Multi-Omics Study of The Salivary Modulation of The Rumen Microbiome

Ruminants are able to produce large quantities of saliva which enter into the rumen. Although previous research has indicated that salivary immunoglobulins can partially modulate the rumen microbial activity, the role of the salivary components other than ions on the rumen microbial ecosystem has not been thoroughly investigated in ruminants. A total of 16 semi-continuous in vitro cultures were used to incubate rumen uid from 4 donor goats inoculated with autoclaved saliva (AUT) as negative control, saliva from the same rumen uid donor (OWN) as positive control, and either GOAT or SHEEP saliva as experimental interventions. Fermentation was monitored throughout the 7 days of incubation and the prokaryotic communities and metabolome were analysed at day 7 of incubation. Characterization of the salivas used prior to incubation showed a high degree of individual variability in terms of the salivary metabolites and proteins, including immunoglobulins. The prokaryotic community composition in AUT incubators was the most divergent across treatments, suggesting a modulatory effect of active salivary components, which were not affected in the other treatments (OWN, GOAT and SHEEP). The differences across treatments in microbial diversity were mostly caused by a greater abundance of Proteobacteria and Rikenellacea and lower of Prevotellaceae, a key rumen bacterium with greater abundance in GOAT and SHEEP treatments. These results suggest that specic salivary components contribute to host-associated role in selecting the rumen commensal microbiota and its activity.


Introduction
The rumen of ruminant animals contains a great diversity of prokaryotic (bacteria, archaea, virus) and eukaryotic (protozoa and fungi) micro-organisms that together breakdown and ferment the feed ingested by the host animal to convert complex plant carbohydrates into short-chain volatile fatty acids (Dehority, 2003). The rumen microbial diversity and function largely in uence many animal traits such as the e ciency of utilization of feeds and the environmental impact through methane emissions (Jami et al., 2014). The digestive microbiomes in most mammals are controlled by host genetic variation (Koskella & Bergelson, 2020) through multiple processes, one of the most crucial being immune modulation, by secreting many substances from epithelial cells (i.e. antimicrobial peptides, immunoglobulins,..) and germline-encoded pattern recognition receptors (Zheng et al., 2020). However, in the rumen no organized lymphoid tissue exists in the epithelium (Sharpe et al., 1977), which includes up to a 15 cell layer, that limit the permeability of large molecules. Saliva has been suggested as main vehicle of introducing immune active metabolites in to the rumen (Yáñez-Ruiz et al., 2015).
Ruminants' saliva is secreted in large amounts and assists the animal in the process of feed lubrication, deglutition and regurgitation. Saliva constituents include a signi cant amount of ions (mainly bicarbonate and phosphate), that help maintain rumen osmotic pressure and pH within physiological range (Warner & Stacy, 1977) providing a buffered medium to allow rumen microorganisms to thrive (Faniyi et al., 2019). The protein fraction of saliva comprises a number of proteins involved in transportation and pH buffering (Cheaib & Lussi, 2013), from which albumin is found in greater amounts (Lamy et al., 2009). However, the ruminant salivary proteome also includes a complex mix of other proteins with a wide range of physiological and enzymatic functions (Ang et al., 2011). Immunoglobulins, especially secretory immunoglobulin A (IgA), modulate the proliferation of symbiotic microbiota (Fouhse et al., 2017), either inhibiting or stimulating their growth (Donaldson et al., 2018). Smaller salivary proteins, which includes a variety of cytokines (Stenken & Poschenrieder, 2015) and antimicrobial peptides (Fábián et al., 2012), have been shown to be the most discriminant in the salivary proteome across individuals and animal species (Lamy et al., 2009). In a recent in vitro batch culture study, we have shown that some speci c protein components have the ability to modulate rumen fermentation in goats (Palma-Hidalgo et al., 2021a). However, due to such speci city, the mechanisms behind the complex and modulatory interaction that takes place between salivary components and host rumen microbiota are still largely unknown.
This work aimed to assess the role of saliva on modulating the rumen fermentation and microbial diversity using a 7-days semi-continuous in vitro trial. A detailed characterization of the protein and metabolites composition of saliva from different individual animals was conducted and rumen uid from goats was incubated with different types of saliva (own animal-saliva, goat-saliva, sheep-saliva and autoclaved-saliva) to elucidate the potential modulatory effect on the rumen microbial ecosystem.
The proteomic MS/MS analysis of the saliva samples resulted in the identi cation of 195 proteins/polypeptides across the 6 samples. The average number of proteins per sample was 59, with Goat 5's saliva having the lowest count (46), and the sheep's saliva having the highest (68). The heatmap of the 30 most abundant proteins across the saliva samples showed a very variable proteomic pro le based on emPAI values ( Figure 1). On average, the protein with the greatest abundance was Thymosin beta, but its emPAI values were very variable, ranging from 0 (Goat 2) to 153 (Goat 3). Even after saliva samples were processed for albumin depletion, albumin (fragment) was still the second most abundant protein in the saliva samples. The 3rd (Beta A globin chain), 4th (II alpha globin) and 5th (I alpha globin) most abundant proteins were all hemoglobin subunits, and their abundance pattern was similar in each saliva sample. The variability in the abundance of the rest of the proteins helped determine, to some extent, the clustering pattern between the 6 samples. Interestingly, not the sheep's but Goat 3's saliva turned out to be, compared with the others, the most different sample with regards to the proteomic pro le. However, unlike all the goats' salivas, sheep's saliva did not contain goat-speci c Capra hircus Akirin 2 mRNA but it did have high values of other proteins such as Glutathione S-transferase and Insulinlike growth factor 1, which were almost not present in goat salivas.
The metabolomic MS/MS analysis identi ed 39 metabolites present in all the saliva samples. The heatmap of the 39 compounds based on mTIC values showed very unique metabolic pro les in each saliva sample (Figure 2). The peak heights of most of the compounds detected in the MS/MS spectra were very variable across the 6 samples, hence the high degree of variability observed in the constructed heatmap. Similar to what was found in the proteomic analysis, the Goat 3's saliva was, again, the one with the most different metabolomic pro le. On average, the 1st (Tetraethylene glycol), 4th (Hexaethylene glycol) and 5th (Deithylene glycol monoethyl ether) most abundant metabolites were ethylene glycol derivatives. Other abundant detected compounds include the aminoacids L-Isoleucine (2nd ) and L-Phenylalanine (8th ), as well as the choline cation (3rd ) and the Tri(3-chloropropyl) phosphate (6th ), all of them with up to 1000 fold mTIC value variability between at least two of the saliva samples. In comparison to GOAT saliva, the SHEEP saliva had lower concentrations of urea and higher concentrations of creatinine and nucleic acids derivatives such as guanine, guanosine or hypoxantine.

In vitro fermentation and microbial abundances
The fermentative activity peak was reached in the rst 12 hours of incubation, as shown by the lowest pH values and greatest gas production and VFA concentration (P < 0.001 according to sampling time). From the rst day of incubation a stable fermentative activity was observed in terms of pH and gas production, with only a slight gradual decrease in total VFA concentration as incubation time progressed (Supplementary Table 1). The incubation of rumen uid from 4 different goats with AUT, OWN, GOAT and SHEEP salivas led to substantial differences in the fermentation pattern (Table 1). Incubation with AUT led to the lowest gas production (-9.4%) and highest pH and butyrate molar proportion (+6.4%), while GOAT samples produced the lowest pH, butyrate molar proportion and acetate: propionate ratio (P < 0.001, P = 0.018 and P = 0.005; respectively). Bottles with SHEEP saliva generated the highest gas production and highest acetate: propionate ratio (+4% and + 6%; respectively) compared with the rest. Table 1 Effect of the incubation with autoclaved saliva (AUT), own saliva (OWN), goat saliva (GOAT), and sheep saliva (SHEEP) on in vitro rumen fermentation and the abundance of the major rumen microbial groups.

Microbial diversity
The sequencing analysis performed on incubation samples generated 41,514 ± 13,383 high quality prokaryotic sequences per sample. The number of sequences was normalized to 28,131 for further processing and analyses. Good's coverage index was 98.8% on average and similar for the 4 saliva treatments, hence a good level of sequencing depth was achieved. The primers used for sequencing mostly targeted bacterial amplicons, however, ~0.75% of the detected reads were identi ed as archaeal sequences. The prokaryotic alpha-diversity in terms of observed ASVs, Chao1, Shannon and Simpson indexes within the incubation bottles was not affected as a consequence of the incubation with the different types of saliva (Supplementary Figure 1).
The Venn diagram ( Figure 3) showed that a majority of the detected ASVs (959) were shared across the 4 saliva treatments. AUT was the treatment with the least overlapping ASVs with the rest of the treatments (1468 vs. 1549 vs. 1517 vs. 1494 in AUT, OWN, GOAT and SHEEP treatments, respectively).
PERMANOVA analysis showed that the differences in the prokaryotic community structure across saliva treatments were signi cant in speci c pair-wise comparisons ( Figure 4). The clearest difference was that observed between the communities in the AUT and the rest of the treatments. The level of dissimilarity between the AUT community structure and in the other three treatments (mainly SHEEP and GOAT) was also very apparent in the subsequent sPLS-DA. The component 1 axis in the sPLS-DA (explaining 5% of the total variation) sorted the AUT samples apart from the rest, whereas the component 2 (explaining 4% of the total variation) disaggregated the OWN from the other two treatments with fresh saliva (GOAT and SHEEP). PERMANOVA analysis showed no signi cant differences in the prokaryotic community structure between OWN, GOAT and SHEEP treatments.
The relative abundance of the identi ed prokaryotic taxa was moderately variable according to the saliva treatment (Supplementary Table 2). At phyla level, 4 out of 18 had signi cantly different abundances across saliva treatments. Actinobacteria (2.73% average relative abundance) was more predominant in bottles incubated with GOAT and SHEEP saliva (P = 0.0385). On the contrary, Proteobacteria (6.88% average relative abundance) was more predominant in bottles incubated with AUT and OWN saliva. Thirteen out of the 32 most abundant prokaryotic families ( Figure 5) and twelve out of the 33 most abundant genera denoted differences across the saliva treatments.

Effects of the type of saliva on the metabolomic composition
The MS-MS metabolomics analysis on in vitro incubation samples identi ed 19 compounds after processing and ltration ( Figure 6, Supplementary Table 3). On average, the most abundant metabolite was 15-Ketoprostaglandin E1, followed by ethyldiethanolamine and N-Methyl-2-pyrrolidone. Eight metabolites presented signi cantly different abundances according to saliva treatment. The heatmap based on mTIC values ( Figure 6) clustered samples rom AUT treatment separately and then separated samples from SHEEP to those from GOAT/OWN saliva treatments. This was further demonstrated, in agreement with microbial diversity results, by the signi cantly distinct metabolomic pro le in AUT treatment compared with the other three treatments and that the SHEEP metabolome was different (P < 0.005) from that of the GOAT and OWN samples.

Discussion
In our study, a thorough description of the protein and metabolite components of the animals´ saliva was achieved prior to incubation with rumen uid. The immunological pro ling of fresh saliva from the ve goats and one sheep revealed relatively low concentrations of IgA compared to the 5.95 mg/ml recently reported in bovine saliva (Fouhse et al., 2017). Even though some previous works where ELISA was not used for quanti cation (Porter & Noakes, 1970) had di culties at detecting IgA even after a 20 fold concentration using dialysis, others such as Mach & Pahud (1971) and Lascelles & McDowell (1974) reported much higher IgA saliva concentrations (560 µg/ml, 157 µg/ml; respectively) than what we detected in our study (37.6 µg/ml). As expected, the average concentration of IgA, which is the major immunoglobulin in ruminants' saliva (Lascelles & McDowell, 1974), was ~4 fold higher than that of IgG. Interestingly, even though this IgA:IgG ratio was maintained in the sheep's saliva, both concentrations were notably lower in comparison with the other salivas.
The total identi ed proteins across the salivas used in our study (195) was much greater than the 33 and 13 proteins annotated in sheep and goat saliva following a two-dimensional gel electrophoresis (2D PAGE) approach with two different spectrometry methods (Lamy et al., 2009(Lamy et al., , 2011respectively). Despite the number of annotated proteins was much higher in our study, we hypothesize that this difference could partially be caused by the utilization of a protein database such as TrEMBL which, unlike SwissProt, contains computationally annotated protein features instead of manually reviewed annotated proteins. A comprehensive study of the bovine salivary proteome where similar nontargeted MS-MS approaches were used (Ang et al., 2011) identi ed an average of 179 proteins across different sample preparation methods, which is similar to our gure and slightly closer to the hundreds of proteins identi ed in human saliva studies (Loo et al., 2010). Like in our study, variability based on different methodologies used and/or animal speci city in previous works played a signi cant role in this high rate of detected proteins. Such a wide array of salivary proteins are involved in numerous physiological functions across the animal kingdom (Mandel, 1987). Despite the inter-and intraspecies variability with regards to salivary protein components in ruminants, these proteins seem to be involved in similar physiological functions (Ang et al., 2011). Indeed, the functional pro le of the salivary proteins detected in cows (Ang et al., 2011) was pretty consistent with that found in goat and sheep proteins identi ed in our study, most of which are involved in nutrient-binding, transport, enzymatic activity and, to a less extent, immune response.
A previous in vitro study revealed that pre-incubation of speci c diets (such as tannins-rich forages) with either sheep or goat saliva had a positive effect on diet degradation when incubated with rumen uid (Ammar et al., 2013). On the other hand, other works have reported that the diet provided to ruminants and their saliva composition (including its protein fraction) have only minor effects on the rumen microbial activity (Ammar et al., 2011) and vice-versa . The lack of substantial effects found in these studies could be caused by the relatively short time of incubation (48 h) but also due to a missing exploration of the salivary proteome and metabolome, which we addressed in our study. In this context, the use of different diets or the inoculation with unique microbial strains have been suggested to induce a number of immunological mechanisms in the GIT (Yáñez-Ruiz et al., 2015). This has been reported to be of particular importance with regards to immunological proteins (mainly Ig), given that their concentration varies signi cantly depending on their rate of secretion through saliva (Subharat et al., 2016), which greatly depends on the presence of speci c microorganisms in the rumen (Sharpe et al., 1977).
The metabolomic pro le of the ruminants' saliva has not been thoroughly explored to date. In general, research on the saliva metabolome has been focused on the identi cation and characterization of salivary biomarkers that could be used as indicators for the detection of a number of diseases (Yoshizawa et al., 2013). Other studies have attempted to better assess the metabolome composition throughout the gut, and the cross-effects that might take place between this and the host microbiota Our semi-continuous incubation system reached a peak of microbial activity in the rst hours of incubation and then remained stable in terms of pH and gas production from 36 hours and thereafter. Through the last days of incubation, gas production was very low in AUT bottles compared with the rest, indicating that untreated saliva from goats or sheep contain bioactive components that enhance fermentative activity. At this stage of incubation, the saliva donor species was the most in uential factor in vitro fermentation as the SHEEP saliva promoted the highest levels of fermentative activity (+4% gas production) as well as the greatest bacterial and protozoal concentrations (+2% and +7%, respectively). (2013) when using tannins-rich substrates, which again suggests that the unique salivary composition of each species or even individuals may modulate microbial activity differently.
Incubation with AUT saliva led to the most divergent rumen microbial community in terms of overlapping ASVs with other treatments and general microbial composition. At phyla level, the relative abundance of the two main bacteria phyla across all treatments was 53% for Bacteroidetes and 30% for Firmicutes, a ratio (1.76) which is almost half (3.25) of what has been previously described in the rumen of goats (Palma-Hidalgo et al., 2021b). We hypothesize that the salivary proteins promoted the growth of distinct salivary components of different species seem to have a strongest in uence on the rumen content as they shape the rumen metabolome differently depending on whether they are constituents of goat's or sheep's saliva. Our data indicate that saliva components may partly responsible for the host species-speci c rumen microbiota and related metabolites, potentially due to the co-evolution of the microbiome and host (Koskella & Bergelson, 2020).
The characterization of sheep and goats saliva showed distinct metabolomic and proteomic pro les across individuals and animal species, though the general functions (enzymatic, transport, immune response) remain consistent. Inactivating these compounds (i.e. autoclaving) exhibit an important change in the function of saliva in shaping rumen fermentation and microbial community. This nding together with the differences observed between species suggest that the cross-talk mechanisms between salivary components and rumen microbiota can be speci c for individuals and/or species and that may contribute to the host selection of the commensal microbiota and its function.
the mouth of the animals with absorbent sponges for 5 min. Saliva was collected from the sponges by centrifuging at 190×g for 10 min, then ltrated through 0.25 µm pore size to remove microorganisms and large particles, pooled per animal and stored in aliquots at -80°C until the start of the in vitro incubation. Additionally, equal volumes of saliva from goats 1-4 were mixed, autoclaved at 121°C for 30 min and stored at -80°C (AUT). Four aliquots from each saliva (goats 1-5 and sheep) were used for immunoglobulins, proteome and metabolome analyses.

Experimental design and in vitro incubation
An in vitro semi-continuous incubation was conducted during 7 days using rumen uid from goats 1-4 sampled before the morning feeding and ltrated through a double layer of cheesecloth. Sixteen Wheaton bottles with 30 ml capacity were used in the incubation. Each rumen uid was incubated with 4 different types of saliva (n=4): saliva from the same rumen uid donor (Goats 1-4) (OWN) as positive control, saliva from goat 5 (GOAT), saliva from sheep (SHEEP) and pooled autoclaved saliva to denature active metabolites in the saliva but keeping the minerals, (AUT) from goats 1-4, as negative control. Incubations consisted in a total volume of 20 ml per bottle composed 6.67 mL of rumen uid, 6.67 mL of saliva and 6.67 mL of bicarbonate buffer (3.5 g NaHCO 3 + 0.4 g (NH 4 )HCO 3 in 100 ml dH 2 O). The same oats hay and commercial concentrate that were offered to the animals were grinded to 1mm size particles and used as incubation substrate (100 mg each).
In order to maintain an active in vitro system, every 12 hours (9.00 and 21.00), gas pressure in the headspace of the bottles was measured using a Wide Range Pressure Meter (SperScienti c LTD, Scottsdale, AZ, USA), which then was transformed into volume units by the ideal gas law. After gas measurement, bottles were opened, the content was homogenized by a gentle horizontal movement, and 1/3 of the incubation volume (6.67 ml) was removed with a syringe and used to measure the pH. The removed incubation volume was replaced by 3.33 ml of the aforementioned bicarbonate buffer, 3.33 ml of the same saliva used in each treatment, and 1/3 of diet (33 mg oats hay and 33 mg commercial concentrate). A continuous ow of CO 2 was applied to each bottle through this process to maintain the anaerobic conditions. One sub-sample (800 µl) of the removed incubation content was taken at 12, 36, 60, 84, 108, 132 and 156 hours, mixed with 800 µL of an acid solution (0.5 mol/l HCl, 200 g/l metaphosphoric acid and 0.8 g/l crotonic acid as internal standard) for volatile fatty acids (VFA) determination by gas chromatography (AutoSystem gas chromatograph, Perkin Elmer, Waltham, MA). A second sub-sample (200 µl) of the removed incubation content at day 7 was used for metabolome analysis following a similar procedure to that described for saliva samples. A third sub-sample (200 µl) taken at days 2, 4, 6 and 7 was used for DNA extraction using a commercial kit (QIAamp DNA Stool Mini Kit, Qiagen Ltd., Barcelona, Spain).

Characterization of proteins and metabolites in saliva
One aliquot of each saliva was thawed to measure the protein content by spectrophotometry using a commercial assay kit (PierceTM BCA Protein Assay Kit, Thermo Fisher Scienti c, Waltham, MA, USA). For salivary immunoglobulins A and G (IgG) quanti cation, aliquots of each saliva were thawed and centrifuged at 3,000×g for 10 min. IgA and IgG concentrations were measured using the Goat Immunoglobulin A and Goat Immunoglobulin G ELISA kits (MyBioSource, San Diego, CA, USA), respectively.
Before conducting saliva proteome analysis, albumin depletion was performed on thawed aliquots of each saliva, using the Pierce™ Albumin Depletion Kit (Thermo Fisher Scienti c

Calculations and statistical analysis
Statistical analyses were carried out using SPSS software (IBM Corp., Version 26.0, New York, USA). To assess the effect of time on the fermentative activity throughout the incubation, rumen fermentation parameters were analysed based on a repeated measures mixed effects ANOVA as follows: where Yijlm is the dependent, continuous variable, µ is the overall population of the mean, Si is the xed effect the type of saliva (i= AUT vs OWN vs GOAT vs SHEEP), Tj is the xed effect of the time (j= 12h vs 36h vs 60h vs 84h vs 108h vs 132h vs 156h sampling times), STij is the interaction term, R k is the random effect of the rumen uid and eijl is the residual error. To assess only the effect of the type of saliva used in the incubation when this became stable, rumen fermentation parameters, quantitative PCR data and microbial taxa abundances at days 6 and 7 and incubation metabolites at day 7 were analysed using an ANOVA test with the saliva treatment (AUT vs OWN vs GOAT vs SHEEP) as xed effect and the sampling times as a block. When signi cant effects were detected, means were compared by Fisher's protected LSD-test. Quantitative PCR data and microbial relative abundances were log10 transformed before the analysis to achieve a normal distribution. Only prokaryotic families & genera with relative abundance > 0.1% across saliva treatments were further considered for taxonomic analyses (in % of sequences). In all analyses, signi cant effects were declared at P < 0.05 and tendency to difference at P < 0.1.
Proteomic and metabolomic heatmaps based on emPAI and mTIC values, respectively, were constructed using RStudio (R Foundation for Statistical Computing, Vienna, Austria) to characterize the salivas before incubation and the effect of saliva on the rumen metabolome at 156h of in vitro incubation. A Permutation based Analysis of Variance (PERMANOVA) with 999 random permutations based on the Bray Curtis Dissimilarity Matrix was performed based on the mTIC values to compare the metabolomes across treatments using PAST software (Hammer et al., 2001). A Venn diagram was performed to illustrate the saliva treatment effects on the microbial community using a multiple list comparator (www.molbiotools.com). To illustrate the treatment impact on the in vitro rumen prokaryotic community, a PERMANOVA based on the Bray Curtis Dissimilarity Matrix was performed on log10 transformed sequencing data with 999 random permutations. Pair-wise comparisons were performed to compare the microbial composition across treatments. Sparse Partial Least Squares Discriminant Analysis (sPLS-DA) was carried out on center log ratio transformed sequencing data to show the effect of the treatment on the prokaryotic communities' structure.
Declarations Figure 1 Heatmap showing the abundance based on emPAI values of the 30 most abundant proteins/polypeptides found across the salivas used for the in vitro incubation.

Figure 2
Heatmap showing the abundance based on mTIC scores of the 39 detected metabolites found across all saliva samples used in the in vitro incubation.   Relative abundance of the ten most abundant prokaryotic families in rumen content incubated with autoclaved saliva (AUT), own saliva (OWN), goat saliva (GOAT) or sheep saliva (SHEEP).

Figure 6
Heatmap showing the abundance based on mTIC scores of the 19 detected metabolites found in rumen content incubated with autoclaved saliva (AUT), own saliva (OWN), goat saliva (GOAT) and sheep saliva (SHEEP).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupplementarymaterialPalmaHidalgoetalsalivaII.docx