Internal ethylene concentration (IEC) in reduced sorbitol genotypes (A4 and A10) compared with that in the wild type (WT) at different storage times with and without 1-methycyclopropene (1-MCP) application. Error bars represent 1 standard error of the mean. Significance letters for each time point were determined using a post hoc Tukey test of the log-adjusted averages of each sample with P < 0.05 used as the cutoff.
Sugar concentrations for surface peel swabs (A) and ground fresh tissue (B) for apple genotypes A4 and A10 compared with those of the wild type (WT). *Significant (P < 0.05) difference from the ck genotype, which was determined using a post hoc Tukey test. Error bars represent 1 standard error of the mean.
Principal coordinate analyses (PCoAs) of bacterial and fungal microbiomes for both years (A, B) and separated by year (C–F). Each PCoA was performed on Bray-Curtis dissimilarity matrices. The 95% confidence ellipses are shown for each year and genotype (A4 and A10) and compared with the wild type (WT).
Shannon diversity of bacterial and fungal communities on the apple surface for each year across genotypes A4 and A10 compared with the wild type (WT). Error bars represent 1 standard error of the mean. Significance letters for each time point were determined using a post hoc Tukey test, with P < 0.05 as a significance threshold.
Genus-level relative abundance for bacterial and fungal communities in each year grouped by genotype and sampling time. Genera with less than 1% abundance are grouped under “Other.”
Click on author name to view affiliation information
Fruit microbiomes are capable of protecting their hosts from harmful pathogens and aiding in biocontrol; therefore, it is important to understand how differences in host genotype shape fruit microbial communities. The fruit species and even cultivars within a species can harbor different fruit microbiomes, but it has been difficult to establish how a single host gene can shape the microbiome structure. We investigated two genotypes of ‘Greensleeves’ apples with reduced sorbitol biosynthesis through antisense suppression of aldose 6-phosphate reductase with the wild type (WT) to assess how sugar composition of the fruit surface impacts microbial communities. We hypothesized that reduced sorbitol genotypes A4 and A10 would show an epiphytic microbiome different from that of the WT that corresponds to a difference in sugar composition on the fruit surface at harvest and during storage with and without postharvest treatment of fruit with 1-methycyclopropene (1-MCP), which is an inhibitor of ethylene perception. Throughout the sampling window (at harvest, 7 weeks storage, 13 weeks storage) across the 2 years of the study, the genotype, but not 1-MCP, was a significant predictor of microbiome composition. The A10 and A4 lines had an increased abundance of the pathogenic fungal genus Acremonium compared with that of the WT in one year. However, while A4 and A10 had different sugar compositions than that of WT in fruit flesh, no differences on the fruit surface were found. In addition, A4 and A10 showed microbiomes that were different from each other as well as different from that of the WT despite having the same reduced sorbitol phenotype, thus making it difficult to link microbiome differences to a specific physiological mechanism. This work represents an important step in showing the first example, to our knowledge, of how the cascading effects resulting from silencing a single gene can impact the assembly of postharvest fruit microbiomes.
Postharvest microbiomes of fruits and vegetables present a valuable opportunity for manipulating natural microbial communities to control postharvest pathogens (Kusstatscher et al. 2020; Spadaro and Droby 2016) that can be developed to promote food security and resilience in food systems. Because microbiome dynamics on fresh produce are postulated to be interconnected with nearly every aspect of plant physiology (Wisniewski and Droby 2019), many studies have investigated factors that shape microbiomes. This includes the microbiome shifting in response to biocontrol applications (Biasi et al. 2021; Duan et al. 2024; Zhao et al. 2023a, 2023b; Zhimo et al. 2021), conventional vs. organic management (Abdelfattah et al. 2016a, 2021; Bartuv et al. 2023; Leff and Fierer 2013; McLaughlin et al. 2023; Schiavon et al. 2022; Shen et al. 2022; Vepštaitė-Monstavičė et al. 2018; Wassermann et al. 2019b; Wicaksono et al. 2023), cold storage (Abdelfattah et al. 2020; Al Riachy et al. 2024; Biasi et al. 2021; Bösch et al. 2021; Lane et al. 2023, 2024a, 2024b; Shen et al. 2018a; Wassermann et al. 2019a; Zhimo et al. 2022), postharvest treatments such as waxing, hot water, and cold plasma (Abdelfattah et al. 2020; Bösch et al. 2021; Fang et al. 2023; Lane et al. 2025; Shen et al. 2018a; Wassermann et al. 2019a; Wicaksono et al. 2022), chemical applications such as fungicides and ethylene inhibitors (Abdelfattah et al. 2016b; Lane et al. 2023, 2024a, 2024b; McLaughlin et al. 2024; Perazzolli et al. 2014), and spatiotemporal factors such as harvest year, harvest time, and global geography (Abdelfattah et al. 2021; Bösch et al. 2021; Goforth et al. 2024; Lane et al. 2024b; McLaughlin et al. 2023; Wicaksono et al. 2023; Zhimo et al. 2022). Of these, spatiotemporal factors and cold storage tend to most consistently result in microbiome shifts, while chemical applications may result in limited or no shifts.
A particular area of interest is how genotype shapes the microbiome, which is often assessed in terms of cultivars within a specific crop. In apples, wild and domesticated Malus species harbor different microbiomes, leading to the postulation of coevolution between host and microbiome (Abdelfattah et al. 2022). This trend of wild and cultivated fruits harboring distinct microbiomes also holds true for papaya, banana, and passionfruit (Cruz et al. 2019), as well as when comparing sour cherries to sweet cherries (Stanevičienė et al. 2021). In addition, microbiome differences between different types of cultivated fruit can be small (Bösch et al. 2021; Campisano et al. 2014; Martins et al. 2022; Sánchez et al. 2025; Sangiorgio et al. 2022) or large (Bokulich et al. 2014; Cho et al. 2018; Malacrinò et al. 2022; Olivieri et al. 2021; Perpetuini et al. 2019). Microbiome differences attributable to fruit genotype may depend on the context of factors such as the crop used and field conditions.
Less is known about how specific targeted genetic changes that affect physiology can shape fruit microbiomes. In poplar trees, suppressing lignin biosynthesis by silencing cinnamoyl-CoA reductase resulted in upregulation of ferulic acid and a shift of the microbiome to a community that can more readily degrade the compound (Beckers et al. 2016). Notably, these shifts were found in aboveground plant tissue, but not in the rhizosphere soil, showing that using genetic engineering to induce physiological changes with specific differences in the chemical environment for microbes can be an effective way of studying how aboveground microbiomes respond to plant genotype.
We investigated the assembly of apple fruit microbiomes in wild type (WT) and reduced sorbitol genotypes. Antisense suppression of aldose 6-phosphate reductase (A6PR), a gene in the sorbitol biosynthesis pathway, results in decreased sorbitol and increased glucose in ‘Greensleeves’ apple fruits (Hu et al. 2024; Li et al. 2018). By using these reduced sorbitol lines, it was shown that sorbitol, in addition to being a key metabolite, acts as a signal modulating its own metabolism (Meng et al. 2023), resistance to a fungal pathogen Alternaria alternata (Meng et al. 2018), and malic acid accumulation in the vacuole (Hu et al. 2024). The current study followed-up this past research by investigating how sorbitol shapes microbiome community assembly on the apple surface. Given the importance of sorbitol in fruit physiology and its potential to alter the sugar composition on the fruit surface, where microbes colonize the fruit, variation in sorbitol was identified as a good candidate for linking precise genetic changes in physiology to the fruit microbiome.
Previous work regarding how ethylene and fruit maturity shape the microbiome was also expanded. Apples are a climacteric fruit producing ethylene that induces ripening (Johnston et al. 2009). To prolong shelf life, postharvest treatment of fruit with 1-methylcycopropene (1-MCP) is often used to inhibit ethylene perception in a variety of crops (Watkins 2006). Previous findings indicated that suppressing ethylene biosynthesis with aminoethoxyvinylglycine (AVG) shifts the microbiome in ‘NY1’ (Snapdragon®) apples stored in air (Lane et al. 2024b), while bacteria, but not fungi, were affected by the low oxygen storage of ‘Gala’ apples (Lane et al. 2023). Importantly, preharvest AVG sprays had a similar early effect on microbiomes as harvesting apples, suggesting that at-harvest maturity has predictable effects on the postharvest microbiome (Lane et al. 2024b). Other work also linked 1-MCP to microbiome composition in pear and an ethylene microenvironment to microbiome assembly in kiwifruit (Xie et al. 2021; Zhang et al. 2021). Therefore, we investigated how the postharvest microbial communities on each genotype respond to postharvest 1-MCP application to fruit.
We assessed the effects of reduced sorbitol genotypes and 1-MCP on the postharvest microbiome and hypothesized the following: reduced sorbitol genotypes A4 and A10 would harbor different genotypes than the WT because of a different sugar composition on the fruit surface; 1-MCP applications would delay natural shifts in cold storage because of slower fruit ripening; and genotype-related shifts would remain consistent between years, reduced sorbitol line, sampling time, and 1-MCP application.
‘Greensleeves’ apples were harvested from Cornell Orchards (Ithaca, NY, USA) on 7 Sep 2021 and 6 Sep 2023. In both years, the reduced sorbitol genotypes A4 and A10 were sampled, along with a WT. These trees were used in previous experiments (Hu et al. 2024), with A4 and A10 genotypes having reduced sorbitol and increased glucose and sucrose levels attributable to antisense suppression of A6PR (Cheng et al. 2005). Each tree was used as a biological replicate, with apples being taken from five trees of each genotype in 2021 and four trees of each genotype in 2023, when fewer trees had the requisite number of fruit for experiments. Each sample from a tree at a specific treatment/time combination consisted of four apples. Microbiome and quality sampling occurred at harvest and after 7 weeks and 13 weeks of storage.
Half of the apples from each tree were treated with 1-MCP, as previously described using SmartFresh tablets in a plastic tent (Al Shoffe et al. 2021). The fruit were stored at 0.5 °C for 7 weeks and 13 weeks. Internal ethylene concentration (IEC) was sampled on individual fruit at harvest and after 24 h at 20 °C for each storage period (Al Shoffe et al. 2021). This resulted in a total of 75 samples in 2021 (3 genotypes × 5 sampling times/1-MCP combinations × 5 replicates) and 60 samples in 2023 (3 genotypes × 5 sampling times/1-MCP combinations × 4 replicates), for a total of 135 samples.
Internal and surface sugars of fruit (sorbitol, glucose, fructose, and sucrose) in 2023 samples were assessed. Internal sugars were quantified by a Dionex DX-500 series chromatograph system as previously described (Cheng and Fuchigami 2002). Briefly, the day after microbiome and IEC sampling, two slices from opposite sides of the fruit were taken from each apple and ground into powder in liquid nitrogen using a mortar and pestle before being stored at −80 °C. Then, 50 mg of fresh-frozen tissue was added to 1.5 mL of 75% ethanol, thermomixed at 70 °C for 30 min, centrifuged at 13,000 gn for 10 min. Then, the supernatant was diluted in water and run through the Dionex system. The sugars on the surface of each sample were also measured before microbiome sequencing by swabbing a set of two 47-mm-diameter circles on each apple. Swabs were dipped in 80% ethanol before swabbing, and one swab was used for all apples in a sample. Then, swabs were left in 1 mL of 80% ethanol for 1 h after swabbing; thereafter, they were removed, with the remaining ethanol containing the swabbed surface sugars. The ethanol was evaporated from tubes using a Savant™ SpeedVac™ vacuum concentrator SPD140DDA (Thermo Fisher Scientific, Waltham, MA, USA) and reconstituted in 1 mL of water.
Microbiome sampling was performed as previously described (Lane et al. 2023), but without pooling DNA from different samples. Briefly, the four apples that comprised each sample were placed in a resealable bag with a 6.5 pH solution of 0.05 M phosphate buffer and 0.1% Tween 80. Samples were sonicated for 20 min and subjected to rotary shaking for 20 min at 120 rpm before being filtered through 0.22-μm filter paper using vacuum filtration. The resulting pellet on the filter paper was stored at −20 °C until DNA was extracted using the Qiagen DNeasy® PowerSoil® DNA extraction kit (Germantown, MD, USA) according to the manufacturer’s instructions. Then, DNA samples were sequenced through the Novogene Illumina Novaseq platform (Sacramento, CA, USA), as previously described (Lane et al. 2023).
Sequences were processed using QIIME2 (Bolyen et al. 2019). First, sequences were imported and filtered through DADA2 using default parameters (Callahan et al. 2016). The resulting amplicon sequencing variants (ASVs) underwent additional chimera removal using uchime (Edgar et al. 2011). Then, ASVs were assigned taxonomy based on a Naive Bayes Classifier trained on the UNITE version 10 for ITS and the Greengenes2 database for 16S (Abarenkov et al. 2024; McDonald et al. 2024). Chloroplast and mitochondrial sequences were then removed from the 16S dataset. The ASV abundance and taxonomy tables were exported along with representative sequences for subsequent analyses using R version 4.1.2 (R Core Team 2021). To maximize the data obtained from each sample and avoid discarding valid data, rarefication was not performed (McMurdie and Holmes 2014).
For the ITS fungal samples, a total of 2313 ASVs were obtained postfiltering, with a minimum final filtered read count in a sample of 27,705 and a maximum count of 82,324. For the 16S bacterial samples, a total of 13,416 ASVs were obtained, with a minimum final filtered read count in a sample of 20,428 and a maximum count of 100,188.
Sugars, IEC, and microbial diversity were assessed using an analysis of variance (ANOVA), while microbiome compositional differences were tested with a permutational analysis of variance (PERMANOVA) through the vegan package (Anderson 2014; Dixon 2003). Bray-Curtis was chosen as the PERMANOVA dissimilarity measure because of its handling of ecological datasets with many rare features that have low total counts (Beals 1984). Shannon’s index was used for diversity because of its balancing of richness (number of observed features) and evenness of those features. For IEC and surface sugars, data were log-transformed to meet ANOVA requirements. Additionally, for IEC, the means of each sample from the four apples were determined with log-transformed values because of the autocatalytic nature of ethylene leading to exponential growth. Pairwise differences were assessed with a post hoc Tukey honestly significant difference test for ANOVAs and pairwise PERMANOVAs using the Bonferroni adjustment for PERMNAOVAs. For taxonomy, phyloseq was used to generate taxa tables and Maaslin2 was used to determine which genera are different between genotypes in each sampling year using the Benjamini-Hochberg correction because of the more exploratory nature of the analysis (Benjamini and Hochberg 1995; Mallick et al. 2021; McMurdie and Holmes 2013). In all cases, P < 0.05 was used as the cutoff for significance.
The IEC increased over storage time and was suppressed by 1-MCP application (Fig. 1). No differences between genotypes were observed, with the exception of A4, which had a higher IEC than that of A10 in 2021. Ethylene patterns were relatively consistent between sampling years. Sugar sampling showed reduced sorbitol and sucrose and increased glucose concentrations in A4 and A10 for fresh tissue samples (Fig. 2B). However, these trends did not apply to surface sugars, for which no differences were found between genotypes (Fig. 2A). Sugar concentrations on the surface were low overall and showed high variation compared with the sugars of fresh tissue. Additionally, the relative makeup of sugars on the surface was different from that in the flesh; sucrose was found only in low levels on the surface but was more common in fresh tissue. Overall, the IEC data highlight the consistent effects of 1-MCP on reducing ethylene production regardless of genotype, while a differential sugar makeup based on genotype was found in flesh tissue but not on the surface.
Citation: J. Amer. Soc. Hort. Sci. 150, 5; 10.21273/JASHS05521-25
Citation: J. Amer. Soc. Hort. Sci. 150, 5; 10.21273/JASHS05521-25
Sampling year and sampling time were the most important factors that explain microbiome variation, although genotype was a significant predictor in both bacteria and fungi (Table 1). No association was found between 1-MCP and microbiome composition, and the significant interaction effects between year × genotype along with year × sampling time prompted an analysis separated by year (Table 2). Sampling time was the strongest predictor of fungal variation, and 1-MCP was associated with fungal microbiome variation in 2021, while genotype was the strongest predictor in 2023. For bacteria, genotype was not a significant predictor in 2021, but it was the strongest predictor in 2023; similarly, in fungi, 1-MCP was also associated with composition in 2021 but not in 2023. These same trends can be visually observed by how samples cluster separately by year and cluster more clearly by genotype in 2023 compared with 2021 (Fig. 3). These results highlight how microbial communities vary by year and that reduced sorbitol genotype is associated with microbiome composition with some context-specific effects depending on sampling year.
Citation: J. Amer. Soc. Hort. Sci. 150, 5; 10.21273/JASHS05521-25
While genotype is a predictor of microbiome variation, microbiomes are not clearly distinguished based on sorbitol-reducing genotype. In 2021, the fungal line in WT fruits harbored a different genotype than A4 but not A10; additionally, in 2023, all genotypes had distinct microbiomes for both bacteria and fungi (Table 3). This can be visually observed by principal coordinate analysis clustering (Fig. 3E, 3F), whereby no single genotype is clustered separately from the other two. This suggests that while a low-sorbitol genotype shapes the apple microbiome, it may not do so in a consistent or predictable way.
Year was the only overall predictor of microbiome diversity (Table 1), which was higher in fungi in 2021 and higher in bacteria in 2023 (Fig. 4). On a year-by-year basis, fungal diversity was associated with storage time in both years, while bacteria showed no trends outside a significant genotype × storage time interaction with no pairwise differences observed (Table 2). For fungi, the nature of shifts over storage time differed by year. In 2021, diversity increased over storage time; however, in 2023, it increased after 7 weeks of storage but decreased to lower than that at harvest by 13 weeks (Fig. 4A, 4C). These results show that diversity tends to remain steady across experimental factors that explain differences in composition, and that differences observed in diversity are not consistent between bacteria and fungi and across different contexts such as sampling year.
Citation: J. Amer. Soc. Hort. Sci. 150, 5; 10.21273/JASHS05521-25
A summary of genus-level bacterial and fungal taxonomy for each genotype at each time point can be seen in Fig. 5. The most abundant bacterial genera across all samples were Methylobacterium (14.45%), Massilia (12.16%), and Curtobacterium (9.48%), while the most abundant fungal genera were Aureobasidium (30.58%), Cladosporium (25.07%), and Sporobolomyces (6.44%). When testing for genera with a differential abundance in the reduced sorbitol lines, only the fungal genus Acremonium showed increased abundance in both A4 and A10 compared with the WT in 2023. In A4, the fungal genera Leucosporoidium, Vishniacozyma, and Aureobasidium had decreased abundance when compared with the WT in 2023. In A10, the fungal genera Phyllozyma and Acaromyces had increased abundance in 2023 compared with the WT. No genus-level differences were found for bacteria in either year or for fungi in 2021. These results show that while genotype is an important predictor of bacterial and fungal microbiomes, few notable genera were found to be associated with such differences.
Citation: J. Amer. Soc. Hort. Sci. 150, 5; 10.21273/JASHS05521-25
Our results confirmed the hypothesis that reduced sorbitol genotype can shape the apple surface microbiome, which is, to our knowledge, the first study to confirm that genetic engineering targeting a single plant gene affects the microbiomes of associated fruit communities. However, the nature of these shifts varies based on the context of these microbial communities. Genotype is an overall predictor of the microbiome (Table 1); however, when looking at individual years, the results are less consistent for 2021 than for 2023 (Table 2). This could be attributable to the large shift in microbial communities present between the years despite no major management differences. This is consistent with previous work (Bösch et al. 2021), suggesting that variability in weather and environmental conditions impacts the resulting communities (McLaughlin et al. 2023). Diversity also showed different patterns in the 2 years while having no observed association with genotype (Fig. 4), thus highlighting similar influences of context on microbial community dynamics. Overall genotype, however, predicted composition in fungi for both years and bacteria for 2023, making it an overall important factor in shaping the microbiome. Given that genetic engineering has been proposed as a tool to control plant–microbiome interactions for positive outcomes (Thakur et al. 2023), this finding is important for establishing that gene modification targeting phenotypes associated with fruit physiology can result in microbiome changes.
However, linking fruit phenotype changes with the composition of the microbiome is less clear. The A4 and A10 genotypes have the same reduced sorbitol phenotype (Cheng et al. 2005) and show only occasional minimal differences in sugar measurements and gene expression of fruit (Hu et al. 2025). Therefore, it was unexpected that they harbored microbiomes different from each other in 2023. Although differences in sugar content were observed between the WT and reduced sorbitol lines within fruit tissues, we found no evidence of genotype influencing the sugar makeup on the skin surface, where microbes reside (Fig. 2A). This could arise from microbes consuming the most abundant sugars until the makeup is similar, or it could result from unmeasured sugar differences in the peel and calyx end of the fruit, where many microbes reside (Abdelfattah et al. 2020, 2021). However, neither of these possibilities explain why A4 and A10 harbor microbiomes different from each other, thus indicating the possibility that microbiome separation by genotype is driven by a factor other than sugar content on the surface. Therefore, although we described an instance of plant genetic engineering impacting the microbiome composition, as has been described with endosphere communities of reduced lignin poplar trees (Beckers et al. 2016), the mechanism of this differential microbiome assembly remains difficult to elucidate.
Nevertheless, previous work has shown that wild and cultivated apple cultivars have large microbiome differences, while the differences within commercial cultivars are smaller (Abdelfattah et al. 2022; Bösch et al. 2021). Therefore, it is still notable for microbial communities to show any difference when the scale of genetic changes is a single gene. Fruit physiology is theorized to be highly interconnected to the microbiome (Wisniewski and Droby 2019), but this has been difficult to show for genetic variations of physiological traits because of an inability to compare a single trait when looking across cultivars. Therefore, this study showed that the link between genetics and the microbiome is not limited to genomic divergence associated with differences between apple cultivar; rather, this link also involves precise genetic changes that represent variation in a single trait and its downstream effects. This evidence indicates that individual plant traits can be linked to the microbiome.
In addition, only a few fungal genera were differentially abundant between the WT and genotypes A4 and A10. The only one shared by both reduced sorbitol lines when compared with the WT was Acremonium, which had higher abundance in those lines in 2023. Acremonium is a postharvest pathogen responsible for brown spot (Hou et al. 2019), yet it is also commonly found in apple microbiome datasets (Abdelfattah et al. 2016a; Shen et al. 2018a, 2018b, 2022; Wassermann et al. 2019a). Reduced sorbitol genotypes have increased susceptibility to A. alternata because of the role of sorbitol as a signaling molecule in plant defense (Meng et al. 2018); therefore, this may be the reason for greater relative Acremonium abundance in our dataset. However, it is unclear why this trend does not hold true for 2021, when Acremonium was also found, or for other fruit pathogens such as Alternaria itself. Additionally, none of the differentially abundant genera across genotype had clear links to sorbitol utilization. Few genera changed overall with genotype; this mirrored how the variance explained by genotype is lower than other factors such as sampling year and storage time (Table 1). This is consistent with the results of previous work that found that cultivar explains less microbiome variation than these factors (Bösch et al. 2021). Because microbiome differences were subtle and yielded few genus-level differences, the functional nature of the microbiome associated with each genotype is unclear.
Because of these limitations, more work is needed to fully understand how reduced sorbitol genotype shapes the microbiome. One possibility is investigating the microbes in fruit flesh, which tend to harbor different microbes than those in the peel, stem, or calyx ends (Wassermann et al. 2019b). Another option is to assess the microbiome of leaves, which, like fruit, show differential sugar makeups in reduced sorbitol lines compared with the WT (Cheng et al. 2005; Li et al. 2018). Leaf microbial abundance and microbiome stability are driven by vein count (Yan et al. 2022), and leaf surface sugars have long been known to affect bacterial colonization (Mercier and Lindow 2000), with bacteria being preferentially associated with veins (Monier and Lindow 2004). Therefore, it is likely that leaf-associated microbes have more access to the differential sugars of the reduced sorbitol lines than fruit-associated microbes because of colonization of leaf veins, which may shape composition to a greater extent.
Our other hypothesis that 1-MCP would shape the microbiome was not supported by our results despite ethylene suppression (Fig. 1). Although we previously found that the ethylene inhibitor AVG influenced the apple microbiome (Lane et al. 2023, 2024b), there were a number of differences in the current study. For example, AVG was applied preharvest, thus delaying fruit maturity at harvest, and the most consistent effects were found after several months rather than in the short term. This raises the possibility that 1-MCP could shift the microbiome in a cultivar with longer storage time. However, the interaction showing greater effects of the ethylene inhibitor on the microbiome over storage time was fairly weak in previous studies, suggesting that effects should be present across all time points. This was not the case in the current study. Another study found that 1-MCP shapes the microbiome in pear, but the clearest differences were found at 15 d after harvest, when the disease incidence was nearly 100% in untreated fruit and the dominant fungal genus was the pathogen Alternaria (Zhang et al. 2021). Therefore, our results suggest that, for healthy fruit, physiological changes associated with the application of a postharvest ethylene inhibitor, such as 1-MCP, do not necessarily result in microbiome shifts.
This study showed that reduced sorbitol genotypes A4 and A10 of ‘Greensleeves’ apples harbor distinct microbiomes compared with the WT, although the differences were modest compared with other factors tested. As such, this study is the first, to our knowledge, to show postharvest fruit microbiome differences resulting from targeting a single plant gene using genetic engineering. However, while A4 and A10 showed the expected sugar composition in the tissues, no differences were found in sugars on the apple surface when compared with the WT. Additionally, A4 and A10 have microbiomes that are different from each other as well as that of the WT, and genotype-specific differences were affected by sampling year, thus making it difficult to elucidate the exact mechanism by which microbiome differentiation occurs. Despite this, A4 and A10 lines were found to have increased abundance of the pathogenic fungal genus Acremonium compared with the WT in 2023, hinting at functional outcomes associated with this variation in microbial community assembly. Overall, these results push the field of fruit microbiomes beyond describing differences in microbial composition associated with cultivars and into exploring how targeted genetic alterations can impact host–microbe interactions on a whole microbiome scale. These findings open the door to investigating how microbiome changes associated with genetic engineering can result in positive or negative effects on plant health and broader food security.
Internal ethylene concentration (IEC) in reduced sorbitol genotypes (A4 and A10) compared with that in the wild type (WT) at different storage times with and without 1-methycyclopropene (1-MCP) application. Error bars represent 1 standard error of the mean. Significance letters for each time point were determined using a post hoc Tukey test of the log-adjusted averages of each sample with P < 0.05 used as the cutoff.
Sugar concentrations for surface peel swabs (A) and ground fresh tissue (B) for apple genotypes A4 and A10 compared with those of the wild type (WT). *Significant (P < 0.05) difference from the ck genotype, which was determined using a post hoc Tukey test. Error bars represent 1 standard error of the mean.
Principal coordinate analyses (PCoAs) of bacterial and fungal microbiomes for both years (A, B) and separated by year (C–F). Each PCoA was performed on Bray-Curtis dissimilarity matrices. The 95% confidence ellipses are shown for each year and genotype (A4 and A10) and compared with the wild type (WT).
Shannon diversity of bacterial and fungal communities on the apple surface for each year across genotypes A4 and A10 compared with the wild type (WT). Error bars represent 1 standard error of the mean. Significance letters for each time point were determined using a post hoc Tukey test, with P < 0.05 as a significance threshold.
Genus-level relative abundance for bacterial and fungal communities in each year grouped by genotype and sampling time. Genera with less than 1% abundance are grouped under “Other.”
Contributor Notes
We thank Kaspar Kuehn for his help in managing the transgenic trees and assisting with harvest, along with William Miller and Rose Harmon for assisting with preliminary methods testing. This research was supported, in part, by the intramural research program of the US Department of Agriculture, National Institute of Food and Agriculture (Hatch number 7008131, “The surface microbiome of fruits and vegetables in relation to postharvest quality and diseases resistance”). The findings and conclusions in this preliminary publication have not been formally disseminated by the US Department of Agriculture and should not be construed to represent any agency determination or policy. Additional support was provided by Cornell Diversity Fellowship and Pomology Endowment funds.
C.B.W. is the corresponding author. E-mail: chris.watkins@cornell.edu.
Internal ethylene concentration (IEC) in reduced sorbitol genotypes (A4 and A10) compared with that in the wild type (WT) at different storage times with and without 1-methycyclopropene (1-MCP) application. Error bars represent 1 standard error of the mean. Significance letters for each time point were determined using a post hoc Tukey test of the log-adjusted averages of each sample with P < 0.05 used as the cutoff.
Sugar concentrations for surface peel swabs (A) and ground fresh tissue (B) for apple genotypes A4 and A10 compared with those of the wild type (WT). *Significant (P < 0.05) difference from the ck genotype, which was determined using a post hoc Tukey test. Error bars represent 1 standard error of the mean.
Principal coordinate analyses (PCoAs) of bacterial and fungal microbiomes for both years (A, B) and separated by year (C–F). Each PCoA was performed on Bray-Curtis dissimilarity matrices. The 95% confidence ellipses are shown for each year and genotype (A4 and A10) and compared with the wild type (WT).
Shannon diversity of bacterial and fungal communities on the apple surface for each year across genotypes A4 and A10 compared with the wild type (WT). Error bars represent 1 standard error of the mean. Significance letters for each time point were determined using a post hoc Tukey test, with P < 0.05 as a significance threshold.
Genus-level relative abundance for bacterial and fungal communities in each year grouped by genotype and sampling time. Genera with less than 1% abundance are grouped under “Other.”
