Investigating flowering time in wheat under controlled environment conditions

Tina Rathjen

Agriculture and Food

Introduction

I have been working as a molecular biologist for over 25 years on both animal and plant systems. The last 10 years I have worked at CSIRO and have been involved with several projects investigating different aspects of wheat physiology, growth and development. This has included carrying out laboratory, glasshouse and field experiments and previously data generated has been entered and analysed using Excel. Prior to Data School I did not know how to code or use R. Data School FOCUS has opened my eyes to a whole new world of data analysis and I hope to use this as a starting point to learn and apply new and exciting methodologies.

My Project

I work on a project within the GRDC’s National Phenology Initiative which aims to predict the flowering time of wheat cultivars in many different growing regions within Australia.

Wheat cultivars have an optimal flowering window. Crops that flower too early can have lower yield due to insufficient biomass accummulation or exposure to cold or frost events. Conversely, crops that flower too late risk being exposed to water stress or heat events which can negatively impact yield. Growers require accurate information to select the correct cultivar and sowing date for their farming conditions. For new cultivars it takes several years conducting field trials to accumulate data to predict flowering time. With new cultivars being released it is essential that growers have accurate information by the time of release.

The major environmental factors influencing flowering time are thermal time, photoperiod and vernalisation. APSIM models have been developed that use parameters based on these factors to model cultivar flowering times across the many cereal growing regions within Australia. This project aims to improve and modify the existing APSIM models of wheat through derivation of parameters in controlled conditions and association with genomic data to more accurately predict flowering time in the field. This study involved four controlled temperature experiments being carried out on 54 Australian wheat cultivars and 15 Wheat NILs (Near Isogenic Lines).

Preliminary results

The controlled experiments were carried out under four environmental conditions, SN (short days, no vernalisation), LN (long days, no vernalisation), SV (short days plus vernalisation) and LV (long days plus vernalisation) to determine the influence of photoperiod and vernalisation on flowering times. Vernalisation was carried out by imbiding seeds at 4oC for 8 weeks prior to planting. The temperature for all experiments was set at 22oC and measured every 30 minutes using a TinyTag data logger. Traits measured include emergence date, flowering date, heading date, final leaf number and spikelet number. In addition Haun stage, a measure of developmental growth stage based on leaf emergence, was recorded every third day from emergence to flowering. The data was initially entered into several sheets within one Excel file. I have used the tidyverse package to clean and arrange the data. I have separated the data into two files, Haun_temp, containing the haun stage scores and a second file all_data_wide, containing all other traits. I used the temperature data to convert dates to accumulated thermal time (degree-days). Tables 1 and 2 show a representation of the two tidied data files.

Tables

Table 1: Controlled Environment - Haun Stage Data
Genotype Type Maturing Environment Rep Haun stage degree-days
EMU_ROCK Spring Fast SN 1 1.5 64.1
EMU_ROCK Spring Fast SN 1 1.9 149.3
EMU_ROCK Spring Fast SN 1 2.1 213.1
EMU_ROCK Spring Fast SN 1 2.7 298.2
EMU_ROCK Spring Fast SN 1 3.1 362.6
Table 2: Controlled Environment - Flowering Time Data
Genotype Type Maturing Environment Rep fwr_MS fwr_t1 half_fwr half_hd hd_MS hd_t1 final_leaf tt_final_leaf spikelet_no
ADV08.0008 Winter Mid LN 1 2163 NA 2420 2377 2124 NA 13 1925 26
ADV08.0008 Winter Mid LN 2 NA 2352 2569 2548 NA 2307 NA NA 26
ADV08.0008 Winter Mid LN 3 2143 NA 2352 2329 2104 NA 13 1883 29
ADV11.9419 Winter Mid LN 1 2396 NA 2700 2700 2352 NA 15 2163 31
ADV11.9419 Winter Mid LN 2 2307 NA 2764 2764 2307 NA 14 1984 31

Plots

Figure 1 : Flowering time for plants grown in four different environments

Figure 1. Flowering time results for all four environments.

Plants usually flower quickest in the LV environment, where all vernalisation requirements are met and the day length is long. Shortening the day length or removing vernalisation can delay flowering time, but there is a genotypic effect. The flowering times of spring lines are least affected by vernalisation, whilst slow winter lines are most affected. The flowering times of fast lines are least affected by photoperiod, whilst slow lines are most affected.

Figure 2 : Vernalisation effect on flowering time

Figure 2. The effect of vernalisation on flowering time of cultivars.

Spring wheats, Axe and Forrest, have no vernalisation requirement and therefore flower at approximately the same time in both environments. Winter wheats, Longsword and Kittyhawk, require vernalisation, and without vernalisation flowering is delayed.

Figure 3 : Photoperiod effect on flowering time of cultivars

Figure 3. The effect of photoperiod on cultivars.

There is a genotyptic response in daylength. Fast maturing cultivars, Axe and Longsword, are photoperiod insensitive, the effect of changing day length on flowering time is small. Conversely, slow cultivars, Forrest and Kittyhawk, are photoperiod sensitive and shorter day length delays flowering. Photoperiod sensitivity and vernalisation requirements have separate genetic controls. The Australian cultivars used in this experiment display a range of vernalisation and photoperiod sensitivities.

Figure 4 : Vernalisation requirement and photoperiod sensitivity of wheat cultivars

Figure 4. The vernalisation requirement and photoperiod sensitivity of the cultivars studied.

Vernalisation effect is calculated as the difference between mean flowering time in LV and LN. Photoperiod sensitivity is calculated as the difference between mean flowering time in LV and SV.

Figure 5: Correlation of flowering time with the traits measured

Figure 5. Correlation of flowering time with other traits measured

Calculating APSIM Parameters

Seven parameters were calculated from the original data. Descriptions of the seven parameters are shown in Table 3. The parameters were calculated using R scripts and a representation of the output is in Table 4.

Table 3: APSIM Parameters description
Parameter Name Calculation
Parameter_1 Minimal_Leaf_Number LV mean final leaf number
Parameter_2 Pp_Sensitivity SV mean final leaf number minus LV mean final leaf number
Parameter_3 Vrn_Sensitivity LN mean final leaf number minus LV mean final leaf number
Parameter_4 Base_Phyllochron Slope of accummulated thermal time against mean leaf number between 3 and 7 in LV
Parameter_5 Phyllochron_Photoperiod_effect Ratio of Phyllochron SV to Phyllochron LV
Parameter_6 Early_Reproductive_Long_Day_Base Accumulated thermal time at flowering divided by BasePhyllochron
Parameter_7 Early_Reproductive_Pp_Sensitivity Difference of accumulated time between LV and SV treatment divided by BasePhyllochron
Table 4: APSIM Parameter results
genotype Parameter 1 Parmeter 2 Parameter 3 Parameter 4 Parameter 5 Parameter 6 Parameter 7
ADV08.0008 8.0 1.7 5.0 98.5 1.4 3.9 2.1
ADV11.9419 8.3 0.7 5.7 89.9 1.5 4.0 4.5
AGT_SCYTHE 6.7 2.0 5.7 81.0 1.7 4.8 7.2
AXE 6.0 1.0 0.0 87.8 1.3 4.4 1.3
BEAUFORT 9.0 1.0 6.3 98.9 1.2 3.8 2.6

My Digital Toolbox

Favourite tool

Tidyverse is my favorite tool as it is so incredibly useful! I also like ggplot but find it frustrating because there are too many options.

My time went …

My time went on understanding and tidying the data so I could attempt to work out what it meant. I also spent a large amount of time working out scripts to calculate the APSIM parameters, especially Parameter 4, BasePhyllochron, which required use of a lm function (Thanks Aswin). I also spent quite a lot of time trying to make the ggplots and improve their appearance.

Next steps

It would be interesting to examine the data further. One possibility is to see how the data looks when considering photo-thermal time, which takes into account daylength in addition to temperature. I am planning to learn how to use Tassel to carry out GWAS (Genome-wide Association Studies) to identify the molecular markers linked to the parameters I have calculated. Recently a version of Tassel based on R, aptly named RTassel, has been released and it would be cool to explore the possibilities of using this for GWAS analysis. I would love to learn more and find more uses for R. Hopefully this is only the start of my data journey.

My Data School Experience

I have learned so much from Data School Focus. I think one of the most important things I have learnt is to keep rawdata and manipulated data separate and to use R and Git to track any changes I make. The other thing I have learnt is to be consistent as to how data is entered, using consistent cultivar names, only having one type of data in each column and having notes and comments in separate columns or files. I have become a convert to using R to manage my data.