CRISPR-mediated accelerated domestication of African rice landraces

African Oryza glaberrima and Oryza sativa landraces are considered valuable resources for breeding traits due to their adaptation to local environmental and soil conditions. They often possess superior resistance to endemic pests and tolerance to drought and nutrient deficiencies when compared to the “imported” high production Asian rice varieties. In contrast, “domestication traits” such as seed shattering, lodging, and seed yield are not well established in these African landraces. Therefore, the use of these African varieties for high production agriculture is limited by unpredictable yield and grain quality. We are addressing this shortcoming by developing protocols for genetically transforming African landraces to allow the use of CRISPR-Cas mediated breeding approaches. Here we use as proof of concept the cultivated African landrace Kabre to target selected known “domestication loci” and improve the agronomic potential of Kabre rice. Stable genetic transformation with CRISPR-Cas9-based vectors generated single and simultaneous multiple gene knockouts. Plants with reduced stature to diminish lodging were generated by disrupting the HTD1 gene. Furthermore, three loci shown to control seed size and/or yield (GS3, GW2 and GN1A) were targeted using a multiplex CRISPR-Cas9 construct. This resulted in mutants with significantly improved seed yield. Our study provides an example of how new breeding technologies can accelerate the development of highly productive African landrace rice varieties, an important advancement considering that Africa is a hotspot for worldwide population growth and therefore prone to food shortage.


Introduction
Rice is the primary food source for half of the world's population [1]. The genus Oryza is comprised of 23 species, two of which, Oryza sativa and Oryza glaberrima, have been systematically cultivated as a food crop [2]. O. sativa and its main two subspecies indica and japonica, were independently domesticated in the northeastern part of India and in South East China, respectively, roughly 10,000 years ago from wild O. rufipogon populations [3]. Cultivation of O. glaberrima began only 3,000 years ago around the upper Niger River Delta in West Africa, following differentiation from its wild ancestor O. barthii [4]. O. sativa varieties were introduced into West Africa by European colonizers around the 15 th and 16 th centuries, when trade between Africa and India was flourishing [5], and over time supplanted the cultivation of O. glaberrima, the native domesticated rice species. Many of these O. sativa varieties were subjected to human selection and evolved into landraces with adaptations beneficial under local conditions, but were never subjected to intensive and systematic breeding [6]. In Africa, both O. sativa and O. glaberrima accessions are grown using low-input, minimally mechanized agriculture, intended for local consumption [5] [7] [8]. The African landraces have been shown to carry several interesting traits that could empower sustainable and less demanding agricultural production. Valuable features present in African landrace germplasm were neglected for many years, as breeding programs focused on increasingly productive Asian rice varieties. Among these untapped traits are weed competitiveness, photoperiod insensitivity, and resistance to biotic and abiotic stresses [6] [9]. On the contrary, low yield, shattering and lodging are negative traits often present in African landraces that are lost in modern Asian rice varieties [10].
Since the O. sativa genome was published in 2002 [11] many yield-related loci repeatedly selected during domestication have been identified, mapped and functionally characterized [12]. Together with recent advances in CRISPR-Cas9 genome editing technology [13] [14], this knowledge creates unprecedented opportunities for crop improvement. A good example is the work done by Zhou and colleagues [15] who obtained a significant yield increase in O. sativa by simultaneous introduction of three trait related QTLs. To demonstrate that CRISPR technology has the potential to accelerate improvement of African landraces, we are developing transformation protocols for both African O. glaberrima and O. sativa landraces. Here we used as a proof of concept the African O. sativa landrace Kabre and targeted well-known domestication loci that are known to control plant height and seed yield.

Agrobacterium mediated transformation of African landraces
The callus transformation protocol was adapted from Toki [16] and fully described in Supplemental File 1.

Genome editing and constructs
Single knockouts were obtained using plasmids kindly provided by Miao et al. [17], following the guidelines reported by the authors [17], pOs-sgRNA was used to clone a single gRNA targeting HTD1. Subsequently, using the Gateway cloning system the cassette was moved to the binary vector pH-Ubi-cas9-7 encoding for the Cas9 endonuclease and hygromycin resistance as plant selection marker.
Four gRNAs of 20 nt, one for each gene to be disrupted HTD1, GN1A, GS3 and GW2, were designed using the online tool CRISPR-P v2.0 (http://crispr.hzau.edu.cn/CRISPR2/). gRNAs proposed by the software were manually screened and selected based on their proximity to the start codon and avoiding potential off-targets.

Targeted loci
Since Kabre was indicated to be an O. glaberrima accession both by the National Small Grain Collection (USDA) repository and in a previous publication [19], we initially designed the gRNA matching HTD1 using the O. glaberrima orthologous gene sequence (ORGLA04G0179300) identified by performing a blast-search with the O. sativa locus (OS04G0550600). Prior to proceed with plant transformation in order to ensure a perfect sequence match between the O. glaberrima annotated gene and Kabre, the gRNA targeting location and neighboring region were cloned and sequenced using primers Osp_1332 and Osp_1333 (S3 Table). Primers Osp1209 and Osp1210 were used to create the gRNA for targeting HTD1 (S3 Table).
As described in the results section, analysis of morphological traits and genotyping revealed that Kabre is most likely an O. sativa accession. For this reason, gRNAs used to target each of the three seed yield regulators GS3, GN1A and GW2 were designed on the O. sativa loci Os03g0407400, Os01g0197700 and Os02g0244100, respectively. PCR amplification and sequencing of the Kabre gRNA-targeting regions was conducted using specific primers (S3 Table) to confirm sequence identity and the absence of mismatches.
Primers Osp1367 to Osp1372 (S3 Table) are the gRNAs for each of the three genes with adapters linked in order to assembly the multiplex CRISPR construct following detailed instructions published by Xie et al. [24] (pRGEB32 vector). Osp1584 and Osp1585, matching Cas9 of pRGEB32, were used to assess the presence of the transgene in plants transformed with the multiplex construct. Atp5706 and Atp5718 instead were adopted to screen for the presence of the transgene in plants transformed with the construct for a single gene knock out (i.e. Kabre htd1) as published by Miao et al [17].

Genomic DNA extraction and PCR analysis
Genomic DNA (gDNA) was extracted following the protocol described in Supplemental File 1. All PCR reactions using gDNA were performed with GoTaq1 (Promega) polymerase according to the company instructions.

Genotyping
Genotyping of Kabre was conducted in order to assess whether it has an O. sativa or O. glaberrima genetic background. We amplified by PCR three different species-specific genetic markers RM197, PROG1 and S1 as described in the results section. PCR was performed using Terra™ PCR Direct Polymerase Mix (Takara) using a small piece of rice leaf directly in the reaction tube according to the Takara instruction manual. The following PCR condition were used: 98˚C 2min, 35 cycles (98˚C 10 sec, 60˚C 15 sec, 68˚C 30 sec), 68˚C 10 min. Samples from O. sativa (IR64 and Gigante accessions) and O. glaberrima (CG14 and TOG5681 accessions) were used as control. All the primers used for this study are listed in S3 Table.

Growth conditions
All the stages of tissue culture, except co-cultivation, were conducted under continuous cold fluorescent light at a temperature of 28˚C. Once regenerants were ready for transplanting they were moved to soil and kept at 27˚C in long day for about two months (18 hrs light). Later, plants were moved to short day conditions (14 hrs light) to induce flowering, after about two months rice seeds were harvested.

Trait measurements
Five different Cas9 T1 negative plants derived from htd1-1 were selected and analyzed for the dwarf phenotype, composed by 3 and 2 plants having -7 and -17 bps homozygous deletion, respectively. Kabre mutants for seed traits were selected among the T2 generation. Seeds were measured for length and width using "SMARTgrain" a free program available online [20]. Six wild type, 10 gs3 and 10 gn1a single mutants, 18 gs3 gn1a double mutants and 11 gs3 gn1a gw2 triple mutant plants were analyzed taking 40 seeds from each plant; results are presented as average values (Fig 1). All selected mutant genotypes (e.g. single and double mutants) were equally composed in numbers by two different mutant alleles as described in the results section (S2 Fig). In contrast, T2 triple mutants were derived from a single T1 line and thus after segregation were comprised of 5 plants having either +1 insertion on gn1a and 6 plants carrying a -1 deletion on the same gene, while all the plants carried the same mutation on the two others loci gs3 and gw2 as reported in the Results section (S2 Fig). The statistical significance of differences was analyzed by variance (ANOVA) and post-hoc Tukey HSD.

Transformation of the African landrace "Kabre"
CRISPR-Cas9 genome editing technology has great potential as a tool for plant improvement. However, one requirement for the use of this system is the ability to express the CRISPR guide RNA (gRNA) and Cas9 protein in plant cells. To investigate which African landraces can be used for Agrobacterium-mediated transformation, the regeneration capacity of embryogenic calli from different accessions was tested using a modified protocol published previously by Toki [16] (S1 File). All the African rice varieties that we tested were reported to be Oryza glaberrima rice species (Table 1). Eight of the thirteen accessions that were selected for this study showed the ability to regenerate plantlets from calli cultured on hormone gradient plates (see examples in S1 Fig). The Kabre landrace from Northern Ghana was one of the most vigorously regenerating varieties and was selected for Agrobacterium-mediated transformation with CRISPR-Cas9 constructs. Notably, when grown in the greenhouse, Kabre plants showed phenotypes typical of landrace rice accessions but not modern cultivars, like elevated plant height, small seeds and small panicles. Kabre was thus a perfect candidate for our pilot experiment to show that by targeting specific "domestication" loci, fast improvement of landraces is possible. Table 1 accessions were chosen based on their different geographical distribution across West Africa where O. glaberrima was domesticated. Accession identifiers can be used to retrieve online general information and features through the General Resources Information Network (GRIN-Global https://training.ars-grin.gov). TOG5681 and CG14 accessions were obtained from IRD (Montpellier, France). All the remaining accessions were obtained from the U.S.D.A. National Small Grains Collection (NSGC, Stuttgart, Arkansas, USA). In the last column those accessions that showed regeneration of plantlets from calli are indicated with "+". The hormonal concentrations for which we obtained an optimal regeneration are reported in S2 Table.

Improving plant height and tillering
African landraces are often lodging-sensitive due to their excessive height. In Asian rice, loss of function oshtd1 mutants were reported to exhibit dwarfism and over-proliferation of tillers [21], making plants less lodging sensitive and able to produce a higher number of panicles per plant. OsHTD1 (OS04G0550600), encodes a carotenoid cleavage dioxygenase. We performed a Genebank search to identify the O. glaberrima HTD1 (OgHTD1) ortholog. The OgHTD1 gene was identified on chromosome 4 (ORGLA04G0179300) sharing over 99% sequence homology with OsHTD1.  Five Cas9 negative plants from the T1 progeny of htd1-1 were analyzed. These plants displayed excessive tiller production and dwarfism (Fig 1A) in agreement with what has been previously shown in O. sativa Nipponbare [21] thus indicating potential use of this mutation to reduce sensitivity to lodging in this African landrace. As compared to the wild type HTD1 locus which encodes for 609 amino acids, deletions of 17 and 7 base pairs would result in truncated proteins of 195 and 243 amino acids, respectively (S2 Fig). The obtained phenotypes suggest that CRISPR-mediated approaches could lead to important traits improvement in African rice landraces.

Kabre has characteristics typical for O. sativa
When carefully analyzing the htd1-1 and htd1-2 mutant phenotypes we observed that the mutant and wild-type Kabre plants showed some typical O. sativa characteristics, despite being registered as an O. glaberrima species. The mature panicle was not erect and open as in O. glaberrima but more like in O. sativa (Fig 2A). Moreover, the ligule, the membranous appendage on the adaxial surface of a leaf, at the junction between the leaf sheath and the leaf blade was typical of O. sativa [22] (Fig 2B). Therefore, we decided to investigate the genetic background of Kabre in more detail and performed a genotyping experiment using different markers discriminating O. sativa and O. glaberrima species (Fig 3). Firstly, SSR marker analysis using the marker RM197 was performed according to Chen et al. [23]. Furthermore, the deletion of the PROG1 gene and a deletion nearby this gene were investigated by PCR analysis since they are reported to be reliable species-specific O. glaberrima markers [8] [24]. Lastly, a marker specific to the S 1 locus corresponding to a postzygotic reproductive barrier-associated locus between O. sativa and O. glaberrima was also used for the analysis [25]. As reference controls for the two different species genomic DNA extracted from O. sativa cv. IR64 and cv. Gigante and from O. glaberrima accessions CG14 and TOG5681 were used. All these analyses suggest that Kabre is an African O. sativa landrace.  Table. https://doi.org/10.1371/journal.pone.0229782.g003

Multiple trait stacking for seed yield improvement in Kabre
To further investigate possible yield improvements in African rice landraces by simultaneously stacking multiple agronomic traits selected during domestication in Asian rice, three genes, GRAIN NUMBER 1A (GN1A), GRAIN SIZE 3 (GS3) and GRAIN WEIGHT 2 (GW2) [12] were selected for mutagenesis. These loci are known to be major QTLs for negatively regulating plant yield. GN1A (OsCXK2), artificially selected during domestication, encodes a cytokinin oxidase/dehydrogenase that negatively controls panicle length and consequently spikelet number [26]. GS3 encodes a Gγ protein subunit containing a transmembrane domain spanning the plasma membrane and is implicated as a negative regulator of seed length and weight in several cultivars [27]. Lastly, GW2 encodes a RING-type E3 ubiquitin ligase and plays a role as negative regulator of grain girth and weight [28].
One gRNA for each of the three target genes was designed based on the O. sativa japonica genome sequence [11]. gRNAs showing ideal features in terms of target specificity and GC content were designed to match the first, fifth and second exon of the GN1A, GS3 and GW2 locus, respectively (S2 Fig). Subsequently the target sequences were verified by sequencing the gRNA matching regions for the three selected loci in the Kabre background to verify the absence of any sequence mismatch (S3 Table). The gRNA sequences were then cloned into the CRISPR multiplex vector pRGEB32 [24]. Following transformation, 76 plants were identified as Cas9 positive by PCR screening. Disrupted loci were verified by sequencing. Four classes of mutants were obtained in T 0 : 3 gn1a and 19 gs3 single mutants, 38 gn1a gs3 double mutants, and 16 gn1a gs3 gw2 triple mutants. Among the 16 triple mutants, only one plant carried the gw2 mutation in homozygosity, all the others were heterozygous for gw2. All the other plants (single and double mutants) had the targeted loci mutated in homozygosity or carried bi-allelic mutations suggesting that the gRNA targeting GW2 had the lowest efficiency. In the T 1 generation, Cas9 negative plants were selected and grown for phenotypic analysis in the T 2 generation. In total, six wild type (WT), 10 gs3 and 10 gn1a single mutants, 18 gs3 gn1a double mutants and 11 gs3 gn1a gw2 triple mutant plants were analyzed. Of these lines we selected two independent mutant lines for further phenotypical analysis. In T 2 selected single mutant gs3 lines carried either a +1 insertion or a -2 deletion, predicted to result in a protein matching the wild type only for the first 150 amino acids and lacking in-frame stop codon thus likely resulting in an aberrant transcript and leading to non-sense mediated decay [29]. Single gn1a mutants analyzed instead carried either a -2 bps deletion or a +1 bp insertion, in both cases the resulting protein were predicted to match wild type only for the first 86 amino acids and contained premature stop codons. gn1a gs3 double mutant lines selected for phenotypic analysis of seed features either carried a -2 or -4 bps deletion in the gs3 locus, the latter possibly translating in an aberrant protein that matched the wild type for the first 150 amino acids, while both genotypes had homozygous -2 bps deletion for gn1a. Finally, Cas9 negative plants were selected among the segregant T 2 progeny obtained from a T 1 triple mutant carrying a -2 bps deletion at the gs3 locus, a biallelic +1/-1 bp mutation for gn1a and a -1 homozygous deletion on the gw2 locus. The latter was predicted to translate to a non-functional truncated protein of 89 amino acids.
The gn1a single mutants showed an important increase in panicle length of approximately 49% relative to wild type. Interestingly, combining the gn1a mutant with gs3 and gw2 partially attenuated this phenotype, even though the overall panicle length increase was maintained (Fig 1B). Grain length and width were measured with SMARTGrain software [20] (Fig 1C and  1D). Analysis of seed length showed that gs3, gs3 gn1a and gs3 gn1a gw2 mutants all produced longer seeds compared to wild type (see also Fig 1F). Among the four classes of mutants, seed width was slightly increased in gn1a and drastically increased in the triple mutant. Moreover, the gn1a gs3 gw2 mutant produced seeds with a significantly enhanced weight of 24% compared to wild type (Fig 1E). This analysis confirmed that, among the three genes, mutating GW2 in Kabre caused the most significant increase in seed width and weight (Fig 1D, 1E and  1G).

Discussion
This study shows that CRISPR-Cas9 mediated mutagenesis promises to be an important tool for improvement of local African landraces. The development of transformation protocols or other technologies to introduce the CRISPR-Cas machinery into living plant cells is therefore of great importance. Our regeneration experiments on hormone gradient plates using 13 different African rice accessions revealed a high variation in regeneration capacity. Therefore, most likely the bottleneck to genome edit different rice accessions adapted to local conditions will at this moment likely not be the CRISPR-Cas technology but more the development of transformation protocols.
We selected Kabre for its high regeneration capacity but especially because it is a typical African landrace with low yield and not subjected to modern breeding. Although being considered an O. glaberrima landrace by the U.S. National Germoplasm System (https://npgsweb. ars-grin.gov) as well as in previous studies [19], analysis of specific loci in its genome suggests that it is an O. sativa accession rather than being an O. glaberrima; nevertheless it has characteristics found in both species. Despite all markers used showed the same results that we obtained for the O. sativa control plants we cannot completely exclude that Kabre is the result of an introgression between O. sativa and O. glaberrima. Complete genome sequencing of Kabre might reveal its origin.
Recently, Zhou et al. [15] CRISPR edited the same three QTLs in Chinese elite rice varieties. In contrast to our observations, the authors showed that the gn1a gs3 gw2 triple mutants overall have longer and bigger panicles than any other class of mutants for the three genes. It is likely that the difference in panicle length caused by the gn1a mutation is due to the genetic background of the rice accessions used. While Zhou and colleagues [15] aimed at a further boost in grain yield from already productive, highly selected, and likely genetically more homogeneous O. sativa varieties, our work was conducted in a more heterogeneous African landrace with a genetic background that was not selected for high productivity.
Importantly, our data show that the accelerated domestication approach by simultaneously stacking multiple mutant combinations through CRISPR-Cas technology provides a means for rapid domestication of African landraces that have traits making them more suitable for sustainable agriculture. Interestingly, by applying the latest CRISPR mediated base-editing technologies [30] [31] it will not only be possible to obtain harvest increases by knocking out negative regulators of yield, as shown in this study, but it will also allow the introduction of any kind of favorable mutation. Altogether if these tools can be introduced in plant cells, they will empower fast adaptation of orphan and neglected crops to facilitate their use in modern high yielding agriculture practices. Furthermore, improvement of 'native' local crops could empower farmers in Africa and elsewhere to grow familiar crops, already adapted to a particular niche, with greater productivity and environmental sustainability. This might well be important to take in consideration since Africa is expected to be the hotspot of population growth (www.un.org).

Conclusions
Climate change and world population growth, especially in Africa, demands for more shortterm solutions to increase crop yield and sustainability of agriculture. CRISPR-Cas based gene editing technology promises to contribute to find solutions, also due to the increased knowledge about gene functions obtained over the last 30 years. Here we demonstrate that knowledge transfer from high productive Asian rice varieties to local African landraces may facilitate accelerated improvement of these crops without losing their specific benefits for the growth of these plants under local conditions. An important requisite to be able to tap into the CRISPR--Cas toolbox is the introduction of this system into plant cells. We foresee a revival of scientists studying tissue culture practices to introduce stably or transiently engineered Cas proteins and gRNA into target plant cells for highly precise gene engineering in vivo.