Squash Team | 2019 CucCAP Progress Report

Team members Michael Mazourek (Cornell University), Linda Wessel-Beaver (University of Puerto Rico), Angela Linares (University of Puerto Rico), and Chris Smart (Cornell University) reported on progress made on the CucCAP grant in the previous year.

Establish core GWAS populations

1.2.1. ,1.2.2 GBS of cucurbit species, establish molecular-informed core populations and 1.2.2. Population genetics and GWAS analysis

The core set of accessions representing Cucurbita pepo diversity in the NPGS has been self-pollinated and is being combined with heirloom cultivars to anchor market classes and enrich for cultivar genetics. Sources of resistance and other representatives of diversity in other species to extend the utility of the panel and being selfed pollinated. The goal of representatives from other species is both because squash improvement often involves crosses between species and to extend the benefits of CucCAP investments to those that work with other species such as C. moschata (Puerto Rico). The second round of self pollination will likely take place with a subcontractor. The process is now including strain purification within the stocks for accessions that do not match their descriptors for hull-less seeded accessions that aren’t, we are creating new selfed stocks from hull-less segregants.

Three projects are already taking advantage of the GWAS population. For CucCAP, given the lack of phenotypic data, we have phenotyped the collection for qualitative traits of bush growth habit and hull-less seeds. Markers were created with this material validated in breeding populations to include as part of the MS. A separate study is mapping cotyledon cucurbitacin content to support results from biparental populations.

Powdery mildew resistance in squash Marker development and verification

(Mazourek lab –K. LaPlant)
Completed in 2017

Virus resistance in squash, Mapping resistance (M. Mazourek lab-K. LaPlant

‘Whitaker’ is a C. pepo summer squash from Cornell that is resistant to PRSV and CMV, as well as powdery mildew and ZYMV. The pedigree of ‘Whitaker’ contains C. ecuadorensis and C. okeechobeensis subsp. martinezii, and therefore it contains several introgressions from each species within its genome. By using ‘Whitaker’ as a guide to common introgressions from C. ecuadorensis, we have tentatively identified a genomic region on chromosome 16 with a length of approximately 1 Mb that may be associated with resistance to PRSV. ‘Whitaker’ has been used extensively in many breeding programs. We have developed ‘Whitaker’-based biparental mapping populations to further refine and validate any identified genomic regions associated with resistance., Introgress resistance into advanced breeding lines

(L. Beaver, A. Linares labs – M. Miranda, W. Seda)

Inheritance of resistance to PRSV:

Two sources of resistance are well known in C. moschata: ‘Nigerian Local’ and ‘Menina’. The inheritance of resistance from ‘Nigerian Local’ has been previously studied, but inheritance studies have not been reported for ‘Menina’, nor is it known if resistance to PRSV in ‘Nigerian Local’ is allelic to that in ‘Menina’. In the inheritance study susceptible genotypes were ‘Verde Luz’, ‘Taina Dorada’ and ‘TP411’. The third to fifth leaf of inoculated seedlings were rated on a 0 to 4 scale for disease severity and scores were combined to convert to a 0 to 12 scale. Resistant x susceptible F2 populations using ‘Nigerian Local’ as the source of resistance (distributions on the left-hand side of Figure 1) had nearly normal distributions with an average disease severity of 5.23 in Nigerian Local x Taína Dorada and 6.25 in Verde Luz x Nigerian Local. In contrast, F2 populations with ‘Menina’ (distributions on the right-hand side of Figure 1) were strongly skewed towards resistance with an average severity of 3.38 in Menina x Taína Dorada, 2.27 in Verde Luz x Menina and 2.80 in TP411 x Menina. The resistant x resistant Nigerian Local x Menina F2 population was very highly skewed, with an average combined severity of 0.840 (Figure 2).

Resistant to susceptible segregations in F2 populations were variable, depending on how severity scores were grouped into the resistant versus susceptible classes. The most consistent results over similar types of crosses occurred when we grouped plants with an overall severity rating of <4 as resistant and grouped plants with an overall severity rating of >5 as susceptible. This grouping system also made biological sense since plants with ratings of >5 had high individual leaf severity scores, especially in leaves 4 and 5. Both F2 crosses made with ‘Nigerian Local’ fit a 7:9 (R:S) genetic model while all three crosses using ’Menina’ fit a 13:3 model (Table 1). The resistant x resistant cross (Nigerian Local x Menina) fit a 15:1 model. These segregations suggest that at least two genes are involved in the inheritance of resistance to PRSV for both ‘Nigerian Local’ and ‘Menina’. The data clearly indicate that at least some of the genes for resistance in ‘Nigerian Local’ and ‘Menina’ are different. The resistance conferred by ‘Menina’ seems to be superior to that of ‘Nigerian Local’.

An important consideration when evaluating disease resistance in the greenhouse is the association between greenhouse readings and readings taken in the field. In breeding for PRSV resistance both symptom severity and ELISA readings can be used as a way to evaluate resistance. In a previous CucCAP report we reported correlations between greenhouse and field ELISA readings to be poor. However, we have since looked at this issue from a different point of view. A high correlation per se is not important as long as plants judged resistant (or susceptible) in the greenhouse are also judged as resistant (or susceptible) in the field. Figure 2 presents greenhouse and field data from 2017. Data from 2016 showed a similar trend. All plants of genotypes known to be susceptible (Mos166, Waltham Butternut and Taína Dorada) fell into the upper right-hand quadrant, meaning they were classified as susceptible in both the greenhouse and field according to their ELISA reading. The results for genotypes known to be resistant (Nigerian Local and Menina) were not as clear. For these genotypes, ELISA readings in the greenhouse were often expectantly high (positioned in the lower right-hand quadrant), while readings in the field were low. However, it should be noted that in this trial greenhouse readings were taken at 18 days post-inoculation (dpi) on the 3rd leaf. Since carrying out this study we have found that greenhouse ELISA readings for PRSV are best taken on the 4th leaf at about 21 dpi (PRSV ELISA readings tend to be high for all genotypes in the first few leaves).

Figure 1. Distributions of severity ratings in F2 populations of tropical pumpkin (Cucurbita moschata) inoculated with Papaya ringspot virus (PRSV). Populations developed with resistant parent ‘Nigerian Local’ are shown on the left; populations developed with resistant parent ‘Menina’ are shown on the right. For each plant, disease severity in leaf position 3, 4 and 5 was evaluated on a 0 to 4 scale (0 = no symptoms). Values were summed to produce an overall severity index of 0 to 12.

Figure 2. Distribution of combined severity ratings of plants (n=238) from the Nigerian Local x Menina F2 population inoculated with Papaya ringspot virus (PRSV). For each plant, disease severity in leaf position 3, 4 and 5 was evaluated on a 0 to 4 scale (0 = no symptoms). Values were summed to produce an overall severity index of 0 to 12.

View both figures on page 56 of the pdf version

Table 1. Number of plants evaluated and observed segregations in parental, F1 and F2 populations. ‘Nigerian Local’ was the resistant parent in the F1 and F2 crosses. Goodness-of-fit in F2 populations was tested with chi-square.



Observed segregation

Tested ratio (R:S)



Nigerian Local Res. parent 34 0
Menina Res. parent 50 0
Taina Dorada Sus. parent 2 18
Verde Luz Sus. parent 4 16
TP411 Sus. parent 0 9
Resistant x susceptible crosses with Nigerian Local as resistant parent:
Nigerian Local x Taína Dorada F1 8 2
Verde Luz x Nigerian Local F1 10 0
Nigerian Local x Taína Dorada F2 47 64 7:9 0.0894 0.795
Verde Luz x Nigerian Local F2 42 68 7:9 1.3859 0.2391
Resistant x susceptible crosses with Menina as resistant parent:
Menina x Taína Dorada F1 10 0
Verde Luz x Menina F1 10 0
TP411 x Menina F1 10 0
Menina x Taína Dorada F2 91 29 13:3 2.311 0.1285
Verde Luz x Menina F2 101 17 13:3 1.4611 0.2268
TP411 x Menina F2 91 20 13:3 0.039 0.8434
Cross between two resistant parents:
Nigerian Local x Menina F1 20 0 n/a n/a n/a
Nigerian Local x Menina F2 224 14 15:1 0.0549 0.8147

¹For each plant, disease severity in leaf position 3, 4 and 5 was evaluated on a 0 to 4 scale (0 = no symptoms). Values were summed to produce an overall severity index of 0 to 12. Plants were then categorized R for resistant (overall severity rating of <4) or S for susceptible (overall severity rating of >5).

Figure 3. Scattergram of enzyme-linked immunosorbent assay (ELISA) readings for Papaya ringspot virus (PRSV) in six genotypes of tropical pumpkin inoculated with PRSV. Each data point represents readings for a single plant at 18 days post-inoculation (dpi) and 54 dpi. . MEN=’Menina’, NL=’Nigerian Local’, MOS=’Mos166’, TD=’Taína Dorada’, SOL=’Soler’, WAL=’Waltham’. Readings above the horizontal line are considered to be positive readings for the virus. Readings to the right of the vertical line (18 dpi) or above the horizontal line (54 dpi) are considered positive for the presence of PRSV.

View figure 3 on page 57 of the pdf

Phytophthora blight resistance in butternut squash Mapping resistance to Phytophthora blight

– Smart, Vogel, Kousic

Raw genotypes calls from TASSEL for the NPGS C. moschata collection were filtered using vcftools to retain just biallelic SNPs. This SNP set was then filtered for minor allele frequency (MAF) > 0.05, sample call rate > 0.20, and mean read depth per site < 44 (corresponds to 95th percentile). The filtered data set included 311 accessions and 40,428 SNPs.

Missing genotypes were imputed with the LD-k nearest neighbors algorithm implemented in TASSEL. 1000 genotype calls with a read depth > 8 were masked to estimate the imputation accuracy rate, which was estimated to be 78%. Any genotypes not imputed by the algorithm were then imputed with the mean. The imputed SNPs were then filtered again for MAF > 0.05, resulting in a final set of 36,568 SNPs.

Ratings at 12 and 41 days post inoculation (dpi) were analyzed separately, only using plots with 3 or more non-missing data points. The mean rating per plot was used as the response variable in a mixed linear model, with accession treated as a random effect because of unbalanced data. Rep was included as a random effect with Dpi12 but not with Dpi41 because the variance explained by Rep was effectively 0. Best linear unbiased predictions (BLUPs) for accession effects were then used in

GWAS. The line mean heritability was 0.66 for Dpi12 average plot rating and 0.29 for Dpi41 average plot rating. GWAS was performed using the rrBLUP R package. Genotype data was available for 265 accessions with BLUPs for Dpi12 and 264 accessions with BLUPs for Dpi41. Population structure and relatedness among accessions were controlled by including the first principal component of the genotype matrix as a fixed effect and treating a random effect for accession and including a relationship matrix to model their effect and treating a random effect for accession and including a relationship matrix to effect and treating a random effect for accession and including a relationship matrix to model their covariance.

Figure 4. Manhattan plot and qqplot for disease ratings for C. moschata GWAS panel inoculated with P. capsici twelve days post inoculation

View figure 4 on page 58 of the pdf

No significant SNPs were identified using the ratings from 12 dpi. The qqplot shows a good fit of the observed p-values to the expected under the null hypothesis. (Figure 4).

With the ratings at 41 dpi, there are three significant SNPs at a false discovery rate of 5% on chromosomes 10 and 18. However, the qqplot shows that the Type I error is inflated. This is likely related to the highly non-normal data used for GWAS which violates model assumptions. These three significant SNPs have low minor allele frequencies (<0.10).. (Figure 6).

Figure 5. Manhattan plot and qqplot for disease ratings for C. moschata GWAS panel inoculated with P. capsici41 days post inoculation

Figure 6. Boxplots of effect of the most significant SNP from the two chromosomes on the average plot rating at both 12 and 41 dpi

Figure 7. Prediction accuracy reported is the correlation between observed phenotypes and predicted phenotypes.

View figures 5, 6, & 7 on page 59 of the pdf

Genomic prediction using a GBLUP model was performed with the rrBLUP package. Eighty percent of the accessions were used as a training set to predict the phenotypes of the remaining twenty percent, whose phenotypic values were masked. This five-fold cross validation scheme was repeated 100 times.

The mean prediction accuracy for dpi12 BLUPs (0.51) was considerably higher than the mean prediction accuracy for dpi41 BLUPs (0.25). (Figure 7).

View the PDF version of this report