Bread wheat (Triticum aestivum L.) is an allohexaploid species grown worldwide and adapted to different latitudes and climatic conditions. This ability is related to a variation in the two main genes: PPD1 (PHOTOPERIOD1, photoperiod response) and VRN1 (VERNALIZATION1, vernalization requirement) (Trevaskis et al., 2003; Yan et al., 2003; Beales et al., 2007). VRN1 belongs to MADS-box (MCM1, AGAMOUS, DEFICIENS, SRF) transcription factors (Yan et al., 2003) and plays a crucial role as an integrator of vernalization-accelerated flowering. Due to the hexaploid nature of bread wheat genome, VRN1 is present as homoeologs (VRN-A1, VRN-B1 and VRN-D1) on chromosomes 5A, 5B and 5D (Snape et al., 2001). Its natural allelic variation is associated with the growth habit - spring or winter (Yan et al., 2004a; Fu et al., 2005). Cold period (=vernalization) accelerates the flowering of winter (autumn-sown) varieties (Chouard, 1960). The length of effective vernalization can range from three to eight weeks (Košner, Pánková 2002; Li et al., 2013). Winter varieties carry recessive vrn1 alleles. Dominant alleles in the spring varieties are expressed without vernalization and carry mutations in the promoter or the first intron of VRN1 (Yan et al., 2004a; Fu et al., 2005). The mutations result in the partial or complete inhibition of vernalization requirement. At least one dominant VRN1 allele confers the spring growth habit (Stelmakh, 1987). In winter wheats, the VRN1 chromatin undergoes histone methylation changes (H3K4me3 and H3K27me3) during vernalization, possibly affecting the VRN1 expression (Xiao et al., 2014). Several models of vernalization mechanism have been proposed so far (Amasino, 2004; Yan et al., 2006; Chen and Dubcovsky, 2012; Xiao et al., 2014; Xu et al., 2021; Debernardi et al., 2022). Nevertheless, we still lack a detailed understanding of vernalization molecular mechanism. This review recapitulates current knowledge of the VRN alleles and reflects on vernalization models. We also discuss VRN1 multiple copies and touch on wheat devernalization.

Vernalization has been known and studied for a long time, but the precise molecular mechanism still waits to be revealed. A growing number of studies kept bringing new findings that were used in developing several models of how vernalization may work (Amasino, 2004; Yan et al., 2006; Chen and Dubcovsky, 2012; Xiao et al., 2014; Debernardi et al., 2022). The models are depicted in Figure 1 . The first simplified model (Amasino, 2004) shows the vernalization pathway with VRN2 repressing VRN1 and several other unknown genes ( Figure 1Ai ). The vernalization was suggested to repress VRN2 expression. The VRN2 expression was also downregulated by a short day (during autumn), while VRN2 repressed both VRN1 and FT1 ( Figure 1Aii ). Shimada et al. (2009) developed transgenic plants overexpressing FT1, causing VRN1 upregulation and VRN2 downregulation. FT1 is highly transcribed even in both Δvrn1-null and Δvrn1-vrn2-null mutants lacking functional VRN1 and VRN2 genes (Chen and Dubcovsky, 2012). They proposed the model where VRN1 represses VRN2, which downregulates FT1, and VRN1 and FT1 mutually upregulate one another, creating a positive feedback loop. The reworked vernalization model was now referred to as “original”, and Shimada’s model as “reverse” (Chen and Dubcovsky, 2012) ( Figure 1Aiii ). The most recent working model presented by Debernardi et al. (2022) supports findings reported by (Chen and Dubcovsky, 2012), showing VRN1 downregulation of VRN2. During long days in the autumn, PPD1 upregulates VRN2, which downregulates FT1, preventing wheat from flowering ( Figure 1Aiv ). Debernardi et al. (2022) also identified a conserved pathway integrating plant age into flowering regulation. This pathway involves the expression of miR172 acting like a flowering promotor but its targets APETALA2-like1 and 5 (AP2L1, AP2L5) function as flowering repressors.

These models describe the interaction of VRN1, VRN2 and FT1. However, they do not explain the molecular mechanism at the DNA level. Dominant Vrn1 alleles have higher basal transcription levels, minimising the vernalization requirement. Indels present in the dominant alleles may have removed or disrupted a putative binding site for an unknown VRN1 repressor. Xiao et al. (2014) described a mechanism of VRN1 induction during vernalization. The proposed model suggests glycine-rich RNA-binding protein (GRP2) as a repressor preventing VRN1 transcript accumulation. Before vernalization, GRP2 directly binds to the VRN1 pre-mRNA in the region characterised as critical. It comprises the VRN1 intron approximately from 1.3 kb to 4.2 kb (from the start codon) (Fu et al., 2005). The cold treatment induces the expression of VER2 and increases the GRP2 O-GlcNAcylation level. VER2 creates a complex with GRP2 that releases VRN1 transcript accumulation and induces flowering promotion ( Figure 1B ). Xiao et al. (2014) also reported histone methylations might participate in the vernalization response. During the cold period, the H3K27me3 levels at VRN1 chromatin decrease while levels of H3K4me3 increase. The high levels of H3K4me3 are associated with active gene transcription, while increased H3K27me3 levels are linked with gene repression (Roh et al., 2006; Wysocka et al., 2006; Zhang et al., 2007). These changes are targeted to the first half of the VRN1 first intron, which is in concordance with the previous model (Alonso-Peral et al., 2011) and findings reported in barley (Oliver et al., 2009). It also supports the significance of the VRN1 critical region in vernalization response. Xiao et al. (2014) developed wheat transgenic lines with GRP2 overexpression (GRP-OE) or GRP2 silencing by RNA interference (RNAi). The results showed that the mean heading time of GRP-OE lines did not significantly differ from the wild type (winter variety JH9). The GRP2-RNAi lines had reduced mean heading time (155 days) compared to the wild type (165 days). These lines were always earlier than wild type, irrespective of the length of the vernalization treatment. Although the difference was statistically significant, it did not approximate the spring varieties’ mean heading time. Depending on the growth conditions, the spring wheat heading time can range from 25 to 90 days (Li et al., 2017; Huang et al., 2018). Thus, we can hypothesise about the presence of another putative, more powerful VRN1 repressor. One could inspire from the study in Brachypodium distachyon where the REPRESSOR OF VERNALIZATION (RVR1) was described (Woods et al., 2017). They showed that mutation in RVR1 bromo-adjacent homology and transcriptional elongation factor S-II domains leads to reduced H3K27me3 levels of VRN1 chromatin and results in accelerated flowering without vernalization.

The histone methylation of VRN1 chromatin observed during vernalization (Xiao et al., 2014) might result from Polycomb repressive complex 2 (PRC2) activity. This complex is a histone methyltransferase consisting of four subunits (Bantignies and Cavalli, 2011). The SET domain in the catalytic subunit Enhancer of zeste [E(z)] is associated with the H3K27 trimethylation. The vernalization requirement is reset in the next sexual generation, probably during seed development (reviewed in (Trevaskis, 2010) ( Figure 1E ). The genes coding for individual PRC2 subunits in bread wheat have been recently reported (Strejčková et al., 2020), but the role of PRC2 in wheat vernalization still has to be unravelled. Lomax et al. (2018) characterised Brachypodium mutant flowering rapidly under non-vernalizing conditions. A single nucleotide polymorphism (SNP) in the ENHANCER OF ZESTE-LIKE 1 (EZL1) was associated with global reduction of H3K27me3, which corresponds with EZL1 function in the PRC2 activity.

Flowering can be also regulated by long non-coding RNAs (lnc RNAs). In Arabidopsis, lnc RNAs derived from both strands of FLOWERING LOCUS C (FLC) affect FLC transcription (Helliwell et al., 2011; Heo and Sung, 2011; Kim and Sung, 2017; Kim et al., 2017). The more recent model in bread wheat describes VRN1 regulation by non-coding RNA transcribed from the VRN1 sense strand (Xu et al., 2021). This alternative transcript (TaVRN1 alternative splicing, VAS) is induced during the first weeks of vernalization and includes the first exon and the first intron. In non-vernalised winter wheat, VRN1 forms a loop due to the activity of unknown proteins. VAS stimulates the production of VRN1 transcripts by engaging other proteins, such as TaRF2a and TaRF2b. The loop is released during vernalization, which leads to the complete transcription of VRN1 ( Figure 1C ). VAS includes the short alternative transcript reported previously (Xiao et al., 2014).

Chromosomal segments are subject to deletions or duplications. Such rearrangements larger than 1 kb are called copy number variation (CNV) (Żmieńko et al., 2014). CNVs played a significant role in human evolution but are also an important factor causing diseases, including cancer (reviewed in (Hastings et al., 2009)). In polyploids like bread wheat, CNV refers to the number of gene copies per haploid genome (Würschum et al., 2015). VRN-A1, VRN-B1 and VRN-D1 homoeologs are located on different chromosomes (Snape et al., 2001), but individual VRN1 genes can also be present in multiple copies on the same chromosome. CNV has been reported mainly for the VRN-A1 gene, which can be present from one to two copies (dominant Vrn-A1a) or up to four copies (recessive vrn-A1) (Díaz et al., 2012; Würschum et al., 2015; Strejčková et al., 2021). Two copies of VRN-B1 were observed in the hexaploid species T. compactum and T. spelta (Muterko and Salina, 2019). No CNV for VRN-D1 have been described so far (Strejčková et al., 2021), but screening more varieties may reveal multiple VRN-D1 copies. Although the word “copy” implies identical sequences of repeated sections, several types of VRN1 copy number variation exist. Actually, this fact helped to identify individual copies. Using current sequencing and assembling techniques, it would be extremely hard to distinguish one copy from another if they were 100% identical. The first type of CNV is the SNP in VRN-A1 exons 4 and 7, reported in wheat accessions carrying two or more copies (Díaz et al., 2012; Muterko and Salina, 2018). An advanced case of CNV displays the Vrn-A1c allele present in the spring Afghanistan land race IL369. The vrn-A1c allele has two copies: one recessive (intact) copy and one dominant copy with the deletion in the first intron (Fu et al., 2005; Díaz et al., 2012). The VRN-D4 gene originated by translocation of ∼ the 290-kb region from the distal part of the long arm of the 5A chromosome to the proximal region of the short arm of chromosome 5D (Kippes et al., 2015). This region included the vrn-A1 gene; therefore, VRN-D4 might be considered an unusual case of CNV as it involves translocation between two haploid subgenomes. The CNV can have a diverse effect on flowering: an extra vrn-A1 copy translocated from 5A to 5D chromosome (= VRN-D4 gene), also carrying SNP (A367C), resulted in the spring growth habit with no vernalization requirement (Kippes et al., 2015). The Vrn-A1c allele confers spring growth habit due to the mutated copy carrying the large deletion within the first VRN1 intron. The higher number of vrn-A1 copies within the same 5A chromosome, associated with SNPs in VRN1 exons 4 and 7, led to an increased vernalization requirement; plants with more than one copy needed a prolonged cold period to start VRN-A1 transcription (Díaz et al., 2012; Li et al., 2013). Recently, a speed vernalization (SV) method was reported (Cha et al., 2022). They showed that the SV effectivity varied among varieties with the different numbers of vrn-A1 copies. Wheat variety Charger (three vrn-A1 copies) flowered quicker when speed-vernalized for four weeks, while variety Claire (one vrn-A1 copy) had a shorter flowering time under two or six weeks of speed vernalization.

Regarding VRN2, Tan and Yan (2016) reported duplication of the VRN-B2 gene in hexaploid wheat, but no significant effect on flowering time was observed.

No increased number of FT1 was described in hexaploid wheat so far. In barley (Hordeum vulgare L.), four HvFT1 copies significantly accelerated flowering time (Nitcher et al., 2013). Wheat and barley are evolutionary close to each other, suggesting a possible, unrevealed existence of multiple FT1 copies in wheat.

Vernalization results in the change from the vegetative to the reproductive growth. The initial metabolic changes lead to morphological changes when the shoot apical meristem begins to produce floral primordia instead of leaf primordia (Yong et al., 2003). The effect of vernalization treatment can be partially or completely removed (in some species) by several days of heat treatment called devernalization. The most effective temperature was considered 30 – 40°C (Bernier, 1981). The heat treatment (around five days) needed to be applied directly after the end of vernalization; it becomes ineffective after a few days of plant growth at ordinary temperatures (Michaels, Amasino 2000). In bread wheat, several studies were performed to determine if vernalization-induced developmental changes could be also reversed (Gregory and Purvis, 1948; Chujo, 1970; Xiu-zhen et al., 1987; Yong et al., 2003; Xu et al., 2019). During devernalization experiments, vernalised wheat plants were exposed to higher temperatures, ranging from 18°C to 35°C. The treatment was associated with delayed flowering, changes gene expression patterns, or protein content changes. The length of the cold treatment reported in the studies varied from 21 to 40 days, which might not be enough to complete vernalization. The vernalization requirement duration can be genotype-dependent (Li et al., 2013). Its genetic nature is not fully understood, but might be linked with VRN-A1 locus or mutations in acetylglucosamine transferase TaOGT1 (Díaz et al., 2012; Li et al., 2013; Kippes et al., 2015; Fan et al., 2021). Study in Arabidopsis demonstrated that the effect of six-week-vernalization might be erased by heat treatment (30°C). It was associated with the elimination of epigenetic mark H3K27me3 from FLC (Bouché et al., 2015). This H3K27me3 removal caused FLC reactivation, which led to delayed flowering. Thus, the devernalization phenomenon in cereals might be putative only and needs further research.

The knowledge of wheat vernalization mechanism has expanded enormously in the last 20 years once VRN genes have been cloned and characterised. The advances in molecular methods enabled us to identify sizeable allelic variability in the VRN1 gene and to update vernalization models. Despite all these achievements, the main question remains: what is the actual molecular mechanism of wheat vernalization? In bread wheat, the VRN1 allelic variation data suggests two different evolutionary events resulting in the spring growth habit – the deletion within the VRN1 first intron and the insertion of the mutator DNA transposon in the VRN1 promoter region. Both deletion and insertion may disturb the binding site for the putative VRN1 repressor(s). The Vrn-A1a allele has the first intron intact (same with recessive allele) but carries the insertion in the promoter region. This allele has the highest basal VRN1 expression level associated with shorter heading time than the dominant Vrn-B1 and Vrn-D1 alleles. However, vernalization increases Vrn-A1a transcription, supporting the previously described role of the first intron. As mentioned previously, the Vrn-A1a allele has duplicated promoter and exon 1. This might be the reason for such high transcript levels, but the function of this duplication remains undiscovered. The current knowledge about the wheat vernalization mechanism supports the hypothesis that there might be two independent putative VRN1 repressors: one targeting the VRN1 promoter region and the other interacting with the first intron (Alonso-Peral et al., 2011) ( Figure 1D ). Besides, the role of multiple vrn-A1 copies in the vernalization response should be studied in more detail. The position of individual copies within the genome is not known and there is no information whether they have the same expression pattern. Finally, our understanding of how vernalization requirement is reset in the next generation is limited. It possibly occurs during the seed development and may involve PRC2-related histone modification of VRN1 chromatin.

We should not forget to mention an integral part of vernalization: how plants sense the duration of cold period. Longer cold treatment increases VRN1 transcript levels suggesting a quantitative character of vernalization (reviewed in (Trevaskis, 2010)). The length of sufficient vernalization varies among winter wheats. Several hypotheses regarding VRN1 were suggested but they were not in concordance: copy number variation, amino acid change or SNP in the putative repressor binding site (Díaz et al., 2012; Li et al., 2013; Xiao et al., 2014). Another explanation was proposed in Arabidopsis (Kyung et al., 2022). The long-term cold-mediated response might employ circadian clock regulators CIRCADIAN CLOCK ASSOCIATED 1 and LATEELONGATED HYPOCOTYL.

Recent technical and methodological advances will further help to untangle vernalization. Increased availability of the genome and transcriptome sequencing, along with the improvements in computational biology, may reveal new molecular mechanisms involved in the vernalization pathway. VanGessel et al. (2022) recently demonstrated transcriptional signatures of inflorescent development in the tetraploid wheat variety Kronos. The gene expression atlas of the floral meristem based on single nucleus RNA-seq data was developed latterly in barley (Neumann et al., 2022). Adapting the new techniques of targeted mutagenesis could help to develop the new alleles for functional studies.

There might be more unknown genes and their interactions participating in wheat vernalization. We could compare this phenomenon to mosaic assembling; once all fragments are in their place, we will see the complete picture.

ZM proposed the manuscript. ZM, BS and JS participated in the research and provided original results. All authors contributed to the article and approved the submitted version.