Genetic diversity in Oryza glumaepatula wild rice populations in Costa Rica and possible gene flow from O. sativa

Introduction

Crop wild or weedy relatives (CWR) are an important source of genetic diversity for modern agriculture (Hajjar & Hodgkin, 2007). Genetic variability in these wild populations is important for breeding programs and genetic crop improvement (Brondani et al., 2005). The study and conservation of CWR populations should be a priority to secure genetic resources for future breeding programs. This is particularly important for critical food staples. Most people on earth use rice (Oryza sativa) as their major staple food and it represents 19% of the world’s per capita energy consumption (Maclean, Hardy & Hettel, 2014). One-fifth of the world’s population depends on rice cultivation for their livelihood (The Sustainable Rice Platform, http://www.sustainablerice.org/). Wild rice species are valuable resources for breeding programs. There are 21 identified wild rice species; six of them are diploid and capable of hybridizing with commercial O. sativa (Khush, 1997; Vaughan, Morishima & Kadowaki, 2003). These wild species have been shown to be important sources of novel and commercially important traits, such as tolerance to acid soils and drought and yield improvements (Brar, 2005; Hajjar & Hodgkin, 2007). Oryza glumaepatula is the only diploid (A^gpA^gp) native rice species in America.

Central American populations of O. glumaepatula are generally small and fragmented due to specific habitat requirements (e.g., rivers and wetlands). Additionally, land use changes have negatively impacted habitat availability (Vaughan et al., 2005), reducing population sizes, which in turn may reduce genetic diversity within populations and increase genetic structure among them. Small, isolated populations are more likely to exhibit lower levels of intrapopulation genetic diversity and higher structure among them, attributes that increase the risk of extinction due to genetic or ecological factors (Newman & Pilson, 1997). Mesoamerican populations have not yet been analyzed; genetic diversity studies have primarily been conducted in Brazil (Karasawa et al., 2007b). These studies have consistently found a larger degree of genetic diversity structured among populations in different rivers, likely due to limited gene flow (Akimoto, Shimamoto & Morishima, 1998; Buso, Rangel & Ferreira, 1998; Karasawa et al., 2007b; Veasey et al., 2008; Veasey et al., 2011). Populations within a river system are less likely to show differences in allele frequencies. Seed dispersal through waterways or river systems in hydrochorous species is likely to homogenize genetic diversity among populations. However, the location of populations along the river may influence genetic diversity estimates. Upriver populations are more likely to contribute seeds and propagules to downriver populations, increasing genetic diversity. Upriver populations receive less gene flow and their genetic diversity should decrease because of genetic drift.

Gene flow between related interfertile species may result in gene introgression (Chu & Oka, 1970). Introgressive hybridization is thought to play an important role in plant evolution (Grant, 1981; Arnold, 1997) by introducing novel genetic variation which allow species to exploit new adaptive landscapes (Ellstrand, Prentice & Hancock, 1999). However, introgression often increases genetic differences between wild and introgressed populations (Jiang et al., 2012) and may pose a threat if hybridization reduces local adaptation (outbreeding depression) by culling autochthonous genetic diversity (Ellstrand, Prentice & Hancock, 1999). Introgression from cultivated rice into wild or weedy rice species has been documented in controlled and natural settings (Song et al., 2003; Chen et al., 2004), and although introgression may lead to a short-term increase in genetic diversity within populations, original or autochthonous diversity is often lost in the process (Lu, 2013). Introgression in O. glumaepatula is highly relevant to biosafety regulations. Gene flow from cultivated species into wild relatives may also pose a biosafety threat if transgenes from genetically modified (GM) crops are able to move via gene flow into wild populations (Snow, 2002; Ellstrand, 2003).

With this study we aim to quantify the distribution of genetic diversity within and among Costa Rican populations of the wild rice species Oryza glumaepatula. We also intend to determine whether populations within a single river are genetically structured and whether river direction influences the magnitude of genetic diversity. Finally, we want to determine if there is indirect evidence of gene flow between wild rice and cultivated varieties of O. sativa, because commercial rice species are planted in the vicinity of wild rice populations. These results have important implications for in situ conservation strategies of the only known populations of O. glumaepatula in Costa Rica.

Methods

Study sites

We analyzed all known populations of O. glumaepatula in Costa Rica. These populations are located in two distinct geographic regions, the first in northwestern Costa Rica (10°57′03.3″N∕85°36′59.2″W) in a seasonal wetland in the Guanacaste Province (GU, Fig. 1). During the rainy season the area floods, allowing O. glumaepatula individuals to grow and form a small population (n < 100) restricted to a small area of less than 1 Ha. The second site is located in the Medio Queso wetland (MQ) in northeastern Costa Rica (11°01′34.1″N∕84°40′42.8″W, Fig. 1). This palustrine wetland of approximately 5,000 ha is irrigated by the Medio Queso River (Jimenez, 2004) and extends for about 10–20 km in length. Gramineous vegetation is the dominant flora with a few tree species that are adapted to a flooded environment (e.g., Pachira aquatic). Along riverbanks, farms with livestock and maize, bean and rice plantations are common. O. glumaepatula is commonly found in MQ, with large patches of individuals on both banks of the river and in multiple sites along the MQ wetland. Along the river unconnected patches of >100 specimens are commonly found. Separation between populations may increase during the rainy season when river volume increases (Jimenez, 2004).

Table 1:

Genetic diversity estimates for Oryza glumaepatula samples from three sub-populations (S1, S2, S3) along the Medio Queso (MQ) river and Guanacaste (GU) in northern Costa Rica.

Population	N	I	Hj	P%
MQ	68	0.178	0.1353	81.82
S1	17	0.160	0.1280	50.3
S2	30	0.169	0.1326	64.04
S3	21	0.175	0.1386	62.02
GU	15	0.148	0.1196	47.27

DOI: 10.7717/peerj.1875/table-1

N: number of individuals; I, Shannon’s index estimated with GenAlEx; H_j, Bayesian estimate of expected heterozygosity; %P, polymorphic loci p > 0.05 estimated with AFLPsurv.

Results

We analyzed a total of 83 O. glumaepatula individuals in MQ and GU. We also resolved AFLP bands on 19 cultivated O. sativa individuals (CO). A total of 444 AFLP bands were scored in all individuals. Genetic diversity estimates for O. glumaepatula are shown in Table 1. Pooled populations had 84.2% polymorphic loci. Heterozygosity estimates were similar across sub-populations and diversity was not correlated with river position (r < 0.05, p > 0.05). Populations located on opposite sides of the Medio Queso River had comparable levels of genetic diversity. GU was the smallest and most isolated, but isolation and population size did not significantly affect genetic diversity. Both Shannon Indexes and Bayesian estimates of heterozygosity were comparable across sites (Table 1).

According to Hickory the full model fit to the data better (DIC = 5248.19) than the model without inbreeding (DIC = 5252.19) or the model without structure (DIC = 5643.89), suggesting a significant departure from Hardy–Weinberg equilibrium (HWE). Average inbreeding was f = 0.9442 ± 0.058 (CI [0.78–0.99]) across O. glumaepatula populations (including GU), suggesting a significant heterozygote deficit. Hickory estimated a low but significant genetic structure among sites (θ = 0.023 ± 0.003). If GU is excluded from the model, genetic structure estimates do not differ significantly. When O. sativa individuals are included in the analysis, as expected, there is an increase in genetic structure (θ = 0.08 ± 0.005), suggesting differences in allele frequencies among wild and cultivated species. Similar results were shown by the neighbor-joining tree (Fig. 2). Commercial rice clustered separately from O. glumaepatula. MQ and GU individuals were mixed in separate clades; however, a large proportion of specimens from sub-population three (S3) and GU were clustered in an intermediate clade close to O. sativa individuals.

Bayesian structuring algorithms found little evidence of structure among O. glumaepatula populations in MQ and grouped individuals into a single cluster (K = 1, data not shown). When Guanacaste was included in the analysis, Structure suggested K = 2 as the most likely number of clusters (Fig. 3A). Several individuals in sub-population 3 (S3) and GU were grouped into a second cluster. Including commercial O. sativa into the analysis resulted in Structure grouping individuals into K = 2 clusters. However in this case, commercial rice individuals (CO) were assigned to one cluster, whereas O. glumaepatula individuals were assigned to a second cluster. Structure indicates that wild rice individuals at S3 and GU appear to share a significant proportion of their genomes with O. sativa (Fig. 3B).

Discussion

We present the first genetic diversity estimates in Mesoamerican populations of O. glumaepatula. Our results suggest intermediate levels of genetic diversity, significant evidence of inbreeding, as well as a lack of structure among populations. Contrary to expectations, we found small differences among populations separated by more than 100 km. We also found indirect evidence of gene flow from cultivated rice in Costa Rican populations of O. glumaepatula. These results are of great importance for this CWR; introgressive hybridization may increase the risk of losing autochthonous genetic diversity.

Previous studies on O. glumaepatula have shown great variability in intrapopulation diversity estimates (reviewed in Karasawa et al., 2007b), which likely respond to differences in population size, connectivity and the nature of markers used to estimate genetic diversity. Generally, studies using hyper-variable markers such as SSR find higher diversity levels (Brondani et al., 2005; Karasawa et al., 2007b; Karasawa et al., 2012). Genetic diversity estimates from Costa Rican populations are comparable to those from Brazil estimated using co-dominant and dominant markers (Buso, Rangel & Ferreira, 1998). This suggests that O. glumaepatula populations in Costa Rica have moderate levels of genetic diversity, which are lower than expected for a monocot species with predominant autogamous mating, but similar to those expected for a mixed annual or short-lived species (Hamrick & Godt, 1996).

Intrapopulation genetic levels in O. glumaepatula are likely influenced by its mating system. As in previous reports we also found a significant deficit of heterozygotes in all populations (Karasawa et al., 2007b). As a predominantly selfing species, with a significant proportion of clonal growth, we expect O. glumaepatula populations to display high rates of inbreeding. In equilibrium our F_IS estimates in MQ and Guanacaste suggest a predominantly selfing mating system (${\hat{t}}_m=\frac{1-F_{IS}}{1+F_{IS}}=0.04$). Progeny arrays and maximum likelihood estimates of O. glumaepatula’s mating system obtained comparable results (tm = 0.01--0.223. Karasawa et al., 2007a). Large interconnected populations are more likely to have higher levels of genetic diversity (Frankham, 1996). The O. glumaepatula populations in this study are the only populations of this wild rice species found in Costa Rica. These populations are limited to very specific areas (inundated wetlands with a characteristic dry season) and are separated by large distances between each other and from other populations in Central America or the Caribbean (Vaughan et al., 2005). This should increase the effects of drift, reducing intrapopulation genetic diversity. However, MQ is a large metapopulation with more than 10K individuals, all of which are interconnected by the Medio Queso River. This population may be large enough to preserve a significant proportion of the original genetic diversity. Similar results were found for other large and similarly interconnected populations in Brazil (Buso, Rangel & Ferreira, 1998; Vaz et al., 2009).

We were unable to detect any effect of river direction on genetic diversity estimates. Upstream gene flow is very likely to occur in the MQ river because it does not have a strong current (Jimenez, 2004), and during the rainy season floating meadows or seeds may easily move upstream, thus homogenizing genetic diversity estimates across populations regardless of river position. Waterfowl are also likely to disperse seeds among rivers (Pollux et al., 2007) and into upstream populations. Populations interconnected via vegetative dispersal, seed-dispersal and pollen flow should have comparable levels of genetic diversity regardless of their position along the river, as observed in our sample.

In a review by Karasawa et al. (2007b) they consistently found significant differences in allele frequencies among populations—as expected for an autogamous, short-lived species (Hamrick & Godt, 1996). In contrast, we found a significant but dismissible structure (θ = 0.03) among sub-populations separated by less than 15 km within the same river. Migration between sub-populations may occur at the beginning of the rainy season (May), when seeds and culms are likely carried along the river. Vaz et al. (2009) reached similar conclusions and did not find evidence of structure among samples collected along the same river separated by ∼10 km. Studies that found structure (Akimoto, Shimamoto & Morishima, 1998; Buso, Rangel & Ferreira, 1998; Brondani et al., 2005; Karasawa et al., 2007b) consistently analyzed populations separated by large distances (>100 km), and in most cases in different unconnected river systems. In MQ all sub-populations likely form a single meta-population with ample levels of gene flow among individuals and sites.

An interesting and somewhat contradictory result is that GU, which is separated by more than 100 km from MQ, shows almost no genetic structure with MQ (Fig. 2). Individuals in GU are probably the result of a recent founder effect that most likely originated from MQ. A detailed analysis of the environmental requirements of O. glumaepatula in Costa Rica shows that it is only likely to occur in the MQ wetland (Gastezzi, Martínez & Villareal, 2012). GU is very small (less than 100 individuals), however, allelic frequencies there do not deviate significantly from those in MQ suggesting that this population may have only been recently founded and genetic drift has not had enough time to reduce its genetic diversity. Guanacaste province has the largest number of rice plantations in Costa Rica and the largest processing plants (Lomas & Herrera, 1985). Therefore, cultivated rice from different parts of the country is transported to facilities in Guanacaste. O. glumaepatula seeds may have been transported from MQ by animal dispersers or humans in the recent past (Mack & Lonsdale, 2001; Nathan et al., 2008).

We found indirect evidence of gene flow between O. glumaepatula and cultivated rice (O. sativa). Our analyses showed that O. glumeapatula individuals appear to be admixed with cultivated O. sativa (Fig. 3). Introgression between O. sativa and O. glumaepatula has been previously documented when the species are planted in close proximity (Brondani et al., 2005). Cultivated rice and O. glumaepatula are sympatric in MQ and in GU. In MQ, farms increasingly invade the pristine wetland and in GU O. glumaepatula frequently grows in rice farms. However, we expected limited cross-pollination because the flowering phenology of cultivated rice and O. glumaepatula rarely overlap. Hybrids between O. glumaepatula and O. sativa are male sterile, however stigmas are still receptive and can produce viable seeds (Yamagata et al., 2010). Therefore, hybrids are able to backcross to either progenitor, which in turn may result in the introgression of parts of the O. sativa genome into O. glumaepatula. During our collecting trips we may have inadvertently collected third or forth generation backcrosses as well as wild rice individuals without introgression, all of which are morphologically similar. Currently, genomic analyses are being conducted to confirm whether hybridization is occurring in wild rice populations in Costa Rica.

In Costa Rica, gene flow between cultivated and wild rice species may pose an important biohazard because it allows commercial traits to be transferred into the only populations of O. glumaepatula in Costa Rica, with a detrimental impact on autochthonous variation. Our results highlight the importance of interspecific gene flow for both in situ and ex situ conservation strategies as well as present and future breeding programs. Recently, introgression has become a major focus of interest in regards to biosafety issues, particularly relative to gene flow from genetically modified (GM) organisms into non-GM cultivars and their wild/weedy relatives (Lu & Snow, 2005; Sanchez-Olguin et al., 2009). Biosafety authorities and regulators may use our results to establish zones of exclusion for the eventual release of GM rice in these areas. The information presented in the present study, should be used to delineate conservation strategies and implement better planting practices to allow for the long-term viability of O. glumaepatula populations in Costa Rica. Although wetlands are under international protection by the Ramsar Convention (http://www.ramsar.org/wetland/costa-rica) this site is currently endangered by the construction of a road along the northern border of Costa Rica, which could imperil the largest O. glumaepatula population in Costa Rica and a likely stepping-stone between Costa Rican and Nicaraguan wild rice populations.

Supplemental Information