 2n = 102–119, including R570 clones (2n = 107–115; D’Hont et al., 1996), the variety most studied at cytological and genomic levels. The genome of R570 is composed of 80% S. officinarum, 10% S. spontaneum, and 10% recombinant chromosomes (D’Hont et al., 1996, 1998). In cultivar NCo376 (2n = 112), the distribution of these chromosome categories was 70:20:10%, confirming for the first time the exchanges between S. officinarum and S. spontaneum chromosomes (D’Hont, 2005). Recently, Piperidis et al. (2010) reported 2n = 106–119 for cultivars and elite clones from Australia (5) and South Africa (2). In this elegant work, the authors used in situ hybridization to show that the percentage of S. officinarum chromosomes varied from 70% to 77%, S. spontaneum chromosomes 11% to 21%, and recombinants 8% to 13%, depending on the genotype.

The classic papers of Moriya (1940), Price (1957, 1963a,b, 1968), and Bremer (1961a,b) provided valuable data on the cytology of the Saccharum species, and studies on chromosome transmission in interspecific crosses involving Saccharum were undertaken following G. Bremer’s observations. The F₁ hybrids, e.g., Kassoer, had 2n = 136 chromosomes, equal to the sum of the maternal diploid (S. officinarum) and the paternal haploid (S. spontaneum) numbers (2n+n = 80+56 = 136). These F₁ hybrids were referred to as the first nobilization of S. spontaneum and the subsequent backcrosses were similarly denoted. Surprisingly, regular meiosis with a large number of bivalents at diakinesis was found in this even number of chromosomes (136). Pictures of diakinesis and anaphase I shown in Bremer’s article (Bremer, 1961a) indicated 66 bivalents and four laggings near the equator of the PMCs of an interspecific F₁ hybrid; others have shown 4–12 univalents. Later, Nair (1975) examining interspecific hybrid cells (S. officinarum × S. spontaneum) at anaphase (I and II) found lagging chromosomes and chromosome bridges, as well as micronuclei in tetrads. Remarkably, our results show similar meiotic behavior, although four generations with 8 crosses following Kassoer were performed to originate SP79-1011, the female parent of IACSP93-3046, which shows redundancy with respect to its ancestors (Figure 1). A question that arises is whether univalents have a pure S. officinarum origin or an interspecific origin (see Jannoo et al., 2004). Recombinants have arisen during the first generations due to the pairing of homeologs.

However, as widely documented for resynthesized allopolyploids (reviewed in Chen and Ni, 2006 and Grusz et al., 2017) and some natural autopolyploids such as Arabidopsis arenosa (Yant et al., 2013), current sugarcane cultivars are assumed to have undergone genome readjustments aimed at achieving stability and re-establishing functional pathways, including the suppression of multivalent associations by a genetic control of chromosome pairing and recombination as occurs in wheat and oat (see Naranjo and Benavente, 2015). In wheat, the locus Ph1 ensures that recombination only occurs between homologous chromosomes. Recent experimental studies suggest that the overall effect of Ph1 on chromosome pairing must be the promotion of homolog pairing rather than the active suppression of pairing between related chromosomes. It clearly blocks recombination or crossover between them³.

A reduction in chiasma number per bivalent has been suggested as a potential mechanism for meiotic diploidization in autopolyploids (as in the case of S. officinarum) because limiting crossovers per chromosome prevents multivalent associations (revisited in Cifuentes et al., 2010; Bomblies et al., 2016). We may speculate that pairing could be confined to homologous chromosomes or even to syndetic segments, albeit carried by recombined chromosomes. These points need to be clarified in sugarcane, and were elegantly discussed in Grandont et al. (2014) with regard to homoeologous chromosome sorting during early prophase I and progression of meiotic recombination in different B. napus accessions.

Burner and Legendre (1993) determined chromosome transmission and meiotic stability in sugarcane progenies derived from crosses of both elite and interspecific clones with S. spontaneum clones (F₁, BC₁, BC₂, BC₃). The frequency of trivalents increased linearly from one generation to the next. Pairing was primarily of bivalents, although variable numbers of univalents and multivalents were observed. Chromosome transmission was strictly n+n, but aneuploid gametes resulted from meiotic abnormalities, which included anaphase bridges and laggards, as well as asynchronous meiosis. Bielig et al. (2003) analyzed the frequency of 2n microspore formation in Saccharum spp. hybrid clones, observing that univalents were rare and multivalents did not occur at all. The authors described the formation of dyads and triads at the end of the meiotic process in the hybrid clones, which produced 2n gametes more frequently than did the parental species, and the production of 2n gametes was attributed to the absence of cytokinesis rather than to a combination of asynchronous division and chromosome non-disjunction. Cytogenetic analysis confirmed that abnormal meiotic behavior existed in Gossypium synthetic hybrids due to non-synchronous chromosome separation at hybrid first meiosis. These abnormalities include the occurrence of a triad (partly from non-synchronized dyad separation), unequal division with abnormal unbalanced micronucleus formation, and various abnormal polyads (Wu et al., 2017).

The prevalence of unreduced gamete involvement in crosses between species and within species’ populations is suggestive of a complex set of underlying mechanisms that further support the role of unreduced gametes in facilitating polyploidy and speciation. The formation of unreduced gametes is prevalent in diverse plant species, including in Gossypium in which it was recently found to occur (Montes et al., 2017).

It is therefore plausible to speculate that the Saccharum genus has experienced a continuum between auto- and allopolyploids (Storme and Mason, 2014; Mason and Pires, 2015). Regarding Saccharum spp., a man-made genome, no excluding possibilities can be envisaged. Homoeologous associations allowed interspecific crossovers and the formation of recombinant chromosomes (D’Hont et al., 1996, 1998), but sugarcane adapted the mechanism to provide mostly bivalent associations during the nobilization process. It should have occurred few backcrosses later the interspecific hybridization, when the hybrid broke down to n+n. The mechanism to suppress multivalent association was possibly inherited from parental species due to the evolutionary history of polyploidy in Saccharum. Furthermore, our hypothesis is that PMCs of S. officinarum and S. spontaneum divide non-synchronously, and follow Saccharum spp. nuclei, explaining the rare formation of dyads and triads at the end of meiosis in the PMCs of current cultivars. Spindle fibers were visible in both asynchronous cells (see Figures 4L,M), and only tetrads were observed at the end of IACSP93-3046 meiosis.

Interspecific chromosomal exchanges have been shown to exist in the genome of modern sugarcane cultivars (D’Hont et al., 1996; Cuadrado et al., 2004; Piperidis et al., 2010), and these exchanges could be the result of homoeologous recombination or even translocation between non-homologous chromosomes. Most contemporary cultivars, including the one we analyzed, seem to be meiotically stable, since bivalent associations predominate. It is reasonable to assume that all the sugarcane chromosomes involved in preferential mutual pairing probably have a pure origin, and this applies in particular to those of S. officinarum. However, there is evidence that some bivalents also include chromosomal regions originating from the two ancestral species (see Jannoo et al., 2004).

None of the other currently available papers has produced high-quality meiotic preparations, as we have shown (700 meiotic cells were examined). Taking this into account and in view of the meiotic abnormalities visualized herein (Table 1), the following scenarios are possible. Firstly, some chromosomes undergo incomplete association leading to univalency at late prophase I. Note that our FISH results revealed the occurrence of one or two univalents at diakinesis (Table 2). These univalents behave as laggards in subsequent meiotic phases and may result in aneuploidy. Secondly, bi-, tri-, or tetravalent associations in early prophase may be ineffective in terms of chiasma formation and this could be part of the mechanism for inhibiting the siting of crossovers as a consequence of a gene controlled system that modifies pairing behavior in subsequent generations after the confrontation of the two genomes. Thirdly, the presence of rod bivalents (up to five visualized in late anaphase) showing the lagging tendency is consistent with a reduction in chiasma frequency (Supplementary Figure S2A). Finally, the presence of chromatin bridges in late anaphase I (Supplementary Figure S2B) is indirect evidence of the presence of paracentric inversions, although acentric chromosomes (or fragments incorporated into a micronucleus or lost) were not clearly recognized. All these mechanisms have been widely discussed in regard to natural and resynthesized polyploids, especially Brassica (Nicolas et al., 2007; Gaeta and Pires, 2010; Szadkowski et al., 2010).

In S. officinarum, univalents are very rare (Bielig et al., 2003). This applies to most established autotetraploids, in which meiotic stabilization and bivalent formation are coupled with a reduction in chiasma frequency. For example, established autotetraploids of Arabidopsis arenosa show a reduction in the number of crossovers per chromosome compared to the diploid ancestor (Bomblies et al., 2016). Diakinesis cells of the Thai KPS 01-01-25 sugarcane exhibited not only true chiasma bonds between homologs/homeologs, but also secondary chromosome associations resulting from loose bivalents in the absence of chiasmata (Thumjamras et al., 2016). This reduction in chiasma frequency may also indicate how sensitive the structural meiotic components are to disturbance, and could be a direct consequence of meiotic adaptation to genomic challenges (Bomblies et al., 2015).

It is worth noting that the use of centromeric probes allowed us to confirm the predominance of bivalent associations, avoiding possible errors due to the small size and high number of sugarcane chromosomes. The methodology and interpretation of results herein could provide a model for analyzing meiotic behavior in other canes, and the possible implications for breeding programs.

Furthermore, it is imperative to elucidate species-derived chromosome behavior during meiosis, and we are aware that GISH techniques combined with high quality meiotic preparations must be used if this is to be achieved. By using these techniques it will be able to trace the origins of uni- and bivalents. We also believe that chromosomal data interpretation should include genealogy information. All irregularities may vary in frequency depending on the variety, as each has its own proportion of S. officinarum, S. spontaneum, and recombinant chromosomes coexisting in the genome (or chromosome set). It would be interesting to discover the origins of univalents, and even bivalents, by GISH analysis, but we would have to select appropriate germplasm bank entries with usable genotypes of S. officinarum and S. spontaneum.

Currently, the global concern to seek sustainable and renewable energy has revived interest in biomass energy as an important worldwide issue (reviewed in Hoang et al., 2015). The demand for higher biomass volume to produce second generation ethanol and electricity has encouraged new directions in research for developing cultivars with higher fiber content, so-called energy cane. One strategy is to combine Saccharum spp. with S. spontaneum genome rich canes (Matsuoka et al., 2014; Hoang et al., 2015) or even S. robustum. It would be very interesting to analyze the consequences of this “reciprocal backcross process” in terms of the meiotic behavior of energy canes.

CA performed all cytological preparations with assistance from CO and LT under the supervision of MV and EF-M. CA, CM, and LC-S isolated and analyzed the sugarcane centromeric sequences under the supervision of CM-V. LP and MX provided the plant material and information. MV conceived the study and wrote the article.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.