In hexaploid wheat, ZIP4 homologues are located within the Ph1 locus on 5B, and also on chromosomes 3A, 3B and 3D. Before undertaking the targeted induced lesion in genomes (TILLING) mutant analysis, we assessed the expression of TaZIP4-B2 to confirm that the TaZIP4-B2 gene within the Ph1 locus is expressed during meiosis; that it has a higher level of expression than the ZIP4 homologues present on chromosome group 3; and finally, that Ph1 deletion significantly reduces overall ZIP4 expression. At the coding DNA sequence and amino acid levels, TaZIP4-B2 (AA1305800.1) showed 95.3 and 89.2% similarity to TaZIP4-B1 (AA0809860.1), 94.1 and 87.5% similarity to TaZIP4-A1 (AA0645950.1) and 94.4 and 87.2% similarity to TaZIP4-D1 (AA0884100.1), respectively (Supplementary Fig. 1). To compare the relative expression of these ZIP4 homologues on chromosomes 3A, 3B, 3D and 5B, RNA samples were collected from anthers of hexaploid wheat (Triticum aestivum cv. Chinese Spring) (WT) and the Ph1 deletion mutant (ph1b) at the late leptotene-early zygotene stage, and six libraries prepared for the RNA-seq study. RNA-seq analysis showed that TaZIP4-B2 exhibited a higher level of expression than the ZIP4 homologues on homoeologous group 3 chromosomes (Fig. 1; Supplementary Fig. 1). Moreover, TaZIP4-B2 also showed three splice variants (Supplementary Fig. 2) in contrast to homoeologous group 3 chromosome ZIP4 homologues. One of these splice variants (splice variant 1) accounted for 97% of the TaZIP4-B2 transcripts. As expected, when the Ph1 locus was deleted, expression of TaZIP4-B2 was also eliminated (p < 0.05), but there was no apparent increase in the transcription of the ZIP4 homologues on homoeologous group 3 chromosomes, to compensate for the absence of ZIP4 on chromosome 5B (p > 0.05) (Fig. 1; Supplementary Table 1). Thus, RNA-seq data revealed that the expression of ZIP4 was derived mainly from the gene (TaZIP4-B2) on chromosome 5B, within the Ph1 locus.
TaZIP4-B2 suppresses homoeologous COs in wheat-Ae. variabilis hybrids
The protein-coding sequences of 1200 EMS mutant lines (Rakszegi et al. 2010) from hexaploid wheat cv. Cadenza have been recently sequenced using exome-capture and displayed to allow the identification of millions of mutations in the sequenced genes using the www.wheat-tilling.com database (Krasileva et al. 2017). The mutations identified are accessible using the wheat survey sequence (Marcussen et al. 2014) via this database, which includes their location within the gene, and the predicted effect that each variant has on its protein. Simply searching the database (www.wheat-tilling.com) reveals those plants possessing mutations in the target genes, as well as a list of all mutations possessed by the plant (Krasileva et al. 2017). We selected seven of the 1200 EMS mutant lines, which possessed potentially interesting mutations within TaZIP4-B2 (Traes_5BL_9663AB85C.1) (Supplementary Table 2). Five of these mutant lines exhibited regular pairing at meiotic metaphase I, so were not taken further. These lines possessed a missense mutation, which indicates that amino acid changes within ZIP4 in these Tazip4-B2 lines did not affect its function. However, two of the mutant lines (Cadenza1691 and Cadenza0348) showed reduced number of COs in cytological analysis, suggesting that their Tazip4-B2 mutations did exhibit a phenotype. Both lines were selected for wide crossing studies with wild relatives to score the effect of their Tazip4-B2 mutations on homoeologous CO frequency in the wheat Tazip4-B2 mutant-wild relative hybrids, as compared to non-mutagenised wheat-wild relative hybrids. Null segregants were not available, so wild-type Cadenza lines were used as controls. Mutations within TaZIP4-B2 were verified by sequencing, and primers were designed to the mutated regions to follow mutated genes during crossing (Supplementary Fig. 3 and Materials and Methods). Tazip4-B2, within the Cadenza1691 mutant line, possessed a missense mutation within the Spo22 domain (C to T change leading to an A167V), shown to be important for ZIP4 function (Perry et al.2005). Tazip4-B2, within the Cadenza0348 mutant line, possessed a nonsense mutation (a premature stop codon: G to A change leading to W612*) (Supplementary Fig. 3). In addition to these Tazip4-B2 mutations, the two mutant lines also possessed mutations (mostly missense, but also splice or stop codons) within the coding sequences of 106 other shared genes. Sixteen of these genes, including TaZIP4-B2, were located on chromosome 5B. However, none of the genes apart from TaZIP4-B2 were located within the 2.5 MB Ph1 region defined in our previous study (Griffiths et al. 2006; Al-Kaff et al. 2008).
Compared to the chromosome 5B deletion mutant—wild relative hybrid—no other wheat chromosome deletion mutants have previously been reported as exhibiting a similar level of homoeologous CO formation at metaphase I (Riley and Chapman 1958; Sears 1977). For example, the 3D locus Ph2 exhibits a four-fold lower level of induction compared to Ph1 (Prieto et al. 2005). Equally, deletion of regions of chromosome 5B apart from the 2.5 MB Ph1 region does not result in homoeologous CO formation at metaphase I when the lines are crossed with wild relatives (Roberts et al. 1999; Griffiths et al. 2006; Al-Kaff et al. 2008). Sears (1977) used such crosses between hexaploid wheat cv. Chinese Spring, both in the presence and absence (ph1b deletion) of Ph1, and the wild relative tetraploid Aegilops kotschyi (also termed Ae. variabilis), to show that homoeologous COs are induced when the Ph1 locus is deleted. Interspecific hybrids of the ph1b mutant and wild relatives have been subsequently used in plant breeding programmes for introgression purposes (Sears 1977). Sears (1977) observed one rod bivalent at metaphase I in the presence of Ph1, and 6.35–7.28 rod bivalents in the Ph1 absent hybrids. Chiasma frequency in the hexaploid wheat-Ae. kotschyi or hexaploid wheat-Ae. variabilis hybrids was between 1 and 3 in the presence of Ph1 and 11–14 in the absence of Ph1 (Sears 1977; Farooq et al. 1990; Fernández-Calvín and Orellana 1991; Kousaka and Endo 2012).
In this study, both Tazip4-B2 mutant Cadenza lines, as well as a wild-type Cadenza (TaZIP4-B2), were crossed with Ae. variabilis. The frequency of univalents, bivalents, multivalents and total chiasma frequency was scored at meiotic metaphase I in the resulting F1 hybrid (Fig. 2). In these hybrids, there were similar numbers of rod bivalents to that reported by Sears (1977), with 6.75 (SE 0.17) (Cadenza1691) and 6.64 (SE 0.18) (Cadenza0348) rod bivalents at metaphase I in the Tazip4-B2 mutants, and 1.48 rod bivalents (SE 0.12) at metaphase I in the wild-type Cadenza. Moreover, the chiasma mean frequency was 1.48 (SE 0.12) in the presence of Ph1 (TaZIP4-B2) and 12.21 (SE 0.19) and 12.23 (SE 0.20) in the Cadenza1691-Ae. Variabilis and Cadenza0348-Ae. variabilis hybrids, respectively. The observed chiasma frequencies at metaphase I, in the two Tazip4-B2 mutant line-Ae. variabilis hybrids, are similar to those previously reported at metaphase I in the Ph1 deletion mutant (ph1b)-Ae. variabilis hybrids. Thus, the data indicate that TaZIP4-B2 within the Ph1 locus is likely to be involved in the suppression of homoeologous COs.
Tazip4-B2 mutant Cadenza lines show no multivalents
The frequencies of meiotic associations at metaphase I in hexaploid wheat and the Ph1 deletion mutant (ph1b) have been reported previously (Martín et al. 2014). Martín et al. (2014) observed 20 ring bivalents and one rod bivalent, with a chiasma frequency of 40.97 in the presence of the Ph1 locus. However, the number of ring bivalents decreased to 14.83, with a reduced chiasma frequency of 35.78, while the number of univalents, rod bivalents, trivalents and tetravalents increased to 0.80, 4.73, 0.20 and 0.37, respectively, when the Ph1 locus was absent. The number of univalents, bivalents, multivalents and chiasma frequency at meiotic metaphase I was also scored in both the Tazip4-B2 mutant lines and in the wild-type Cadenza (Fig. 3). The Tazip4-B2 mutant lines exhibited a reduction in the number of ring bivalents at metaphase I, and a slight increase in the number of rod bivalents, from a mean of 1.30 (SE 0.17) in the wild-type Cadenza, to 3.29 (SE 0.16) in Cadenza1691 and 3.63 (SE 0.18) in Cadenza0348. This indicates a slight reduction in homologous COs in these Tazip4-B2 mutant lines. CO frequency was a mean of 40.50 (SE 0.21) in the wild-type Cadenza, 38.13 (SE 0.20) in Cadenza1691 and 37.30 (SE 0.23) in Cadenza0348. These observed chiasma frequencies at metaphase I in the two mutant Cadenza lines are again similar to those previously reported at metaphase I in wheat in the absence of the Ph1 locus. However, no multivalents were observed, and there was no significant increase in the number of univalents at metaphase I in the Tazip4-B2 mutant lines, as is normally observed in Ph1 deletion mutants (Roberts et al. 1999).
If Tazip4-B2 mutants had enabled homoeologues to synapse while failing to CO, then a significant increase in univalents would be expected, but this was not observed. This suggests that homoeologous synapsis may not be significantly affected by TaZIP4-B2. On the other hand, the lack of multivalents at metaphase I suggests that both mutant lines will exhibit a reduced level of homoeologous exchange or chromosome translocation to that observed in the CS ph1b mutant. The ph1b mutant line has been reported to accumulate extensive background translocations over multiple generations due to homoeologous synapsis and COs (Sánchez-Morán et al. 2001). Thus, the apparent lack of multivalents in the Tazip4-B2 mutant lines could allow their exploitation for introgression purposes during plant breeding programmes, rather than the current ph1b line.
Thus, seven lines carrying mutations within the TaZIP4-B2 gene were screened for a phenotype with reduced homologous crossover at metaphase I. Of these, two lines were identified with this phenotype, one carrying a nonsense mutation within TaZip4-B2, and the other carrying a mutation in one of the key functional domains of TaZip4-B2. When crossed with Ae. variabilis, both of these lines also exhibited increased homoeologous crossover at metaphase I in the resulting hybrid, suggesting that the two phenotypes were linked. This is consistent with our previous study, which scored for the absence of the Ph1 locus by the occurrence of reduced homologous crossover in wheat and increased homoeologous crossover in wheat-rye hybrids (Roberts et al. 1999; Al-Kaff et al. 2008). Therefore, lines with increased homoeologous crossover were identified without an initial screen for the desired phenotype. Until now, the only way by which homoeologous crossover at metaphase I could be increased to this extent in wheat-wild relative hybrids was by deletion of the Ph1 locus, defined as a deletion effect phenotype specific to chromosome 5B. However, an alternative way of reproducing the Ph1 deletion effect would be to use EMS treatment to generate nonsense or truncation mutations in the homoeologous crossover-suppressing gene within the Ph1 locus. Analysis of the 1200-line TILLING population revealed that in any given mutant line, 1.5% of genes will have a truncation allele and 2% a missense allele (Krasileva et al. 2017). Thus, the probability of two mutant lines both sharing a truncation or missense mutation by chance in a second gene is P < 0.0005. The probability that two mutated genes will be located on the same chromosome is extremely low (2.4 × 10−5), and the probability that they will both be located within the Ph1 region is even lower (2.4 × 10−7). Thus, it is extremely unlikely that the increased homoeologous crossover phenotype found in both Tazip4-B2 mutant lines results from a nonsense mutation in a further gene independently linked with Tazip4-B2 within the Ph1 locus.
In terms of follow-on studies based around the observations reported here, we are currently backcrossing (BC) both Tazip4-B2 lines to clean up background mutations, as well as transferring Tazip4-B2 into the highly crossable hexaploid wheat cv. Chinese Spring. We are currently at BC3. BC lines will be made available once this exercise is complete. We have also complemented the approach of identifying chemically induced Tazip4-B2 mutants, by exploiting CRISPR to generate a large deletion within TaZIP4-B2. Initial analysis reveals that this Tazip4-B2 line has a similar phenotype to the two Tazip4-B2 Cadenza lines described above. We are currently segregating the transgenes away. TaZIP4-B2 is also being over-expressed to assess whether increased ZIP4 levels reduce homologous crossover. Our current hypothesis is that a higher ZIP4 level is optimal for homologous crossover, and a lower ZIP4 level is optimal for homoeologous crossover. Reducing or increasing ZIP4 around the optimal levels reduces the frequency of crossover. Interestingly, increasing the copy number of 5B chromosomes carrying TaZIP4-B2 reduces homologous crossover (Feldman 1966). In addition, a recent study identified another recombination pathway gene exhibiting dosage-dependent control on crossover (Ziolkowski et al. 2017). Thus, ZIP4 levels optimal for homologous crossover may be too high for homoeologous crossover, while the lower ZIP4 levels optimal for homoeologous crossover are too low for optimum homologous crossover. Finally, we are exploiting deletions of the less complex, homoeologous Cdk/methyl transferase locus on chromosome 5D, to shed further light on the role of these genes in the regulation of chromosome synapsis.
In summary, two Tazip4-B2 mutants were identified through a non-GM route, which can be exploited as an alternative to the CS ph1b mutant. Seeds for both mutants have been deposited with the Germplasm Resource Unit at the John Innes Centre (www.jic.ac.uk/research/germplasm-resources-unit). The accession number of seeds for Cad0348 is W10336 and for Cad1691 is W10337. The seeds for both lines are available on request, free of intellectual property restrictions.