Crop, Host, and Gut Microbiome Variation Influence Precision Nutrition: An Example of Blueberries

Epidemiological studies have shown associations between polyphenol-rich fruit intake and bone health, and preclinical studies have shown that blueberries improve bone health. To determine the genotype and dose of blueberries that are effective in ameliorating age-related bone loss, a multi-institutional team of investigators performed in vitro, preclinical, and clinical studies on blueberry varieties that differed in flavonoid profiles. Principal component analysis was used to select blueberry genotypes that varied in anthocyanin profiles. Total phenolic content did not predict the bioavailability of polyphenolic compounds in rats. A range in bioavailability was observed in individual polyphenolic compounds across genotypes. Both alpha and beta diversity analyses indicated that gut microbiome profiles varied with blueberry dose in rats. Additionally, the identification of specific taxa, such as Prevotellaceae_UCG-001 and Coriobacteriales, increasing after blueberry consumption adds to the mounting evidence of their role in polyphenol metabolism. All of the sources of variation can inform blueberry breeding practices to influence precision nutrition.


Introduction
Precision nutrition is of great current interest and is a growing focus of nutrition research funding. The concept is aimed at predicting an individual's response to a food or diet to make personalized dietary recommendations for optimal health and disease prevention for that individual. This contrasts with traditional dietary guidelines for public health recommendations that are framed for generally healthy people. However, evaluating individual diversity in response to diet is just one part of the complexity that is involved. Diversity occurs along the whole spectrum, from the food, beverage, or supplement being consumed to the physiology of the individual consuming the food and the spectrum of metabolites generated by that individual's gut microbiome and host metabolizing systems.
Fruit and vegetable consumption is associated with a decreased risk of developing many chronic diseases, including obesity, diabetes, cardiovascular disease (CVD), and neurocognitive diseases [1][2][3][4]. These benefits are commonly attributed to bioactive phytochemicals produced by plants as non-nutritive, secondary metabolites that possess biological activities in humans, including antioxidant, anti-inflammatory, anti-cancer, antimicrobial, and neuroprotective effects [5][6][7]. Of the thousands of phytochemicals that Oregon Wild Harvest, Sandy, OR, USA), cranberries (V. macrocarpon, Cranberry Network, Wisconsin Rapids, WI, USA), and lowbush blueberries (V. angustifolium) from both a single field collection of wild blueberries in Ellsworth Maine (designated as Wild Blueberries) and a commercially prepared wild blueberry composite from all major growing regions in Maine and maritime Canada designated as a Lowbush Blueberry (LB) composite (Wild Blueberry Association of North America, Old Town, ME, USA), were also analyzed. All animal studies used LB composite in the treatment arm.
Phenolics were detected and quantified using a Waters Micromass ZQ 2000 mass spectrometer, as shown in Table 1. Positive mode electrospray ionization (ESI+) was used for detection and quantitation of anthocyanins, with the following settings: source temperature 150 • C, desolvation temperature 250 • C, nitrogen desolvation gas flow rate 250 L/h, nitrogen cone gas flow rate 25 L/h, capillary voltage 2500 V, cone voltage 50 V, and extractor voltage 3 V. Negative mode ESI (ESI-) was used to detect and quantify phenolic acids, stilbenoids, flavan-3-ols, and flavonols, with the following settings: source temperature 150 • C, desolvation temperature 250 • C, nitrogen desolvation gas flow rate 400 L/h, nitrogen cone gas flow rate 60 L/h, capillary voltage 3000 V, cone voltage 40 V, and extractor voltage 5 V. Total monomeric anthocyanins were quantified using the pH differential method [26] and expressed in cyanidin-3-glucoside equivalents.

Statistical Analysis
Polyphenolic characterization from blueberry raw material data are presented as mg polyphenol/100 g of berries (dry weight).

Diversity of Bioavailability of Blueberry Polyphenols-Acute PK Study
All animal studies were conducted under the guidance of, and with protocols approved by, the Purdue University Animal Care and Use Committee (Protocol 1405001076). Twenty 3-month-old rats were ovariectomized (as a model for postmenopausal women) and, upon arrival, individually housed in stainless steel wire-bottom cages with a 12 h on:off light cycle and fed a chow diet and deionized water ad libitum. Rats were acclimated to their new housing and stabilized after their surgery for two months. Nine days prior to the pharmacokinetic study, rats were put on a polyphenol-free diet (modified AIN93M diet where soybean oil was replaced with corn oil) and randomized to one of the five blueberry genotype treatment groups: Montgomery, Ira, Onslow, SHF2B1-21:3, and LB composite. The study design is shown in Figure 1.  Whole, freeze-dried blueberries were milled to a ~0.5 mm particle size. TP content was used to determine the individual dose of blueberry for each rat. Each dose was normalized to provide 25 mg TP/kg of body weight (BW). Individual doses were suspended in 2 mL of ultrapure water, vortexed, and administered to the rats on the day of the phar- Whole, freeze-dried blueberries were milled to a~0.5 mm particle size. TP content was used to determine the individual dose of blueberry for each rat. Each dose was normalized to provide 25 mg TP/kg of body weight (BW). Individual doses were suspended in 2 mL of ultrapure water, vortexed, and administered to the rats on the day of the pharmacokinetic study via oral gavage.
Two days prior to the study, rats were anesthetized with isofluorane (3-5%) in an anesthesia chamber and maintained under anesthesia with a mask (1.5-3% isofluorane). A silastic catheter was implanted into the right jugular vein and they were given buprenorphine (0.1 mg/kg BW) to alleviate pain prior to regaining consciousness. Rats were allowed to rest for 48 h after surgery and catheters were kept patent by flushing with heparinized saline (20 U/mL) every 12 h. Rats were fasted 8 h prior to the study and~400 µL of whole blood was collected at baseline (prior to gavage with their respective blueberry dose) and 0.25, 0.5, 1, 2, 4, and 8 h post-gavage from the jugular catheter into lithium-heparinized tubes. Whole blood was processed to plasma by centrifugation at 4 • C and 6000 rpm for 10 min. Two aliquots of 80 µL of plasma each were combined with 20 µL of acidified saline (1% ascorbic acid (AA), w/v %), flushed with N2 gas, and stored at −80 • C until analysis. Rats were euthanized at the end of the 8 h pharmacokinetic study by carbon dioxide asphyxiation. Uterine horns were excised and weighed to confirm ovariectomy.
For anthocyanin metabolites, binary mobile phases were A: 2% FA (aq.) (% v/v) and B: 0.1% FA in acetonitrile (% v/v). The column was heated to 30 • C and the system flow rate was 0.3 mL/min. The binary solvent gradient was: 5% B at 0 min, 25% B at 15 min, 30% B at 19 min, and back to 5% B at 19-22.5 min. Fragmentor voltage was set at 135 V and collision energy was 17 eV in positive ESI for all compounds.

Statistical Analysis
Data from the bioavailability study were expressed as mean ± standard error of the mean (SEM). Pharmacokinetic parameter area under the curve from 0 to 8 h (AUC 0-8 h ) was calculated via the linear trapezoidal method using Microsoft Excel 2013 v.15. Maximum plasma concentration (C max ) and the time of the maximum plasma concentration (T max ) were obtained from the pharmacokinetic curve. Both AUC 0-8 h and C max were normalized by dose of parent anthocyanin, flavonol, and flavan-3-ol present within each individual dose of blueberry treatment. Statistical analyses were performed using JMP 17.0 and SAS 9.3 (SAS Institute, Cary, NC, USA) for bioavailability and phenolic characterization, respectively. Data were analyzed via two-way ANOVA and one-way ANOVA for phenolic characterization and bioavailability, respectively. Significant differences were evaluated via the Tukey-Kramer HSD post hoc pairwise comparison test. Significance was accepted at α < 0.05. Microbiome analysis was performed on 4-month-old female ovariectomized Sprague Dawley rats (Harlan Laboratories) designed to determine the effect of blueberries on bone calcium retention as previously described [28]. Briefly, twenty 5-month-old ovariectomized Sprague Dawley rats were put on a polyphenol-free diet to determine baseline to determine baseline measures then randomized to four 10-day treatment periods on a LB compositeenriched diet (2.5%, 5%, 10% or 15% LB composite diet). Each treatment was followed by a 10-day washout period on a polyphenol-free diet. Fecal samples for microbiome analysis were collected on day 10 of each of the treatments and washout periods and stored at −80 • C until analysis.

Microbiota Extraction and Analysis
Fecal DNA extraction and amplicon sequencing. DNA for microbiota analysis was extracted from 160 fecal samples collected from 20 rats at the end of baseline, ethe nd of each of the 10-day blueberry treatments, and the end of each washout phase. DNA was extracted using the FastDNA SPIN Kit for Soil (MP Biomedicals) following the manufacturer's instructions, and DNA quality and quantity were determined as previously described [29]. PCR amplification of the V3-V4 (343F: TACGGRAGGCAGCAG to 804R: CTACCRGGGTATCTAATCC) region of the 16S rRNA gene was performed as previously described [29] and sequenced using MiSeq Illumina 2 × 250 paired-end sequencing.
Sequence analysis. Sequences were analyzed using the QIIME2 pipeline [30] (version 2-2022.11). Sequences were paired, trimmed for quality, and sorted into representative sequences (amplified sequence variants, ASV) using DADA2 [31]. Representative sequences were assigned taxonomies using the q2-feature-classifier [32] trained using the SILVA dataset [33] (version 138_99). ASVs were filtered from the dataset if they represented less than 0.1% frequency of the mean sample depth, were found in only two samples, matched chloroplast or mitochondria sequences, or had non-assigned taxonomies. For an equal number of reads for alpha and beta diversity analyses, sample sets were rarified to 4300 reads. Alpha diversity was analyzed using Faith's phylogenetic diversity (PD), Shannon, observed features, and Pielou's evenness. Beta diversity was analyzed using both phylogenetic Unifrac distances [34] (weighted and un-weighted) and non-phylogenetic distances (Jaccard and Bray-Curtis).

Microbiota Statistics
For all analyses, microbiome data from baseline and washout phases were combined to represent the 0% blueberry treatment. Data were first tested for normal distribution using the Shapiro-Wilk W test. Statistical differences of alpha-diversity measurements among all groups were determined using the Kruskal-Wallis test, as well as pairwise comparisons followed by Benjamini-Hochberg FDR correction. Statistical differences among beta-diversity measures among communities were determined using 999 permutations of PERMANOVA [35] and pairwise comparisons with Benjamini-Hochberg FDR correction. PERMDISP (permutational analysis of multivariate dispersions) was performed to ensure significance was not due to dispersion differences among samples [36]. Differences were considered significant if p < 0.05. Differential abundances in taxonomy across all samples were determined using Analysis of Composition of Microbiomes (ANCOM) [37].

Diversity of Anthocyanin Profiles across Blueberry Genotypes
Raw data available for 17 anthocyanins, measured in mg/100 g wet weight (frozen fruit) for 1086 individual plants from 267 genotypes of blueberries [17], were used for analysis to identify priority genotypes for study that represent the broader diversity of the blueberry germplasm. In this instance, measures of plants from the same genotype were averaged. The anthocyanins were cyanidin 3-O-arabiniside, cyanidin 3-O-galactoside,

Polyphenol Content of Blueberries Used in Bioavailability Studies
Using the 6 blueberry genotypes with divergent anthocyanin profiles from the PCA analysis ( Figure 2) and the 4 additional members of the Vaccinium genus, total phenolic content (TP) ranged from 1951-4627 mg/100 g berry (DW) and total monomeric anthocy anins ranged from 369-1722 mg/100 g berry (DW), with bilberry and the LB composite having higher levels of phenolics and total monomeric anthocyanins than their highbush counterparts ( Figure 3).

Diversity of Blueberry Polyphenolic Profiles Polyphenol Content of Blueberries Used in Bioavailability Studies
Using the 6 blueberry genotypes with divergent anthocyanin profiles from the PCA analysis ( Figure 2) and the 4 additional members of the Vaccinium genus, total phenolic content (TP) ranged from 1951-4627 mg/100 g berry (DW) and total monomeric anthocyanins ranged from 369-1722 mg/100 g berry (DW), with bilberry and the LB composite having higher levels of phenolics and total monomeric anthocyanins than their highbush counterparts ( Figure 3).
Using the 6 blueberry genotypes with divergent anthocyanin profiles from the PCA analysis ( Figure 2) and the 4 additional members of the Vaccinium genus, total phenolic content (TP) ranged from 1951-4627 mg/100 g berry (DW) and total monomeric anthocyanins ranged from 369-1722 mg/100 g berry (DW), with bilberry and the LB composite having higher levels of phenolics and total monomeric anthocyanins than their highbush counterparts ( Figure 3).

Figure 3.
Total phenolics (in gallic acid equivalents) and total monomeric anthocyanins in 10 genotypes of Vaccinium spp. berries. Total phenolics were quantified via the Folin-Ciocalteu method and significant differences between genotypes in total phenolic content are denoted by capital letters. Total monomeric anthocyanins were quantified via the pH differential method and significant differences between genotypes in total monomeric anthocyanin content are denoted by lower case Figure 3. Total phenolics (in gallic acid equivalents) and total monomeric anthocyanins in 10 genotypes of Vaccinium spp. berries. Total phenolics were quantified via the Folin-Ciocalteu method and significant differences between genotypes in total phenolic content are denoted by capital letters. Total monomeric anthocyanins were quantified via the pH differential method and significant differences between genotypes in total monomeric anthocyanin content are denoted by lower case letters. Data are shown as means ± SDs. Significant differences between genotypes were determined using Tukey's HSD test (p < 0.05).
Anthocyanins were the most abundant phenolic class present in Vaccinium species, accounting for approximately half of all phenolics in blueberries (Table 2, Figure 4). Considerable variation in the amounts and proportions of different anthocyanins was present across the 10 genotypes we examined. Consistent with the higher levels of total phenolics and total monomeric anthocyanins observed in bilberry and LB composite, these same genotypes had the highest levels of anthocyanins. Bilberry had the highest levels of delphinidin and cyanidin species, and the LB composite had the highest levels of malvidin and acylated anthocyanins. Cranberry had a significantly different anthocyanin profile compared with other berries, with the highest levels of peonidin but the lowest levels of delphinidin, malvidin, and petunidin species. Glycosylation also varied among the genotypes tested, with most exhibiting substantial amounts of arabinoside and galactoside derivates. Some genotypes (e.g., Ira, Legacy, and Sampson) had very low levels of glucosidic derivatives, while other genotypes (e.g., Onslow, SHF2B1-21:3, WBB, bilberry, and LB composite) had levels of glucosidic derivatives equal to or greater than the arabinosidic and galactosidic derivatives. letters. Data are shown as means ± SDs. Significant differences between genotypes were determined using Tukey's HSD test (p < 0.05).
Anthocyanins were the most abundant phenolic class present in Vaccinium species, accounting for approximately half of all phenolics in blueberries (Table 2, Figure 4). Considerable variation in the amounts and proportions of different anthocyanins was present across the 10 genotypes we examined. Consistent with the higher levels of total phenolics and total monomeric anthocyanins observed in bilberry and LB composite, these same genotypes had the highest levels of anthocyanins. Bilberry had the highest levels of delphinidin and cyanidin species, and the LB composite had the highest levels of malvidin and acylated anthocyanins. Cranberry had a significantly different anthocyanin profile compared with other berries, with the highest levels of peonidin but the lowest levels of delphinidin, malvidin, and petunidin species. Glycosylation also varied among the genotypes tested, with most exhibiting substantial amounts of arabinoside and galactoside derivates. Some genotypes (e.g., Ira, Legacy, and Sampson) had very low levels of glucosidic derivatives, while other genotypes (e.g., Onslow, SHF2B1-21:3, WBB, bilberry, and LB composite) had levels of glucosidic derivatives equal to or greater than the arabinosidic and galactosidic derivatives. The most prominent phenolics are presented in Tables 3 and 4, with chlorogenic acid (20-755 mg/100 g berry (DW)) and quercetin (93-418 mg/100 g berry (DW)) exhibiting the highest concentrations in berries. Flavan-3-ols were present in comparatively low amounts for all genotypes. As with anthocyanins, cranberry had a very different phenolic The most prominent phenolics are presented in Tables 3 and 4, with chlorogenic acid (20-755 mg/100 g berry (DW)) and quercetin (93-418 mg/100 g berry (DW)) exhibiting the highest concentrations in berries. Flavan-3-ols were present in comparatively low amounts for all genotypes. As with anthocyanins, cranberry had a very different phenolic profile than the other Vaccinium species evaluated here, with the lowest level of chlorogenic acid and the highest levels of epicatechin, quercetin-3-O-arabinoside, and myricetin-3-O-glycosides. Table 2. Anthocyanin content of 10 genotypes of Vaccinium spp. (mg phenolic/100 g berry (dw)), from HPLC-MS analysis † .

Phenolic Acids
Flavan-3-ols 5.44 ± 0.62 c † Letters represent significant differences in phenolics between genotypes. Significant differences between genotypes were determined using Tukey's HSD test (p < 0.05). "Trace" indicates compounds that were detected but below the LOQ. Figure S1 and Table S1. Of the anthocyanin metabolites, Cy-3-glcs and Mal-3-glcs from the Montgomery blueberry had significantly higher bioavailability in comparison to the other genotypes of berries (p < 0.01) ( Figure 5A). For Del-3-glcs, Montgomery had significantly higher bioavailability than SHF2B1-21:3 and Onslow (p = 0.025), but was no different than Ira and LB composite. The C max of anthocyanins did not differ among genotypes, but when these values were normalized by dose, C max for Cy-3-glcs was significantly higher in Montgomery than in Ira, Onslow, and SHF2B1-21:3, reaching a maximum concentration in plasma of 24.96 ± 3.72 nM/mg of Cy-3-glcs dose (p < 0.001). Montgomery showed a higher C max normalized by dose for Del-3-gls and Mal-3-glcs, with 14.65 ± 2.16 nM/mg Mal-3-glcs dosed and 14.84 ± 3.55 nM/mg Del-3-glcs, than the other genotypes (p < 0.01), respectively. All anthocyanins showed a similar T max between 0.25-1 h, regardless of blueberry genotype (Table S1). Table 4. Flavonols in 10 genotypes of Vaccinium spp. (mg phenolic/100 g berry (dw)) † .   Figure 5B and Supplemental Figure S2, respectively. As shown in Supplemental Table S2, Ira reached a significantly higher plasma concentration of 0.040 ± 0.006 µM of C-glcr in comparison to LB composite, which only reached a C max of 0.015 ± 0.005 µM (p < 0.05). Ira, Onslow, and Montgomery showed a higher AUC 0-8 h than SHF2B1-21:3 and LB composite, but when the AUC 0-8 h was normalized by dose to determine bioavailability, there were no significant differences between the AUC 0-8 h /dose of flavan-3-ols, regardless of blueberry genotype.

Gut Microbiota
In total, 8,599,413 high-quality merged sequences were obtained after processing of MiSeq Illumina sequencing results by DADA2 and filtering out low-incident and non-target reads. Per sample, there was a median of 50,996.5 reads (range: 4321-134,668). To maintain all samples, 4300 reads were chosen by rarefaction for alpha and beta diversity analyses.

Alpha Diversity
Comparisons of the of gut microbiome diversity within each sample indicated a significantly higher diversity in samples with higher blueberry treatments using the Shannon diversity metric (measure of richness and evenness) (p = 0.001) ( Figure 6) and Pielou's evenness measure (p = 2.28 × 10 −10 ). However, no significant differences in using the richness measures (i.e., number of taxa present) were observed when using Faith's PD or observed features. This indicates that the change in diversity results from changes in the quantities of some taxa and not changes in specific taxa composing the communities.

Beta Diversity
Comparison of microbial communities among the blueberry treatments also indicated significant differences using all four metrics tested (p = 0.001). The clearest distribution of samples was seen using a PCoA scatterplot of Bray-Curtis metrics (Figure 7), a measure comparing the presence, absence, and quantities of taxa present among the samples used. Samples separated along PCoA axis 1 accounted for 11.16% of variation. Figure 7. Principal coordinate analysis illustrating differences unweighted Bray Curtis beta diversity metrics among bacterial communities after blueberry diets. Data points colored by blueberry dosages, 2.5 and 5% combined, 10 and 15% combined and no blueberry control. perMANOVA comparisons of samples blueberry dosage, overall differences significant p = 0.001 and all pairwise comparisons q = 0.001. Overall and pairwise permDISP comparisons not significant, indicating differences are not due to differences in variation.

Beta Diversity
Comparison of microbial communities among the blueberry treatments also indicated significant differences using all four metrics tested (p = 0.001). The clearest distribution of samples was seen using a PCoA scatterplot of Bray-Curtis metrics (Figure 7), a measure comparing the presence, absence, and quantities of taxa present among the samples used. Samples separated along PCoA axis 1 accounted for 11.16% of variation.

Beta Diversity
Comparison of microbial communities among the blueberry treatments also indicated significant differences using all four metrics tested (p = 0.001). The clearest distribution of samples was seen using a PCoA scatterplot of Bray-Curtis metrics (Figure 7), a measure comparing the presence, absence, and quantities of taxa present among the samples used. Samples separated along PCoA axis 1 accounted for 11.16% of variation. Figure 7. Principal coordinate analysis illustrating differences unweighted Bray Curtis beta diversity metrics among bacterial communities after blueberry diets. Data points colored by blueberry dosages, 2.5 and 5% combined, 10 and 15% combined and no blueberry control. perMANOVA comparisons of samples blueberry dosage, overall differences significant p = 0.001 and all pairwise comparisons q = 0.001. Overall and pairwise permDISP comparisons not significant, indicating differences are not due to differences in variation. Figure 7. Principal coordinate analysis illustrating differences unweighted Bray Curtis beta diversity metrics among bacterial communities after blueberry diets. Data points colored by blueberry dosages, 2.5 and 5% combined, 10 and 15% combined and no blueberry control. perMANOVA comparisons of samples blueberry dosage, overall differences significant p = 0.001 and all pairwise comparisons q = 0.001. Overall and pairwise permDISP comparisons not significant, indicating differences are not due to differences in variation.
Pairwise statistical comparisons indicated that 0% blueberry differed significantly from all the blueberry treatments (2.5, 5, 10, and 15%) (PERMANOVA, p = 0.001) and dispersion was not significantly different. As seen in the figure, pairwise comparisons of the 2.5 and 5% combined dosage communities and the 10 and 15% combined communities differed significantly from each other and 0% (q = 0.001), whereas in pairwise comparisons of each dose, the 2.5% blueberry treatment differed significantly from the 10 and 15% blueberry treatment, and the 5% treatment differed significantly from the 15% blueberry (figure not shown). Similar trends were found using the other beta diversity metrics tested, but the blueberry treatments separated along PCoA axis 2.

Differentially Abundant Taxa
ANCOM identified specific taxa contributing to the differences among communities based on blueberry dosage treatment. ANCOM identified eight taxa that differed significantly, but on examination of the volcano plot of the data ( Figure S4), only four of the taxa had high W (>100) and clr (>40) values. None of these taxa have been classified at the genus level. Two are from the phylum Actinobacteriota and listed as uncultured from the order Coriobacteriales and the other incertae sedis; one is from the phylum Bacteroidota family Prevotellaceae_UCG-001, and one from the Firmicutes family Anaerovoracaceae XIII_UCG-001. Boxplots of these taxa indicated that all their abundances were higher in samples after blueberry treatments (Figure 8). Pairwise statistical comparisons indicated that 0% blueberry differed significantly from all the blueberry treatments (2.5, 5, 10, and 15%) (PERMANOVA, p = 0.001) and dispersion was not significantly different. As seen in the figure, pairwise comparisons of the 2.5 and 5% combined dosage communities and the 10 and 15% combined communities differed significantly from each other and 0% (q = 0.001), whereas in pairwise comparisons of each dose, the 2.5% blueberry treatment differed significantly from the 10 and 15% blueberry treatment, and the 5% treatment differed significantly from the 15% blueberry (figure not shown). Similar trends were found using the other beta diversity metrics tested, but the blueberry treatments separated along PCoA axis 2.

Differentially Abundant Taxa
ANCOM identified specific taxa contributing to the differences among communities based on blueberry dosage treatment. ANCOM identified eight taxa that differed significantly, but on examination of the volcano plot of the data ( Figure S4), only four of the taxa had high W (>100) and clr (>40) values. None of these taxa have been classified at the genus level. Two are from the phylum Actinobacteriota and listed as uncultured from the order Coriobacteriales and the other incertae sedis; one is from the phylum Bacteroidota family Prevotellaceae_UCG-001, and one from the Firmicutes family Anaerovoracaceae XIII_UCG-001. Boxplots of these taxa indicated that all their abundances were higher in samples after blueberry treatments (Figure 8).

Discussion
Blueberry was used as a model for studying the potential impact of crop genetic diversity on the bioavailability of polyphenolic constituents and variation in the gut microbiome, two factors associated with chronic disease outcomes.

Variation in Blueberry Anthocyanin Profiles
PCA, a statistical method that considers a large number of variables, such as anthocyanin profiles, for a large number of blueberry genotypes, allowed us to identify a small number of genotypes that effectively could serve as a way to capture the broad diversity present in the germplasm in further study. The ability to refine biomedical testing to a limited array of genotypes captures the broader diversity of the collection and allows for identification of bioactives unique to individual genotypes or common functions that can be generalized across the broader germplasm. Identifying functional traits with specific genotypes can inform breeding programs toward improving the nutritional quality of blueberries overall.

Variation in Bioavailability of Polyphenolics
Phenolic content of LB composite was twice that of highbush blueberries, consistent with previous studies [18]. This is likely explained both by genetics, as well as the response to the harsher environmental conditions where lowbush blueberries are grown. However, parent compounds are rarely observed circulating in plasma, suggesting that the bioactive compounds in blueberry are the multiple metabolites produced after first-pass metabolism that will eventually reach target tissues and exert their health benefits.
The anthocyanin metabolites for cyanidin and malvidin were significantly more bioavailable from the Montgomery genotype. Consistent with anthocyanin bioavailability studies in humans [38] and rats [39,40], all anthocyanins reached a maximal plasma concentration 0.25-1 h post-dose. A small part of anthocyanin absorption may begin early in the stomach, as evidenced in this study by the high concentration of circulating anthocyanin metabolites as early as the first blood draw at 15 min post gavage, but the greatest absorption occurred through the small intestine. The rate and extent at which they are absorbed may be affected by the aglycone, sugar moiety, or acylated groups [41]. The acylation of anthocyanins highly influences their bioavailability. Charron et al. showed that nonacylated anthocyanins were 4-fold and 4.7-fold more bioavailable than the acylated anthocyanins in red cabbage and purple carrot juice, respectively [42,43]. Only 5% of Montgomery's total anthocyanin content were acylated compared to~9-28% acylated anthocyanins for other genotypes.
Both flavan-3-ols and flavonols varied widely across genotypes. Once normalized for dose, there were few differences in flavan-3-ol pharmacokinetics. In contrast, the rabbiteyes, Montgomery, Ira, and Onslow had the lowest total quercetin content, but had higher bioavailability for quercetin glucuronide (p < 0.05) than the other berries. For its methylated conjugate, MeQ-glcr, Ira had the lowest bioavailability (p < 0.05) of all berries. This may be due to differences in the glycosylation of quercetin across the different blueberry genotypes. The sugar moiety highly influences quercetin bioavailability as it requires cleavage for full absorption by lactase phlorizin hydrolase (LPH) or cytosolic β-glucosidase (CBG) within the enterocyte. Once the aglycone is released, quercetin is then subject to host metabolism by phase II enzymatic conjugation systems including glucuronide/sulfates/methyl groups [44]. Except for glucose, all other glycosides are poor substrates for enzymatic cleavage, limiting the release of quercetin aglycone for absorption through the small intestine. Quercetin glycosides then travel to the lower gut, where they can be metabolized by lower gut bacteria [45]; this early and late absorption can be observed in Figure S3, where quercetin exhibits a pharmacokinetic curve with double peaks for its MeQ-gclr metabolite, which is not uncommon and is consistent with previous observations. Yang et al. reported a similar phenomenon with Q-glcr displaying peaks at 2 h and 8 h post dose, which, in their study, was possibly caused by enterohepatic recirculation [46]. In our study, the peaks for MeQ-glcr occurred at 0.5 h and 1-2 h post-dose, which could have been caused by an early and more rapid passive absorption of the quercetin aglycone in comparison to the slower active absorption of quercetin glucoside.
The rabbiteye genotypes Montgomery, Ira, and Onslow have a significantly lower concentration of myricetin in comparison to the highbush and lowbush genotypes, with a 23-fold difference in concentration between Ira and SHF2B1-21:3. Although Ira and Montgomery had significantly lower content of myricetin (p < 0.05), they reached significantly higher plasma concentrations of 7-and 11-fold higher than SHF2B1-21:3 (p < 0.05), respectively. The bioavailability from Ira was higher than all genotypes (p < 0.05), followed by Montgomery and Onslow, and a very low bioavailability was observed for SHF2B1-21:3 and LB composite. As with quercetin, myricetin absorption rate may be low because its absorption might have been influenced by the type of glycosidic linkages. Although the aglycone and glycosylated forms of myricetin were not differentiated in our LC method, it is possible that the differences in myricetin bioavailability could have been due to differences in content of myricetin aglycone and the glycosylated form.

Variation in Gut Microbiome in Response to Blueberry
Both alpha and beta diversity analyses indicated there were dose-dependent effects of blueberry on the gut microbiome in this rat model of postmenopausal estrogen deficiency. The increase in alpha diversity is suggested to be related to better microbiome gut health and to cognitive function [47]. Four taxa were found to be significantly higher in blueberry treatments compared to no blueberry controls, i.e., two Actinobacteriota and listed as uncultured from the order Coriobacteriales, one from the phylum Bacteroidota family Prevotellaceae_UCG-001, and one from the Firmicutes family Anaerovoracaceae XIII_UCG-001. The taxa listed as Prevotellaceae_UCG-001 and Coriobacteriales incertae sedis were found to be significantly associated with blueberry in the OVX mouse study described by Sato et al. [48]. The Coriobacteriaceae family are thought to be to polyphenol degraders [49].
The limitations of our studies varied. In the PCA, the sample size was large, but was limited in the bioavailability studies. Alternative approaches could lead to different conclusions. In the PCA analysis, clustering of genotypes did not consider categorical characteristics, such as type or region of origin. For studying bioavailability, we opted to give the same dose of TP (25 mg/kg BW), resulting in wide differences in the amount of freeze-dried blueberry powder dosed from 189 mg to 285 mg of blueberry powder. Genotype comparisons should be made for the gut microbiome, as was performed in our previous report of a mouse study [48].

Conclusions
Blueberries of different genetic backgrounds contain different phenolic profiles that vary in the extent of bioavailability and metabolism of their polyphenols. PCA of 17 anthocyanins from 267 genotypes allowed selection of a few genotypes with widely differing profiles for further characterization. The high bioavailability of anthocyanins and phenolics in certain genotypes is a unique contribution of this work. Further, evidence of a gut microbiome response to blueberry dose was observed. This rich variation encountered in each system from crop to consumer to gut microbiome can be used to improve crop selection, breeding practices, and identification of lead genotypes, which may be leveraged to understand functional responses for health and develop precision nutrition practices.  Figure S1: Plasma pharmacokinetic response of anthocyanin metabolites from different blueberry genotypes. Data represented as mean ± SEM (n = 4 rats/group); Supplementary Figure S2: Plasma pharmacokinetic response of flavan-3-ols me-tabolites from different blueberry genotypes. Data represented as mean ± SEM (n = 4 rats/group). EC-glcr, epicate-chin-5-glucuronide; C-glcr, catechin-5-glucuronide; EC-gclr, 3 -O-methylepicatechin-5-glcr; MeC-glcr, 3 -O-methylcatechin-5-glcr; Supplementary Figure S3: Plasma pharmacokinetic response of flavonol metabolites from different blueberry genotypes. Data represented as mean ± SEM (n = 4 rats/group). Q-3-glcr, querce-tin-3-glucuronide; Me-Q-glucr, methylquercetin-3-glucuronide; Myr-glucr, myricetin glucuronide; Figure S4: Vol-cano plots of taxa identified by Analysis of Composition of Microbiomes (ANCOM) as differentially abundant in different doses of blueberry diet. Taxa in the upper right side (clr >40 and W >100) of the figure are illustrated us-ing box-plots in Figure 7. X-axis represented by center log ratio (clr) and y-axis is W is number of time null hy-pothesis was rejected.

Informed Consent Statement: Not applicable.
Data Availability Statement: Data will be made available upon request to senior author. Microbiome sequence data are deposited in NCBI-SRA https://www.ncbi.nlm.nih.gov/ under accession number BioProject ID: PRJNA954427. The data on composition of blueberries were collected at the North Carolina State University Piedmont Research Station (PRS), Salisbury, NC, USA (https://cals.ncsu. edu/food-bioprocessing-and-nutrition-sciences/). More information about the data and its use is available from coauthors Mary Ann Lila and Mario G. Ferruzzi.

Conflicts of Interest:
The authors declare no conflict of interest.