Accurate staging of reproduction development in Cadenza wheat by non-destructive spike analysis

Abstract Wheat is one of the most important crops in the world; however, loss of genetic variability and abiotic stress caused by variable climatic conditions threaten future productivity. Reproduction is critical for wheat yield; however, pollen development is amongst the developmental stages most sensitive to stresses such as heat, cold, or drought. A better understanding of how anther and pollen development is regulated is needed to help produce more resilient crops and ensure future yield increases. However, in cereals such as wheat, barley, and rice, flowers form within the developing pseudostem and therefore accurate staging of floral materials is extremely challenging. This makes detailed phenotypic and molecular analysis of floral development very difficult, particularly when limited plant material is available, for example with mutant or transgenic lines. Here we present an accurate approach to overcome this problem, by non-destructive staging of reproduction development in Cadenza, the widely used spring wheat research variety. This uses a double-scale system whereby anther and pollen development can be predicted in relation to spike size and spike position within the pseudostem. This system provides an easy, reproducible method that facilitates accurate sampling and analysis of floral materials, to enable anther and pollen developmental research.


INTRODUCTION
Wheat is one of the main staple foods worldwide and is the most abundant crop cultivated in the developing world, providing 20% of total dietary calories and proteins worldwide (Shiferaw et al., 2013), however loss of genetic variability due to domestication and abiotic stress caused by variable climatic conditions threaten future productivity (Smýkal, 2018). Wheat yearly production reached 14.1 mTns, with a total market value of 2.63 billion dollars in 2016 (Rabobank, 2016).
However, increasing populations and a steady growth in wheat consumption alongside climatic variability, means that wheat yields need to increase by an estimated 2.6% per annum to meet ongoing demand (Shiferaw et al., 2013), which poses major challenges for breeders, researchers and governments. In addition, the exhaustion of the effects that increased wheat yield during the green revolution (Whitford et al., 2013) and new environmental policies, mean that novel varieties capable of contributing to more sustainable and resilient crops are needed (Shiferaw et al., 2013).
Multiple factors contribute to yield loss, from impoverishing of genetic variability due to intensive farming and domestication, and a/biotic stresses. To help mitigate these adverse effects, introgression programmes are ongoing to introduce beneficial traits that were lost due to intensive farming and domestication, from ancestors into modern elite lines (Grewal et al., 2018). In addition, new genomic tools have been developed to exploit the potential within the wheat genome. For example, genome sequencing, transformation, gene editing and genomic tools, are facilitating more ambitious research into wheat yield sensitivity to a/biotic stresses (Dolferus et al., 2011;Liu et al., 2015;Parish et al., 2012). Research carried out in Arabidopsis and rice has shown that stress can impair reproduction, particularly during meiosis, gametogenesis and fertilization (Barnabas et al., 2008); for instance, drought (Ding et al., 2018), heat or cold stress (Zampieri et al., 2017) causes abnormal pollen development reducing fertility (Shimono et al., 2016). The impact of stress damage is also particularly dependant on the stage when it occurs and the species. For instance, stress during pollen development in rice, when the tapetum activity is high during the transition between tetrads to early microspore release, produces major damage (Chaturvedi et al., 2017;Oliver et al., 2005).
However, research on flower development, fertility regulation and associated sensitivity to environmental stress in wheat, is significantly limited by the accurate staging of anther and pollen development. From early reproductive development until spike emergence floral development occurs within the pesudostem, and thus is not visible without destructive analysis. This means that it is extremely difficult to accurately stage and collect floral samples for phenotypic or molecular analysis without damaging the plant, this is particularly important when limited plant resources are available, for example for mutant or transgenic lines. It is therefore necessary to develop staging systems based on overall plant morphology and development that can be linked to anther and pollen development, instead of using samples collected based on plant age (Ninkovic and Åhman, 2009;Simmons et al., 2006). Non-destructive prediction methods are preferable to conventional anther or floret measurement that require plant dissection (Gómez and Wilson, 2012).
The prediction of anther and pollen development has been approached in different manners; generally anther and pollen stages appear to be tightly linked to spike and anther size in barley and wheat (Kirby, 1988;Waddington et al., 1983). Waddington et al. (1983) developed a scoring system for barley and wheat that correlates spike development with anther length. In addition, other studies correlate floret size and anther length with stages of anther development in rice (Raghavan, 1988). Recently a study of four Australian wheat varieties has shown that spikelet and anther size can be used to accurately predict anther stages (Browne et al., 2018). However, these systems require the dissection of plants to measure the parameters used to predict the anther and pollen development stages and therefore are not suitable for ongoing developmental analyses. Gomez and Wilson (2012) previously developed a non-destructive staging system in barley based upon the Zadok stages, that combined node number, Last Flag Elongation (LFE) and ear position within the plant to predict spike size in barley, which was then linked to anther and pollen development stages (Gómez and Wilson, 2012). In addition, a non-destructive method of measuring the auricula distance (distance between the auricles of the flag leaf and the penultimate leaf) has been used in wheat and rice to predict the young microspore stage (Ji et al., 2010;Morgan, 1980;Oliver et al., 2005), however this approach is varietal specific and has limitations in its ability to predict all anther stages.
Here we present a non-destructive anther and pollen development staging system in the wheat variety Cadenza. Cadenza is a widely used research variety that was the genotype selected to generate the wheat Tilling Exon Capture Population (King et al., 2015). This publicly available population contains thousands of mutant lines, which have annotated gene mutations that are available for public use (https://www.seedstor.ac.uk/shopping-cart-tilling.php). The development of this prediction system will enable accurate analysis of anther and pollen development which combined, with the availability of this mutant population, will allow ambitious research programmes on reproductive development.

Morphological Analysis
Cadenza morphological development was studied from the onset of the elongation stage to anther anthesis to gain better understanding of Cadenza development and to establish clearly recognisable features that could be related to anther development. Zadoks decimal code was initially used to identify key developmental points (Zadoks et al., 1974), however, new time points were used to establish a correlation between morphological and reproductive development.
Five plants, with four tillers per plant, were used. Material used for data collection was restricted to tillers that were already within the elongation stage when the main shoot was around booting stage. The florets analysed were collected from spikelet position 1 and 2, and always in the central positions within the spike (Figure 1c).
Data was collected every three days and consisted of: total height, number of nodes, internode elongation, flag leaf emergence (shoot axis elongation), spike position within the pseudostem, spike emergence, peduncule elongation after total spike emergence and spike size ( Figure 1).
In addition, a separate replicate set of plants growing under the same conditions were used for spike size analysis. All the data parameters were collected, however, these plants were dissected to measure the spike and to collect anther samples for sectioning and qRT-PCR. Further generations of material were also subsequently grown for analysis and confirmation of the staging system.

Histological analysis of anther development
Florets from the middle zone of the spike were collected from different spike developmental stages. Floral buds from different stages were fixed overnight in 4ºC in 4% (v/v) paraformaldehyde. Tissues were washed twice (30 min each) with 1xPBS. Fixed panicles were immediately dehydrated with ethanol (30-100% (v/v)) prior to embedding in Spurr resin (TAAB Laboratories Equipment, Ltd). After 100% ethanol, samples had a mixture of ethanol 100% and Ox.propilene (Sigma) 1:1 added, which was replaced by 100% Ox.propilene after 20 min. As a preinfiltration step, samples were mixed with Ox.propilene/Spurr resin (1:1) for 1.5 hours, the solution was then replaced with 100% Spurr resin and kept at 4°C overnight. After this, samples were embedded in molds using Spurr resin and incubated at 70-75°C for 10-12hr. Ultrathin sections (0.5µm) were produced using an ultramicrotome Leica EM UC6. Sections were stained with 0.25% (w/v in 1% Sodium Borate) toluidine blue prior to imaging.

Anther specific gene expression
Due to the high specificity of some of the genes expressed in the anther development network, expression analysis was used as an additional approach to confirm the correlation between the samples collected and the expected stages.
Samples were collected using spike size/position scale covering all spike stages.
RNA was purified using RNeasy spin columns (Qiagen). First-strand cDNAs were synthesized from1.5 µg total RNA using Superscript III reverse transcriptase (Invitrogen) and an oligo (dT) primer (Invitrogen) according to the manufacturer's instructions. Three anther specific genes were used for staging, TaDYT1, TaMS1 and TaMYB26. TaDYT1 and TaMS1 are essential tapetum specific transcription factors, whereas TaMYB26 plays a critical role in anther dehiscence.

X-ray µCT Scanning
X-ray µCT scanning of wheat plants were performed using the v|tome|x M 240 kV X-ray µCT scanner (GE Sensing & Inspection Technologies GmbH, Wunstorf, Germany). Scanning conditions were 80kV voltage, 250 current, 250 timing, 32 average, 0 skips, 1x1 bining and 4.0 sensitivity. A single radiograph consisting of 32 integrated images was taken to limit the X-ray dose to the plant to 10s per radiograph (total exposure 15 minutes over all scans). Spikes and nodes were easily identifiable by this method, and close up images of each spike being monitored were taken, as well as an overall picture. Three plants were taken from early flowering, and individual tillers were identified and monitored over 2 weeks (Day 0, 7 and 14) to monitor spike growth and position (Tracy et al., 2017).

Plant morphology development in wheat variety Cadenza
Plant development was studied to establish a correlation between external morphology and anther developmental stages for the wheat cultivar Cadenza  Table 1).

Internode and Last Flag Leaf (LFE) elongation development
Nodes and internode elongation, total tiller height and last flag elongation (LFE) data were collected every three days from the two detectable nodes until spikes were out and the peduncule extended ( Figure 2). Data collected permitted the creation of 5 internode stages based on nodes visability and internode elongation

Spike upward movement and development within spike
From the start of the reproductive stage (first node detectable separated >1cm from the ground) (Zadoks et al., 1974), the spike starts an upward movement driven by internode elongation that continues until the spike is completely emerged and the peduncle stops elongating (Figure 4a). This spike upward movement was divided into 16 identifiable points (spike stages) ( Figure 4a; Table 1 (Table 1).
Spike stage correlation with spike size was determined by measuring spikes collected using our spike staging system (Figure 4b; Table 1). Spikes showed a continuous growth from Zadok stage 30, spike around 0.5 cm (sample day 1, Figure   3b) and could be easily linked to spike position from spike stages 1 (1.5+/-0.3 cm) until stages 9-10 (9.5+/-0.4 cm) (Figure 4b; Table 1). After stage 9 the spike reached its maximum size (10-11 cm), remaining unchanged until the end of development at stage 16 ( Figure 4b).

The combination of spike size and its position within the pseudostem can be used to predict anther development
To establish a correlation between spike size/position and anther development, spike staging was used to collect different spike samples for ultra-thin microscopy.
Due to the spike ceasing to grow from stage 9 onward (Figure 4b), spike size prediction was used to collect samples from stage 1-9, whereas spike position was used between stages 10-16 (Figure 4 This combined staging approach allowed the identification of critical anther development stages using easily identifiable spike size and positioning within the pseudostem stages (Figure 4) ( Table 1). The system was divided into two parts due to the spike ceasing to grow after stage 9. From stages 1 to 9, spike size as determined using spike position, was used to predict anther development progression. This allowed the identification of critical stages for male fertility such as tapetum formation (stages 2-3), meiosis (stages 4-5) and tetrad formation (stages  Table 1).
MS1 in Arabidopsis, rice and barley is expressed in the tapetum just before microspore release as the callose surrounding the tetrad starts to breakdown until the free microspore stage (Fernandez Gomez and Wilson, 2014;Li et al., 2011;Wilson et al., 2001). MYB26, a transcription factor that regulates secondary thickening in the endothecium and is essential for anther dehiscence, is expressed just before Mitosis I until Bicellular pollen stage (Yang et al., 2007). Finally UDT1, the Arabidopsis DYT1 orthologue in rice, shows a bimodal expression pattern with two expression peaks, at meiosis and at early tetrad stage (Jung et al., 2005). The expression patterns of the putative wheat orthologues showed similar profiles to those observed in Arabidopsis, rice and barley. TaMS1 expression was observed in anthers between early tetrad to vacuolated pollen stage, reaching maximum expression at early microspore release (Figure 6a). Furthermore, TaMYB26 showed maximum expression between vacuolated pollen and Mitosis I stage (Figure 6b). TaDYT1 showed low expression from sporogenous cells (SC) to sporogenous cells-tapetum generation (SCt) stages, increasing at the free microspore stage (Ms), then decreasing again at the microspore-tapetum transition stage (MC-Tt). From MC-Tt, expression reaches its peak at early tetrad stages (ETd), decreasing again around young microspores (Ym) increasing slightly at Mitosis I (Figure 6c). These results confirm the accuracy of the non-destructive staging system for collection of stage specific samples, enabling higher precision in sample collection which is needed for detailed molecular and physiological analysis of pollen development.

Cadenza development
Predicting anther and pollen development stages is key to investigate plant fertility, the sensitivity of this critical trait to environmental stress and the potential generation of hybrid seeds. Facilitating effective sample collection can also contribute to more homogeneous sampling sets in experiments such as RNAseq, or expression analysis, that will facilitate the identification and characterisation of genes of interest and their manipulation.
Wheat cultivar Cadenza has been extensively used as a research tool, and recently, a multinational effort has generated an Exon Capture Population that is contributing to accelerate wheat research (King et al., 2015). Therefore, this variety is a good model for developmental studies and the generation of a system for accurate anther staging will capitalise on the existent mutant population and contribute to male fertility gene characterisation. Physiological measurements, destructive staging using sectioning, and non-destructive staging were all used to generate combined data on the progression of Cadenza flower development. X-ray µCT contributed to the analysis of spikes at a very early stage in development (spike < 1 cm; undifferentiated tissue) to assist in the developmental assessment and in additional validation of our final staging system (Figure 3).
Wheat anther staging has been studied in detail by Mizelle et al (1989), however, no relationship between external development and anther development was established (Mizelle et al., 1989). Nonetheless, attempts to establish relationships between spikelet and anther sizes have proven very accurate in predicting these stages (Browne et al., 2018;Kirby, 1988;Waddington et al., 1983), however, these systems require plant dissection and therefore destruction of the plant materials.
Non-destructive methods based on auricle distance (AD) have been used in rice (Oliver et al., 2005) and wheat (Ji et al., 2010) to predict young microspore stage in cold and drought stresses treatments. However, the relation between auricle distance (AD) and anther stages in wheat proved highly varietal specific (Browne et al., 2018). This was confirmed in Cadenza where LFE (denomination used instead of AD in this paper) proved to be inadequate to predict all anther stages due to the early cessation of elongation observed in the last flag around spike stage 9-10 ( Figure 2).
Parallelism between pseudostem and spike development has been described previously during wheat elongation stages (Kirby, 1988;Reynolds et al., 2009) and a correlation was made between increase in spike size and floret development scores (Waddington et al., 1983). However, establishing a relationship between external morphological changes, spike size and anther stages is difficult due to varieties differences and by the ongoing spike development within the pseudostem.  Table 1).

Anther stages prediction system
Internode elongation stages were initially targetted as the principal morphological parameter to link with anther development (Figure 2a), however this meant that predictions of internode maximum elongation were needed to establish the internode stages, which proved to be complex. Nonetheless, internode stages contributed to increasing the knowledge regarding Cadenza development and morphology. This knowledge facilitated the generation of a spike position staging which proved highly consistent and easy to follow as stated here and in previous investigations (Gómez and Wilson, 2012;Kirby, 1988;Reynolds et al., 2009) ( Figure 4; Table 1).
Spike size and its correlation with spike position resulted in a very accurate prediction of anther development stages (Table 1) Table 1). Therefore, this double scale facilitated the identification of the stages from the onset of plant reproduction until anthesis (Figures 4-5); this provides a tool for accurate sample collection aimed at analysing gene expression and developmental characterisation ( Figure 6; Table 1). This staging system was confirmed using µCT and qRT-PCR expression analysis of key anther-specific genes; this validated the staging system, but is not required for subsequent accurate analysis staging of materials.
Similar anther and pollen development progress was observed in wheat and barley (Figure 5b), however differences were observed in their timing and duration. For instance, in barley, stages such as tetrads and young microspore release occurred earlier in plant development, before all nodes were detected and extended (barley stage 33-34), and spikes were between 2.5-4.5 cm (Gómez and Wilson, 2012). In addition, these important stages all occurred before the last flag had started elongating (Gómez and Wilson, 2012) (Figure 5b). On the other hand, in wheat, early tetrad stage occurred at spike stage 5-6 when spikes were over 6 cm long and the spike was about to enter the last flag sheath, which is already elongated to over 8 cm (Figure 4) (Table 1). Moreover, it was observed that wheat tetrad stage lasted longer than in barley (Figure 5b). In wheat, meiosis was observed at spike stage 4-5 (spike around 4.5cm), with tetrads evident soon after (stage 5-6, spike around 6.2cm) (Table 1), and lasting until stage 8 (spike 7-8 cm). These stages corresponded to spike stages just before entering the last sheath (stage 6) until the spike was half within the spike (stage 8) (Figure 4). Whereas, the tetrad stage in barley was much quicker, lasting from spike 3-4 cm (Gómez and Wilson, 2012), or Zadoks stages 33-33.5 (Zadoks et al., 1974) (Figure 5b). Environmental conditions and genotype is likely to impact upon developmental progression and slight variation has been seen between different wheat genotypes (data not shown). Our floral staging system was developed for Cadenza grown under controlled environment condition since it is extensively used as a model for wheat molecular genetic analysis, nevertheless this staging system provides a basis for direct application to other wheat genotypes and growth conditions, and can be readily correlated to other wheat genotypes/growth conditions. This system of using spike size and position to determine anther development stage, was further validated by expression analysis of three anther/stage specific transcription factors, TaMS1, TaMYB26 and TaDYT1 (Jung et al., 2005;Wilson et al., 2001;Yang et al., 2007). Samples collected using the spike size/position staging system were analysed by RT-qPCR ( Figure 6) to confirm the stages.
Some differences were observed in the MYB26 expression pattern that is restricted to early Mitosis I until Bicellular pollen in Arabidopsis (Yang et al., 2007), whereas in wheat, expression seems to be more widespread through the development ( Figure 6b). This confirms the accuracy of this staging system to predict anther development by a non-destructive method in Cadenza. This method, although focussed upon Cadenza growing under control conditions, provides a valuable tool for anther and pollen stage prediction that will contribute to increased understanding of male fertility by enabling correct sampling of materials to facilitate gene expression analysis, RNAseq sampling and gene characterisation.          Spike measurement started on day 48 after sowing (Zadock stage 30; spikes were around 0.5 cm long). a, c-e) µCT analysis of Cadenza spikelets. Spike size increase was observed over the following 14 days, spikes were up to 1.2 cm and >3 cm after 7 and 14 days respectively. c-e) Spike observed after c) day 1, d) day 7, and e) day 14. a) Spike position was divided into 16 stages, from early development, just before spike enters the "prior to previous last sheath" (PLS1) spike stage 1, until the spike is completely emerged, peduncule is elongated and anthers enter anthesis (spike stage 16). These 16 stages were linked to spike position within the pseudostem. b) Spike stage/spike size correlation. Spike position and spike size showed a close correlation from stage 1 until stages 9-10. From stage 9-10 the spike ceased growing and remained unchanged.