Fine mapping of hybrid necrosis gene Ne1 in common wheat (Triticum aestivum L.)

Hybrid necrosis gene Ne1 was delimited into an approximate 4.06 Mb region on chromosome arm 5BL and an InDel marker that co-segregated with Ne1 alleles was developed. Hybrid necrosis in wheat, characterized by progressive chlorosis and necrosis of plant leaves, tillers or whole plants in certain hybrids, is caused by complementary genes Ne1 and Ne2 located on chromosome arms 5BL and 2BS, respectively. Hybrid necrosis can be a barrier in combining desirable traits from various wheat genotypes. In this study, we fine mapped Ne1 on chromosome arm 5BL, and delimited it to a 4.06 Mb region using large segregating recombinant inbred line families from cross ‘Zhengnong 17’ × ‘Yangbaimai’. Genetic characterization confirmed that the ne1 allele was closely associated with a 2.89 Mb deletion in Zhengnong 17. A tightly linked InDel marker, 5B-InDel385, for Ne1 was developed and was used to predict the presence of Ne1 in a diverse panel of 501 common wheat accessions. Among those accessions, 122 (61%) of 200 landraces were predicted to carry the Ne1 allele, whereas only 79 (26%) of 301 modern cultivars were predicted to carry Ne1. The significant decrease in Ne1 frequency in modern cultivars indicated that the Ne1 allele had been negatively selected in wheat breeding. This study provides a foundation for marker-assisted selection, gene cloning and functional studies of Ne1 in wheat.


Introduction
Hybrid necrosis in wheat is characterized by premature senescence or death of leaves, tillers and even the whole plants in certain hybrids (Caldwell and Compton 1943;Hermsen 1963a). Hybrid necrosis can be lethal or semilethal resulting in gradual death of vegetative tissues or loss of seed productivity (Chu et al. 2006;Tsunewaki 1960). Thus, hybrid necrosis can be a barrier to gene transfer in breeding and can hinder the genetic improvement of wheat (Bizimungu et al. 1998;Galaiev 2016;Tomar et al. 1991).
Hybrid necrosis is caused by interaction of complementary dominant genes Ne1 and Ne2, located in chromosome arms 5BL and 2BS, respectively (Chu et al. 2006;Kandel et al. 2017;Nishikawa et al. 1974;Pukhalskiy et al. 2018;Zeven 1972). Necrosis-inducing alleles of these genes are widely distributed in tetraploid and hexaploid wheat (Tsunewaki 1992;Vikas et al. 2013). Based on the relative strengths of necrosis symptoms, three dominant alleles (Ne1 w , Ne1 m , and Ne1 s ) were identified at the Ne1 locus, and five dominant alleles (Ne2 w , Ne2 wm , Ne2 m , Ne2 ms , and Ne2 s ) were reported for the Ne2 locus (Hermsen 1960(Hermsen , 1963aPukhalskiy et al. 2019). A wheat genotype carrying only the Ne1 allele (Ne1Ne1ne2ne2) or the Ne2 allele (ne1ne1Ne2Ne2), or neither (ne1ne1ne2ne2), has a normal green phenotype. When Ne1 and Ne2 alleles are both present (Ne1-Ne2-), hybrid necrosis appears, and the degree of necrosis is determined by the particular combination of alleles at the Ne1 and Ne2 loci (Hermsen 1963a). However, the molecular basis of wheat hybrid necrosis is still largely unknown. Communicated by Urmil Bansal. Due to the widespread occurrence of Ne1 and Ne2 alleles in wheat accessions, it is useful to know which alleles might be present in order to avoid the occurrence of necrotic F 1 and thus to ensure useful progeny (Chu et al. 2006;Galaiev 2016). In the past, researchers relied on hybridizations with fixed testers containing genotype ne1ne1Ne2 s Ne2 s or Ne1 s Ne1 s ne2ne2 to test if a genotype carried a particular necrosis allele (Hermsen 1963a). This approach is timeconsuming, laborious, and limited in accuracy as the allele prediction is a qualitative assessment and the symptoms of hybrid necrosis are affected by degree of dominance, genetic background and environmental factors (Hermsen 1963a). Chu et al. (2006) mapped the hybrid necrosis loci Ne1 and Ne2 using SSR markers and demonstrated that markers Xbarc74-5B and Xbarc55-2B were linked to Ne1 and Ne2 at genetic distances of 2.0 cM and 3.2 cM, respectively. Due to the relatively loose genetic distances between the markers and Ne genes, incorrect predictions were likely and use of markers was adopted only in a limited way (e.g., Galaiev 2016;Kandel et al. 2017). Therefore, it is necessary to develop a more convenient and efficient molecular marker for prediction of Ne alleles in wheat germplasm.
Although hybrid necrosis was first reported in the 1940's (Caldwell and Compton 1943), Ne loci have not been finemapped or isolated. In this study, we developed a recombinant inbred line (RIL) population segregating at the Ne loci, and used selected segregating lines to fine map the Ne1 gene. We developed an InDel marker that co-segregated with Ne1 in a large mapping population. We also determined the distribution of Ne1 in bread wheat accessions using the InDel marker.

Plant materials
A RIL (F 8 ) population containing 188 lines was developed from an F 2 population derived from the cross 'Zhengnong 17' (ZN17, ne1ne1Ne2Ne2) × 'Yangbaimai' (YBM,Ne1Ne1ne2ne2). The two parents exhibited normal growth, and 24 RILs showed variable levels of hybrid necrosis. This RIL population was used for mapping Ne1 and Ne2. Two lines  with Ne1ne1Ne2Ne2 genotype in this population showed segregation for normal and necrotic plants. Homozygous F 9 individuals (ne1 YBM n-e1 YBM Ne2Ne2 or Ne1 ZN17 Ne ZN17 Ne2Ne2) from RIL-45 and RIL-128 were identified with markers and self-pollinated to produce near-isogenic lines (NIL-Ne1 YBM and NIL-ne ZN17 ).
An additional RIL population (F 8

Field experiments and phenotypic evaluation
The necrosis of the RIL population was evaluated in three replicates at two locations, Beijing (40° 16′ N, 116° 24′ E) and Zhaoxian county in Hebei province (37° 50′ N, 114° 49′ E), during two growing seasons (2016-17 and 2017-18). The populations used for fine mapping of Ne1 were grown at the latter site in 2017-2018. Each line was represented as a single one-meter row plot with 11 plants, 25 cm between rows, and 10 cm between plants. Necrosis was scored at the grain fill stage.

Association analysis for Ne loci
Genomic DNA was extracted from young leaves by the CTAB method (Chatterjee et al. 2002). The parents and 188 RILs were genotyped with the wheat660K SNP array (https:// wheat. pw. usda. gov/ ggpag es/ topics/ Wheat 660_ SNP_ array_ devel oped_ by_ CAAS. pdf). An association analysis between polymorphic SNPs and necrosis scores was conducted with the function of GLM (Generalized Linear Model) implemented in the software TASSEL Version 5.0 (https:// tassel. bitbu cket. io/). The threshold P-value was set to 3.42 × 10 −7 to select associated SNPs based on the Bonferroni adjustment of independent SNPs. Genome coordinates presented in the study were based on the Chinese Spring RefSeq v1.0 assembly, and gene annotations were based on the updated RefSeq v1.1 annotation (IWGSC 2018).

Ne1 induced the necrotic phenotype and yield decrease in the presence of Ne2
The necrotic phenotype in NIL-Ne1 YBM first appeared on the lower leaves at the tillering stage ( Fig. 1a) and became more evident at the heading stage in the field. More detailed observations indicated that leaf necrosis initiated from the bottom leaves and progressed gradually to the upper leaves, and advanced from the tip to base of the leaf (Fig. 1b). Necrosis formation in NIL-Ne1 YBM leaves appeared to follow a developmental pattern, in which the severity of necrosis was positively correlated with the age of leaves. At the grain fill stage, only the flag leaf or upper leaves remained partially green, and the rest of leaves were necrotic or dried in the NIL-Ne1 YBM , while at the same stage the leaves of NIL-ne1 ZN17 remained green and healthy (Fig. 1c). In addition, the NIL-Ne1 YBM displayed smaller leaves ( Fig. 1b) and fewer tillers, with a shorter plant height ( Fig. 1c) compared to those of the NIL-ne1 ZN17 . The premature senescence of leaves in NIL-Ne1 YBM resulted in significantly yield decline compared to NIL-ne1 ZN17 (Fig. 1d).

Genetic mapping of Ne1
Necrosis was very evident at the grain fill stage, twentyfour of 188 RILs from ZN17 × YBM showed necrosis symptoms in all environments, whereas all the other 164 lines showed normal phenotypes. Therefore, association analysis was conducted using data from three replications grown in Zhaoxian county during 2016-2017.
The parents and 188 RILs of the ZN17 × YBM population were genotyped with the wheat660K SNP array. After removing the SNPs with > 5% data missing values and multiple mapping sites on RefSeq v1.0 (Keeble-Gagnere et al. 2018), 146,351 high-quality polymorphic SNPs were obtained. The associations of SNPs and the necrotic phenotype were made with the GLM function in TASSEL Version 5.0 software. There were 1,844 SNPs significantly associated with the phenotype at the threshold of P-value (3.42 × 10 −7 ). Among these SNPs, 916 (49.67%) and 870 (47.18%) were on chromosome arms 2BS and 5BL, respectively, and the remaining 58 (3.15%) were distributed across chromosomes 2A, 3A, 5A, 2D, 4D and 5D (Fig. 2a). These results were consistent with previous reports showing that hybrid necrosis in common wheat was controlled by Ne1 on 5B and Ne2 on 2B (Chu et al. 2006;Nishikawa et al. 1974;Zeven 1972). The significant SNPs on 5BL were in the ~ 297 Mb to 399 Mb region and those on 2BS located between ~ 110 Mb and 180 Mb (Fig. 2b).

NIL-Ne1
These results indicated that the necrotic phenotype in this population was controlled by a semi-dominant gene. Based on the IWGSC RefSeq v1.0 and flanking sequences of the associated SNPs, 50 molecular markers were developed for the Ne1 region (~ 290-400 Mb) on chromosome arm 5BL; among them, four were polymorphic between ZN17 and YBM. These four markers were genotyped on 682 individuals in RIL-128. Ne1 was delimited to a region flanked by markers 5B-378 and 5B-388 (Fig. 2c). A further 3,402 individuals from this population were analyzed to screen for additional recombinants; 23 recombinants were identified between the markers 5B-378 and 5B-388 (Fig. 2d). Based on re-sequencing data of both parents, we developed new polymorphic InDel and SNP markers and genotyped the recombinant individuals to narrow down the candidate interval of Ne1, which was eventually mapped to a 4.06 Mb physical interval delimited by markers 5B-383 and SN-2142 (Fig. 2c, d).
The ne1 allele in ZN17 was associated with a 2.89 Mb deletion on chromosome arm 5BL In order to clone Ne1, we designed 57 SSR markers in the interval of ~ 383 Mb to 387 Mb between markers 5B-383 and SN-2142 on 5BL based on IWGSC RefSeq v1.0. Most of the SSR marker variations between YBM and ZN17 were presence/absence. We speculated that this was caused by a deletion in ZN17. For confirmation, we re-sequenced ZN17 and YBM at 10X sequencing depths. Sequence reads were aligned to the reference genome of Chinese Spring (RefSeq v1.0), and SNPs and InDels located in the region between markers 5B-383 and SN-2142 were called by the Haplo-typeCaller module (Chai et al. 2018). We found a continuously missing sequence from ~ 383,441,497 to 386,325,646 (RefSeq v1.0) in ZN17, but not in YBM (Supplementary  Table S4). This suggested that there was an approximate 2.89 Mb deletion between markers 5B-383 and SN-2142 in ZN17. To verify the missing segment, we designed primers 1 3 to amplify six high confidence genes located in the putative missing segment of chromosome arm 5BL in the NILs and parents ZN17 and YBM (Supplementary Table S3). All six genes were detected in YBM and NIL-Ne1 YBM , but not in ZN17 and NIL-ne1 ZN17 . These results further confirmed deletion of the Ne1 candidate region in ZN17. Therefore, it was not possible to clone Ne1 with the population developed from cross between ZN17 × YBM.

Development, validation and application of a selection marker for Ne1
To confirm that the missing 2.89 Mb fragment of chromosome arm 5BL was closely associated with Ne1, we screened a new segregating RIL (RIL-368, Ne1ne1Ne2Ne2) from the cross 'Zhengzhou 6903' × 'Yumai 14'. This population segregated for the necrotic phenotype in the field and showed the same PCR product pattern as RIL-128 with marker 5B-378. We also identified all six genes located in the 2.89 Mb deletion in Yumai 14 (carrying Ne1), but not in Zhengzhou 6903 (carrying ne1). This indicated that deletion in ne1 varieties might not be rare. We chose 5B-InDel385 as a predictive DNA marker for Ne1 and divided wheat accessions into genotype 5B-InDel385_YBM and 5B-InDel385_ZN17 based on the presence or absence, respectively, of the 2.89 Mb fragment.
To test the efficiency of the selection marker 5B-InDel385, 1,034 individuals from RIL-368 were genotyped and analyzed. As expected, all plants with normal phenotype were 5B-InDel385_ZN17 and all necrotic individuals were 5B-InDel385_YBM. These data confirmed that Ne1 co-segregated with the marker 5B-InDel385 on chromosome arm 5BL.
The 5B-InDel385 was also validated in some carriers of known alleles of Ne1 and Ne2/Lr13; including Kubanka (T. durum, Ne1 s ), Chinese Spring (Ne1 w ) and Sonalika, Manitou and Frontana (Ne2 m /Lr13). The marker 5B-InDel385 assay showed that the cultivars carrying Ne2 m /Lr13 had the 5B-InDel385_ZN17 associated with the ne1 allele. Kubanka and Chinese Spring, known to possess Ne1 alleles, were 5B-InDel385_YBM type. Based on these results, we conclude that the marker 5B-InDel385 could accurately distinguish Ne1 from ne1 which if included within the deletion would be a null allele.
Using the selection marker, we characterized 259 Chinese common wheat accessions (29 landraces and 230 cultivars). The Ne1 allele was frequent in landraces (62%, Fig. 3a). Conversely, its frequency was much lower in modern cultivated wheat varieties (32%, Fig. 3b). We studied the geographical distribution of the Ne1 allele in China and found that the frequency of wheat varieties containing the Ne1 allele was highest in Henan (41.51%), followed by Shaanxi (41.38%), Shanxi (33.33%), Beijing (31.25%), Hebei (28.57%), Jiangsu (29.41%), and Shandong (5.13%; Table 1). Furthermore, 242 common wheat landraces/cultivars from diverse origin were detected with 5B-InDel385. The results further confirmed that the 2.89 Mb deletion of the Ne1 candidate region was widespread throughout the world. The Ne1 allele was  Fig. 3c). On the contrary, the Ne1 allele was present only in 3 (7%) of a subset of 71 common wheat cultivars from the USA (Fig. 3d). In total, 122 landraces (61%) had Ne1 (Fig. 3e) compared to 79 modern cultivars (26%) (Fig. 3f). This result suggested that the Ne1 allele had been subjected to negative selection pressure in breeding. In addition, we also predicted the presence Ne1 alleles in 40 tetraploid wheat varieties and found that 12 accessions were positive (Supplementary Table S2). All current results indicate that the deletion is frequent in both hexaploid and tetraploid wheat.

Putative Ne1 candidate genes
We analyzed the predicted genes on the chromosome arm 5BL from 383.03 Mb to 387.10 Mb of the Chinese Spring RefSeq v.1.0 sequence to identify candidate gene for Ne1. Fifty-four genes were identified in the region; 28 were in the deleted 2.89 Mb region, 6 were in the region between marker 5B-383 and the deletion, and 20 were between marker SN-2142 and the deletion (Supplementary Table S5). Re-sequencing analysis of ZN17 and YBM showed that the main difference between these parents in the Ne1 candidate region was the 2.89 Mb deletion and no amino acid difference was identified in the 26 genes outside of the deletion bin although there were 385 SNPs/InDels present in the entire candidate interval (Supplementary Table S6).
To predict the Ne1 gene, we analyzed the expression profiles of genes in the candidate region of Chinese Spring using the wheat expVIP expression platform (http:// www. wheat-expre ssion. com/). It was known that the necrotic phenotype was caused by interaction of the Ne1 and Ne2 alleles, and that 'Chinese Spring' carried the Ne1 w allele (Hermsen 1963a;Zhang et al. 2016). Since symptoms of hybrid necrosis begin from an early seedling stage, we hypothesized that the Ne1 allele should express in leaves and shoots for the entire growth period. Eighteen of 54 candidate genes expressed (above two transcripts per million) in at least three RNA-seq samples of Chinese Spring (leaves/ shoots and roots, n = 40) at different developmental stages (Supplementary Table S7). Nine of the 18 expressed genes were high-confidence genes, two encoding auxin-responsive SAUR proteins, two encoding serine protease HtrA-like proteins, two encoding alpha/beta-hydrolase superfamily proteins, one encoding an initiation factor 4F subunit (DUF1350), one encoding a RING/U-box superfamily protein, and one encoding a trypsin-like serine protease (Supplementary Table S8).

The hybrid necrotic phenotype is influenced not only by gene dosage (dominance), but also by the genetic background
Wheat hybrid necrosis phenotypically manifests itself as premature gradual perishing of leaves, sheaths and even entire plants (Caldwell and Compton 1943;Hermsen 1963a). Hybrid necrosis is caused by the interaction of functional alleles at the complementary Ne1 and Ne2 loci (Hermsen 1963a), located on chromosomes 5B and 2B, respectively (Chu et al. 2006;Nishikawa et al. 1974). Previously, Ne1 alleles were considered to be dominant based on the segregation ratio of necrotic: normal plants in F 2 populations derived from parents with genotypes Ne1Ne1ne2ne2 and ne1ne1Ne2Ne2 (Hermsen 1963a). Ne1 had dosage effects (dominance) and Ne1Ne1Ne2Ne2 genotypes developed more severe necrosis than Ne1ne1Ne2Ne2 genotypes (Hermsen 1963b). In this study, we observed that heterozygous Ne1ne1Ne2Ne2 individuals in segregating RILs were significantly different from homozygous Ne1Ne1Ne2Ne2 sibs ( Supplementary Fig. S1), confirming that Ne1 was an incompletely dominant allele.
In addition, we found that the 24 of 188 RILs from ZN17 × YBM showed different degrees of necrosis indicating minor effects from other genes. Most of the necrotic lines exhibited necrosis at the leaf tips during the seeding stage; necrosis gradually progressed to the middle and bottom of the leaves at the jointing and flowering stages, and leaves  Fig. S2a-c). Three lines showed weak necrosis at the flag leaf tips at late jointing to flowering stage (Supplementary Fig. S2d). At the grain fill stage, six lines showed severe necrosis and shriveled grains caused by premature death of the leaves (Supplementary Fig. S2e); and two lines had yellow leaves ( Supplementary Fig. S2f). This result implied that other genes were causing variation in symptoms other than Ne-gene interaction.
Allele ne1 is likely to be null due to location of Ne1 in a 2.89 Mb deletion Microsatellite marker Xbarc74-5B was previously located 2.0 cM from Ne1 (Chu et al. 2006 (Zhang et al. 2016), both were classified as genotype 5B-InDel385_YBM. Thus, we can select many germplasms with ne1 using 5B-InDel385 to predict crosses that avoid hybrid necrosis. We mapped Ne1 to 383.03-387.10 Mb on chromosome arm 5BL (Chinese Spring RefSeq v.1.0). There were 54 annotated genes in this region, including 36 low confidence genes and 18 high confidence genes (IWGSC 2018). Resequencing of ZN17 and YBM revealed that 28 of the 54 genes were in a 2.89 Mb deletion in ZN17 and that another 385 SNPs/InDels caused no amino acid changes between the parents. Thus, it was likely that the expression differences of candidate genes led to the necrotic phenotype. We examined the expression profiles of all 54 genes in the 4.06 Mb genomic region of Chinese Spring and identified 9 highconfidence genes that were expressed in leaves/shoots or roots (Supplementary Tables S7, 8). We hypothesized that these 9 genes were candidates for Ne1. Gene annotation of the corresponding region in Chinese Spring revealed two auxin-responsive SAUR proteins and one initiation factor 4F subunit (DUF1350) (Supplementary Table S8). Rice plants overexpressing the SAUR39 (one small auxin-up RNA) gene exhibited senescence of lower leaves beginning prior to initiation of the reproductive stage but extending upward with plant development (Surya et al. 2019). eIF4F is a complex formed by three proteins: eIF4A, eIF4E, and eIF4G and is tightly regulated to provide protection from abiotic and biotic stress. In Arabidopsis, double mutant eukaryotic translation initiation factors eIFiso4G1 (i4g1) and eIFiso4G2 (i4g2) exhibited pronounced premature senescence of primary rosette leaves long before completion of the life cycle (Lellis et al. 2010). However, based on the available data we could not determine the gene underlying Ne1. Therefore, quickest way to identify Ne1 might be to individually knock out the nine most likely candidate genes in the presence of an Ne2 allele. Disruption of the candidate gene should produce a normal phenotype. Although we associate the ne1 phenotype with a large deletion that appears to be frequent, a search for an ne1 haplotype that does not involve the deletion could enable map based cloning. An alternative procedure would be to identify a chemically induced mutant of Ne1. Once Ne1 is cloned the basis of its allelic variation can be addressed.

The frequency of the 2.89 Mb deletion has increased during wheat breeding
Five hundred and one landraces/cultivars from diverse origin were analyzed with marker 5B-InDel385. The frequency of Ne1 carriers was higher in landraces, both from China and the USDA-ARS National Small Grains Collection (from Central, West and South Asia; Fig. 3a, c). In contrast, the frequency of Ne1 was significantly lower (32%) in modern cultivated wheat varieties in China (Fig. 3b) and was only 7% in cultivars from the USA (Fig. 3d). This was consistent with Pukhalskiy et al. (2000), who reported that 9.1% of wheat cultivars in North and South-Central USA were Ne1 carriers. Therefore, the frequency of Ne1 was non-random due to negative selection pressure in breeding programs (Pukhalskiy et al. 2008(Pukhalskiy et al. , 2000. In addition, we found the 2.89 Mb deletion was also frequent in wild emmer wheat and durum wheat; an Ne1 allele with the intact fragment was present in 12 (30%) of a panel of 40 tetraploid wheat accessions (Supplementary Table S2). This agreed with other researhers (Maccaferri et al. 2019;Zhu et al. 2019).
Since hybrid necrosis is the result of interaction between Ne1 and Ne2 alleles, the frequency of Ne1 can be influenced by the presence of Ne2 in wheat breeding programs. It is well known that the Ne2 gene is tightly linked to the Lr13 gene (a widely distributed leaf rust resistance gene in wheat) located on chromosome 2BS, and genetic and mutational analyses indicate that they are the same gene (McIntosh et al. 1995;Zhang et al. 2016). Thus, the fact that Ne2 and Lr13 represent a pleiotropic locus led to a sharp increase in the frequency of Ne2 m in wheat in the USA, as Lr13 was favored by selection for leaf rust resistance (Pukhalskiy et al. 2000;Zhang et al. 2016). The long-term use of Lr13/Ne2 m may explain the significant decrease in frequency of Ne1 among USA cultivars that may have been selected to avoid hybrid necrosis.

Conclusion
Using large populations of selected RILs segregating at the Ne1 locus and development of new molecular markers, the hybrid necrosis gene Ne1 was mapped to a 4.06 Mb region (383.03 Mb-387.10 Mb) on chromosome arm 5BL and cosegregated with InDel marker 5B-InDel385. Significant differences in Ne1 frequency between landraces and modern cultivars predicted by the marker indicate that Ne1 alleles have been subjected to strong negative selection in wheat breeding.