Demography as a confounding factor to explain highly diverged loci between cultivated and wild rice

Abstract

The domestication of rice increases the divergence between cultivated rice and its wild progenitor because of artificial selection. However, it remains unknown whether highly diverged loci in rice can be explained by neutral demographic scenarios alone. In this study, we genotyped 45 InDels (insertion/deletion) in two subspecies of Asian cultivated rice (Oryza sativa ssp. japonica and Oryza sativa ssp. indica) and their wild progenitor (O. rufipogon/O. nivara). Among them, 17 loci are highly diverged (F_ST > 0.4) between rice cultivars and their ancestor. We performed coalescent-based simulations on neutral demographic scenarios and found that neutral demography alone could explain the polymorphic profiles on those highly diverged loci between cultivated and wild rice. Therefore, more signatures of selection should be considered when detecting artificial selection in rice.

Introduction

Artificial selection is a selection process by which human instead of nature select animals or plants with desirable traits to reproduce. After thousands of years, domesticated animals and plants often have distinct characteristics comparing with their wild progenitors. To study the molecular mechanism of artificial selection, it is important to map genes affected by artificial selection during domestication because these genes are commonly related to economic traits or biological processes associated with development and reproduction (Gepts, 2014). To date, there are different methods and software available to detect artificial selection (Li and Stephan, 2006; Sabeti et al., 2006; Li 2011; Lin et al., 2011; Fu and Akey, 2013; Waldmann et al., 2020). When artificial selection occurs in one of populations, it increases the fixation index (F_ST) on the selected locus. F_ST is a measurement of population differentiation (Hartl and Clark, 1997), and a locus with high F_ST indicates that two populations are highly diverged on the locus. Therefore, high F_ST has been frequently used as one of the signatures of selection in rice and many other species (Gao, 2004; Yu et al., 2008; Fabian et al., 2012; Huang et al., 2012; Daub et al., 2013; Shen et al., 2014; Kumagai et al., 2016; Islam et al., 2018; Cheng et al., 2019).

It has been well-known that demographic factors, such as population structure and varying population size, may generate selection-like signatures in the genetic variation of domesticated animals or plants (Catriona and Emma, 2006; Li and Stephan, 2006). Thus, it may hinder the efforts to detect artificial selection. To examine the confounding effects of demography, we investigated whether the polymorphism on highly diverged loci (i.e. high F_ST loci) could be accounted by neutral demographic scenarios alone, without invoking the hypothesis of artificial selection.

We used Asian rice to examine the confounding effects of demography, including population structure and varying population size. Asian rice (Oryza sativa L.) has two subspecies, indica and japonica, and it domesticated from its wild rice ancestor O. rufipogon Griff. (He et al., 2011; House et al., 2014; Zhang et al., 2016). An entangled history of rice domestication has been revealed (Huang et al., 2012). As common variants in rice (Shen et al., 2004), InDels (insertion/deletion) may be good neutral markers and are easy to be genotyped. Polymorphisms of InDels have gained long-term interests in the rice research community (Shen et al., 2004; Liu et al., 2015a; Lu et al., 2015; Moonsap et al., 2019), and these markers were used to conduct QTL mapping (Kim et al., 2016) and study genetic differentiation in rice (Liu et al., 2012; Sahu et al., 2017). In this study, we selected the highly diverged InDels between the rice cultivars (japonica or indica) and their wild ancestors, O. rufipogon and O. nivara (also recognized as the annual type of O. rufipogon) (Yamanaka et al., 2003; Zheng and Ge 2010; Liu et al., 2015b) and genotyped these loci in 172 accessions of rice collected through the East, South and Southeast Asia. Then we conducted large-scale coalescent-based simulations on neutral demographic scenarios to examine whether the polymorphic profiles on the highly diverged loci can be explained by demography alone. This study would shed lights on how genes affected by artificial selection could be detected in Asian cultivated rice and other domesticated animals and plants.

Materials and methods

Sampling and InDels genotyping

In this study, total 172 accessions of japonica (n = 63), indica (n = 66), O. rufipogon (n = 25) and O. nivara (n = 18) were collected through the East, South and Southeast Asia (Table S1). These samples were provided by the International Rice Research Institute, the Shanghai Center for Agricultural Biological Genetics, Zhejiang University and Yunnan Agricultural University, respectively, and most of their genomes have not been sequenced. The 45 InDels were genotyped in the collected accessions (Table S2). Those InDels were initially discovered by comparing the genomic sequences of indica (93–11) and japonica rice (Nipponbare) (Shen et al., 2004; Lu et al., 2009). They are randomly distributed along the rice genome.

Genomic DNA extraction and InDels genotyping was conducted according to our previous published methods (Lu et al., 2009). Seedlings of rice samples were germinated in an incubator at 37 °C and then transferred into a glasshouse at around 25 °C. About 0.5 g of fresh leaf samples were collected from seedlings at about the 3–4-leaf stage and then placed in a plastic bag containing silica gel for fast drying. Genomic DNA was extracted using the hot CTAB procedure (Murray and Thompson, 1980). PCR analysis and electrophoresis were performed. The primers for the 45 InDels were obtained by the previous research (Shen et al., 2004) and listed in Supplemental Table S3. PCR products were resolved on a 4% denaturing polyacrylamide gel. After electrophoresis, bands were revealed using the silver-staining procedure. The electrophoretic banding patterns were used for genotyping and subsequent calculation of allele frequency.

Heterozygosity and F_ST calculation follows the textbook (Hartl and Clark, 1997). Heterozygosity of wild rice, japonica and indica at each InDel was calculated from allele frequency. To calculate F_ST, we used the following formula: $F_{ST} = {{{\rm H}_{\rm T}-{\rm H}_{\rm S}} \over {{\rm H}_{\rm T}}}$, where H_T is expected heterozygosity for total population and H_S is average expected heterozygosity in subpopulations.

Model selection and simulated neutral demographic scenarios

Four demographic models were considered here (Fig. 1). Two of them represent the single evolutionary origin of rice, and the other two models stand for the multiple origins. Here, the single origin means that O. sativa was domesticated once and diverged to japonica and indica (Sang and Ge, 2007; Gross and Zhao, 2014). The multiple origins indicate that japonica and indica have independent origins from distinct wild rice subpopulations (Oka, 1988).

Figure 1. Four demographic scenarios of rice populations considered in this study. (a) The single origin of rice: the model I: japonica rice was first domesticated; the model II: indica rice was first domesticated. (b) The two origins of rice: the model III: japonica rice was first domesticated; the model IV: indica rice was first domesticated.

By using a software ms (Hudson, 2002) we simulated the neutral data under the demographic scenarios. The parameters of the four demographic models were obtained from the previous study (Molina et al., 2011). Thus, to simulate the data, the ms command lines are as following:

Model I (single origin, japonica domesticated first):

$$\eqalign{& {\rm ms\ 344\ 100000000\ }\hbox{-}{\rm s\ 1\ }\hbox{-}{\rm I\ 3\ 86\ 126\ 132\ }\hbox{-}{\rm n\ 2\ 0}{\rm.07\ }\hbox{-}{\rm en\ 1\ 2\ 0}{\rm.01\ }\cr& \hbox{-}{\rm ej\ 1\ 2\ 1\ }\hbox{-}{\rm n\ 3\ 0}{\rm.07\ }\hbox{-}{\rm en\ 0\ 3\ 0}{\rm.01\ }\hbox{-}{\rm ej\ 0}{\rm.07\ 3\ 2\ }\hbox{-}{\rm m\ 1\ 2\ 2}{\rm.01\ }\hbox{-}{\rm m}\cr& {\ 1\ 3\ 5}{\rm.07\ }\hbox{-}{\rm m\ 2\ 3\ 2}.35}$$

Model II (single origin, indica domesticated first):

$$\eqalign{& {\rm ms\ 344\ 100000000\ }\hbox{-}{\rm s\ 1\ }\hbox{-}{\rm I\ 3\ 86\ 126\ 132\ }\hbox{-}{\rm n\ 2\ 0}{\rm.07\ }\hbox{-}{\rm en\ 0\ 2\ 0}{\rm.01\ }\cr& \hbox{-}{\rm ej\ 0}{\rm.04\ 2\ 3\ }\hbox{-}{\rm n\ 3\ 0}{\rm.09\ }\hbox{-}{\rm en\ 1\ 3\ 0}{\rm.01\ }\hbox{-}{\rm ej\ 1\ 3\ 1\ }\hbox{-}{\rm m\ 1\ 2\ 1}{\rm.57\ }\hbox{-}{\rm m}\cr& {\ 1\ 3\ 3}{\rm.84\ }\hbox{-}{\rm m\ 2\ 3\ 2}.16}$$

Model III (multiple origins, japonica domesticated first):

$$\eqalign{& {\rm ms\ 344\ 100000000\ }\hbox{-}{\rm s\ 1\ }\hbox{-}{\rm I\ 3\ 86\ 126\ 132\ }\hbox{-}{\rm n\ 2\ 0}{\rm.08\ }\hbox{-}{\rm en\ 1\ 2\ 0}{\rm.01\ }\cr& \hbox{-}{\rm ej\ 1}{\rm.01\ 2\ 1\ }\hbox{-}{\rm n\ 3\ 0}{\rm.1\ }\hbox{-}{\rm en\ 0\ 3\ 0}{\rm.01\ }\hbox{-}{\rm ej\ 0\ 3\ 1\ }\hbox{-}{\rm m\ 1\ 2\ 1}{\rm.85\ }\hbox{-}{\rm m}\cr& {\ 1\ 3\ 4}{\rm.02\ }\hbox{-}{\rm m\ 2\ 3\ 1}.37}$$

Model IV (multiple origins, indica domesticated first):

$$\eqalign{& {\rm ms\ 344\ 100000000\ }\hbox{-}{\rm s\ 1\ }\hbox{-}{\rm I\ 3\ 86\ 126\ 132\ }\hbox{-}{\rm n\ 2\ 0}{\rm.1\ }\cr& \hbox{-}{\rm en\ 0\ 2\ 0}{\rm.01\ }\hbox{-}{\rm ej\ 0}{\rm.01\ 2\ 1\ }\hbox{-}{\rm n\ 3\ 0}{\rm.11\ }\hbox{-}{\rm en\ 1\ 3\ 0}{\rm.01\ }\hbox{-}{\rm ej}\cr& {\ 1}{\rm.01\ 3\ 1\ }\hbox{-}{\rm m\ 1\ 2\ 1}{\rm.41\ }\hbox{-}{\rm m\ 1\ 3\ 3}{\rm.51\ }\hbox{-}{\rm m\ 2\ 3\ 1}.59}$$

Neutrality test

The null hypothesis is one of the neutral demographic scenarios described above. The alternative hypothesis is that, a locus evolved neutrally in the progenitor population, but was subject to artificial selection in one of the cultivated rice populations. As expected, artificial selection could cause a high F_ST between progenitor-japonica, progenitor-indica or japonica-indica. We pooled simulated results as a sample by rejection-sampling algorithm (Tavare et al., 1997) that has been commonly used to compare simulated and observed data (Li et al., 2016). The neutrality test was conducted as $P( f_j > f_{j, obs}{\rm \vert }\;\vert {f_p-f_{p, obs}} \vert < \varepsilon )$ and $P( f_i > f_{i, obs}{\rm \vert }\;\vert {f_p-f_{p, obs}} \vert < \varepsilon )$, where f _p, f _j and f _i are the allele frequency of mutant in the progenitor, japonica and indica population, and $\varepsilon$ is a fixed tolerance.

When $\varepsilon$ is very small, the computational load would be very large to estimate the probability, whereas the precision of the probability will be poor when $\varepsilon$ is large. Our experience suggests that $\varepsilon = 0.02$ works well.

Results

In this study, we surveyed the genetic diversity of 129 accessions of Asian cultivated rice (O. sativa ssp. japonica and O. sativa ssp. indica) and 43 accessions of wild rice (O. rufipogon and O. nivara). F_ST at 45 InDels loci was calculated among wild rice and Asian cultivated rice (Table S2). F_ST was widely used to measure the level of population differentiation and detect artificial selection in rice populations. Twelve of the 45 loci between wild rice and japonica (Table 1), 5 of the 45 loci between wild rice and indica (Table 2), and 35 of the 45 loci between japonica and indica showed a F_ST value greater than 0.4. We defined highly diverged locus between japonica (or indica) and wild rice with F_ST > 0.4 as previous study (He et al., 2011). So 17 loci are highly diverged between japonica (or indica) and wild rice. These loci are R1M7, R1M37, R1M47, R2M10, R2M24, R2M26, R2M50, M3M10, R3M23, R3M30, R4M17, R4M43, R4M50, R6M44, R7M37, R8M23 and R9M42. Because neutral demographic scenarios could affect F_ST, we tested whether the polymorphism pattern at these highly diverged loci can be explained by neutral demographic scenarios alone.

Table 1. Probability value for the test of neutrality in the 12 highly diverged InDel loci between japonica and wild rice

^a Model I (single origin, japonica domesticated first).

^b Model II (single origin, indica domesticated first).

^c Model III (multiple origins, japonica domesticated first).

^d Model IV (multiple origins, indica domesticated first).

Table 2. Probability value for the test of neutrality in the five highly diverged InDel loci between indica and wild rice

^a Model I (single origin, japonica domesticated first).

^b Model II (single origin, indica domesticated first).

^c Model III (multiple origins, japonica domesticated first).

^d Model IV (multiple origins, indica domesticated first).

The rice domestication process has been studied well (Gross and Zhao, 2014; Choi et al., 2017; Fornasiero et al., 2022; Izawa, 2022; Shang et al., 2022). It was suggested that the first domestication of Asian cultivated rice occurred in the Yangtze River basin at about 9000 years ago (Fornasiero et al., 2022), but the two subspecies (japonica and indica) originated independently (Choi et al., 2017), indicating that different loci may have different evolutionary histories due to migration and introgression. Therefore, we simulated four different neutral demographic scenarios (the models I, II, II and IV) of the single and the multiple origins (Fig. 1), which represent the possible neutral evolutionary histories of genes in two cultivated rice subspecies.

We then compared the simulated and observed data using the proposed neutrality test (see Materials and Methods). At 17 highly diverged loci in the four models, we only observed R6M44 in the model I (single origin, japonica being domesticated first) with a marginally significant difference (p = 0.042) (Tables 1 and 2). The value becomes insignificantly different when the multiple testing is considered. Other p values are between 0.101 and 1.000. Therefore, the high F_ST at 17 loci may have resulted from demography other than selection during japonica and indica domestication.

Discussion

In this study, we genotyped 45 InDels in the privately collected 172 accessions and identified 17 highly diverged InDels between the rice cultivars and their wild ancestors. Large-scale coalescent-based simulations were conducted on four neutral demographic scenarios to test the neutral evolutionary hypothesis of the 17 InDels. The results of neutrality test suggested that demography alone could explain the polymorphic profiles on the highly diverged InDels, irrespective of which domestication hypothesis was used and which subspecies was first domesticated. Therefore, F_ST is informative for detecting artificial selection in rice, but high F_ST should not be used as conclusive evidence that candidate loci are affected by artificial selection.

To examine these 17 highly diverged loci, the new statistical test was developed. It has a number of advantages. First, the test was conducted conditional on a mutation, thus evolutionary rate heterogeneity in the rice genome (Zhao et al., 2018) does not affect our analyses. Second, the test can be used to analyse a single-point mutation because we did not limit the mutation type as InDels (Yu et al., 2022). Third, the test is based on a single locus assuming no recombination occurred within the locus (Yang et al., 2018). Therefore, the test is ready to be applied to other sexual and asexual species.

Though demography alone could explain the polymorphic profiles on the 17 loci with high F_ST, our results do not indicate that there is no artificial selection during the domestication of rice. The number of loci examined in this study is small, comparing with those in genome-wide surveys. Our results indicate that it is extremely important to consider the demography as a confounding factor when detecting artificial selection in rice (Gao and Innan, 2008; He et al., 2011; Molina et al., 2011). When genome-wide scans are conducted for searching the signal of artificial selection, the number of windows examined increases and it is very likely to observe a number of loci with high F_ST due to random demographic effects, such as demography.

Because the detection of selection may be hindered by the demographic effects, it is further suggested that the combination with different signatures of selection (Lin et al., 2011; Horscroft et al., 2019) could be considered when detecting artificial selection. Moreover, the confounding effects of demography can be addressed by controlling its false-positive rate, which is generally composed by two-step analysis (Li and Stephan, 2006; Koropoulis et al., 2020). Genome-wide DNA polymorphism is used to infer the demography (Speidel et al., 2019; Hu et al., 2023) and the detection of positive selection is followed conditional on the inferred demography. The alternative approaches could be to use neutrality tests that are insensitive to demographic events (Li, 2011; Hunter-Zinck and Clark, 2015). Overall, our results suggest that it should be more cautious about the explanation of highly diverged loci when studying domestication of indica and japonica rice varieties.