Animal care and procedures were in accordance with the Chinese guidelines for animal welfare and approved by Animal Care and Use Committee of the Chinese Academy of Agricultural Sciences.

Six Chinese Holstein dairy cows (627 ± 16.9 kg BW; 180 ± 6 DIM) in second-parity fitted with 10 cm ruminal cannulas (Bar Diamond, Parma, ID) were allocated to a replicated 3 × 3 Latin square design. Treatments included a control diet (CON; 20% starch, DM basis), a high grain diet (HG; 33.2% starch, DM basis), and high grain diet supplemented with 180 mg thiamine/kg DMI (HG+T). The diets were formulated according to NRC (2001) to meet or exceed the energy requirements of Holstein dairy cows yielding 20 kg of milk/d with 3.5% milk fat and 3.0% true protein. Details of ingredient analysis and chemical composition of dietary ingredients were given in Table S1. Cows were fed at 06:00 and 18:00 h, one-half of the allowed daily ration at each feeding, and thiamine (thiamine hydrochloride, purity ≥99%; Wanrong Science and Technology Development Co., Ltd., Wuhan, China) was administered via the rumen cannula twice daily after diets were supplied. Three periods were included, each experimental period consisted of 21 days, with a 14-day washout between periods, during which cows were offered CON diet. Throughout the whole experimental periods, the cows were housed in individual stalls and were allowed to fresh water freely during the trial.

On day 21 of each period, approximately 500 ml of rumen contents were sampled from cranial, caudal, dorsal, and ventral aspects of the rumen at 3 h after morning feeding. Collected samples were strained through four layers of cheesecloth to obtain rumen fluid. The rumen fluid was divided into two aliquots. One aliquot (approximately 100 ml) was processed to analyze the pH value, the concentration of volatile fatty acids (VFA) and lactate as described by Pan et al. (2016). Thiamine concentration in rumen fluid was detected by high performance liquid chromatography according to Analytical Methods Committee (2000). The other aliquot (approximately 100 ml) was stored at −80°C immediately for microbial DNA extraction and further analysis.

Ruminal fluid samples were thawed at room temperature and a 1-mL aliquot was centrifuged at 10,000 × g for 1 min at 4°C and the supernatant was discarded. The pellet was used for DNA extraction using a QIAamp DNA Stool Mini Kit (Qiagen, Hilden, Germany) with the addition of a bead-beating step. Briefly, the pellet samples were homogenized with 0.5 g zirconium beads (0.5 mm in diameter) and 1.4 mL ASL buffer using a Mixer Mill MM 400 (Retsch, Haan, Germany) with vibrational frequency of 1,800 rpm and grinding time of 60 s at room temperature. Then the mixture was incubated at 70°C for 5 min to increase DNA yield. The supernatant was further processed using QIAamp kits according to the manufacturer's instructions. Extracted DNA was assessed by agarose gel (1%) electrophoresis and quantified using a NanoDrop spectrophotometer ND-1000 (Thermo Scientific, Waltham, MA, USA). DNA was stored at −80°C until further analysis.

The bacteria 16S rRNA gene was amplified using the barcoded universal primers 338F (5′-barcorde-ACTCCTRCGGGAGGCAGCAG-3′) and 806R (5′-GGACTACCVGGGTATCTAAT-3′) spanning the V3–V4 hyper variable region (Ye et al., 2016), where barcode is an eight-base sequence unique to each sample. Polymerized chain reactions (PCR) were carried out in a triplicate 20 μL mixture containing 4 μL of 5× FastPfu Buffer, 2 μL of 2.5 mM dNTPs, 0.8 μL of each primer (5 μM), 0.4 μL of FastPfu Polymerase and 10 ng of template DNA. The thermal cycling conditions was 95°C for 3 min, followed by 27 cycles at 95°C for 30 s, 55°C for 30 s, and 72°C for 45 s and a final extension at 72°C for 10 min. Amplicons were extracted from 2% agarose gels and purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA) according to the manufacturer's instructions and quantified using QuantiFluor™-ST (Promega, USA). Amplicon libraries were generated using TruSeq™ DNA Sample Prep Kit (TransGen Biotech, China) following the manufacturer's recommendations. The paired-end sequenced (2 × 250 bp) was conducted on an Illumina MiSeq platform according to standard protocols (Caporaso et al., 2010).

Raw fastq files were demultiplexed, quality-filtered using QIIME (version 1.17) with the following criteria: (i) The 300 bp reads were truncated at any site receiving an average quality score <20 over a 50 bp sliding window, discarding the truncated reads that were shorter than 50 bp. (ii) Exact barcode matching, nucleotide mismatch in primer matching and reads containing ambiguous characters were removed. (iii) Only sequences that overlap longer than 10 bp were assembled according to their overlap sequence. Reads which could not be assembled were discarded. Operational Units (OTUs) were clustered with 97% similarity cutoff using UPARSE (Edgar et al., 2011) and chimeric sequences were identified and removed using UCHIME (Haas et al, 2011). Taxonomy was aligned by RDP classifier and compared with the SILVA (SILVA version 115) 16S rRNA database (Pruesse et al., 2007) using confidence threshold of 70% (Amato et al., 2013). Community diversity was estimated with the normalized reads using the based coverage estimator (ACE), Chao1, and Shannon indices. The unweighted UniFrac distance method (Lozupone et al., 2007) was used to perform a principal coordinates analysis (PCoA), and an analysis of similarity (ANOSIM) in QIIME with 999 permutations (R Core Team, 2013) was conducted to assess significant differences between samples.

Data were checked for normal distribution and homogeneity by Shapiro-Wilk's and Levene's tests in SAS 9.2 (SAS Institute Inc., Cary, NC). Ruminal pH, VFA, thiamine content, bacterial abundance, and diversity index were analyzed using PROC MIXED of SAS 9.2 as shown in the following model: Y_ijklm = μ + T_i + P_j + S_k + C_l(S_k) + O_m + T_i × P_j + T_i × S_k + e_ijklm, where Y_ijklm is the dependent variable, μ is the overall mean, T_i the fixed effect of treatment (i = 1–3), P_j is the fixed effect of period (j = 1–3), S_k is the random effect of Latin square (k = 1–2), C_l(S_k) is the random effect of cow nested in square (l = 1–6), O_m is the fixed carryover effect from the previous period (O = 0 if period = 1), T_i × P_j is the interaction of treatment and period, T_i × S_k is the interaction between treatment and Latin square replicate, and e_ijklm is the random residual error. In these analyses, the carryover effects, and the interactions between treatment and period or square weren't detected for all variables, and they were finally removed from the model. Significance was declared at P < 0.05 and a tendency was considered at 0.05 < P < 0.10.

Pearson correlations between bacterial communities and ruminal fermentation variables or thiamine content were assessed using the PROC CORR procedure of SAS 9.2. Only those bacterial taxa with an abundance ≥0.1% of the total community in at least one ruminal sample were included in the analysis. A correlation matrix was created and visualized in a heatmap format by the HemI software (Deng et al., 2014). The abundance of bacterial communities at the genus level and ruminal variables were considered to be correlated with each other when the correlation coefficients (r) were above 0.55 and P value below 0.05 (Wang et al., 2016).

All the raw sequences were submitted to the NCBI Sequence Read Archive (SRA; http://www.ncbi.nlm.nih.gov/Traces/sra/), under accession number SRP114812.

As shown in Table 1, high grain feeding decreased ruminal pH value, and the concentrations of thiamine and acetate in rumen fluid (P < 0.05), whereas the concentrations of propionate, isobutyrate, and lactate were increased significantly (P < 0.05) compared with CON cows. The changes in ruminal pH, lactate, thiamine and propionate caused by HG-diet were inversed by thiamine supplementation (P < 0.05). Butyrate, valerate, isovalerate and total VFA weren't affected by dietary treatments (P > 0.05).

Bacterial sequencing generated a total of 929,581 raw reads. Quality filtering at 97% similarity resulted in 697,247 high quality sequences, which clustered in 1,800 OTUs with 30,384 reads per sample after normalization, with an average of 1,374 ± 95 OTUs per sample. Within the bacterial population, 21 phyla were identified across all samples. Bacteroidetes, Firmicutes, and Proteobacteria were the three dominant phyla, representing 53.41, 35.10, and 3.07% of the total sequences, respectively (Figure 1). Spirochaetae, SR1, and Fibrobacteres represented average percentages of 2.91, 2.25, and 1.02%, respectively, of the total sequences. The proportion of some phyla (Tenericutes, Saccharibacteria, Cyanobacteria, Lentisphaerae, Synergistetes, and Elusimicrobia) was less than 1% of total microbial community, and other phyla such as Actinobacteria, Fusobacteria, Chloroflexi were not consistently present in all of the ruminal samples.

In terms of alpha bacterial diversity (Table 2), no differences were observed across treatments for the OTU numbers or Good's coverage, indicating that the sequencing depth was comparable across treatments. High grain diet decreased the abundance-based coverage estimator (ACE), Chao 1 richness and Shannon diversity index (P < 0.05) compared with CON diet. Thiamine supplementation had no significant effect on alpha bacterial diversity in respect to HG diet (P > 0.05). Principal coordinates analysis (PCoA) plots based on unweighted UniFrac distance metrics were conducted to compare the three treatments. The PCoA result exhibited that cows fed with HG diet were distinctly separated from cows in the CON and HG+T groups (Figure 2). The further ANOSIM analysis revealed a strong effect of high grain feeding on the structure of the bacterial community (R = 0.35, P = 0.01); Significant difference in bacterial community composition between HG and HG+T group was observed (R = 0.25, P = 0.02). Principal coordinate 1 and 2 accounted for 33.65 and 17.9% of the total variation, respectively.

The effects of high grain feeding and thiamine supplementation on main phyla were illustrated in Figure S1. High grain feeding increased the abundance of Firmicutes and decreased the abundance of Proteobacteria compared with other two treatments (P < 0.05). The decrease in Proteobacteria phyla under HG treatment was reversed by thiamine supplementation (P < 0.05), and the abundance of Spirochaetae was increased by thiamine supplementation significantly (P < 0.05). The other predominate phyla including Bacteroidetes, SR1 and Fibrobacteres weren't affected by high grain feeding and thiamine supplementation (P > 0.05).

At the genus level, a total of 218 taxa were examined. The taxa with a relative abundance of ≥0.1% in at least one sample were further analyzed and the influenced genera by treatments were listed in Table 3. A total of 26 genera were affected (P < 0.05) by treatments; 19 genera of these were changed by HG feeding in respect to CON diet, and 15 genera were affected by thiamine supplementation compared with HG group. Specifically, compared with CON group, HG feeding decreased (P < 0.05) the proportions of Bacteroides, unclassified Prevotellaceae, unclassified Bacteroidetes, Erysipelotrichaceae UCG-004, Fibrobacter, uncultured Lachnospiraceae, Eubacterium ventriosum, Ruminococcus 1, Ruminiclostridium 5, Selenomonas 1, Acetobacter, Succinivibrionaceae UCG-002, unclassified Succinivibrionaceae, Succinivibrio, Ruminobacter, and Pyramidobacter. On the contrary, the relative abundance of Succiniclasticum (8.89 vs. 12.45%), Ruminococcaceae NK4A214 (1.71 vs. 2.73%), and Lachnospira (0.06 vs. 0.13%) were significantly increased in HG group (P < 0.05). The decrease in genera of family Succinivibrionaceae and Bacteroidetes, Ruminococcus 1 and Pyramidobacter caused by high grain feeding were reversed by thiamine supplementation (P < 0.05), and the proportion of Treponema 2 was also significantly increased in HG+T group (2.74 vs. 4.22%; P < 0.05). Conversely, there was a significant decrease in relative abundance of Succiniclasticum, Christensenellaceae R7, Ruminococcaceae NK4A214, Saccharofermentans and Papillibacter (P < 0.05) associated with thiamine supplementation when compared with HG group.

As shown in Figure 3, the relative abundance of genera Bacteroides, Ruminococcus 1, Ruminobacter, Pyramidobacter, and Fibrobacter were positively correlated to ruminal pH and thiamine concentration (r > 0.55, P < 0.05), but negatively correlated to ruminal lactate content (r < −0.55, P < 0.05) except for Ruminobacter and Fibrobacter. Conversely, the genus Succiniclastium was negatively correlated with both ruminal pH and thiamine content (r < −0.55, P < 0.05), and positively correlated with lactate concentration (r > 0.55, P < 0.05). The acetate concentration was positively related (P < 0.05) with unclassified Prevotellaceae (r = 0.65), Oribacterium (r = 0.59), Succinivibrionaceae UCG-002 (r = 0.73), unclassified Succinivibrionaceae (r = 0.68), Succinivibrio (r = 0.63), and Ruminobacter (r = 0.55), but negatively related with Succiniclasticum (r = −0.69). The relative abundance levels of 13 taxa were correlated (P < 0.05) with the ruminal propionate concentration [five positive (Succiniclastium, Christensenellaceae R7, Schwartzia, Treponema 2, Ruminococcaceae NK4A214) and eight negative (Bacteroides, Eubacterium ventriosum group, Ruminococcus 1, Ruminiclostridium 5, Succinivibrionaceae UCG-002, Ruminobacter, Pyramidobacter, Fibrobacter)]. The concentration of isobutyrate was negatively correlated (P < 0.05) with Eubacterium ventriosum group (r = −0.60), Ruminococcus 1 (r = −0.69), Acetobacter (r = −0.76), unclassified Succinivibrionaceae (r = −0.59), and Ruminobacter (r = −0.61). No significant correlations were found (|r| < 0.55, P > 0.05) between the relative abundance of genera and butyrate, isovalerate, or total VFA.

In this study, we found that thiamine supplementation in HG diet could help improve rumen fermentation through increasing ruminal pH and acetate content, and reducing the accumulation of lactate in the rumen (Table 1). The improved rumen environment is mainly attributed to the stabilized ruminal microbial community (Pinloche et al., 2013). To understand the underlying mechanism of thiamine on rumen fermentation, the response of bacterial community to thiamine supplementation under high grain diet were investigated using high-throughput sequencing.

At phyla level, the high grain fed cows had higher abundance of Firmicutes and lower Proteobacteria compared with CON cows, since Proteobacteria are sensitive toward low pH and Firmicutes can still degrade easily fermentable carbohydrates under low pH (Wetzels et al., 2015). While the phyla changes above were reversed by thiamine supplementation, indicating thiamine could stabilize the bacterial community in the rumen. Besides, we found that thiamine supplementation significantly increased the relative abundance of Spirochaetae in respect to HG group. The possible reason is that most bacteria in Spirochaetae lack de novo thiamine biosynthesis pathway and need exogenous thiamine for growth (Bian et al., 2011).

At the genus levels, the thiamine treatment group had a higher proportion of the cellulolytic bacteria, including Bacteroides, Ruminococcus 1 group, Succinivibrio and Pyramidobacter, which may enhance the fiber degradation in the rumen. The possible reason for the enrichment of Bacteroides and Ruminococcus 1 group is that thiamine is an essential factor for their growth. Macy and Probst (1979) had certified that thiamine participated in the formation of α-ketoglutarate from succinic acid in Bacteroides ruminicola. Thiamine is essential for the growth of some strains in Ruminococcus, such as Ruminococcus albus (Bryant and Robinson, 1961) and Ruminococcus flavefasciens (Ungerfeld et al., 2009). Succinivibrio could promote the digestion of cellulose and hemicellulose (Sun et al., 2016), the increase of Succinivibrio may be caused by the increasing ruminal pH with thiamine supplementation, since Succinivibrio is positively correlated to the ruminal pH (Wetzels et al., 2016). Besides, Pyramidobacter species are vital cellulolytic bacteria and produce acetate as the main fermentation product (Bainbridge et al., 2016), the enhanced Pyramidobacter by thiamine could support fiber degradation, which explained the increasing ruminal acetate content in HG+T group. Moreover, Selenomonas 1 group in HG-fed cows tended to be stimulated by thiamine supplementation (P = 0.13). As a result, the increasing Selenomonas promoted the degradation of lactate from high starch diet, and thus helped to increase ruminal pH and prevented the occurrence of SARA in cattle. Above all, thiamine supplementation stimulated the growth of cellulolytic and lactate utilizing bacteria, thus increased ruminal pH and acetate production and subsequently improved rumen fermentation.

As the primary succinate utilizing bacteria, Succiniclasticum accounted for 12.45% of total bacterial community in HG group and increased significantly compared with CON group. The increase of Succiniclasticum was associated with the more production of succinate from starch degradation (Hook et al., 2011). Interestingly, thiamine supplementation reduced the proportions of Succiniclasticum. The possible reason was that thiamine, as the essential cofactor of pyruvate decarboxylation, stimulated the decarboxylation of pyruvate to produce acetate (Rastogi, 2010), thus reduced the production of succinate and subsequently decreased the proportion of Succiniclasticum in rumen content. In addition, thiamine supplementation decreased the abundance of Christensenellaceae R7 and Ruminococcaceae NK4A214. However, little information about these two genera has been reported in the literature, the reasons for the altered status of genera by thiamine supplementation are unclear.

As the coenzyme of pyruvate dehydrogenase and α-ketoglutarate dehydrogenase, thiamine plays a critical role in carbohydrate metabolism and is essential for normal cellular functions and growth (Said et al., 1999). The requirement of thiamine in dairy cows may enhance with the increasing carbohydrate levels in high grain diet. In the present study, the concentration of ruminal thiamine in HG-fed cows was lower than its content in CON cows (Table 1), which further proved that overfeeding high grain diet alters thiamine status (Pan et al., 2016). In order to explain the altered thiamine status during high grain feeding and to reveal possible bacteria related to thiamine metabolism, we conducted the correlation analysis between ruminal thiamine concentrations and microbial composition. The results showed that ruminal thiamine content was positively correlated with the genera Bacteroides, Ruminococcus 1, Ruminobacter, Pyramidobacter, and Fibrobacter, suggesting those genera may involve in thiamine synthesis, or thiamine is required for their growth and metabolism. Magnusdottir et al. (2015) found that Firmicutes can't synthesis thiamine monophosphate, and thiamine synthesis is most prevalent in Bacteroidetes and Fusobacteria. Silverman and Werkman (1939) reported that certain propionate—producing bacteria make thiamine or its intermediates, and Louis et al. (2014) pointed out that the main propionate production pathway is the succinate pathway, which is used by Bacteroidetes to generate propionate from carbohydrates. The literatures above proved that Bacteroidetes is highly related to thiamine synthesis, and thiamine biosynthesis gene has been identified in Bacteroidetes fragilis 638R (Veeranagouda et al., 2014). Fibrobacter succinogenes, as a major species in genus Fibrobacter, possess genes that encoded proteins involved in thiamine synthesis (Qi et al., 2008). For Pyramidobacter, the species Pyramidobacter piscolens strain W5455 can synthesize thiamine by salvage pathway (BioCyc database; Caspi et al., 2015). Taken together, we deduced that Bacteroidetes, Fibrobacter, and Pyramidobacter play important roles in ruminal thiamine biosynthesis, and the altered thiamine status by high grain diet maybe related to the decreasing proportions of genera Bacteroidetes, Fibrobacter, and Pyramidobacter in the rumen.

Conversion of succinate to propionate was identified in genera Succiniclasticum (Van Gylswyk, 1995). During high grain feeding, Succiniclasticum increased to stabilize the rumen environment by degrading succinate to propionate (Hook et al., 2011; Liu et al., 2015). In our study, the increasing proportions of Succiniclasticum and decreasing thiamine concentrations leaded to their negative correlations, however, little information about thiamine effect on Succiniclasticum was reported in literature, their negative relationship is difficult to elucidate. The possible reason was that thiamine is the essential cofactor during pyruvate decarboxylation (Rastogi, 2010), the increasing thiamine content could promote the conversion of pyruvate to acetyl CoA, thus reduced the amount of succinate from pyruvate and subsequently the decreasing abundance of Succiniclasticum. Overall, alterations in bacterial populations under high grain feeding can interfere with thiamine synthesis in the rumen, and thiamine supplementation in HG diet helps to stabilize the structure of bacterial community in rumen fluid.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.