Milk permeate was obtained from the agricultural cooperative “Pienas LT” (Biruliskes, Lithuania) and stored at −18°C before fermentation. Zokaityte et al. (7) reported characteristics of the fermented MP, which are given in Supplementary material S1 Characteristics of the fermented milk permeate (Supplementary Tables S1–S3).

The study was conducted on a pig breeding farm in the Klaipeda district (Vanagai, Lithuania) and the Institute of Animal Rearing Technologies, Lithuanian University of Health Sciences (Kaunas, Lithuania). The experiment followed the guidelines outlined in the Republic of Lithuania Act (19). A 25-day experiment was conducted using 36 1-day-old Topigs Norsvin Yorkshire piglets, with 12 piglets assigned to each group. Piglets were selected from three litters of third-parous sows with very similar reproductive performance. The 36 piglets were mixed and allocated to each group based on their weight to ensure uniformity. Each group included three smaller, six medium, and three larger piglets housed in farrowing pens with third-parous sows. Three dietary treatments were compared: (i) C – control group, after day 7th fed with full-fledged combined pre-starter feed for piglets PANTO® pre (Hamburger Leistungsfutter GmbH, Hamburg, Germany); treated groups, from the 1st day of life and, additionally to the full-fledged combined pre-starter feed for piglets, after the 7th day of life, received 25 mL of fermented MP: (ii) group MPPp received MP fermented with P. pentosaceus strain; (iii) group MPPa received MP fermented with P. acidilactici strain. Wet feed was prepared by diluting the feed with water (ratio 1:3), and, for both treated groups, for each piglet, additionally, 25 mL of fermented MP was added to a water–feed mixture. The farrowing crates (containing a heated creep mat for piglets) were 4.0 m² (1.6 m × 2.5 m), of which 0.20 m² was for piglets and 1.50 m² was for sows. The groups were formed on an analogue basis. The sows were fed with compound feed (in accordance with NRC requirements) (20). Drinking water was available ad libitum throughout the trial. The trial started with piglets at an initial body weight of 1.56 kg in all (control and both treatment) groups.

The piglet’s diet before the trial consisted of 17.5% crude protein, 4.20% crude fibre, 10.8% crude fats, 1.45% av. lysine, 0.55% av. methionine, 0.60% Ca, and 0.60% av. P. No antibiotic treatment was administered. The principal scheme of the experiment is shown in Figure 1.

The basal feed was formulated according to the nutritional requirements described in (20). The basal feed composition and nutritional value are shown in Table 1. Dietary contents were analysed according to the AOAC recommendations (21).

Individual body weight (BW) gain was recorded every day of age manually with an electronic MS Schippers weighing platform 100 kg scale (model type: PSSST, BOSCHE GmbH & Co. KG, Germany). The feed conversion ratio (FCR) was calculated based on feed intake (87% of dry matter) and BW gain, which was recorded on the same day as BW gain.

Only six samples were taken from each group to avoid stressing newborn piglets. Plasma biochemical variables were evaluated on the 2nd and 25th days of the piglets’ lives. Before the morning feeding, piglets were bled from the jugular vein into vacuum blood tubes (BD Vacutainer, United Kingdom). Tubes with clot activators were used for biochemical examination. The parameters included aspartate aminotransferase, alanine aminotransferase, and immunoglobulins IgA, IgM, and IgG, which were analysed with an automatic biochemistry analyser in the accredited laboratory ‘Anteja’ (Klaipeda, Lithuania).

Evaluation of faecal pH and colour characteristics was performed at the end of the experiment because, on the 1st day of piglets’ lives, there was not enough sample to perform the above-mentioned analyses. The faecal pH was analysed with a pH meter (Inolab 3, Hanna Instruments, Italy). The dry matter (DM) of the faeces was evaluated after drying the samples at 103 ± 2°C to a constant weight. Texture hardness was measured as the energy required for sample deformation (CT3 Texture Analyzer, Brookfield, Middleboro, MA, United States).

Faecal samples were collected from 12 piglets from each group before (at day 1) and after (at day 25) the experiment, stored in vials (+4°C) with transport medium (Faecal Enteric Plus, Oxoid, Basingstoke, UK) and analysed on the same day. LAB, total enterobacteria (TEC), and yeast/mould (Y/M) counts were evaluated following the methods described by Zavistanaviciute et al. (22).

After the piglets were divided into groups, faeces from 12 piglets from each group were taken, and a single pooled sample was prepared for microbiome profiling using next-generation targeted sequencing of 16S rRNA. After the experiment, faeces from 12 piglets from the control and both treated groups were collected (36 samples in total), and three pooled samples representing each group were prepared. Samples were stored in DNA/RNA Shield (1:10 dilution; R1100-250, Zymo Research, United States) at −70°C until DNA extraction. ZymoBIOMICS^®-96 MagBead DNA Kit (Zymo Research, Irvine, CA) was used to extract DNA using an automated platform according to manufacturer instructions. Bacterial 16S ribosomal RNA gene-targeted sequencing was performed using the Quick-16S^™ NGS Library Prep Kit (Zymo Research, Irvine, CA). Primers of the V3–V4 region, 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806R (5′-GGACTA CNNGGGTATCTAAT-3′), were used to amplify 16S rDNA to investigate microbial composition. The sequencing library was prepared using real-time PCR. The final PCR products were quantified with qPCR fluorescence readings and pooled together based on equal molarity. The final pooled library was cleaned with the Select-a-Size DNA Clean & Concentrator^™ (Zymo Research, Irvine, CA), then quantified with TapeStation^® (Agilent Technologies, Santa Clara, CA) and Qubit^® (Thermo Fisher Scientific, Waltham, WA). The final library was sequenced on Illumina^® Nextseq^™ with a P1 reagent kit (600 cycles). The sequencing was performed with a 30% PhiX spike-in. Unique amplicon sequence variants were inferred from raw reads using the DADA2 pipeline (23). Potential sequencing errors and chimeric sequences were also removed with the Dada2 pipeline. The taxonomy assignment was performed using Uclust from Qiime v.1.9.1 with the Zymo Research Database, a 16S database that was internally designed and curated as a reference. Sequences were deposited at the NCBI database by the access number PRJNA 1205751. Composition visualisation and alpha-diversity analysis were performed with Qiime v.1.9.1 (24). The number of genome copies per microlitre DNA sample was calculated by dividing the gene copy number by an assumed number of gene copies per genome.

Faeces were prepared for gas chromatography (GC) analysis using solid-phase microextraction (SPME). An SPME device with Stableflex (TM) fibre, coated with a 50-μm DVB-PDMS-Carboxen™ layer (Supelco, United States), was used for sample preparation. For gas chromatography–mass spectrometry (GC–MS), a GCMS-QP2010 (Shimadzu, Japan) was used. The gas chromatograph was equipped with an AOC-5000 Plus Shimadzu autosampler, upgraded with an SPME analysis kit. Analysis was performed according to the procedure described by Vadopalas et al. (25).

A paired samples t-test was used to compare differences in parameter means between the groups (C, MPPp, MPPa). For growth performance, daily data were collected (n = 12 from each group); for plasma parameters, samples were analysed at the beginning and at the end of the experiment (n = 6 from each group). The pH and dry matter of piglet faeces were analysed at the end of the experiment (n = 12 from each group). Microbiological parameters of faecal samples were evaluated at the beginning and at the end of the experiment (n = 12 from each group).

For metataxonomic analysis of bacterial composition in the faecal samples, samples were collected from 12 piglets per group at the beginning of the experiment, with a single pooled sample prepared for microbiome profiling. At the end of the experiment, faecal samples from 12 piglets from each group (C, MPPp, and MPPa) were collected, and three pooled samples (one per group) were prepared and analysed. For the faecal sample VC profile, individual analysis was conducted on 12 samples from 36 piglets at the beginning of the experiment, and at the end, 12 samples from each group were analysed. The influence of diet was determined using multivariate tests of between-subjects effects, with baseline measurements used as covariates to account for experimental conditions. Mean values were compared using Duncan’s multiple range post hoc test, with significance set at a p-value of ≤0.05. Results are presented in the tables as mean values with pooled standard errors. Additionally, Pearson’s correlations between characteristics were calculated, and the strength of correlations was interpreted following Evans (26).

Correlations were deemed significant at a p-value of ≤ 0.05. Differences in bacterial genera/species between the groups were assessed using the Z-test calculator for two population proportions (27). The number of reads for each genus/species was counted from the total reads in the samples, and the relative abundance was compared between the groups. A two-tailed hypothesis was applied.

A standard (D6300, Zymo Research, Murphy Ave., Irvine, CA, United States) of mixed known bacterial cultures was sequenced alongside sample sequencing for the quality control of taxonomical identification. Statistical comparisons were considered significant at a p-value of ≤ 0.05. Heatmap visualisation and analysis were conducted using the R statistical programming software package “ComplexHeatmap” (version 2.14.0). Partial least squares discriminant analysis (PLS-DA) and variable importance of projection (VIP) analyses were conducted using the “mixOmics” package (version 6.22.0). Volcano plots were generated with the “EnhancedVolcano” package (version 1.16.0).

The piglets’ body weight, measured every day of their life, is shown in Figure 2. As could be seen from the data, despite that, significant differences in their body weight were not found from the 1st to the 7th day of piglets’ lives. From the 8th day of piglets’ lives till the end of the experiment, the highest body weight showed MPPa group piglets compared to the control and MPPp groups. Significant body weight differences were not established throughout the whole experiment period between the control and MPPp group piglets.

The highest feed conversion ratio (0.18, calculated on day 25) showed the MPPp group (Figure 3). Control and MPPa group’s feed conversion ratios, in comparison with MPPp, were found to be lower (0.15 and 0.16, respectively). Finally, further studies were performed to identify differences in piglet plasma and faecal parameters and their possible relations with the growth performance in different animal groups.

Plasma parameters of 1- and 25-day-old piglets are shown in Table 2. In comparison, IgA concentration in all groups at the beginning and the end of the experiment was <0.33. In comparison, IgM concentration in piglets’ plasma at the beginning of the experiment, significantly lower IgM concentration in MPPa piglets’ group plasma was found compared with control and MPPp groups. At the end of the experiment, the highest IgM concentration showed MPPp group piglet plasma samples, and the diet was a significant factor in IgM concentration in piglet plasma (p = 0.046). At the experiment’s beginning and end, the differences in IgG concentration between the different group samples were not established. However, in comparison, IgG concentration at the beginning and the end of the experiment, at the end of the experiment, IgG concentration in control, MPPp, and MPPa groups plasma samples were, on average, 7.30, 7.34, and 6.24 times, respectively, lower. Significant differences between the concentration of alanine aminotransferase and aspartate aminotransferase at the experiment’s beginning and end were not found, and the diet was not a significant factor in these plasma parameters.

The pH, dry matter, and texture hardness of piglet faeces at the end of the experiment (on the 25th day of life) are shown in Table 3. Significant differences between the faeces’ pH at the end of the experiment were not found, and the diet was not a significant factor in the faeces’ pH. However, significantly higher dry matter content and, on the opposite end, significantly lower faeces texture hardness were found in both treated groups in comparison with the control samples. A moderate negative correlation was established between faeces texture hardness and dry matter (r = −0.471, p = 0.004).

Microbiological parameters of piglets’ faecal samples are shown in Table 4. Diet was a significant factor in all analysed microorganisms’ group numbers in piglets’ faeces (p < 0.001).

At the beginning of the experiment, the highest TEC was found in the faeces of the MP-Pp group (7.87 log₁₀ CFU/g), while the daeces of the control and MPPa groups showed, on average, 26.2 and 18.4%, lower TEC, respectively. However, at the end of the experiment, in comparison to TEC in the control and both treated groups, higher TEC was established in both treated group faeces (on average, by 15.5% higher than that in the control group). At the end of the experiment, TEC showed a very strong negative correlation with faeces texture hardness (r = −0.832, p < 0.001). The MPPp group exhibited the highest LAB count, increasing from 4.74 log10 CFU/g at the start to 8.13 log10 CFU/g at the end of the experiment. At the end of the experiment, in the control group, faeces LAB count was, on average, 37.2 and 30.0%, respectively, lower than that in the MPPp and MPPa groups. A very strong positive correlation was established between TEC and LAB count at the end of the experiment (r = 0.827, p < 0.001). Also, LAB count showed a positive moderate correlation with faeces dry matter content (r = 0.450, p = 0.006) and a very strong negative correlation with faeces texture hardness (r = −0.849, p < 0.001). At the beginning of the experiment, the highest Y/M count was found in MPPp group faeces (4.93 log₁₀ CFU/g). However, at the end of the experiment, this group showed the lowest number of Y/M (on average, 41.5% lower than that in the control group and, on average, 22.2% lower than that in the MP-Pa group). A very strong negative correlation was found between the Y/M and LAB counts at the end of the experiment (r = −0.953, p < 0.001). Moreover, a strong negative correlation between Y/M and TEC was established (r = −0.744, p < 0.001).

Further metataxonomic studies were performed to identify differences in the faeces’ microbial composition and their possible relations with the growth performance in different animal groups.

Figure 4 demonstrates bacterial composition up to the species level in 1-day-old piglet faeces before the feeding trial.

As can be seen from the data in Figure 4, two species – Clostridium perfringens and Escherichia coli – were the most prevalent in the 1st-day-old piglets, independent of the group. Those two species had a prevalence of 60.2 and 78.6%, respectively. The other most prevalent bacterial species were Streptococcus hyointestinalis, Lachnoclostridium sp., Clostridium cadaveris, C. celatum, Streptococcus sp., and Terrisporobacter sp. One species of lactobacilli (Lactobacillus mucoasae) had a prevalence of 2.6% among all bacteria only in one group (the control group) of the piglets, whereas there were no separate Lactobacillus species with a prevalence of ≥1% in the experimental groups. Overall, the Lactobacillus genus had a prevalence of 0.3% (MPPa group) and 2.8% (control group). Before the experiment, alpha diversity in all groups of piglets was low compared to the diversity observed after the experiment (Figure 5). At the end of the experiment, higher alpha diversity was observed in the MPPa group compared to the control and MPPp groups.

Microbial diversity between the groups (beta diversity) is presented in Figure 6.

Before the feeding experiment, bacterial diversity was very similar between the control and MPPp groups but rather different from the MPPa group. At the end of the experiment, beta diversity between the control group and both experimental groups demonstrated different taxonomic patterns. However, the differences between the MPPa and MPPp groups were also obvious. More detailed taxonomic composition at a genus level after the feeding experiment (day 25th) is presented in Table 5.

On the 25th day, in all groups, the most prevalent bacteria were from the family Ruminococcaceae with an unestablished genus and had a prevalence from 14.9 to 21.0% from all bacterial compositions. The second most prevalent family was Christensenellaceae, but only in the control and MPPa groups, which had prevalences of 10.8 and 10.6%, respectively. In the MPPp group, the second most prevalent genus was Bacteroides, with a prevalence of 12.8%. Bacteria from this genus also had a high prevalence in other groups as well. High differences (p < 0.05) were detected among the amounts of Lactobacillus, where the highest numbers (6.8% from the total bacterial amount) were recorded in the MPPp group, whereas the prevalence of this genus was 3 and 5.7 times lower in the MPPa and C groups, respectively. When comparing bacterial genera variety in control and experimental groups, statistically significant differences were mostly detected between the control and MPPp group but not the control and MPPa group (Table 5). The most prevalent bacteria detected at a species level on day 25 are presented in Figure 7.

Metataxonomic data showed complex bacterial communities in 25-day-old piglets without clearly dominant species compared to newborn piglets; however, the overall composition differs in each animal group. Statistically significant results on bacterial prevalence at a species level between the control and both experimental groups were detected towards Bacteroides fragilis, Subdoligranulum sp., and Ruminococcaceae sp., with the highest prevalence of these microorganisms in the control group, whereas Bacteroides vulgatus has significantly lower prevalence in comparison with both experimental groups. When comparing bacterial species prevalence in the control group with separate experimental groups, the main differences were observed between the control and MPPp groups, where statistically higher prevalence was detected towards Clostridium septicum, Methanobrevibacter smithii, Marvinbryanthia sp., Christensenellaceae sp. and Lachnobacteriaceae sp., and the lower prevalence of Bacteroides pyogenes, B. massilensis, Pyramidobacter piscolens, and Lactobacillus gallinarum in the control group.

When comparing the MPPa and MPPp groups, statistically significant results were observed towards Parabacteroides sp., Methanobrevibacter smithii, Clostridium septicum, Marvinbryanthia sp., Terrisporobacter sp., Christensenellaceae spp., and Lachnospiraceae spp. with the higher prevalence in the MPPa group, whereas Bacteroides pyogenes, B. massilensis, Lachnoclostridium sp., E. coli, Pyramidobacter piscolens, Ruminococcus sp., and Lactobacillus gallinarum were more prevalent in the MPPp group.

The VC profiles of piglet faeces are shown in Figure 8 (VC distribution in the faeces of all piglet groups on days 1 and 25) and Figure 9 (a – PLS-DA plot of the detected VCs in piglet faecal sample groups; b – a VIP score bar plot of detected VCs in piglet faecal sample groups; c – VC distribution in control piglets on day 1 vs control piglets on day 25; d – VC distribution in control vs MPPp groups on day 25; e – VC distribution in control vs MPPa groups on day 25; f – VC distribution in MPPa vs MPPp groups on day 25).

As shown in Figure 8, a greater variety of VC in piglet faeces was identified at the end of the experiment. At the beginning of the experiment, several VCs were not detected in the faeces of all piglet groups, including acetic acid, propionic acid, 4-methylvaleric acid, hexanoic acid, heptanoic acid, 4-ethylphenol, octanoic acid, benzeneacetic acid, 1,3-di-tert-butylbenzene; non-anoic acid, hydrocinnamic acid, and β-gurjunene.

The main compounds in the piglet faeces VC profile included the following (which content was higher then 10% from the total VC):

1) Butanoic acid: Predominant at the beginning of the experiment, 2) 2-methylbutyric acid: At the end of the experiment, its content exceeded 20% of the total VC in all groups, 3) p-Cresol: Represented 10% of the total VC at the beginning of the experiment and increased to over 35% at the end of the experiment across all groups, 4) Skatole: Identified in all groups at the end of the experiment, and 5) Hexadecanal: Predominant at the beginning of the experiment.

PLS-DA analysis (Figure 9A) clearly separated the C1d (control group day 1) sample group from the other sample groups. Moreover, the MPPp25 (day 25) and MPPa25 (day 25) groups were notably separated. Phenolic metabolites, such as p-cresol in this instance, are associated with bacterial degradation of specific amino acids, such as tyrosine. The C25d (control group day 25) samples, however, exhibited much greater scattering, and their group overlapped with the aforementioned sample groups. This higher scattering in the C1d and C25d sample groups suggests higher variability in the individual VCs. In contrast, the MPPp25 and MPPa25 sample groups showed the opposite trend, suggesting a much more stable volatile profile in the faecal samples.

The VIP score analysis (Figure 9B) identified 19 analytes (VIP ≥ 1) as the most significant contributors to the variation between the sample groups. The top three VCs contributing most to group separation were p-cresol, pentadecanol, and skatole. Other significant VCs included various alkanes (such as tetradecane and hexadecane), aldehydes (such as decanal and tetradecanal), acids (such as acetic acid and butanoic acid), and esters (such as n-amyl isovalerate).

At the end of the experiment, the faecal VC profile of the control piglet group showed lower quantities of dodecanal, butanoic acid, hexadecane, decanal, 2-ethyl-3-hydroxyhexyl 2-methylpropanoate, octadecane, cis-9-hexadecenal, hexadecanal, nonanal, tetradecane, tetradecanal, dodecane, and pentadecanol (Figure 9C). In contrast, the quantities of 2-methylbutyric acid, 2,6-di-tert-butyl-4-methylphenol, skatole, and p-cresol were increased.

In comparison, no significant differences in individual VCs were observed between the control and MPPp groups (Figure 9D). However, when comparing the VC profiles of the control and MPPp groups with the MP-Pa group at the end of the experiment, significant differences were observed in the faecal VC profile of the MPPa (Figures 9E,F). Specifically, when comparing the control and MPPa groups, the MPPa group showed higher butanoic acid content but lower quantities of 11-dodecen-1-yl acetate, 2-ethyl-3-hydroxyhexyl 2-methylpropanoate, hexanoic acid, pentadecanal, decanal, hexadecanal, tridecanal <n->, tetradecane, heptadecane, tetradecanal, hexadecane; dodecanal, and 2,6-di-tert-butyl-4-methylphenol (Figure 9E).

In the comparison of MPPp and MPPa groups, the MPPa group exhibited higher butanoic acid and indole contents but lower quantities of decanal, nonanoic acid, octadecane, tetradecanal, 11-dodecen-1-yl acetate, dodecanal, tridecanal <n->, tetradecane, heptadecane, 2-ethyl-3-hydroxyhexyl 2-methylpropanoate, 2,6-di-tert-butyl-4-methylphenol, pentadecanal, and hexadecane (Figure 9F).

In the MPPa group, significantly lower quantities of decanal, tetradecanal, 11-dodecen-1-yl acetate, dodecanal, tridecanal <n->, tetradecane, heptadecane, 2-ethyl-3-hydroxyhexyl 2-methylpropanoate, 2,6-di-tert-butyl-4-methylphenol, pentadecanal, and hexadecane were observed compared to both the control and MPPp groups. Moreover, the MPPa group showed lower quantities of hexanoic acid and hexadecanal compared to the control group and lower quantities of nonanoic acid and octadecane compared to the MPPp group. Despite these significant differences in VC compounds between the MPPa and the other groups, further studies are warranted to clarify the metabolic pathways associated with these compounds and their relationship to the improved growth performance observed in the MPPa group.