Plant material.

All plant materials used in the screenings were named cultivars that are available either through nurseries or the USDA-ARS National Clonal Germplasm Repository. All plants were 5 to 6 years old and were maintained in 2.84-L pots in a 1:1 mixture (by volume) of sand and peat. Plants were maintained in coldframes, pruned to the crown each winter, and allowed to regrow several times. Detailed background and pedigree information on the cultivars used in this study may be obtained by request from the authors. Additional information on some cultivars is available in the Germplasm Resources Information Network (http://www.ars-grin.gov/npgs) online database.

Blight screening materials.

Nine cultivars were selected as standards for blight screening based on earlier testing (Ehlenfeldt and Stretch, 2000; Stretch et al., 1995). The standards were chosen to represent a broad range of resistance (from low to high) as well as to represent both highbush (HB, Vaccinium corymbosum L.) and rabbiteye (RE, Vaccinium ashei Reade) species. The cultivars Bluejay (HB), Brightwell (RE), Callaway (RE), Coastal (RE), Coville (HB), Northblue (a half-high type, i.e., a lowbush × highbush hybrid), Sunrise (HB), and Toro (HB) were included from 1996 to 2007. ‘Climax’ (RE) was included from 1996 to 2006. All cultivars were typically screened for 3 years along with the cultivar standards. Additional years of screening were added if the results from previous years appeared to be particularly variable. Approximately 35 cultivars were screened in each year. A total of 125 cultivars were screened for 2 to 6 years (Table 1). More than 75% were screened for either 3 or 4 years. Those screened for only 2 years were limited by plant availability.

Table 1. Mummy blight estimates, ranks, and raw percentages compared with blueberry cultivar standards (bold) listed in order of increasing susceptibility.

View Fullscreen

Fruit infection screening materials.

Six cultivars were selected as standards for fruit rot screening based on earlier testing data (Stretch and Ehlenfeldt, 2000) and the fact that they represented a range of resistance responses from low to high. The six “standard” cultivars used in all screenings from 1995 to 2007 were: Atlantic (HB), Bluejay (HB), Blueray (HB), Northsky (half-high), Rancocas (HB), and Sierra (HB). A total of 110 cultivars were coscreened with the standards for 2 to 5 years (Table 2). More than 90% of the cultivars were screened for 3 to 4 years.

Table 2. Mummy fruit infection estimates, ranks, and raw percentages compared with blueberry cultivar standards (bold) listed in order of increasing susceptibility.

View Fullscreen

Blight screening methods.

The tests were conducted outdoors in a randomized complete block design with five blocks and one potted plant of each cultivar per block. Potted plants were arranged in a grid pattern in each block with the specific dimensions of the grid determined by the number of entries (cultivars) and the space allocated each year. The space between plants averaged 0.45 m and blocks were laid out linearly. Plants for blight evaluation were typically 30 to 60 cm in height and were maintained as noted previously. Pseudosclerotia (mummies) were collected in September and October of each year from the field at the P.E. Marucci Center for Blueberry and Cranberry Research, Chatsworth, NJ. Mummified fruit were planted (≈5 per pot) into soil-filled 0.9-L plastic pots and overwintered in plastic-covered coldframes. Pots were watered as needed to maintain moisture. Emergence of apothecia typically started in early April. Pots with newly emerged apothecia were removed from the coldframes and evenly spaced within and around the periphery of the experiment area to provide a uniformly distributed source of inoculum. To enhance the infection process, the entire plot received supplemental misting for 1 h every evening during ascospore release (applied volume ≈5 mm/day).

Trials were typically set up starting the first week of April with the timing of plant exposure to inoculum determined by initiation of shoot development in the cultivars. Once exposed to inoculum, plants were examined for blighted shoots on a weekly basis between early May, when symptoms first became visible, and late May/early June, when no further new blighting appeared. Blighted shoots were tabulated and removed at each inspection to avoid overestimating blight incidence. At the end of the blight period, remaining healthy shoots were counted on each plant so that the percentage of blighted shoots could be calculated. The mean number of shoots evaluated per cultivar per year was 77 with a sd of 41.

Fruit infection screening methods.

Plants for fruit infection evaluation were typically 50 to 80 cm tall and maintained as noted previously. Flower/fruit infection screening was conducted in a shadecloth-covered (55% transmission) coldframe. Plants were arranged in a randomized complete block design of five blocks with one pot of each cultivar per block. The space between plants averaged 0.45 m between rows and 0.36 m between plants within rows. In all years, heavily blighted plants of ‘Bluehaven’ and ‘Blueray’ (in 2.8-L plastic pots) with sporulating infections were evenly distributed between every two rows of plants to serve as inoculum sources located within 0.5 m or less of the plants to be tested. In previous studies, ‘Bluehaven’ and ‘Blueray’ had averaged 64% and 36% blight, respectively (considered highly susceptible) (Ehlenfeldt et al., 1996). Additional plants of ‘Bluehaven’ and ‘Blueray’ were added over time, as necessary, to maintain inoculum levels. To facilitate transfer of conidia from the blighted tissue to the stigmata, two hives of honeybees were placed in the netted coldframe, one at each end of the experiment area. Open flowers were removed from the test plants before setting the inoculum plants and bees in the coldframe. The experiment area was sprinkler-irrigated (22 mm/day) every evening during conidia production to enhance infection. All developing fruit were picked while still green, at 4 to 6 weeks after flowering, and stored at 4 °C until evaluated. Fruit were sliced in half crosswise and observed for the distinctive white fungal growth pattern of the mummy berry pathogen. The incidence of mummy berry fruit infection was expressed as a percentage of the total number of fruit on each plant. The mean number of fruit evaluated per cultivar per year was 91 with a sd of 61.

Cultivar ranking methodology.

Previous research determined that year-to-year variability in manifestation of both phases of mummy berry disease is high, requiring 6 to 8 years of testing to establish reliable resistance estimates (Ehlenfeldt et al., 2010). Because it is not practical to test a large number of cultivars for the needed number of years, we ranked cultivars for the years in which data were available.

There are many methods available that might be used to create rankings when the items to be ranked occur together in combinations that leave large gaps, like in our data in which only a few cultivars were present in most years. Thus, any ranking system needs to work when most cultivars have only a spotty representation in the data set. This is a relatively easy exercise if there is no year × cultivar interaction. However, in the presence of a year × cultivar interaction, like in our data, there is considerable uncertainty in how a cultivar would have performed in a year for which there are no data for that cultivar, which makes ranking problematic. We tried a number of different published ranking methods [e.g., best linear unbiased predictors using mixed models, e.g., Hill and Rosenberger (1985); generalization of the Bradley-Terry model, e.g., Graves et al. (2003); Hunter (2004)] and found them to produce contradictory rankings with each other (not only for our data set, but also for the demonstration data sets in the publications), and none appeared to effectively capture our intuitive understanding of the data. The method we developed produces stable rankings that match our intuitive understanding of the data and also produces estimates of uncertainty about the ranking. The method is based on principal component analyses and resampling. Essentially, it resolves year-to-year disease incidence contradictions by averaging but qualifies this process in that less “stable” cultivars will have greater uncertainty associated with their final ranking. The use of principal components allows the scale (i.e., fluctuations in yearly average disease incidence) to differ from year to year. This allows cultivars only present in “low” disease incidence years to be compared with those present in “high” disease incidence years. Because most cultivars have fewer than the requisite number of years required for reliable estimates of disease incidence, the rankings we produced should be viewed as preliminary, although the positions of standards may be considered as better established. Our analysis begins by calculating a mean disease incidence (on the arcsine transformed scale) for each cultivar–year combination and by removing any cultivar that is not present for at least 2 years of testing. We then randomly sample 2 years from our data set and stop when we find 2 years that share at least two cultivars (because standards were present in most years for both mummy blight and fruit rot data sets, typically many cultivars were shared for randomly selected pairs of years). We use a principal components analysis to determine an axis that best goes through the data sets for the 2 years (this is analogous to a Type II regression finding the line that minimizes the sums for squares for both years equally as opposed to the typical Type I regression that minimizes sums of squares in the vertical direction only) and store this result as the first column in a matrix of 1000 columns (each row is a cultivar). This column will contain a score (a projected value) for each of the cultivars present in both years. Cultivars present in 1 year, but not the other, are regressed on the principal component axis and these predictions used for their values. This process is done repeatedly (999 additional times) for randomly sampled pairs of years. The matrix holding the projected scores then replaces the matrix of original data and the same process is then used for randomly sampled pairs of these projected scores (i.e., two columns from the matrix of 1000 columns are repeatedly resampled). Although a pair of years may have only a few cultivars in common, its resulting projection will have values for all unique cultivars in each year, so there will be fewer “gaps” at this stage. The resulting matrix is again subjected to selection of randomly selected pairs of columns, and this process is continued until all 1000 columns are essentially identical (usually ≈20 total iterations for this data set). This produces a set of scores (i.e., a single score for each cultivar) that can be converted to ranks, if desired. However, we repeat the entire process 30 times so we have 30 scores for each cultivar and use the mean of the 30 scores to produce a ranking. Using the 30 independent estimates of scores for each cultivar, we can also estimate the uncertainty of the scores (or rankings).