Skip to main content
NCBI home page
Horticulture Research logoLink to Horticulture Research
. 2025 Mar 21;12(7):uhaf086. doi: 10.1093/hr/uhaf086

Genomic prediction and association analyses for breeding parthenocarpic blueberries

Juliana Cromie 1, Ryan P Cullen 2, Camila Ferreira Azevedo 3,4, Luis Felipe V Ferrão 5, Felix Enciso-Rodriguez 6, Juliana Benevenuto 7, Patricio R Muñoz 8,✉
PMCID: PMC12064955  PMID: 40352286

Abstract

Parthenocarpy is a desirable trait that enables fruit set in the absence of fertilization. While blueberries typically depend on pollination for optimal yield, certain genotypes can produce seedless fruits through facultative parthenocarpy, eliminating the need for pollination. However, the development of parthenocarpic cultivars has remained limited by the challenge of evaluating large breeding populations. Thus, establishing molecular breeding tools can greatly accelerate genetic gain for this trait. In the present study, we evaluated two blueberry breeding populations for parthenocarpic fruit set and performed genome-wide association studies (GWAS) to identify markers and candidate genes associated with parthenocarpy. We also compared the predictive ability (PA) of three molecular breeding approaches, including (i) genomic selection (GS); (ii) GS de novo GWAS (GSdnGWAS), which incorporates significant GWAS markers into the GS model as prior information; and (iii) in silico marker-assisted selection (MAS), where markers from GWAS were fitted as fixed effects with no additional marker information. GWAS analyses identified 55 marker–trait associations, revealing candidate genes related to phytohormones, cell cycle regulation, and seed development. Predictive analysis showed that GSdnGWAS consistently outperformed GS and MAS, with PAs ranging from 0.21 to 0.36 depending on the population of study and the specific markers utilized. MAS showed PAs comparable to GS in some cases, suggesting it could be a cost-effective alternative to genome-wide sequencing. Together, these findings demonstrate that molecular breeding techniques can be used to improve facultative parthenocarpy, offering new avenues to develop high-yielding blueberry varieties that are less reliant on pollination.

Introduction

As the cultivation of pollinator-dependent crops continues to outpace the availability of managed pollinators, concerns regarding the strain on pollination services and its potential impact on agricultural productivity have heightened [1–3]. This issue is further compounded by climate change, declines in pollinator health, and the loss of habitat for native pollinators [4–6]. Parthenocarpy, which enables fruit development in the absence of pollination and fertilization, has drawn attention from plant breeders for its potential to ensure yield in pollinator-limited or stressful abiotic conditions [7–10].

Blueberry (Vaccinium spp.) is a globally important berry crop [11], which typically relies on cross-pollination to achieve optimal yield and fruit quality [12–15]. Despite the regular use of managed pollinators, the crop frequently suffers from pollen limitation and subsequent yield loss [16–20]. Various factors contribute to pollination deficits, including adverse weather during the brief bloom period [21], suboptimal pollination by honey bees compared to native ‘buzz’ pollinators [22–24], and uneven visitation to cultivars based on pollinators’ floral preferences [25, 26]. Thus, breeding parthenocarpic blueberry varieties can help to supplement yields in cases of pollination deficit [27–29].

Parthenocarpy has been utilized in several crops to obtain seedless fruits [7, 30]. There are three primary classes of parthenocarpy, including stenospermocarpy, which relies on fertilization and subsequent seed abortion, obligate parthenocarpy, in which all fruits are seedless due to sterility or parthenogenesis, and facultative parthenocarpy, where plants can produce either seedless or seeded berries depending on the incidence of pollination [10, 31]. Blueberry exhibits natural facultative parthenocarpy, which allows for parthenocarpic genotypes to be both selected and used as parents in a breeding program [32, 33]. While pollination generally improves blueberry fruit quality, certain genotypes can yield seedless parthenocarpic fruits of comparable size and quality to those that are open-pollinated [34, 35]. However, the development of parthenocarpic cultivars is limited by the challenge of evaluating large breeding populations, which involves the prevention of pollination through flower emasculation and the exclusion of pollinators. Thus, the establishment a molecular breeding pipeline for this trait can help to minimize the need for intensive phenotyping efforts and expedite genetic gains.

Molecular breeding tools have been used to advanced parthenocarpy in various crops [36–39]. In tomato, more than 20 sources of parthenocarpy have been identified, with few instances of facultative parthenocarpy [10, 40, 41]. Candidate genes and variants linked to phytohormones like auxin, gibberellic acid, and cytokinin have been uncovered through quantitative trait loci (QTL) mapping and genome-wide association studies (GWAS) in tomato, eggplant, and cucumber [42–46]. While marker-assisted selection (MAS) has been implemented to improve parthenocarpy in some cases [36, 47], numerous studies indicate a complex genetic architecture with several important QTL involved [46, 48, 49]. Despite this, comprehensive strategies tailored for quantitative traits, such as genomic selection, have yet to be applied for facultative parthenocarpy to our knowledge.

As reported in other fruit species, we hypothesized that parthenocarpy is highly genetically controlled in blueberries and therefore molecular tools could be designed to support breeding decisions. To this end, we evaluated the diversity and potential for facultative parthenocarpy in two southern highbush blueberry breeding populations comprising a total of 505 genotypes. To aid in the development of molecular breeding tools, we identified markers linked to parthenocarpy through GWAS analyses and investigated the potential candidate genes within their surrounding genomic regions. Furthermore, we compared the predictive ability (PA) of three molecular breeding strategies, including genomic selection (GS), GS de novo GWAS (GSdnGWAS), and in silico marker-assisted selection (MAS), considering an interpopulation cross-validation framework to simulate implementation within a real breeding program. Together, we present the largest genomic study discussing new tools for addressing pollination issues in the Vaccinum genus.

Results

Parthenocarpy phenotype

A diverse collection of 505 southern highbush blueberry genotypes, comprising two breeding populations, were assessed for parthenocarpic fruit set (Fig. 1A–D). Phenotypic values ranged from 0% to 100% out of 10 unpollinated flowers, with 44% of the genotypes exhibiting at least 10% parthenocarpic fruit set (Fig. 1E). The mean parthenocarpic fruit set for each population was 15% and 17%, respectively. Narrow-sense heritability for parthenocarpy considering both populations was calculated as 0.22.

Figure 1.

Figure 1

Open in a new tab

Assessment of parthenocarpy in two southern highbush blueberry breeding populations. (A) The parthenocarpy treatment included emasculation of flowers and exclusion of pollinators for 10 flowers of each genotype to prevent any possibility of pollination or fertilization. (B) Facultative parthenocarpic genotypes yielded seedless fruits after emasculation treatment. (C) The same genotype yielded seeded fruits from open pollination in untreated flowers. (D) The raw phenotypic distribution of parthenocarpy for independent and combined populations is presented in a violin plot. (E) Pie charts illustrate the proportion of individuals in each population displaying parthenocarpic fruit set of ≥10%.

Genome-wide association studies for parthenocarpy

A total of 59 912 SNPs across 12 chromosome-scale scaffolds were tested for association with parthenocarpy (Figs S1–S3). Manhattan plots displaying the results of the GWAS are presented for each population separately and combined (Fig. 2). Independent analysis of the first population (POP1) revealed 29 significant marker–trait associations on chromosomes 2, 3, 5, 6, 7, 8, 9, 10, 11, and 12. The second population (POP2) exhibited nine associations on chromosomes 1, 3, 4, and 6. In the combined analysis (POP1 and POP2), 17 significant SNPs were detected on chromosomes 5, 7, 8, and 11. In all cases, individual markers explained <15% of the phenotypic variation for parthenocarpic fruit set (Table 1).

Figure 2.

Figure 2

Open in a new tab

GWAS results for parthenocarpic fruit set. Manhattan plots show marker–trait associations for each population (color) and tetraploid gene action model (shape). The most stringent significance threshold for all GWAS analyses is indicated by the dashed horizontal line, and was calculated using the ‘M.eff’ multiple-testing correction (α = 0.05) considering an additive gene action model.

Table 1.

All significant SNPs associated with parthenocarpy identified from GWAS analyses.

Population Chr Position Ref/Alt Model Thresh. Score Effect R 2 P-val MAF
Population 1 2 34110377 C/T General 5.78 5.83 NA 0.01 0.199 0.03
Population 1 2 36035768 C/T General 5.78 6.71 NA 0.08 0.000 0.07
Population 1 3 34991045 C/T General 5.78 6.85 NA 0.09 0.000 0.21
Population 1 3 37811043 C/T 2-dom-ref 5.6 5.82 −0.31 0.00 0.871 0.09
Population 1 5 5776698 C/G General 5.78 6.02 NA 0.04 0.012 0.08
Population 1 6 2060864 A/T 2-dom-alt 5.42 5.63 −0.34 0.03 0.005 0.20
Population 1 6 10348201 A/G 1-dom-ref 5.73 6.08 0.17 0.06 0.000 0.34
Population 1 6 15569023 A/G 2-dom-ref 5.6 5.71 −0.32 0.00 0.465 0.06
Population 1 7 8421613 G/A General 5.78 6.58 NA 0.05 0.004 0.17
Population 1 7 9545305 TC/TT General 5.78 5.88 NA 0.00 0.935 0.01
Population 1 7 29690597 G/A General 5.78 6.61 NA 0.03 0.047 0.19
Population 1 8 1744735 G/C 2-dom-alt 5.42 5.52 −0.27 0.01 0.221 0.20
Population 1 8 5542221 C/T 1-dom-alt 5.22 5.31 −0.37 0.03 0.002 0.23
Population 1 9 16819984 A/G 1-dom-alt 5.22 5.68 −0.34 0.02 0.013 0.18
Population 1 9 35167278 G/A 2-dom-ref 5.6 5.72 −0.26 0.01 0.219 0.08
Population 1 9 35452313 C/T General 5.78 5.84 NA 0.02 0.219 0.10
Population 1 9 36318057 G/A 1-dom-alt 5.22 5.47 −0.15 0.01 0.156 0.42
Population 1 10 24262910 C/A 1-dom-alt 5.22 6.39 −0.31 0.04 0.001 0.49
Population 1 11 1833631 G/C General 5.78 6.34 NA 0.02 0.037 0.04
Population 1 11 2581098 G/A General 5.78 9.04 NA 0.01 0.641 0.12
Population 1 11 2581098 G/A 2-dom-ref 5.6 7.97 −0.29 0.00 1.000 0.12
Population 1 11 4611151 T/C 2-dom-ref 5.6 5.88 −0.35 0.00 0.790 0.06
Population 1 11 5603261 C/G 2-dom-alt 5.42 5.53 0.28 0.00 0.429 0.07
Population 1 11 20780807 A/C Additive 5.78 6.01 −0.17 0.01 0.169 0.02
Population 1 11 20780807 A/C 1-dom-ref 5.73 6.43 −0.19 0.00 0.343 0.02
Population 1 11 21099188 T/C 1-dom-ref 5.73 5.87 −0.18 0.00 1.000 0.02
Population 1 11 21099188 T/C Additive 5.78 5.87 −0.18 0.00 1.000 0.02
Population 1 11 21099188 T/C General 5.78 5.87 NA 0.00 1.000 0.02
Population 1 12 14473779 C/T 2-dom-ref 5.6 5.91 0.12 0.02 0.016 0.46
Population 2 1 24763902 G/A 1-dom-alt 5.17 5.92 −0.51 0.15 0.000 0.32
Population 2 1 34549718 C/T 2-dom-ref 5.56 5.65 −0.44 0.12 0.000 0.12
Population 2 3 12605530 A/T General 5.76 6.13 NA 0.09 0.000 0.02
Population 2 3 12605530 A/T Additive 5.76 6.59 −0.36 0.00 1.000 0.02
Population 2 3 12832095 A/G Additive 5.76 5.93 −0.33 0.00 0.601 0.02
Population 2 3 18284138 C/T 2-dom-ref 5.56 5.57 −0.41 0.07 0.000 0.09
Population 2 4 8103119 T/C General 5.76 6.88 NA 0.00 0.804 0.08
Population 2 4 9662988 T/A General 5.76 6.49 NA 0.10 0.000 0.08
Population 2 6 17667007 C/T General 5.76 6.07 NA 0.06 0.003 0.06
Combined 5 21298294 A/G 1-dom-ref 5.74 6.05 0.11 0.02 0.002 0.10
Combined 5 21719536 C/T General 5.78 6.75 NA 0.06 0.000 0.06
Combined 5 32652631 AT/AG 1-dom-ref 5.74 5.81 0.35 0.01 0.065 0.32
Combined 5 32652671 C/T General 5.78 6.14 NA 0.00 0.873 0.34
Combined 5 32652671 C/T 1-dom-ref 5.74 7.24 0.41 0.00 1.000 0.34
Combined 7 29690597 G/A General 5.78 7.52 NA 0.06 0.000 0.19
Combined 8 723946 A/G General 5. 78 6.06 NA 0.02 0.009 0.14
Combined 11 21099188 T/C 1-dom-ref 5.74 6.55 −0.19 0.00 1.000 0.02
Combined 11 21099188 T/C Additive 5.78 6.55 −0.19 0.00 1.000 0.02
Combined 11 21099188 T/C General 5.78 6.55 NA 0.00 1.000 0.02
Combined 11 22139008 G/A Additive 5. 78 6.09 −0.17 0.00 0.270 0.02
Combined 11 22139008 G/A 1-dom-ref 5.74 6.56 −0.20 0.00 0.410 0. 02
Combined 11 22248486 T/A 1-dom-ref 5.74 6.97 −0.21 0.00 1.000 0.02
Combined 11 22248486 T/A Additive 5.78 7.21 −0.20 0.00 1.000 0.02
Combined 11 22248486 T/A General 5.78 6.36 NA 0.00 1.000 0.02
Combined 11 22561347 A/G Additive 5. 78 6.49 −0.19 0.00 0.735 0.02
Combined 11 22561347 A/G 1-dom-ref 5.74 6.31 −0.19 0.000 0.828 0.02
Open in a new tab

Thresh is the significance threshold for a given gene action model within each population; score indicates the -log10(p) value; effect is the coefficient of the marker; R2 is the proportion of phenotypic variance explained by a multiple QTL model using backward elimination; P-val is the P-value attributed to each R2 value; MAF is minor allele frequency in the population.

Altogether, 55 marker–trait associations were detected using tetraploid genetic parameterizations, including additive (8), simplex dominance (15), duplex dominance (11), and general (21) models. Notably, the majority of significant marker–trait associations were found to be population-specific, with many exhibiting different allele frequencies in each independent population (Fig. S6). However, significant marker–trait associations were identified on chromosome 6 in both breeding populations, with markers positioned proximally within a 2 Mb window. Additionally, two loci on chromosomes 7 and 11 were found to be significant in both POP1 and the combined analysis.

Candidate gene mining

Genomic regions (±100 kb) flanking the significant SNPs identified from GWAS analyses were investigated for candidate genes related to parthenocarpy. Of the 390 genes within these windows, 232 (59%), and 226 (58%) were functionally annotated in silico by the PANZZER2 and eggNOG-mapper servers, respectively. Candidate genes related to phytohormone regulation, cell division, embryogenesis, and seed development were found in these regions. A selected list of top candidate genes with literature support is presented in Table 2. The potential roles of these genes in parthenocarpy are further presented in the discussion section.

Table 2.

A list of selected candidate genes related to parthenocarpy.

Population Gene name Chr Start End Description PFAMs References
Population 2 VaccDscaff1-snap-gene-247.18 1 24700091 24702355 Peroxidase Peroxidase [68], [69], [70], [71]
Combined VaccDscaff7-augustus-gene-327.28 5 32695014 32697554 Autophagy-related 2 - [72], [73]
Population 2 VaccDscaff11-augustus-gene-177.31 6 17668793 17677635 Calmodulin-binding protein 60 B-like Calmodulin_bind [74], [75], [76]
Population 1 VaccDscaff12-processed-gene-83.4 7 8357513 8361527 Cytochrome P450 p450 [77], [78], [79], [80]
Population 1 VaccDscaff12-augustus-gene-296.27 7 29618208 29620941 Heavy metal-associated isoprenylated plant protein 39 HMA [81]
Combined & Population 1 VaccDscaff12-snap-gene-296.36 7 29637070 29639862 WRKY domain-containing protein WRKY [82], [83, 84], [85]
Combined VaccDscaff12-augustus-gene-296.30 7 29651276 29654098 Desiccation protectant protein Lea14 homolog LEA_2 [74], [86]
Combined VaccDscaff12-processed-gene-297.0 7 29676797 29681425 S-adenosyl-L-methionine-dependent methyltransferases Methyltransf_29; zfNARING_2 [87], [88]
Combined VaccDscaff13-augustus-gene-7.21 8 699713 701850 Ethylene-responsive transcription factor AP2 [74], [89], [88]
Population 1 VaccDscaff13-snap-gene-17.60 8 1775028 1781304 Polygalacturonase-like protein Glyco_hydro_28 [90]
Population 1 VaccDscaff17-snap-gene-167.20 9 16733085 16735477 Knotted1-like homeodomain protein liguleless4b (Fragment) ELK,Homeobox_KN [91], [92], [93]
Population 1 VaccDscaff17-augustus-gene-355.30 9 35494747 35499415 Auxin-responsive protein AUX_IAA [94], [95], [96]
Population 1 VaccDscaff21-augustus-gene-207.21 11 20744582 20753840 Glucomannan 4-beta-mannosyltransferase 9 Glyco_trans_2_3 [97]
Population 1 VaccDscaff21-snap-gene-208.21 11 20830135 20845455 Transcriptional activator DEMETER PermNACXXC,RRM_DME [98], [99]
Open in a new tab

Gene descriptions and annotations are provided from the PANZZER2 software. Protein families from the Pfam database are also indicated. Literature sources are provided for studies that identified these genes as candidates for parthenocarpy.

Among the most noteworthy candidate genes, we identified an ‘auxin-responsive protein’ (VaccDscaff17-augustus-gene-355.30) ~42 kb downstream of marker 17_35452313 on chromosome 9 (Table 2). The encoded protein contains an Aux/IAA domain (IPR033389) and shares 47.12% sequence identity with Arabidopsis thaliana IAA27 (AT4G29080), an auxin-related gene involved in fertility and fruit enlargement (Fig. S5) [100]. Moreover, we observed a substantial shared amino acid sequence identity (61.34%) between VaccDscaff17-augustus-gene-355.30 and a previously characterized VcIAA27 (VaccDscaff30-augustus-gene-75.28). Shared conserved domains were also observed between these proteins, suggesting they could possess similar functions. Individuals containing a single copy of the alternate allele of this marker exhibited a mean parthenocarpic fruit set of 45%, while duplex, triplex, and quadruplex genotypes had an average of 13% (Fig. S6).

On the same chromosome, we also discovered a ‘knotted1-like homeodomain protein’ (VaccDscaff17-snap-gene-167.20) located ~85 kb downstream of the significant marker 17_16819984. Individuals that were homozygous for the reference allele had an average fruit set of 32%, more than twice that of all other genotypes.

In both populations, we identified significant marker–trait associations located within 2 Mb of each other on chromosome 6. In this region, we identified a marker 11_17667007, located 1786 bp upstream of a ‘calmodulin 60B-like protein’, named ‘VaccDscaff11-augustus-gene-177.31’. Genotypes duplex for the alternative allele exhibited a mean parthenocarpic fruit set of 50%, while other allelic dosages had a mean fruit set of 13%.

Evaluation of predictive breeding methods

To compare the efficacy of genome-wide and targeted molecular breeding approaches, we evaluated the PA of three models: GS, GSdnGWAS, and in silico MAS (Fig. 3). Overall, genomic prediction showed moderate accuracy, achieving a PA of 0.24 when trained and tested on the combined populations. However, performance declined slightly when the model was tested on the independent populations. In both interpopulation cross-validation schemes, GSdnGWAS outperformed GS methodology, although overall predictive abilities were moderate. In the first training–testing scenario, the inclusion of multiple markers as fixed effects resulted in the greatest improvement in PA, with four markers (21_21099188, 12_29690597, 2_34110377, and 2_36035768) exhibiting a PA of 0.36 ± 0.10 (+57%) compared to GS. In the reciprocal training–testing scenario, GSdnGWAS modestly improved PA to 0.21 (+5%) when using a single marker exhibiting a general gene action. However, including multiple markers as fixed effects decreased PA to 0.08 ± (−60%). MAS of both a single marker and a set of multiple markers resulted in higher predictive abilities than GS in the first training–testing scenario, with predictive abilities of 0.35 ± 0.08 (+52%) and 0.27 ± 0.11 (+17%), respectively. However, improvements in the reciprocal direction were not observed. Still, the PA obtained using a single general marker exhibited a PA of 0.07 ± 0.12, almost half that obtained by GS using the full marker set.

Figure 3.

Figure 3

Open in a new tab

PA of molecular breeding strategies. A bar plot showing PA of different breeding and cross-validation schemes. (A) Cross-validation of the GS model trained on the combined population and tested both within the combined population and across each independent population. (B) Interpopulation cross-validation GS, GSdnGWAS, and MAS by training on POP1 and predicting POP2. (C) Interpopulation cross-validation by training on POP2 and predicting POP1. GS includes the full marker panel of ~59 k SNPs, GSdnGWAS includes the top-performing additive (Add), dominance (Dom), or general (Gnr) markers from GWAS analyses. GSdnGWAS Set includes the best set of 2–5 markers from any genetic parameterization with the highest PA. MAS includes a single marker as a fixed effect, with no other marker information. MAS set is the PA of the top-performing marker set identified from GSdnGWAS tested within a MAS framework.

Discussion

Parthenocarpy is an important objective for pollination-dependent plant breeders as it reduces reliance on pollination services while securing yields in adverse weather conditions or pollinator-limited environments, such as controlled agriculture systems. Additionally, the facultative parthenocarpy exhibited by blueberry enables elite parthenocarpic genotypes to be both selected and used as parents in the breeding program due to the development of viable seeds upon cross-pollination. In the present study, we observed considerable variability and moderate heritability for natural parthenocarpy in a diverse southern highbush blueberry breeding program, underscoring the potential for integration of this trait into selection criteria. Moreover, through GWAS, we identified significant loci contributing to parthenocarpy, along with the candidate genes within the associated genomic regions. Finally, we compared three common molecular breeding approaches to enhance our understanding of their applicability within a breeding program.

Numerous studies have explored the genetic basis of parthenocarpy, revealing various mechanisms that range from simple to complex. Ehlendeldt and Vorsa (2007) proposed a simple genetic model for northern highbush blueberry, suggesting that parthenocarpy is controlled by a single recessive gene with incomplete dominance [101]. However, F2 segregation ratios showed poor fit to a tetrasomic inheritance model. Similar genetic models for parthenocarpy have been reported in cucumber, oil palm, eggplant, and tomato, with mapping studies revealing a small number of regulatory genes exhibiting nonadditive actions [36, 43, 45, 102, 103]. However, some studies indicate that parthenocarpy is instead a quantitative trait, where multiple loci contribute a minor fraction of the phenotypic variance [43, 48, 49]. Our experiment uncovered multiple population-specific marker–trait associations distributed throughout the genome, each accounting for <15% of the variability for parthenocarpic fruit set. Upon further investigation, we found that many of these markers exhibited differences in allele frequencies within each breeding population, limiting our ability to detect the same marker–trait associations across populations using a GWAS analysis. Despite this, subsequent cross-validation of molecular breeding models showed that these markers still enhanced the prediction of parthenocarpy across populations within the context of GSdnGWAS, although the extent of improvement depended on the training–testing populations. This finding further supports that the markers identified in GWAS analyses maintain some biological importance for the regulation of parthenocarpy, even though they explained only a small portion of phenotypic variance in each population.

Interesting candidate genes related to phytohormones, seed development, embryogenesis, and cell cycle regulation were also discovered from GWAS analyses (Table 2). We identified VaccDscaff17-augustus-gene-355.30, which belongs to the Aux/Indole-3-Acetic Acid (Aux/IAA) family, known for its pivotal role in auxin signaling during fruit development [95, 104]. Furthermore, mutants of AUX/IAA genes have been shown to induce parthenocarpy [94, 105, 106]. Phylogenetic analysis of Arabidopsis AUX/IAA proteins revealed that the blueberry candidate gene VaccDscaff17-augustus-gene-355.30 clustered closely with auxin-responsive protein IAA27 (AtIAA27) and VaccDscaff30-augustus-gene-75.28, a functionally validated VcIAA27. Importantly, these genes exhibit substantial similarity within the conserved domains essential for the functionality of AUX/IAA proteins. Recent studies on VcIAA27 demonstrated that heterologous overexpression in Arabidopsis led to the development of seedless siliques [100]. Moreover, in tomato, silencing of SlIAA27 resulted in higher auxin sensitivity and seedless fruit development [107], providing further evidence for the involvement of this gene in parthenocarpy.

A ‘Knotted1-like (KNOX) homeodomain gene’, VaccDscaff17-snap-gene-167.20, was also discovered. KNOX genes encode a group of transcription factors that negatively regulate gibberellic acid biosynthesis, affecting fruit set through the direct repression of gibberellin 20-oxidase (GA20ox) [108, 109]. In tomato, KNOX genes have been implicated for their involvement in parthenocarpy due to their decreased expression in parthenocarpic fruit (pat) mutants, which exhibit constitutive expression of GA20ox [91, 93]. An investigation of pat double mutants found that lines also overexpressing Tomato knotted 2 (Tkn2) showed reduced parthenocarpic ability [92]. This finding emphasizes the role of KNOX genes in the ‘GA-overdose’ phenotype attributed to pat mutants and contributes to the mechanistic understanding of facultative parthenocarpy.

Additionally, both populations revealed neighboring associations on chromosome 6. This region contained a candidate gene VaccDscaff11-augustus-gene-177.31, which encodes a ‘calmodulin-binding 60B-like’ protein. Calmodulin-like proteins (CML) have been shown to mediate plant responses to abiotic and biotic stresses, and auxin signaling [110–112]. Analysis of the ~3-kb promoter region of the cucumber CsCML25 revealed four MYB-binding sites, one WUS-binding site, and several cis-acting elements known to interact with auxins [113]. An experiment involving overexpression of this gene increased the rate of parthenocarpy by 86% [114]. Calmodulin-associated genes have also been associated with tomato parthenocarpy through differential gene expression analysis [76].

In the pursuit of an appropriate molecular breeding strategy, we next evaluated the PA of three marker-assisted approaches within a ridge–regression framework. Each approach varies in the molecular information it utilizes and is tailored to a specific genetic architecture [115]. For example, GS leverages genome-wide marker data and is well suited for complex traits influenced by numerous loci with small effects [116, 117]. On the other hand, GSdnGWAS provides the same marker information but incorporates a limited number of known QTL regions as fixed effects. This strategy prevents the shrinkage of these marker estimates and enhances the capture of genetic variance guided by biological priors [118]. Conversely, MAS relies on a single marker, or small set of markers, making it ideal for traits controlled by a few major-effect QTLs [119]. To our knowledge, the present study represents the first application of genomic prediction and the largest phenotypic dataset collected for facultative parthenocarpy among horticultural crops.

Results from GS on the combined populations yielded a baseline PA of 0.24, while interpopulation cross-validation exhibited predictive abilities of 0.23 and 0.20, for each training–testing scenario. These values are comparable to those obtained for various fruit quality traits in tetraploid blueberries [120], and other fruit trees [121, 122]. The inclusion of significant GWAS hits as fixed effects within a GSdnGWAS framework enhanced predictions of parthenocarpy in both training–testing scenarios, although the extent of these benefits varied depending on the GWAS hits utilized and the allele frequencies of these markers within the training–testing populations. For example, GSdnGWAS resulted in substantially higher predictive abilities when training on POP1 to predict POP2, compared to the reciprocal direction. This finding could be explained by the higher number of significant markers associated with parthenocarpy that were identified in the first population (29 versus 11 markers), and the greater strength of their signal. GSdnGWAS has shown similar improvements to the prediction of horticultural traits, such as blueberry volatile organic compounds [67], and of agronomic traits in rice [123] and wheat [124]. However, in these cases, the relative performance of GSdnGWAS models also varied depending on the species, population, the trait of interest, and its genetic architecture.

We observed similar population-specific outcomes using in silico MAS. When training on POP1 to predict POP2, we identified single markers and sets of markers that could predict parthenocarpic fruit set with greater accuracy than GS. However, when MAS models were fit in the reciprocal training–testing scenario, we observed substantially lower predictive abilities. For several markers identified in GWAS of POP1, we observed lower allele frequencies in POP2, which could have enabled more accurate predictions of this population. Additionally, we suspect that genotype-by-environment interactions could influence the expression of parthenocarpy, affecting our ability to detect common significant marker–trait associations in both populations. Still, one single marker identified in POP2 exhibited a PA of nearly half that of GS, demonstrating the relative importance of these markers in the regulation of parthenocarpy. Therefore, while GSdnGWAS provides the most robust prediction of this trait, the development of diagnostic marker assays could also support molecular breeding efforts without significant investments in genome-wide sequencing.

Conclusions

Through the utilization of various molecular breeding tools, we identified several loci and candidate genes for facultative parthenocarpy and determined the optimal predictive breeding strategy for its improvement. We show that although single markers may be useful to predict parthenocarpy in a MAS approach, optimal predictions were obtained using GSdnGWAS. These findings underscore the advantage of leveraging prior biological information from GWAS to improve the prediction of complex traits. Furthermore, we highlight the utility of small-effect QTL identified from GWAS in the prediction of new breeding populations. Overall, this work represents a step forward in our understanding of facultative parthenocarpy, exhibited by blueberry, and offers a robust prediction framework for implementation in modern breeding programs.

Materials and Methods

Population and plant materials

In this study, 505 southern highbush blueberry genotypes were analyzed. These genotypes were sourced from two groups of advanced selections developed as part of the clonal selection trials from the University of Florida Blueberry Breeding program planted in Waldo, FL, USA (29.792664, −82.132109). All individuals belong to the same breeding pool and are related by pedigree at different levels. For the validation of genomic analyses, these populations are designated as Population 1 (POP1) and Population 2 (POP2), and they include 311 and 194 individuals, respectively. Phenotypic evaluation of POP1 and POP2 was conducted during the 2021 and 2022 seasons when plants were 7 and 8 years of age, respectively. Additionally, 50 genotypes from POP1 were evaluated in 2022 to connect both datasets and correct for eventual confounding effects of the year and age of the plants.

Measuring parthenocarpic capacity

For each genotype, a single raceme was bagged with clear plastic mesh netting during the bud stage. Prior to anthesis, 10 flowers were emasculated by peeling back the corolla and removing stamens with a pair of forceps. All untreated flowers and remaining flower buds were removed, and the mesh bag was reinstalled to prevent pollination. Parthenocarpic capacity was measured as the proportion of ripe fruit set out of 10 flowers in response to emasculation and pollinator exclusion. All fruits were checked for seeds to ensure the absence of fertilization.

Genotypic data

Young leaves were collected from each genotype and submitted for DNA extraction and targeted sequence capture genotyping at RAPiD Genom10372ics (Gainesville, FL, USA), as described by Ferrão et al. [50]. A total of 10 K probes of 120-mer were used for target enrichment. Sequencing was conducted via 150-cycle paired-end runs using the IlluminaHiSeq2000 platform [51]. Trimmed and filtered reads were then aligned to the primary haplotype of the reference genome Vaccinium corymbosum cv. ‘Draper’ [52] using Mosaik v.2.2.3 software [53]. Variant calling was performed via FreeBayes v.1.3.2 [54], targeting the probe regions and considering a minimum mapping quality of 10. Single nucleotide polyphormisms (SNPs) were then filtered for biallelic markers only with a maximum of 50% missing data. Tetraploid allelic dosage was assigned based on the proportion of alternative and reference allele counts using the ‘updog’ R package [55]. Missing genotypes were imputed by the mean of each locus. The final SNP set included 59 912 markers that were used for all subsequent genome-wide association and genomic prediction analyses.

Phenotypic analysis

Best linear unbiased estimators (BLUEs) were used to connect phenotypic information across years, including both populations and considering checks, as follows: Inline graphic (1), where Inline graphic is the vector of phenotypes for individuals collected in POP1 and POP2, Inline graphic is the population phenotypic mean, and Inline graphic is a vector of the same dimension of Inline graphic being all elements equal to 1. Following a similar model reported by Pastina et al. [56] and De Lara et al. [57], we separated the genotypic effect Inline graphic into two groups, where Inline graphic is a fixed effect vector of checks connecting both years of evaluation with an incidence matrix (Inline graphic), Inline graphic is a fixed effect vector of test genotypes, with an incidence matrix of (Inline graphic), and Inline graphic is the fixed effect of the year, with an incidence matrix (Inline graphic). The residual term was assumed as Inline graphic, where Inline graphic is the identity matrix and Inline graphic is the error variance.

To estimate genetic parameters, we fit a similar model where the year and check genotypes were considered as fixed effects, and the genetic effect of the test genotypes was instead considered as a random effect. For the random genetic term, we computed a pedigree-based heritability value by assuming Inline graphic, where Inline graphic is the additive relationship matrix constructed using the ‘AGHmatrix’ R package [58] following statistical methodology for autotetraploid species from [59], and Inline graphic is the additive genetic variance. Narrow-sense heritability (h2) was computed by Inline graphic (2). All models were fitted using the ‘ASReml’ R package [60].

Genome-wide association studies

GWAS were conducted on each population independently and combined. Individual GWAS considered raw phenotypic values, and the combined analysis utilized BLUE values corrected for the year effect, as described in the previous section. For the GWAS analysis, a Q + K model was implemented using the GWASpoly R package [61]. The Q matrix incorporated the first five principal components as covariates to address population structure, and the K covariance matrix was generated using the leave-one-chromosome out (LOCO) method to account for polygenic effects. All tetraploid gene action models were tested, including additive, dominance (simplex, ‘1-dom’, and duplex, ‘2-dom’), and general models. Significance thresholds for marker–trait associations were adjusted using the ‘M.eff’ multiple-testing correction option, considering α = 0.05 and minimum minor allele frequency (MAF) of 1%. This method estimates an effective number of markers using linkage disequilibrium (LD) between markers [62]. Quantile–quantile (Q–Q) plots were generated to verify model fit, and LD decay was visualized using the ‘LD.plot’ function.

Candidate gene identification

Genomic windows of ±100 kb surrounding each significant marker discovered in the GWAS were mined for candidate genes. Gene homology and functional annotation data were gathered from NCBI BLAST, Gene Ontology (GO), KEGG pathway, Conserved Domain Database (CDD), and protein family database (PFAM) using the ‘PANZZER2’ [63] and eggNOG-mapper [64] software. Genes related to cell cycle, fruit set, fruit development, phytohormone production and signaling, and those previously identified in the literature to be involved in parthenocarpy were included.

To further investigate the structure and function of candidate genes, blueberry gene and protein sequences were obtained from the Genome Database for Vaccinium (https://www.vaccinium.org/), and homologous genes were obtained from NCBI (http://www.ncbi.nlm.nih.gov). Amino acid sequences were aligned using the MEGA 11 software [65].

Predictive breeding strategies (GS, GSdnGWAS, and MAS)

Three molecular breeding strategies were compared for their ability to predict parthenocarpic capacity: GS, GSdnGWAS, and in silico MAS. As a benchmark, GS was performed on the combined populations using random cross-validation. In all other scenarios, an interpopulation cross-validation scheme was used in which one population was used for model training, and the other for model testing. Additionally, we conducted cross-validation by training on the combined populations and testing on each independent population separately. To account for differences in population size, all models were fit using a random sample of 175 individuals from the training population to predict a random 50 individuals from the testing population. Cross-validation was repeated for 30 iterations.

All prediction models utilized a ridge–regression model to obtain genomic estimated breeding values. The following statistical model was considered: Inline graphic (1) where y is the vector of phenotypic observations, Inline graphicis the overall mean, Inline graphic is the vector of fixed effects with incidence matrix Inline graphic,  Inline graphic is the vector of random marker effects with incidence matrix Inline graphic, and Inline graphic is the residual vector with Inline graphic, where Inline graphic is the identity matrix, and Inline graphic is the residual variance. All models were implemented using the ‘rrBLUP’ R package [66]. For GS, genome-wide SNP markers were considered as random effects, and no fixed effects were included (Inline graphic). For GSdnGWAS, up to five SNP markers were included as a fixed effect, with all other markers as random effects. The rationale behind GSdnGWAS is to avoid shrinkage of those SNP effects identified as significant in GWAS, providing more flexibility for the estimation of markers with a larger effect. The selection of SNPs with significant effects from GWAS analyses for inclusion as predictors in GSdnGWAS models was determined through bidirectional stepwise regression. In this process, markers included in the final model were tested in all possible combinations in sets of two to five markers. In the MAS regression model, each significant marker from GWAS analyses was tested individually and considered as a fixed effect with no additional marker information (Inline graphic). Furthermore, the top-performing marker set from GSdnGWAS was also evaluated within a MAS framework. Importantly, to avoid overfitting, all tested markers for GSdnGWAS and MAS were derived from GWAS of the independent populations. Similar molecular testing strategies were reported by Ferrão et al. [67]. PAs were computed as the Pearson’s correlation between the predicted and observed phenotypic values.

Supplementary Material

Web_Material_uhaf086
web_material_uhaf086.zip (11.4MB, zip)

Acknowledgements

We are grateful to Straughn Farms LLC for their continued support of the University of Florida Blueberry Breeding Program. We also thank Ben Covert, Haley Dabbs, Andrew Komatz, Jordyn Lind, Taylor Sawyer, and Chaz Wilhoite for their assistance in phenotyping efforts. This work was supported by royalties from the University of Florida Blueberry Breeding program.

Contributor Information

Juliana Cromie, Blueberry Breeding and Genomics Laboratory, Horticultural Sciences Department, University of Florida, 2560 Hull Rd, Gainesville, FL 32611, USA.

Ryan P Cullen, Blueberry Breeding and Genomics Laboratory, Horticultural Sciences Department, University of Florida, 2560 Hull Rd, Gainesville, FL 32611, USA.

Camila Ferreira Azevedo, Blueberry Breeding and Genomics Laboratory, Horticultural Sciences Department, University of Florida, 2560 Hull Rd, Gainesville, FL 32611, USA; Statistics Department, Federal University of Viçosa, Av. Peter Henry Rolfs, s/n, Viçosa, MG 36570-900, Brazil.

Luis Felipe V Ferrão, Blueberry Breeding and Genomics Laboratory, Horticultural Sciences Department, University of Florida, 2560 Hull Rd, Gainesville, FL 32611, USA.

Felix Enciso-Rodriguez, Blueberry Breeding and Genomics Laboratory, Horticultural Sciences Department, University of Florida, 2560 Hull Rd, Gainesville, FL 32611, USA.

Juliana Benevenuto, Blueberry Breeding and Genomics Laboratory, Horticultural Sciences Department, University of Florida, 2560 Hull Rd, Gainesville, FL 32611, USA.

Patricio R Muñoz, Blueberry Breeding and Genomics Laboratory, Horticultural Sciences Department, University of Florida, 2560 Hull Rd, Gainesville, FL 32611, USA.

Author contributions

J.B. and P.R.M. designed and supervised the experiment. J.C. and R.P.C. led phenotyping efforts. J.B. and L.F.V.F. prepared the genomic dataset. J.C., L.F.V.F., C.F.A., and F.E.R. devised the methodology for data analysis. J.C. conducted the analysis, candidate gene mining, and wrote the paper with guidance and revisions from all authors.

Data availability

The genomic and phenotypic data are available as supplementary material.

Conflict of interest statement

The authors declare no competing interests.

Supplementary data

Supplementary data is available at Horticulture Research online.

References

  • 1. Aizen  MA, Aguiar  S, Biesmeijer  JC. et al.  Global agricultural productivity is threatened by increasing pollinator dependence without a parallel increase in crop diversification. Glob Chang Biol. 2019;25:3516–27 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2. Aizen  MA, Harder  LD. The global stock of domesticated honey bees is growing slower than agricultural demand for pollination. Curr Biol. 2009;19:915–8 [DOI] [PubMed] [Google Scholar]
  • 3. Smith  MR, Singh  GM, Mozaffarian  D. et al.  Effects of decreases of animal pollinators on human nutrition and global health: a modelling analysis. Lancet. 2015;386:1964–72 [DOI] [PubMed] [Google Scholar]
  • 4. Goulson  D, Nicholls  E, Botías  C. et al.  Bee declines driven by combined stress from parasites, pesticides, and lack of flowers. Science. 2015;347:347. [DOI] [PubMed] [Google Scholar]
  • 5. Potts  SG, Biesmeijer  JC, Kremen  C. et al.  Global pollinator declines: trends, impacts and drivers. Trends Ecol Evol. 2010;25:345–53 [DOI] [PubMed] [Google Scholar]
  • 6. Settele  J, Bishop  J, Potts  SG. Climate change impacts on pollination. Nat Plants. 2016;2:1–3 [DOI] [PubMed] [Google Scholar]
  • 7. Dhatt  AS, Kaur  G. Parthenocarpy: a potential trait to exploit in vegetable crops: a review. Agric Rev. 2016;37:300–8 [Google Scholar]
  • 8. Knapp  JL, Bartlett  LJ, Osborne  JL. Re-evaluating strategies for pollinator-dependent crops: how useful is parthenocarpy?  J Appl Ecol. 2017;54:1171–9 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Pandolfini  T, Molesini  B, Spena  A. Parthenocarpy in crop plants. In: Ostergaard  L (ed.), Fruit Development and Seed Dispersal. Oxford, UK: Wiley-Blackwell, 2009;38:326–45 [Google Scholar]
  • 10. Picarella  ME, Mazzucato  A. The occurrence of seedlessness in higher plants; insights on roles and mechanisms of parthenocarpy. Front Plant Sci. 2019;9:1997. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11. Brazelton  C. Global state of the blueberry industry report. International Blueberry Organization. Santiago, Chile, 2023. Available at:  https://www.internationalblueberry.org/2023-report [Google Scholar]
  • 12. Campbell  JW, Kimmel  CB, Bammer  M. et al.  Managed and wild bee flower visitors and their potential contribution to pollination services of low-chill highbush blueberry (Vaccinium corymbosum L.; Ericales: Ericaceae). J Econ Entomol. 2018;111:2011–6 [DOI] [PubMed] [Google Scholar]
  • 13. Dogterom  MH, Winston  ML, Mukai  A. Effect of pollen load size and source (self, outcross) on seed and fruit production in highbush blueberry cv. “Bluecrop” (Vaccinium corymbosum; Ericaceae). Am J Bot. 2000;87:1584–91 [PubMed] [Google Scholar]
  • 14. Mallinger  RE, Ternest  JJ, Weaver  SA. et al.  Importance of insect pollinators for Florida agriculture: a systematic review of the literature. Fla Entomol. 2021;104:222–9 [Google Scholar]
  • 15. Taber  SK, Olmstead  JW. Impact of cross- and self-pollination on fruit set, fruit size, seed number, and harvest timing among 13 southern highbush blueberry cultivars. HortTechnology. 2016;26:213–9 [Google Scholar]
  • 16. Benjamin  FE, Winfree  R. Lack of pollinators limits fruit production in commercial blueberry (Vaccinium corymbosum). Environ Entomol. 2014;43:1574–83 [DOI] [PubMed] [Google Scholar]
  • 17. Cavigliasso  P, Negri  P, Viel  M. et al.  Precision management of pollination services to blueberry crops. Sci Rep. 2021;11:1–8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. DeVetter  LW, Chabert  S, Milbrath  MO. et al.  Toward evidence-based decision support systems to optimize pollination and yields in highbush blueberry. Front Sustain Food Syst. 2022;6:1006201 [Google Scholar]
  • 19. Eeraerts  M, Devetter  LW, Batáry  P. et al.  Synthesis of highbush blueberry pollination research reveals region-specific differences in the contributions of honeybees and wild bees. J Appl Ecol. 2023;60:2528–39 [Google Scholar]
  • 20. Nicholson  CC, Ricketts  TH. Wild pollinators improve production, uniformity, and timing of blueberry crops. Agric Ecosyst Environ. 2019;272:29–37 [Google Scholar]
  • 21. Tuell  JK, Isaacs  R. Weather during bloom affects pollination and yield of highbush blueberry. J Econ Entomol. 2010;103:557–62 [DOI] [PubMed] [Google Scholar]
  • 22. Javorek  SK, Mackenzie  KE, Vander  SP. Comparative pollination effectiveness among bees (Hymenoptera: Apoidea) on lowbush blueberry (Ericaceae: Vaccinium angustifolium). Ann Entomol Soc Am. 2002;95:345–51 [Google Scholar]
  • 23. Ramírez-Mejía  AF, Lomáscolo  S, Blendinger  PG. Hummingbirds, honeybees, and wild insect pollinators affect yield and berry quality of blueberries depending on cultivar and farm’s spatial context. Agric Ecosyst Environ. 2023;342:108229 [Google Scholar]
  • 24. Sun  Q, Zhao  X, Wu  L. et al.  Differences in pollination efficiency among three bee species in a greenhouse and their effects on yield and fruit quality of northern highbush ‘Bluecrop’ blueberry. HortScience. 2021;56:603–7 [Google Scholar]
  • 25. Courcelles  DMM, Button  L, Elle  E. Bee visit rates vary with floral morphology among highbush blueberry cultivars (Vaccinium corymbosum L.). J Appl Entomol. 2013;137:693–701 [Google Scholar]
  • 26. Cromie  J, Ternest  JJ, Komatz  AP. et al.  Genotypic variation in blueberry flower morphology and nectar reward content affects pollinator attraction in a diverse breeding population. BMC Plant Biol. 2024;24:814–3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Ehlenfeldt  MK. Breeding blueberries for the new century. Acta Hortic. 1997;446:227–34 [Google Scholar]
  • 28. Ehlenfeldt  MK. Breeding for parthenocarpic fruit development in blueberry. 2012;12:261–8 [Google Scholar]
  • 29. Lyrene  PM. Some problems and opportunities in blueberry breeding. Acta Hortic. 1993;346:63–71 [Google Scholar]
  • 30. Joldersma  D, Liu  Z. The making of virgin fruit: the molecular and genetic basis of parthenocarpy. J Exp Bot. 2018;69:955–62 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31. Subbaraya  U, Rajendran  S, Simeon  S. et al.  Unravelling the regulatory network of transcription factors in parthenocarpy. Sci Hortic. 2020;261:108920 [Google Scholar]
  • 32. Ehlenfeldt  MK, Stringer  SJ. Evaluations of parthenocarpy in Vaccinium virgatum Aiton (Rabbiteye) blueberry cultivars. J Am Pomol Soc. 2018;72:2–11 [Google Scholar]
  • 33. MacKenzie  KE. Pollination requirements of three highbush blueberry (Vaccinium corymbosum L.) cultivars. J Am Soc Hortic Sci. 1997;122:891–6 [Google Scholar]
  • 34. Cullen  R, Cromie  J, Sawyer  T. et al.  Parthenocarpic fruit quality and production under pollinator-exclusion in southern highbush blueberry. Sci Hortic. 2024;328:112935 [Google Scholar]
  • 35. Miyashita  C, Koito  Y, Ogiwara  I. Utility of parthenocarpic interspecific hybrids between Vaccinium corymbosum and Vaccinium virgatum for breeding blueberry cultivars suitable for cluster harvesting. Hort J. 2019;88:180 [Google Scholar]
  • 36. Miyatake  K, Saito  T, Negoro  S. et al.  Development of selective markers linked to a major QTL for parthenocarpy in eggplant (Solanum melongena L.). Theor Appl Genet. 2012;124:1403–13 [DOI] [PubMed] [Google Scholar]
  • 37. Shinozaki  Y, Ezura  K. Tomato fruit set and its modification using molecular breeding techniques. In: Ezura  H, Ariizumi  T, Garcia-Mas  J, Rose  J (eds.), Biotechnology in Agriculture and Forestry. Berlin, Heidelberg: Springer, 2016, 93–112 [Google Scholar]
  • 38. Tian  S, Zhang  Z, Qin  G. et al.  Parthenocarpy in Cucurbitaceae: advances for economic and environmental sustainability. Plan Theory. 2023, 2023;12:3462. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39. Wang  X, Liu  Z, Zhang  F. et al.  Integrative genomics reveals the polygenic basis of seedlessness in grapevine. Curr Biol. 2024;34:3763–3777.e5 [DOI] [PubMed] [Google Scholar]
  • 40. Klap  C, Yeshayahou  E, Bolger  AM. et al.  Tomato facultative parthenocarpy results from SlAGAMOUS-LIKE 6 loss of function. Plant Biotechnol J. 2017;15:634–47 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41. Martí  C, Orzáez  D, Ellul  P. et al.  Silencing of DELLA induces facultative parthenocarpy in tomato fruits. Plant J. 2007;52:865–76 [DOI] [PubMed] [Google Scholar]
  • 42. Beraldi  D, Picarella  ME, Soressi  GP. et al.  Fine mapping of the parthenocarpic fruit (pat) mutation in tomato. Theor Appl Genet. 2004;108:209–16 [DOI] [PubMed] [Google Scholar]
  • 43. Devi  S, Sharma  PK, Behera  TK. et al.  Identification of a major QTL, Parth6.1 associated with parthenocarpic fruit development in slicing cucumber genotype, Pusa Parthenocarpic Cucumber-6. Front Plant Sci. 2022;13:1064556. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44. Gorguet  B, Eggink  PM, Ocaña  J. et al.  Mapping and characterization of novel parthenocarpy QTLs in tomato. Theor Appl Genet. 2008;116:755–67 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45. Lietzow  CD, Zhu  H, Pandey  S. et al.  QTL mapping of parthenocarpic fruit set in North American processing cucumber. Theor Appl Genet. 2016;129:2387–401 [DOI] [PubMed] [Google Scholar]
  • 46. Wu  Z, Zhang  T, Li  L. et al.  Identification of a stable major-effect QTL (Parth 2.1) controlling parthenocarpy in cucumber and associated candidate gene analysis via whole genome re-sequencing. BMC Plant Biol. 2016;16:182–14 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47. Yoshioka  Y, Nashiki  A, Zaelani  A. et al.  Identification of major quantitative trait loci for parthenocarpic ability in East Asian melon. bioRxiv. 2023;2023.02.16.528896 [Google Scholar]
  • 48. Gou  C, Zhu  P, Meng  Y. et al.  Evaluation and genetic analysis of parthenocarpic germplasms in cucumber. Genes (Basel). 2022;13:225. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49. Sun  Z, Staub  JE, Chung  SM. et al.  Identification and comparative analysis of quantitative trait loci associated with parthenocarpy in processing cucumber. Plant Breed. 2006;125:281–7 [Google Scholar]
  • 50. Ferrão  LFV, Amadeu  RR, Benevenuto  J. et al.  Genomic selection in an outcrossing autotetraploid fruit crop: lessons from blueberry breeding. Front Plant Sci. 2021;12:1075. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51. Benevenuto  J, Ferrão  LFV, Amadeu  RR. et al.  How can a high-quality genome assembly help plant breeders?  Gigascience. 2019;8:1–4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52. Colle  M, Leisner  CP, Wai  CM. et al.  Haplotype-phased genome and evolution of phytonutrient pathways of tetraploid blueberry. GigaScience, 2019;8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53. Lee  WP, Stromberg  MP, Ward  A. et al.  MOSAIK: a hash-based algorithm for accurate next-generation sequencing short-read mapping. PLoS One. 2014;9:e90581 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54. Garrison  E, Marth  G. Haplotype-based variant detection from short-read sequencing. arxiv [pre-print]. 2012 [Google Scholar]
  • 55. Gerard  D, Ferrão  LFV, Garcia  AAF. et al.  Genotyping Polyploids from Messy Sequencing Data. Genetics, 2018;210:789–807 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 56. Pastina  MM, Malosetti  M, Gazaffi  R. et al.  A mixed model QTL analysis for sugarcane multiple-harvest-location trial data. Theor Appl Genet. 2012;124:835–49 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 57. De Lara  LAC, Santos  MF, Jank  L. et al.  Genomic selection with allele dosage in Panicum maximum Jacq. G3 Genes|Genomes|Genetics. 2019;9:2463–75 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 58. Amadeu, R. R., Garcia, A. A. F., Munoz, P. R., & Ferrão, L. F. V. (2023). AGHmatrix: genetic relationship matrices in R. Bioinformatics, 39(7). https://doi.org/10.1093/bioinformatics/btad445 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 59. Kerr  RJ, Li  L, Tier  B. et al.  Use of the numerator relationship matrix in genetic analysis of autopolyploid species. Theor Appl Genet. 2012;124:1271–82 [DOI] [PubMed] [Google Scholar]
  • 60. Butler  DG, Cullis  BR, Gilmour  AR. et al.  ASReml-R Reference Manual Version 4.2 ASReml Estimates Variance Components Under a General Linear Mixed Model by Residual Maximum Likelihood (REML). Hemel Hempstead, HP2 4TP, UK: VSN International Ltd., 2023 [Google Scholar]
  • 61. Rosyara  UR, De Jong  WS, Douches  DS. et al.  Software for genome-wide association studies in autopolyploids and its application to potato. Plant Genome. 2016;9: [DOI] [PubMed] [Google Scholar]
  • 62. Moskvina  V, Schmidt  KM. On multiple-testing correction in genome-wide association studies. Genet Epidemiol. 2008;32:567–73 [DOI] [PubMed] [Google Scholar]
  • 63. Törönen  P, Medlar  A, Holm  L. PANNZER2: a rapid functional annotation web server. Nucleic Acids Res. 2018;46:W84–8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 64. Huerta-Cepas  J, Forslund  K, Coelho  LP. et al.  Fast genome-wide functional annotation through orthology assignment by eggNOG-mapper. Mol Biol Evol. 2017;34:2115–22 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 65. Tamura  K, Stecher  G, Kumar  S. MEGA11: molecular evolutionary genetics analysis version 11. Mol Biol Evol. 2021;38:3022–7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 66. Endelman  JB. Ridge regression and other kernels for genomic selection with R package rrBLUP. Plant Genome. 2011;4:250–5 [Google Scholar]
  • 67. Ferrão  LFV, Johnson  TS, Benevenuto  J. et al.  Genome-wide association of volatiles reveals candidate loci for blueberry flavor. New Phytol. 2020;226:1725–37 [DOI] [PubMed] [Google Scholar]
  • 68. Quesada  MA, Sánchez-Roldán  C, Heredia  A. et al.  Peroxidase and IAA oxidase activities and peroxidase isoenzymes in the pericarp of seeded and seedless “Redhaven” peach fruit. J Plant Growth Regul. 1992;11:1–6 [Google Scholar]
  • 69. Li  Y, Chen  X, Jingquan  Y. Changes in activities of ascorbate peroxidase and ascorbate oxidase in N-(2-Chloropyridyl)-N’-Phenylurea(CPPU)-induced parthenocarpic fruits of Lagenaria leucantha. Acta Hortic. 2008;105–112:105–12 [Google Scholar]
  • 70. Tsaniklidis  G, Delis  C, Nikoloudakis  N. et al.  L-ascorbic acid metabolism in parthenocarpic and seeded cherry tomatoes. Plant Growth Regul. 2014;72:141–53 [Google Scholar]
  • 71. Mubarok  S, Nuraini  A, Hamdani  JS. et al.  Antioxidative response of parthenocarpic tomato, iaa9-3 and iaa9-5, under heat stress condition. Plant Physiol Biochem. 2024;207:108333 [DOI] [PubMed] [Google Scholar]
  • 72.72  Melnick  RL, Marelli  JP, Sicher  RC. et al.  The interaction of Theobroma cacao and Moniliophthora perniciosa, the causal agent of witches’ broom disease, during parthenocarpy. Tree Genet Genomes. 2012;8:1261–79 [Google Scholar]
  • 73. Huang  C, Wang  Q, Zhang  C. et al.  Identification and expression analysis of autophagy-related gene family associated with embryonic development in seeded and seedless grapes (Vitis vinifera). N Z J Crop Hortic Sci. 2023;1–18 [Google Scholar]
  • 74. Martinelli  F, Uratsu  SL, Reagan  RL. et al.  Gene regulation in parthenocarpic tomato fruit. J Exp Bot. 2009;60:3873–90 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 75. Meng  J, Li  Y, Xu  J. et al.  Cloning and expression analysis of the Expansin gene CsEXPb1 in cucumber. Acta Hortic Sin. 2015;42:679 [Google Scholar]
  • 76. Ruiu  F, Picarella  ME, Imanishi  S. et al.  A transcriptomic approach to identify regulatory genes involved in fruit set of wild-type and parthenocarpic tomato genotypes. Plant Mol Biol. 2015;89:263–78 [DOI] [PubMed] [Google Scholar]
  • 77. Ito  T, Meyerowitz  EM. Overexpression of a gene encoding a cytochrome P450, CYP78A9, induces large and seedless fruit in Arabidopsis. Plant Cell. 2000;12:1541–50 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 78. Sotelo-Silveira  M, Cucinotta  M, Chauvin  AL. et al.  Cytochrome P450 CYP78A9 is involved in Arabidopsis reproductive development. Plant Physiol. 2013;162:779–99 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 79. Gupta  SK, Barg  R, Arazi  T. Tomato agamous-like6 parthenocarpy is facilitated by ovule integument reprogramming involving the growth regulator KLUH. Plant Physiol. 2021;185:969–84 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 80. Zhang  H, Han  W, Wang  H. et al.  Downstream of GA4, PbCYP78A6 participates in regulating cell cycle-related genes and parthenogenesis in pear (Pyrus bretshneideri Rehd.). BMC Plant Biol. 2021;21:292–13 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 81. Wang  L, Zhang  SL, Jiao  C. et al.  QTL-seq analysis of the seed size trait in grape provides new molecular insights on seedlessness. J Integr Agric. 2022;21:2910–25 [Google Scholar]
  • 82. Cheng  C, Jiao  C, Singer  SD. et al.  Gibberellin-induced changes in the transcriptome of grapevine (Vitis labrusca × V. vinifera) cv. Kyoho flowers. BMC Genomics. 2015;16:1–16 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 83. Liu  W, Chen  M, Bai  L. et al.  Comprehensive transcriptomics and proteomics analyses of pollinated and parthenocarpic litchi (Litchi chinensis Sonn.) fruits during early development. Sci Rep. 2017, 2017;7:1–16 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 84. Liu  Y, Sun  J, Zhang  M. et al.  Identification of key genes related to seedlessness by genome-wide detection of structural variation and transcriptome analysis in ‘Shijiwuhe’ pear. Gene. 2020;738:144480 [DOI] [PubMed] [Google Scholar]
  • 85. Jia  Y, Wu  YP, Wang  FW. et al.  Full-length transcriptome sequencing analysis and characterization of gene isoforms involved in flavonoid biosynthesis in the seedless kiwifruit cultivar ‘Chengxiang’ (Actinidia arguta). Diversity (Basel). 2022;14:424 [Google Scholar]
  • 86. Yin  P, Ding  W, Zhang  H. et al.  Morphological, physiological and molecular characteristics of the seedless ‘Hongjiangcheng’ sweet orange. Hortic Plant J. 2023;9:437–49 [Google Scholar]
  • 87. Jung  CJ, Hur  YY, Jung  SM. et al.  Transcriptional changes of gibberellin oxidase genes in grapevines with or without gibberellin application during inflorescence development. J Plant Res. 2014;127:359–71 [DOI] [PubMed] [Google Scholar]
  • 88. Pomares-Viciana  T, Del Río-Celestino  M, Román  B. et al.  First RNA-seq approach to study fruit set and parthenocarpy in zucchini (Cucurbita pepo L.). BMC Plant Biol. 2019;19:61–20 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 89. Domingos  S, Fino  J, Paulo  OS. et al.  Molecular candidates for early-stage flower-to-fruit transition in stenospermocarpic table grape (Vitis vinifera L.) inflorescences ascribed by differential transcriptome and metabolome profiles. Plant Sci. 2016;244:40–56 [DOI] [PubMed] [Google Scholar]
  • 90. Nishitani  C, Yamaguchi-Nakamura  A, Hosaka  F. et al.  Parthenocarpic genetic resources and gene expression related to parthenocarpy among four species in pear (Pyrus spp.). Sci Hortic. 2012;136:101–9 [Google Scholar]
  • 91. Olimpieri  I, Siligato  F, Caccia  R. et al.  Tomato fruit set driven by pollination or by the parthenocarpic fruit allele are mediated by transcriptionally regulated gibberellin biosynthesis. Planta. 2007;226:877–88 [DOI] [PubMed] [Google Scholar]
  • 92. Mazzucato  A, Picarella  ME, Ruiu  F. The interaction bewteen parthenocarpic fruit and curl indicates a role of class I KNOX genes in tomato fruit set. In: Proceedings of the 56th Italian Society of Agricultural Genetics Annual Congress. 2012; 51
  • 93. Tang  N, Deng  W, Hu  G. et al.  Transcriptome profiling reveals the regulatory mechanism underlying pollination dependent and parthenocarpic fruit set mainly mediated by auxin and gibberellin. PLoS One. 2015;10:e0125355 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 94. Goetz  M, Hooper  LC, Johnson  SD. et al.  Expression of aberrant forms of AUXIN RESPONSE FACTOR8 stimulates parthenocarpy in Arabidopsis and tomato. Plant Physiol. 2007;145:351–66 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 95. Kumar  R, Khurana  A, Sharma  AK. Role of plant hormones and their interplay in development and ripening of fleshy fruits. J Exp Bot. 2014;65:4561–75 [DOI] [PubMed] [Google Scholar]
  • 96. Su  L, Rahat  S, Ren  N. et al.  Cytokinin and auxin modulate cucumber parthenocarpy fruit development. Sci Hortic. 2021;282:110026 [Google Scholar]
  • 97. Liu  L, Wang  Z, Liu  J. et al.  Histological, hormonal and transcriptomic reveal the changes upon gibberellin-induced parthenocarpy in pear fruit. Hortic Res. 2018a, 2018;5:1–13 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 98. Liu  R, How-Kit  A, Stammitti  L. et al.  A DEMETER-like DNA demethylase governs tomato fruit ripening. Proc Natl Acad Sci USA. 2015;112:10804–9 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 99. Li  Y, Zhang  S, Dong  R. et al.  The grapevine homeobox gene VvHB58 influences seed and fruit development through multiple hormonal signaling pathways. BMC Plant Biol. 2019;19:1–18 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 100. Hou  Y, Li  H, Zhai  L. et al.  Identification and functional characterization of the Aux/IAA gene VcIAA27 in blueberry. Plant Signal Behav. 2020;15:1700327 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 101. Ehlenfeldt, M. K., & Vorsa, N. (2007). Inheritance Patterns of Parthenocarpic Fruit Development in Highbush Blueberry (Vaccinium corymbosum L.). HortScience, 42(5), 1127–1130. https://doi.org/10.21273/hortsci.42.5.1127 [Google Scholar]
  • 102. Barcelos  E, De Almeida Rios  S, Cunha  RNV. et al.  Oil palm natural diversity and the potential for yield improvement. Front Plant Sci. 2015;6:1–16 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 103. Damayanti  F, Lombardo  F, Masuda  JI. et al.  Functional disruption of the tomato putative ortholog of HAWAIIAN SKIRT results in facultative parthenocarpy, reduced fertility and leaf morphological defects. Front Plant Sci. 2019;10:472730 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 104. Liu  M, Chen  Y, Chen  Y. et al.  The tomato ethylene response factor Sl-ERF.B3 integrates ethylene and auxin signaling via direct regulation of Sl-Aux/IAA27. New Phytol. 2018b;219:631–40 [DOI] [PubMed] [Google Scholar]
  • 105. Su  LY, Audran  C, Bouzayen  M. et al.  The Aux/IAA, Sl-IAA17 regulates quality parameters over tomato fruit development. Plant Signal Behav. 2015;10:e1071001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 106. Wang  H, Schauer  N, Usadel  B. et al.  Regulatory features underlying pollination-dependent and -independent tomato fruit set revealed by transcript and primary metabolite profiling. Plant Cell. 2009;21:1428–52 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 107. Bassa  C, Mila  I, Bouzayen  M. et al.  Phenotypes associated with down-regulation of Sl-IAA27 support functional diversity among Aux/IAA family members in tomato. Plant Cell Physiol. 2012;53:1583–95 [DOI] [PubMed] [Google Scholar]
  • 108. Hay  A, Tsiantis  M. KNOX genes: versatile regulators of plant development and diversity. Development. 2010;137:3153–65 [DOI] [PubMed] [Google Scholar]
  • 109. Tanaka-Ueguchib  M, Itoh  H, Oyama  N. et al.  Over-expression of a tobacco homeobox gene, NTH15, decreases the expression of a gibberellin biosynthetic gene encoding GA 20-oxidase. Plant J. 1998;15:391–400 [DOI] [PubMed] [Google Scholar]
  • 110. Dieter  P. Calmodulin and calmodulin-mediated processes in plants. Plant Cell Environ. 1984;7:371–80 [Google Scholar]
  • 111. Galon  Y, Snir  O, Fromm  H. How calmodulin binding transcription activators (CAMTAs) mediate auxin responses. Plant Signal Behav. 2010;5:1311–4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 112. Yang  T, Poovaiah  BW. Molecular and biochemical evidence for the involvement of calcium/calmodulin in auxin action. J Biol Chem. 2000;275:3137–43 [DOI] [PubMed] [Google Scholar]
  • 113. Wang  X, Yang  F, Zhu  P. et al.  Cloning and functional analysis of the promoter of CsCML25 gene related to determinate growth in cucumber. J Nanjing Agric Univ. 2020;43:621–8 [Google Scholar]
  • 114. Liu  J, Zhang  L, Meng  Y. et al.  Study on cucumber parthenocarpy induced by over-expressing of calmodulin-like protein gene CML25-like. J Nanjing Agric Univ. 2019;42:421–9 [Google Scholar]
  • 115. Bernardo  R. Breeding for Quantitative Traits in Plants. 3rd ed. Woodbury, MN: Stemma Press; 2020: [Google Scholar]
  • 116. Goddard  ME, Hayes  BJ. Genomic selection. J Anim Breed Genet. 2007;124:323–30 [DOI] [PubMed] [Google Scholar]
  • 117. Meuwissen  THE, Hayes  BJ, Goddard  ME. Prediction of total genetic value using genome-wide dense marker maps. Genetics. 2001;157:1819–29 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 118. Bernardo  R. Genomewide selection when major genes are known. Crop Sci. 2014;54:68–75 [Google Scholar]
  • 119. Dekkers  JC. Commercial application of marker- and gene-assisted selection in livestock: strategies and lessons. J Anim Sci. 2004;82 E-Suppl:E313–28 [DOI] [PubMed] [Google Scholar]
  • 120. de  Bem Oliveira  I, Amadeu  RR, Ferrão  LFV. et al.  Optimizing whole-genomic prediction for autotetraploid blueberry breeding. Heredity. 2020;125:437–48 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 121. Minamikawa  MF, Kunihisa  M, Moriya  S. et al.  Genomic prediction and genome-wide association study using combined genotypic data from different genotyping systems: application to apple fruit quality traits. Hortic Res. 2024;11:uhae131 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 122. Sun  M, Zhang  M, Kumar  S. et al.  Genomic selection of eight fruit traits in pear. Hortic Plant J. 2024;10:318–26 [Google Scholar]
  • 123. Spindel  JE, Begum  H, Akdemir  D. et al.  Genome-wide prediction models that incorporate de novo GWAS are a powerful new tool for tropical rice improvement. Heredity. 2016, 2016;116:395–408 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 124. Haile  TA, Walkowiak  S, N’Diaye  A. et al.  Genomic prediction of agronomic traits in wheat using different models and cross-validation designs. Theor Appl Genet. 2021;134:381–98 [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Web_Material_uhaf086
web_material_uhaf086.zip (11.4MB, zip)

Data Availability Statement

The genomic and phenotypic data are available as supplementary material.

Articles from Horticulture Research are provided here courtesy of Oxford University Press

RESOURCES