The Horse Gut Microbiome Responds in a Highly Individualized Manner to Forage Lignication

Background: Although contributions of the equine gut microbiome to forage utilization are well recognized, the impact of alfalfa lignication on the equine gut microbiome remains unknown. Here, we characterized microbial community dynamics in the equine distal gut when feeding reduced lignin (RL) and reference alfalfa hays (CON-control) ( Medicago sativa L.) to adult stock-type horses. Hay from RL and CON cultivars were similar in crude protein, neutral detergent ber, and equine digestible energy, but differed in acid detergent lignin content (RL:74 g kg -1 vs. CON: 81 g kg -1 ). Dietary treatments were fed to six horses in a crossover study. Experimental periods consisted of a 9-d dietary adaptation phase followed by a 5-d total fecal collection phase, during which horses were housed in individual box stalls and manure was removed on a continuous 24-h basis. At 12-h intervals, feces were thoroughly mixed, frozen, and used for bacterial community composition analyses via V4, 16S rRNA amplicon MiSeq sequencing. Results: RL alfalfa did not result in specic fecal microbiome composition across all horses. However, upon incorporating individual horse in the model, it was shown that the microbiome of each subject did respond to hay lignin content in an individualized manner over time, in terms of alpha and beta diversity. Closer inspection of specic taxonomic changes upon feeding the two diets also revealed horse-specic trends, with unique amplicon sequence variants classied as Akkermansia , Fibrobacter succinogenes , Treponema, and Paludibacter uctuating signicantly in abundance when RL alfalfa was fed, depending on horse. Along these lines, horse-specic associations between individual gut microbiome traits and characteristics of the digested CON or RL alfalfa were observed, mainly in regards to dry matter digestibility and mean feed particle size. Conclusions: These results indicate that the horse gut microbiome responds in an individualized manner to small changes in the amount of acid detergent lignin in alfalfa hay, potentially impacting several feed digestibility

alfa are now commercially available, and eld research evaluating RL alfalfa under different harvest frequencies has demonstrated a reduction in total herbage acid detergent lignin (ADL) and an increase in 48-h neutral detergent ber digestibility (NDFD48) and relative forage quality (RFQ) compared to reference (or CONventional) cultivars [11].When feeding RL alfalfa hay to lambs, researchers found that dry matter digestibility (DMD) and NDFD48 was greater for the RL hay [12].

Similarly, when RL alfalfa was included as 50% of the ration for lactating dairy cows, NDFD48 was increased and resulted in 1.3 kg more milk roduction per head per day [13].Indeed, Grev et al. (2019) reported a 4% increase in apparent DMD when horses consumed RL alfalfa hay compared to CON alfalfa hay.While RL alfalfa has been shown to improve forage digestibility for sheep and cattle, it is unknown if changes in lignin will impact the equine microbiome.Therefore, the objective of this study was to evaluate the equine fecal microbiome when feeding RL or CON alfalfa hay to adult horses.


Materials And Methods

All experimental procedures were conduct d according to those approved by the University of Minnesota Institutional Animal Care and Use Committee (1710-35228A).


Horses

Dietary treatments consisted in two commercially available alfalfa cultivars, including RL alfalfa ('54HVX41', Forage Genetics, Napa, ID) and CON alfalfa ('WL355.RR', W-L Alfalfa, Ozark, MO).Both alfalfa types were harvested into small square bales prior to the start of the study.The experiment was completed using a crossover design with two treatments (alfalfa cultivar) and two periods.Six adult (20 ± 4 years), stock-type horses with an average body weight (BW)

544 kg (SE ± 36 kg) a
d body condition score (BCS; Henneke et al., 1983) of 5.7 (SE ± 1.0) were blocked by BW and divided into two similar herds with three horses each.Prior to the st

t of th
study, horses were acclimated to their herd and paddock and given free-choice access to legume-grass mixed hay and water.Herds remained together for the duration of the study.Each experimental period consisted of a 9-d dietary adaptation phase (d 1 to 9) followed by a 5-d total fecal collection phase (d 10 to 14; [14]).A graphical representation of the experimental design can be seen in Fig. 1.

At the beginning of each adaptation phase (d 1), horses were weighed using a livestock platform scale and BCS was assessed (Henneke et al., 1983).For the duration of the adaptation phase, horse herds were housed in dry lots with access to shelter and water.Horses received their experimental diet on an ad libitum basis, with hay fed twice daily at 0800 and 1900 h.With each morning feeding, horses were also given 0.9 kg of a commercially prepared ration balancer (vitamin and mineral mix) to ensure that all nutritional requirements were met for adult horses at maintenance (NRC, 2007).At the beginning of each fecal collection phase (d 10), horses were moved to indi idual rubber-matted boxstalls (3.6 × 3.6 m), where they were housed for the duration of the fecal collection phase (Fig. 1).Each morning of the fecal collection phase, representative forage samples were obtained by randomly sampling hay bales from each dietary treatment using a core-sampler (Penn State Forage Sampler, University Park, PA).Hay cores for each dietary treatment were combined by day and stored at -20 °C for later analysis.Hay was offered at 2% BW from hay nets (Half Bale Net, Hay Chix, Taylor Falls, MN) and was fed in two equal portions at 0800 and 1900 h.Horses had free-choice access to water throughout the fecal collection phase.


Samples

Manure was removed from the stalls on a continuous 24-h basis to allow for determination of total daily fecal output and to reduce any possible contamination with hay or urine.Feces for each horse were collected individually into large plastic containers lined with plastic bags that remained closed throughout the day to retain moisture.Cumulative feces weight was recorded every 12 h, at which time the collected feces was thoroughly mixed and subsampled in duplicate.Subsamples were placed in sealed collection bags and stored at -20 °C until DNA extraction.DNA from each fecal sample was extracted using the Qiagen power soil extraction kit, following manufacture

s instru
tions.Extracted DNA underwent MiSeq sequencing of the 16S rRNA V4 variable region (dual index approach; [15]) at the University of Minnesota Genomic Center (St.Paul, MN, USA).


Feed digesta analyses

Hay samples were thawed, dried in a forced-air oven at 60 °C for 48 h, and ground to pass through a 1mm screen in a Cyclotec (Foss, Hillerod, Denmark).Dry matter was determined by dividing the weight of the sample after drying by the wet weight of the sample.Ground samples were analyzed for forage nutritive value via wet chemistry by a commercial forage testing laboratory (Equi-Analytical, Ithaca, NY) using the following methods: crude protein was calculated as the percentage of nitrogen multiplied by 6.25 [16] neutral and acid detergent bers and acid detergent lignin were measured using lter bag techniques [17,18], and equine DE was calculated using

equation developed by
Pagan [19].


Bioinformatics and data analyses of amplicon sequence data

High quality (Q > 30, using fastx_toolkit) paired-end reads were considered for downstream analysis using Dada2, mafft, and fasttree plugins of Qiime2 to obtain unique amplicon sequence variants (ASVs), align representative sequences, and generate rooted and unrooted phylogenetic trees [20][21][22][23].Greengenes reference sequences (clustered at 99% sequence identity) were downloaded and trained using naïve Bayes classi er, and this trained classi er was further used for taxonomic assignment of representative ASVs using the feature-classi er classify-sklearn plugin within Qime2 [24].All downstream microbial communit ecology analyses were performed in the R statistical interface [25].Relative abundances of each ASV and phylogenetic tree (rooted) were used for calculating Bray-Curtis and UniFrac distance matrices, principal coordinate and alpha diversity analysis using the vegan, ape, and phyloseq packages in R [26][27][28].Permutational Multivariate Analysis of Variance (PERMANOVA) was calculated using the Adonis function of the vegan package.Selection of signi cantly discriminating taxonomic markers was carried out using indicator species analysis on the labdsv package, Wilcoxon rank sum tests and fold changes of each discriminating ASV between CON or RL alfalfa for a given horse [29].Generalized linear models were constructed using the glm function in R to mine for associations between different feed digestibility and microbiome traits [30].All graphs were plotted using boxplot, ggplots, and EnhancedVolcano functions within the R [31,32].


Results


Nutrient composition of RL and CON alfalfa hays

With the exception of ADL, nutrient content was similar between alfalfa cultivars, and concentrations for all nutrients fell within the normal range for alfalfa hay (Table 1; [11,[33][34][35]).The nutrient pro les for both alfalfa hays in the present study were su cient to meet the daily DE and CP requirements for adult horses at maintenance (NRC, 2007).In the current study, the RL alfalfa hay demonstrated a 9% reduction in ADL compared to CON alfalfa hay.Previous research comparing RL and CON alfalfa hays has also reported a reduction in ADL with little to no change in other nutrient components (Mertens and McCaslin, 2008

RL alfalfa cultivars have shown a 6 to 24% decr
ase in total herbage ADL compared to CON alfalfa cultivars [11,36].


Shifting between RL or CON alfalfa triggered unique microbiome changes in each individual horse

After bioinformatic processing for amplicon quality control and taxonomic assignment of unique sequence variants, an average of 21,836 ± 6,052 reads per fecal sample were obtained.Feeding RL alfalfa resulted in only minor effects on fecal microbiome composition across all horses (Bray-Curtis distance PERMANOVA, R2 = 0.01, P = 0.002; Fig. 2a).However, upon closer inspection, the data showed that each horse's microbiome was unique, and that each horse's microbiome responded to dietary lignin content in an individualized manner (horse effect: R2 = 0.11, P < 0.001; sampling date effect: R2 = 0.10, P < 0.001; interaction horse*diet: R2 = 0.04, P < 0.001; interaction diet*sampling time: R2 = 0.07, P = 0.02; interaction horse*sampling time: R2 = 0.3, P < 0.001; Fig. 2b).That is, feeding CON vs. RL alfalfa signi cantly shifted each horse's fecal microbiome composition in a unique fashion over time (Fig. 2c-h).

For instance, the fecal microbiome of horse An, who started the feeding trial on CON alfalfa, drastically changed when fed the RL alfalfa (treatment effect: R2 = 0.09, P < 0.001; sampling date effect: R2 = 0.53, P = 0.009; Fig. 2c).Within each horse, this microbiome composition shift showed unique dynamic trends across the 5 sampling days, indicating that the distal microbiome of the sampled horses not only changed according to forage lignin content but also across each day and even within a day depending on time of sample collection (morning or afternoon).For instance, horse microbiome variation within sampling days was also unique to each horse, with some ndividuals showing greater variation between AM and PM pro les in the same day (e.g.Be and Zi, Fig. 3a-f).

Similar trends were observed when evaluating microbiome diversity and richness.For example, feeding CON or RL alfalfa did not have an effect on mean gut microbiome diversity (Shannon index) when horses were grouped, but diversity seemed to be unique to speci c horses (ANOVA, treatment effect: F = 0.5, P = 0.5; horse effect: F = 2.7, P = 0.03; date affect: F = 2.3, P = 0.02; interaction horse*sampling time: F = 1.9, P = 0.01; Fig. 4a).The number of observed amplicon sequence variants (ASV richness) also followed that trend (Fig. 4b).Interestingly, however, patterns of alpha diversity across each horse were not consistent; that is, diversity increased upon feedin RL alfalfa in only one horse (An), while in others, the effect was opposite and/or not supported by the data (Figs. 4 a and b).

To further investigate factors that explaining the individual patterns observed, we applied an indicator species analysis and identi ed bacterial taxa that increased or diminished in abundance when feeding CON or RL alfalfa.As expected, the taxa that dominated the gut microbiome when feeding each alfalfa hay were unique in each individual horse (Additional le 1: Table S1-S6).For instance, in horse An, feeding RL alfalfa mainly resulted in signi cant increases in the abundance of several ASVs a liated to Akkermansia, while feeding CON alfalfa was associated with increased abundance of unknown ASVs from the Bacteroidales order (Indicator value > 0.7, Wilcoxon test's P < 0.001, Additional le 1: Table S1).

In contrast, these speci c taxa did not uctuate in the same way in other horses, which showed different abundance dynamics of other indicator taxa (e.g.horse Ba, Fig. 5a and b, Table S2-S6).Furthermore, feeding CON or RL alfalfa also resulted in changes at very speci c taxonomic levels within the same broad taxonomic group.For instance, other ASVs a liated to the genus Akkermansia were also observed in higher abundance in horse An when feeding CON alfalfa, and other Bacteroidales were abundant when feeding RL alfalfa, but with a lower magnitude change (Fig. 5b).

This phenomenon was seen with many of the indicator taxa identi ed as enriched in either the CON or RL d et, indicating that the microbiome changes observed upon feeding the two diets operate beyond the genus level in the horse gut microbiome (e.g.different species and strains within the same genera were observed to uctuate; Fig. 5 c-h).To test this hypothesis, we looked for compositional changes when feeding CON or RL alfalfa at the genus level in one of the horses (An).Although differences were still detectable and signi cant, the magnitude of the change was smaller and microbiome distinctions between the two diets were less clear compared with the ASV (species/strai ) level analysis (treatment effect: R2 = 0.08, P = 0.04; sampling date effect: R2 = 0.63, P = 0.02; Additional le 2: Figure S1a, Additional le 1: Table S7).Of note, the main genera discriminating between the RL and CON diets in horse An were Lachnospira and genera a liated to the Paraprevorellaceae family, respectively (Indicator value = 0.7, Wilcoxon test's P < 0.01, Additional le 2: Figure S1b), which did not come out as markers of the CON or RL diet at the strain level analyses.Furthermore, we used a second distance matrix that considers phylogenetic distances between taxa (weighted UniFrac on ASV relative abundances), and also observed less distinction in the microbiomes of CON-and RL-fed horses, with no signi cant differences in half of the cases (Additional le 3: Figure S2 c-h).However, individual horse was still the main driver of the microbiome patterns observed (Additional le 3: Figure S2a).

Despite the strain level and horse speci c heterogeneity detected, most of the taxa that uctuated in response to the CON or RL alfalfa have been typically associated with brolytic or glycolytic functions in mammals that primarily depend on forage for subsistence.For example, Akkermansia ASVs, which mostly increased in horse An under the RL alfalfa diet, were close in identity to A. glycaniphila (Basic Local Alignment Search Tool-BLAST: % identity = 93.5%,coverage 97%, E-value = 5e-109).Furthermore, the Akkermansia sequences identi ed are closer to 16S rRNA clones isolated from the feces of black rhinos (BLAST: % identity = 99.6%,coverage 97%, E-value = 8e-132).Other ASVs uct ating the most in the gut microbiome of each horse (indicator value > 0.7) under RL alfalfa diets were a liated to unclassi ed Paludibacter, Ruminococcaceae, Bacteroidales, Treponema, and Fibrobacter succinogenes.Different taxa that increased under the CON alfalfa diet included Solibacillus, Acinetobacter lwo i, and Methanocorpusculum (Fig. 5c-h).However, these patterns were not uniform across all horses.

Forage digestibility traits are associated with individual microbiome pro les Next, we sought to investigate if the microbiome patterns seen in each horse were associated with data on forage digestibility for the CON and RL alfalfa hay diets.Speci c digesta characteristics did co-vary with particular microbiome traits, but with covariation patterns unique to each individual horse.For example, in horse An, apparent dry matter digestibility (DMD) was signi cantly associated with principal coordinate scores along axis 1 (PCo.1),increasing under the RL alfalfa diet (generalized linear model-GLM: t=2.5, P= 0.02, Figure 6a).We also observed that mean FPS in horse An and Fa positively corr lated with scores along PCo.1 (GLM: t=3.78-4.38,P< 0.01, Figure 6b,e) and abundances of Akkermansia and Treponema ASVs (GLM: t=2.89-5.6,P< 0.01, Figure 6c,f), both indicator taxa in these two horses.No associations between apparent DMD and microbial traits were observed in horse Fa (GLM: t=1, P= 0.3, Figure 6d); In other horses, some of these trends were also observed, although in an opposite trend.For instance particularly in horses Da and Ba lower mean feed particle size (FPS) was observed under the RL treatment, and in association with PCo.1 scores.The same horses also showed lower FPS associated with the RL hay, along with increased abundances of Paludibacter (GLM, P<0.05, Figure 6g-l).No associations were found between alpha diversity (Shannon's H or ASV richness) metrics and individual feed digestibility traits.


Discussion

Herein, we report changes in the equine distal gut microbiome upon feeding alfalfa hay with a reduced lignin concentration (9% reduction).The compositional changes observed were not uniform across all horses, but were signi cant and unique to each individual, involved different groups of bacteria at very speci c taxonomic levels (species or strains), and showed different dynamism over time.The data also showed horse-speci c associations between microbiome composition and speci c digestibility traits, including apparent DMD and mean FPS.


Individualized responses to alfalfa ligni cation

One of the most intriguing ndings reported here is the high individualization of the equine d

tal gut mic
obiome upon the CON or RL feeding intervention.This level of individualization in response to a feeding intervention or baseline diets over time is a common characteristic of the human gut microbiome [37].For example, evidence has shown that the introduction of diverse dietary polysaccharides that aim at stimulating gut microbial communities (prebiotic bers) triggers unique compositional changes in speci c individuals [38,39].Microbiome patterns at a very speci c taxonomic level (e.g.strains), as reported here, have also been shown in resp

se to single antibiotics or antibiotic cocktails
40,41].

Likewise, interventions with speci c probiotic strains in humans are reported to yield highly individualized engraftment responses, likely pre-determined by individual microbial repertoires previous to intervention [42,43].

Thus, this level of individualization in the equine microbiome may be expected, except for the fact that such individuality to a feeding intervention is not commonly reported in domestic or wild animals.In other domesticated species, dogs have shown individual responses to antibiotics [44,45], and there have been reports of individualized responses of bovines to rumen transfaunation, with rumen microbial communities returning to pre-transfaunation states in an individual manner [46].Detection of intraindividual variation is facilitated by pro ling serial samples from the same individual, which has previously revealed horse-speci c microbiome pro les at the stomach mucosal level in adults [47] and ndividual-speci c fecal microbiome composition in ponies [48].However, individualized microbiome responses to a particular intervention in equines have only been reported for antibiotics [49].To the best of our knowledge, this is the rst report that emphasizes individual equine microbiome pro les in response to a forage feeding intervention, which is remarkable considering that this change consisted of a small reduction (9%) in the concentration of lignin in alfalfa hay.Moreover, the responses to each hay type were also highly individualized day to day and even within a single sampling day (Fig. 3).

Along these lines, the observation that taxonomic changes associated with hay ligni cation were associated speci c ASVs within the same core taxa (Akkermansia, Treponema, Paludibacter, Bacteroidales, Ruminococcaceae, and Fibrobacter, among others) is remarkable.This observation not only underscores the speci city and heterogeneity of ber metabolism in the gut microbiome of herbivores [50], but also may indicate that e ciency of hay degradation may differ across individual horses.It is possible that such heterogeneity may be linked to individual genetic backgrounds.For example, host genetic control on the mammalian microbiome should operate on deeper branches of the bacterial phylogenetic tree across ma mals (species and strains), as opposed to on broader taxonomic groups [51], with bacterial communities further modulated by genetic diversity and individual life history (e.g.physiological changes associated with individual growth, development, and health status) [52].Indeed, the modern American Quarter Horse exhibits high genetic diversity, which is the result of increased gene ow during domestication and wide geographical spread due to its historical use in transportation [53,54].Thus, just as in humans, diverse genetic variants in the American Quarter Horse may be linked to signi cant variation of gut microbiome pro les between individuals at very speci c taxonomic groups [55,56].This linked variation between host genetics and gut microbiome may determine subject-speci c immune and metabolic responses to diets or disease challenges [57].


Individual microbiome pro les and feed e cency

However, from the data available, it is not possible to determine what these horse-speci c microbiome responses to minor differences in alfalfa ligni cation mean, from a functional or physiological perspective.All ASVs that uctuated in response to hay ligni cation are a liated to core taxa in the equine gut and have been associated with hay consumption [58][59][60].These taxa were differently associated with markers of RL or CON alfalfa for each of the six horses (Figure 5), with different ASVs from the same genera or species (e.g.Akkermansia, Treponema, Paludibacter) arising in response to RL alfalfa consumption in some horses and blooming when consuming CON a

alfa in others.Thus, it could be assumed that t
ese unique taxonomic responses are overridden by functional redundancy, as previously shown for the gut microbiome of heterogeneous human populations [55].

Nonetheless, discriminant microbiome pro les were uniquely associated with apparent DMD and mean FPS for both types of hay.This observation may indicate that not all horses were equally e cient at metabolizing each hay type, and that their gut microbiome may be an important determinant of this individuality.For instance, less ligni ed hay would allow for increased substrate availability for degrading/fermenting microbes [12], resulting in blooms of specialized brolytic or fermentative bacteria such as F. succiniogenes and Akkermansia, as seen in a some individuals.But while in some horses microbiome changes associated with RL alfalfa were associated with higher DMD, in others, unique microbiome pro les were correl ted with lower DMD values (Figure 5).The same phenomenon was seen with mean FPS, whose lower values are associated with increased digestible energy [61].Thus, the data show that the effect of feeding less ligni ed hay with the intent of improving the e ciency of forage utilization in horses might have to be considered on an individual basis.

Thus, these results also open a debate on individualized nutritional interventions in equines that can target the gut microbiome to improve performance and health.In humans, the individualized nutrition concept considers that each individual digests and metabolizes identical foods in a different way, depending on genomic and microbiome traits unique to each subject [37,62].The individualized nutrition framework in humans considers information about an individual's lifestyle, dietary habits, genetics, functional genomic landscape, metabolome, and microbiome as critical data to understand mechanisms and treatments behind a myriad of diseases, with promising results for ameliorating symptoms associated with metabolic disea e [63,64].Given the high rates of metabolic disorders in domestic animals, including companion, work, and endurance horses [65][66][67], and the economic losses linked to these health issues, an individualized nutrition approach for metabolic and other diseases that considers individualized microbiome pro les and speci c metabolic responses to particular diets may prove to be promising.


Conclusion

In summary, we report highly individualized gut microbiome responses in horses to feeding alfalfa hay with reduced lignin concentrations, and show that horse-speci c patterns are also associated with feed digestibility traits.The ndings underscore the importance of considering individual and temporal factors  (a) PCoA based on Bray-Curtis distance matrices show no apparent fecal microbiome differences when feeding RL or CON.However, clustering of samples seems to be driven mainly by individual horses: An, Ba, Be, Da, Fa and Zi (b).Thus, an analysis of microbiome changes when feeding RL or CON alfalfa was conducted within each horse, revealing that ligni cation of forage signi cantly shits the equine microbiome

iate Anal
sis of VarianceDeclarationsEthics approval and consent to participateThe experimental procedure of the present study was reviewed and approved by the University of Minnesota Institutional Animal Care and Use Committee (1710-35228A).Consent for publicationNot ApplicableCompeting interestsThe authors declare that they have no competing interests.Author's contributions: KM, AGr and CS made substantial contributions to the conception and design of the study.KM, and AGr performed the experiment.AGo and AKS performed the bioinformatics and statist