Participatory breeding of high-value wheat for the Northeast

1. Conduct participatory breeding of wheat, spelt, and einkorn on farms and research stations according to traits that are most important to farmers;

During the summers of 2015 and 2016, the graduate researcher facilitated farmer selection of superior winter wheat varieties. Five participating farmers from Oechsner Farms, Lakeview Organic Grain, Threshold Farms, Gleason Grains, and White Frost Farms planted bulked F3 biparental families from winter wheat and einkorn crosses in the fall of 2014. Each farm established two replicates of five biparental family plots and one check variety commonly grown by organic farmers in the region: ‘Warthog’. Plot sizes varied from 4.65 to 9m², depending on the size of regional planting equipment. Farm-specific selection protocols facilitated the selection of phenotypes best suited to the priority traits of each farmer. With the help of the graduate investigator, the farmer visually separated each biparental family plot into four quadrants and selected the 10% of plants or spikes that best met priority traits in each quadrant. The graduate investigator also randomly collected the same number of spikes from each biparental family plot to form F3:F4 winter wheat and F4:F5 spring wheat baseline populations to track gains in selection. In addition to farmers selecting based on their priority traits in the field, we were able to select for protein content, which was the second highest rated trait by participating winter wheat farmer breeders. For farmers who selected protein as a priority trait, a single seed non-destructive Near-infrared spectroscopy (NIR) machine allowed the selection of 50% of spikes with highest protein. The graduate researcher calibrated the NIR instrument using a partial least squares model and a destructive nitrogen test (LECO TrueMac N). One hundred and ninety-two seeds each from six wheat varieties from a range of protein contents and color classes comprised the calibration set. The r² of the calibration was high, at 0.87. This successful calibration allowed us to use the single seed NIR to advance of genotypes with highest protein content at an early stage in the breeding program, when seed is scarce. Using the NIR, we selected the 50% of farmer-selected wheat heads with the highest protein content. Selected seed was pooled from the two replicate plots of each biparental family and planted in 2016. Selection was repeated in 2016.

In tandem with on-farm selection, researchers selected biparental families on research stations for the most important regional traits. For winter wheat, research station selection protocols focused on traits of most importance to regional winter wheat farmers: Fusarium Head Blight (FHB) tolerance, high protein content, and desirable baking and sensory qualities. From the same winter wheat biparental families planted on farms, researchers evaluated 312 headrows of F2:F3 individuals and parental varieties at Freeville, NY and at a Fusarium head blight nursery in Ithaca, NY in 2015. At the disease nursery, headrows were inoculated at anthesis with Fusarium, and screened for FHB index (infection rate x severity). Other metrics of headrow evaluation included average height, leaf disease severity (1-9), glume blotch severity (0-3), and date at which 50% of heads emerged. Index selection provided a quantitative way to select the 30% best headrows from the field, weighted strongly against fusarium index, and to a lesser extent, against late heading, very tall height, leaf disease, and glume blotch (Equation 1). For baking and sensory quality, the third most important trait ranked by winter wheat farmers, the index added points to crosses with parents that have desirable quality characteristics. Researchers collected seed from 10 individual plants in each selected headrow. Eight seeds from each plant entered NIR evaluation for protein content. The 50% of plants with highest protein from each selected headrow, assed through NIR, advanced to the next year of selection. In 2016, 466 winter wheat F2:F4 individuals were assessed using the same method as 2015.

Equation 1. Index Score = IF(FHB index>10,-20,0)+(50/FHB index))+(IF(Heading Day after May1<38,0, Heading Day after May1*-0.5))+IF(Height<110,0,(Height*-1/20))+IF(RC(leaf disease<7,0,leaf disease*-1)+IF(leaf disease<5,(8/leaf disease),0)+IF(glume blotch=3,-5,IF(glume blotch=0,5,IF(glume blotch=1,3,0)))+(5 to 10 for parents with good qualities) (Excel 2013)

For spring wheat, researchers screened 210 headrows of F2:F3 individuals for the trait of most importance to spring wheat farmers in the region: weed competitive ability. Flagged plants and headrows with the highest early vigor (1-5) and leaf width (1-5) at third to fifth leaf stage advanced to the second year of selection in 2016. 90 headrows of F3:F4 individuals and F2:F4 individuals were grown in two locations in 2016. F4 lines were again ranked for early vigor (1-5) at Freeville, NY. Moreover, lines were inoculated in the Fusarium headblight nursery and scored for incidence and severity. Due to unprecedented drought conditions, the spring wheat F4 lines on research stations suffered greatly, and late season measurements were not feasible or representative of the crop. The best 30 individuals based on early vigor and FHB index were harvested.

1. Breed free-threshing populations of emmer, einkorn, and spelt;

In the spring of 2015, superior varieties ‘Lucille’, ‘ND Common’, and ‘Red Vernal’ were crossed with the free-threshing emmer variety, ‘Debra’, that is poorly adapted to the Northeast, in addition to a free-threshing emmer relative, Kamut®. In the greenhouse, F1 progeny were backcrossed to each parent during flowering. The seed from the BC1 generation were selected that threshed free of glumes. The plants were allowed to self to F2 and F3 and assessed for free threshing.

In the fall of 2014, einkorn and spelt seeds from various F3 biparental populations that most easily separated from the hull were selected. After rolling in a simple rubber tube, small portions of the hulled seed broke free of the glume. These seeds were planted in F2:F3 headrows at two sites for winter hardiness, heading date, height, and FHB tolerance. In the fall of 2015, we repeated this process, selecting the most easily dehulled seed from superior performing headrows. This selected seed was replanted as F2:F4 in the fall of 2015 for another year of agronomic evaluation. In the fall of 2015, the most easily dehulled F5 seed from top-performing headrows were selected and planted on a research station in Freeville, NY and on farm at Lakeview Organic Grain. After recognizing that our selection methods for free-threshing einkorn may not be producing the desired results, we sought germplasm of free-threshing einkorn to establish new breeding populations.

Free threshing einkorn lines were acquired from Canada, the USDA, and a breeding program in Germany.  These lines were grown in the field in 2016 to test their field performance.  Simultaneously, these free threshing einkorns were crossed with our top-performing agronomic F4 breeding lines from Lakeview Organic Grain and Cornell University.

1. Test homozygous F7 farmer-selected wheat populations for performance throughout the Northeast, and compare selected populations for local vs. broad adaptation;

In preparation for yield testing during the summer of 2016, the graduate researcher sent the farmer-selected populations to an off-season nursery for seed increase. Unfortunately, the nursery that we chose to work with in California destroyed the majority of seed as a result of poor land management. We supplemented seed by planting 533 plants for seed increase in a greenhouse at Cornell University. One site (Willsboro, NY) was removed from the multienvironment trials due to lower than expected amounts of seed from the increase. 

In 2016, farmer-selected F₇ populations of spring wheat and baseline randomly-collected F₄:F₇ populations were planted in split plot pairs to measure gains in selection at five research station locations (Old Town, ME; Borderview, VT; Fusarium Headblight Nursery in Ithaca, NY; Ketola in Ithaca, NY; Helfer in Ithaca, NY). Recorded data at Borderview and Old Town included yield at 12% moisture, test weight at 12% moisture, height, lodging (1 to 9), and early vigor between 4th and 5th leaf stage (1 to 9). At Old Town, ME, plots were overseeded with the surrogate weed Sinapsis alba cv. ‘Idagold’ using a Brillion seeder at a rate of 75 live seeds per m2. Weed and wheat biomass were measured in each plot by sampling two 0.25 m2 quadrats (0.5 m2 total sample size), separating out wheat from weed biomass, drying at 55°C, and weighing. At Ketola and Helfer, six-row one-meter miniplots were assessed by two to three evaluators for early vigor at 4^th leaf stage (1 to 9). Ground cover was also assessed at Ketola and Helfer using a 16 Megapixel camera and Canopeo App (Patrignani and Ochsner 2015) at one meter height during 3^rd, 4^th, and 5^th leaf stages. Due to errors incurred in the Canopeo processing software during high light conditions, shade cloth was added to camera equipment for the 4^th and 5^th leaf stages. A calibration trial was conducted to assess the consistency of visual early vigor genotype rankings among evaluators. Six graduate student evaluators were trained on visual rating of early vigor for five minutes, and then asked to rate 20 spring wheat genotypes for early vigor over three replicates. Through the package ‘lme4’ [version1.1-10] (Bates et al. 2015), the random effects of genotype, replicate, evaluator, and the interaction between genotype and evaluator were tested for variance in early vigor scores.

To determine if selection was effective, an F-test compared the fixed effect of population type (F₇ and F₄:F₇) for WCA traits using R [version 3.2.2] (R Core Team 2015), package ‘lme4’ [version1.1-10] (Bates et al. 2015) (Equation 2). The model included family and farm where selection took place as fixed effects, and block as a random effect.

Equation 2.     Y_ijkl = µ + α_i + β_j + γ_k + d_l+ε_ijkl 

H₀: µ_F7= µ_F4:F7; α≤0.05

y_ijkl: trait of interest for type i, family j, farmer k, and replicate l;

µ: overall mean response;

α_i: fixed effect of type i (e.g. F₇, F₄:F₇);

β_j: fixed effect of family j;

γ_k: fixed effect of farmer k; d_l: random effect of block l;

ε_ijkl: experimental error associated with response i,j,k,l

To measure local adaptation of selected populations, the F₇ populations were placed on four farms in a replacement experiment in 2016: Butterworks Farm (Westfield, VT), Grange Corner Farm (Lincolnville Center, ME), Essex Farm (Essex, NY), Adirondack Organic Grains (Westport, NY). Trials followed a randomized complete block design with three replicates. Each farmer grew the five to six “local” populations that he had selected in addition to three to four populations selected by each of the other farmers, for a total of 10 to 12 “introduced” comparison populations. Plots were screened for visual early vigor on a 1 to 9 scale, and visual measures of WCA. We were not able to seed a surrogate weed species on participating farms, due to concerns about increasing farm weed seedbanks. However, visual measures of WCA have been shown to correlate highly with biomass measurements (r=0.87), while saving considerable time and effort (Worthington et al., 2013). Measures of WCA varied by farm due to different levels of weed competition at each farm. At Essex and Adirondack, the visual ratio of wheat to weed biomass was measured. At Butterworks, intense natural weed competition prevented an accurate visual estimate of wheat to weed biomass, and heads of wheat were counted as a more reliable measure. At Grange Corner, where there was little to no weed presence, visual ratings of weed competition were not possible. An ANOVA with a random blocking factor tested differences in mean values of WCA between local and introduced populations using R [version 3.2.2] (R Core Team, 2015), package ‘lme4’ [version1.1-10] (Bates et al. 2015) (Equation 3). Data were plotted using ‘forestplot’ (Gordon 2014).

Equation 3.     Y_ijklm = µ + α_i +β_j + γ_k + d_l(f_m)+ ε_ijklm

H₀: µ_local = µ_introduced; α≤0.05

y_ijklm: trait of interest for type i, family j, farm k, rep k, trial m;

µ: overall mean response;

α_i: fixed effect of type i (local or introduced); β_j: fixed effect of family j;

γ_k: fixed effect of farm where selected k;

d_l(f_m): random effect of rep l, nested in trial m;

ε_ijklm: experimental error associated with response i,j,k,l,m

Seeding rates in all trials were adjusted based on germination rates and thousand kernel weight to obtain 376 viable seeds per square meter. Right-skewed responses were log transformed to obtain normal distributions for ANOVA models (e.g., data for the number of wheat heads at Butterworks and weed to wheat ratio at Old Town).

1. Develop a roadmap with farmers and the seed industry to release top-performing varieties.

The graduate researcher collaborated with Cornell University, NOFA-NY, and Greenmarket-GrowNYC to host an event titled “Sowing the Future of Organic Wheat” on June 10^th, 2016 (see Table 1). We collaborated with regional bakers to make varietal-specific breads and pastries for tasting.  Researchers provided a comprehensive overview of organic wheat research that has taken place up to this point in the Northeast.  Participants engaged in selecting wheat populations in the field.  To end the day, participants entered a focus group session to vision priorities, needs, and structure for releasing varieties and future breeding efforts.

Table 1. Agenda for “Sowing the Future of Organic Wheat”

12:30-1:00 Registration, tasting of Wegmans breads and varietal pastries

1:00-3:30 Review of regional grains research

Northern New England Bread Wheat Project – Heather Darby, University of Vermont


Value Added Grains for Local and Regional Food Systems – Lisa Kissing Kucek, Cornell University

Perennial Grains Update – Sandra Wayman, Cornell University

Regional Grain Marketing and Economics - June Russell, Greenmarket, GrowNYC


Restoring the Culture of Heritage Wheat – Eli Rogosa, Heritage Grain Conservancy

Hudson Valley Grain Research - Justin O’Dea, Cornell Cooperative Extension of Ulster County

3:30-4:00 Sample varietal breads made by Brookyln Bread Lab

4:00-5:00 Tour organic wheat breeding nurseries

Integrated Management of Fusarium Head Blight and Mycotoxins – Gary Bergstrom, Cornell University

Introduction to Organic Breeding Populations

Quality Seed for Organic Production – Phil Atkins, New York Seed Improvement Program

5:00-6:00 Focus group and listening session on the future of organic grains research

6:00-7:00 Celebration with wood fired pizza and regional beer sampling