This study aimed at elucidating the genetic basis of blackspot bruising in potato tubers in combination with several other important agronomical traits. The approach was to associate DNA polymorphisms at multiple functional candidate loci with natural phenotypic variation in advanced tetraploid varieties and breeding materials. We identified highly significant (P < 0.001) associations between DNA variants at candidate as well as non-candidate (SSR) loci for tuber bruising, specific gravity and yield. To the best of our knowledge, the association mapping experiment described in this paper is amongst most comprehensive and first that investigates the molecular basis of the natural variation of enzymatic browning in plants. Association genetics in model and non-model plants substantially improved the understanding of the molecular basis of various complex plant phenotypes and their impact on plant performance and economically essential traits [35, 40–44]. The human genetics field provides the models how association studies influence the development of novel strategies in medicine [45–48]. In crop genetics, the combination of genomic tools with phenotypic evaluation as used in the traditional breeding process can identify molecular markers useful in "Marker-Assisted Selection" (MAS) and at the same time genetic factors that influence complex traits such as pathogen resistance and crop quality [35, 38, 49, 50].
The experimental design (Additional file. 5) comprised the phenotypic evaluation of tuber bruising, specific gravity, yield and shape as well as plant maturity with standard methods used for variety selection at six different locations in two consecutive years. Similar to a previous association mapping experiment , proprietary breeding clones were evaluated at the corresponding singular breeding station, whereas a common set of potato varieties was evaluated at all six breeding stations, which facilitated joint data analysis. This experimental design makes possible to perform studies including proprietary breeding materials that otherwise would not be available, insuring that the associations found are relevant for elite germplasm. Marker-trait associations were detected using a mixed model approach, which took into account field design, environment, genotype by environment interaction, population structure and kinship. An SSR marker based genetic similarity matrix was included in the mixed model to correct for population substructure that was not detectable by principal coordinate analysis. Stepwise forward regression led to a combination of associated markers that collectively explained between 7.5 and 38.2 percent of the genetic variation of the evaluated traits.
The phenotypic data analysis during this study led to several correlations that were observed between the investigated traits. In this respect, the strong positive correlation between bruising susceptibility and specific gravity which is well known in breeding practice was confirmed. In addition, the effects of all markers associated with both traits always had the same direction. A comparable connection between tuber traits exists in the case of specific gravity and chip color, the latter being determined by the amount of reducing sugars in the tubers . One explanation for the correlation between tuber bruising susceptibility and SG is close genetic linkage between different genes that control both traits independently. This might be the case for some QTL but is rather unlikely to be the case for all QTL with concerted effects on both traits that were tagged by the same marker. An alternative explanation concerning the impact of specific gravity upon bruising might be the interaction of starch grains with the surrounding internal membrane ultrastructure. Thus, a high amyloplastic load with starch grains provokes a faster and easier rupture of amyloplastic membranes upon mechanical stress. The loss of membrane integrity results on the one hand in the release of PPO and on the other hand in a mix-up of cytoplasmic and vacuolar compounds leading to the encounter of PPO with phenolic substrates [51–53]. Indeed, associations with tuber bruising and specific gravity were identified with markers derived from PPO genes as well as starch phosphorylase and lipase genes in concordance with this hypothesis. L-type starch phosphorylases (PHO1) are starch degrading enzymes present in plastids. The gene PHO1A (Stp23) on potato chromosome III is expressed in the amyloplasts, whereas PHO1B (StpL) on chromosome V is expressed mainly in the chloroplasts . The net amount of starch in the plastids is the result of anabolic and katabolic enzymatic reactions, and carbon flux from source leaves to sink tubers. Markers originating from PHO1A increased mainly bruising susceptibility and to less extent specific gravity. Vice versa, markers derived from PHO1B were associated primarily with specific gravity and secondary with bruising. The effects of PHO1A and PHO1B allelic variants on SG verify associations found previously at these two loci in an independent association mapping experiment . The observed concerted allele effects on bruising susceptibility and specific gravity also fit the hypothesis outlined above, which predicts that alleles increasing SG will indirectly increase bruising susceptibility due to a higher load of starch grains and thereby higher mechanical stress susceptibility. Sequence differences both at DNA and/or amino acid level could alter biochemical or molecular characteristics of starch phosphorylases and in consequence lead to variation in specific gravity of genotypes characterized by a given allelic composition.
Polyphenol oxidases are the major functional candidate genes for the enzymatic browning reaction. Accordingly, markers derived from at least three paralogous PPO genes clustered at the PPO locus on potato chromosome VIII were associated with tuber bruising. The markers POLOXA and POT32PS1-f were highly significantly associated with SCB, which makes them particularly interesting for MAS aiming at increased bruising resistance not compromised by low specific gravity of tubers. The association of the POLOXA marker with increased bruising susceptibility (BI and SCB) is especially intriguing, regarding the reported tissue specific expression of the POTP1 gene that is high in leaf tissue but restricted in tubers . Sequencing the POLOXA amplicon revealed that the corresponding primer pair tags more than one PPO gene or allele (unpublished data). The observed association might therefore result from a PPO gene other than POTP1, which is expressed in tubers. Selection against the rather frequent (35%) POLOXA marker should improve the average bruising resistance in a breeding population. The single strand conformation polymorphism (SSCP) fragment POT32PS1-f, was derived from the POT32 gene by subjecting the CAPS marker for allele 1 described by Werij
et al.  to single strand conformation analysis. POT32 allele 1 co-localized with a large effect QTL for enzymatic discoloration on chromosome VIII and its presence increased enzymatic discoloration . The fact that SCB is virtually uncoupled from the structural components of tuber bruising makes it more comparable to enzymatic discoloration catalyzed by PPO. The POT32PS1-f marker decreased SCB and therefore likely represents a novel PPO allele that was detected due to the higher resolution of SSCP analysis and the use of tetraploid breeding material instead of the diploid mapping population employed by Werij et al. .
Furthermore, a particularly interesting candidate gene is a putative class III lipase, which has been discovered de novo by comparing the tuber proteome of bruising resistant and susceptible cultivars (Urbany et al., in preparation). SSCP markers derived from this gene showed significant associations with either BI, SCB and TS (LIPIII-27-1h) or BI and SG (LIPIII-27-1e) or SG and TY (LIPIII-27-a) (Table 4). The marker LIPIII-27-1h, present in 80% of the tetraploid individuals, was associated with reduced SCB and a trend towards round shaped tubers. A functional role of lipases in tuber bruising, specific gravity or shape is not obvious, apart from membrane modification by altering lipid composition. Recent work on the closest homolog of potato class III lipases, the CaPLA1 gene encoding a hot pepper phospholipase A1, points to a possible connection with tuber bruising. Seo et al.  proposed that CaPLA1 is involved in the regulation of cell shape and number as well as the control of carbon flux through gluconeogenesis and β-oxidation. In addition, it was postulated that the enzyme is involved in lipid signalling and thereby regulates cellular and biochemical functions in heterotrophic plant tissue. The CaPLA1 homologous potato unigene LIPIII-27 (SGN-U269327) might fulfil analogous functions in heterotrophic tuber tissue, thereby indirectly influencing tuber bruising susceptibility and specific gravity. A putative role in the regulation of cell shape and number could explain why the marker LIPIII-27-1h associates with tuber shape as well.
The remaining candidate markers associated with tuber bruising and specific gravity can be classified as genes involved in polyphenol synthesis (St4Cl, Hqt, Hct, C3h, C4h), redox homeostasis (Cat, Zep) and general carbohydrate and energy metabolism (PHA1, AGPaseB).
Populations of cultivated, tetraploid potato are characterized by large haplotype blocs, resulting from a limited number of meiotic recombination events separating the individuals and/or selection . This is demonstrated by the finding of highly significant associations with SSR markers that did not encode any candidate gene. Large haplotype blocs are on the one hand favourable for the identification of diagnostic markers for breeding purposes but on the other hand prevent the verification of candidate genes by association genetics. By using linear regression for the identification of significant marker trait association, marker redundancy could be limited. The results concerning bruising validate the main candidate associations for polymorphic fragments derived from starch phosphorylase and 4-coumarate CoA ligase genes as well as the non-coding SSR locus StI013 previously identified by the mixed model approach (Table 5). PPO markers significantly associated with BI using mixed model statistics were also tagged by linear regression when the probability threshold was increased to P < 0.005 (data not shown). In the case of SG, markers with a significant impact on this trait by linear regression originate either from the PHO1B gene or the non coding SSR locus StI013, which also influenced BI. Regarding tuber yield and SCB, linear regression resulted in the repeated selection of significantly associated marker fragments from the L-galactono-1,4-lactone dehydrogenase and hydroxycinnamoyl CoA quinate transferase genes, respectively. These results strongly support the impact of candidate genes identified by both association analysis and linear regression on the analyzed traits. The ambiguity whether a candidate gene variant is causal for the trait effect or whether the association observed is due to LD with the causal gene has to be resolved by further functional analysis of candidate gene alleles. Candidate gene alleles associated with positive or negative trait values can be isolated and functionally characterized in heterologous model systems such as yeast, Arabidopsis or tobacco [38, 57–60].
Multiple correlations between traits were evident in the investigated population. Most of these correlations can be explained on a physiological or genetic basis, as described above for tuber bruising susceptibility and specific gravity. Plant maturity was correlated with specific gravity, yield and bruising susceptibility. This might be due to a longer vegetation period (later maturity) that results consequently in higher tuber yield and specific gravity, and indirectly in higher bruising. The reason for the negative correlation between tuber shape and bruising is probably physical. The surface of round shaped tubers is more exposed to mechanical impact than oblong tubers.
The interconnection of the traits investigated and the associated marker fragments identified are a valuable resource for breeders and molecular biologists dealing with enzymatic tissue discoloration and in particular potato tuber bruising. Functional studies will have to provide further evidence on whether and how sequence diversity of candidate genes results in functional differences and thereby explains phenotypic variation. The present work is the result of a collaboration between the breeding industry and basic research and accentuates the efforts to transfer the genetic and molecular knowledge gained during such studies into the field.