Changes in the Intestinal Microbiome during a Multispecies Probiotic Intervention in Compensated Cirrhosis

Probiotics have been used in trials to therapeutically modulate the gut microbiome and have shown beneficial effects in cirrhosis. However, their effect on the microbiome of cirrhosis patients is not fully understood yet. Here, we tested the effects of a multispecies probiotic on microbiome composition in compensated cirrhosis. The gut microbiome composition of 58 patients with compensated cirrhosis from a randomized controlled trial who received a daily dose of multispecies probiotics or placebo for six months was analysed by 16S rRNA gene sequencing. Microbiome composition of patients who received probiotics was enriched with probiotic strains and the abundance of Faecalibacterium prausnitzii, Syntrophococcus sucromutans, Bacteroides vulgatus, Alistipes shahii and a Prevotella species was increased in the probiotic group compared to the placebo group. Patients who had microbiome changes in response to probiotic treatment also showed a significant increase in neopterin and a significant decrease in faecal zonulin levels after intervention, which was not observed in placebo-treated patients or patients with unchanged microbiome compositions. In conclusion, multispecies probiotics may enrich the microbiome of compensated cirrhotic patients with probiotic bacteria during a six-month intervention and beneficially change the residential microbiome and gut barrier function.


Introduction
Probiotics is a collective term used for live organisms (usually bacteria or fungi) that can exert health benefits on their host if consumed in sufficient amount [1]. Probiotic bacteria in fermented foods have long been a part of human nutrition but have been neglected in Western style diets [2]. end, the most statistically significant balances were examined to define what taxa constitute those partitions. Associations between the microbiome and clinical parameters were evaluated with Spearman-correlations and between-group differences were assessed with Kruskal-Wallis test and Bonferroni multiplicity correction, performed in SPSS for Windows Version 23 (SPSS Inc., Chicago, IL, USA). p-value < 0.05 was considered to be statistically significant. Visualization was done in GraphPad Prism 6 (GraphPad Software, San Diego, CA, USA).

Results
For this study, patients with compensated cirrhosis were selected from a well characterized cohort that underwent a six-month probiotic intervention and were then followed up for another six months without intervention [26]. In total, 58 patients were analysed, 26 were allocated to the probiotic group and 32 to the placebo group. Patients' characteristics are given in Table 1, enrolment is summarized in Figure 1. constitute those partitions. Associations between the microbiome and clinical parameters were evaluated with Spearman-correlations and between-group differences were assessed with Kruskal-Wallis test and Bonferroni multiplicity correction, performed in SPSS for Windows Version 23 (SPSS Inc., Chicago, IL, USA). p-value < 0.05 was considered to be statistically significant. Visualization was done in GraphPad Prism 6 (GraphPad Software, San Diego, CA, USA).

Results
For this study, patients with compensated cirrhosis were selected from a well characterized cohort that underwent a six-month probiotic intervention and were then followed up for another six months without intervention [26]. In total, 58 patients were analysed, 26 were allocated to the probiotic group and 32 to the placebo group. Patients' characteristics are given in Table 1, enrolment is summarized in Figure 1. . Figure 1. Enrolment scheme.
In total 10.998.425 and 14.107.620 sequencing reads were analysed in the probiotic and placebo group, respectively. After quality checking and denoising, the number of sequences was reduced to 6.967.396 in probiotic and 8.947.492 in placebo group. Around 17% of the reads in each group were not merged during the merging paired-end reads step. Approximately 20% of sequences in each group were chimera and therefore removed. Ultimately, 4.616.745 and 5.834.621 sequences per group entered the taxonomic assignment step, and an average of 60.746 ± 15.792 and 61.417 ± 30.305 sequences per sample were analysed in the probiotic and placebo group, respectively. Since diet is a potential confounder in the analysis of microbiome composition, patients' nutrition was monitored with a 36-item food frequency questionnaire; no changes in nutrition were observed. In total 10.998.425 and 14.107.620 sequencing reads were analysed in the probiotic and placebo group, respectively. After quality checking and denoising, the number of sequences was reduced to 6.967.396 in probiotic and 8.947.492 in placebo group. Around 17% of the reads in each group were not merged during the merging paired-end reads step. Approximately 20% of sequences in each group were chimera and therefore removed. Ultimately, 4.616.745 and 5.834.621 sequences per group entered the taxonomic assignment step, and an average of 60.746 ± 15.792 and 61.417 ± 30.305 sequences per sample were analysed in the probiotic and placebo group, respectively. Since diet is a Nutrients 2020, 12, 1874 5 of 11 potential confounder in the analysis of microbiome composition, patients' nutrition was monitored with a 36-item food frequency questionnaire; no changes in nutrition were observed.
Beta diversity remained stable throughout the intervention in both subject groups. Bray Curtis dissimilarity showed that microbiome compositions were comparable in both groups before and after intervention. The intra-individual distance between samples before and after intervention was similar in both groups, and significantly lower than the distance to other individuals ( Figure 2b). Accordingly, NMDS based on Bray Curtis dissimilarity and consecutive ANOSIM revealed no specific clustering of groups or time points (p = 0.516), as shown in Figure 2c. RDA showed no significant correlation between the overall microbiome composition and probiotic intervention p > 0.999 (Figure 2d). Beta diversity remained stable throughout the intervention in both subject groups. Bray Curtis dissimilarity showed that microbiome compositions were comparable in both groups before and after intervention. The intra-individual distance between samples before and after intervention was similar Nutrients 2020, 12, 1874 6 of 11 in both groups, and significantly lower than the distance to other individuals ( Figure 2b). Accordingly, NMDS based on Bray Curtis dissimilarity and consecutive ANOSIM revealed no specific clustering of groups or time points (p = 0.516), as shown in Figure 2c. RDA showed no significant correlation between the overall microbiome composition and probiotic intervention p > 0.999 (Figure 2d).

Taxon Comparison
Changes in unique OTUs were analysed using ANCOM and differential abundance analysis with Gneiss and were deemed of interest if a significant difference was observed between baseline and end of intervention as well as between groups at the end of treatment. While ANCOM did not show any differences between groups or timepoints, 11 out of 16,032 observed OTUs met these criteria when using differential abundance analysis with Gneiss in the probiotic group and none in the placebo group.
Three differentially abundant OTUs were identical with the sequences of the probiotic bacteria Lactobacillus salivarius W24, Lactobacillus brevis W63 and Lactococcus lactis W19/W58, which were ingested with the study product. Those OTUs were more abundant after the intervention in the probiotic group, while patients in the placebo group showed little to no abundance. In the probiotic group, increased levels were detected only immediately after the intervention, not at baseline or six months after the intervention had ended (Figure 2e). The remaining probiotic strains ingested with the study product were not indicated in the a priori analysis. A separate search for these sequences showed that Bifidobacterium bifidum, Lactobacillus acidophilus and Lactobacillus casei remained on their endogenous levels, Bifidobacterium lactis could not be found.
Furthermore, an OTU identified as Faecalibacterium prausnitzii was significantly increased by the probiotic intervention, resulting in a significantly higher abundance after six months compared to the placebo group. It almost doubled in abundance and remained elevated after the intervention had ended. OTUs identified as Syntrophococcus sucromutans, Bacteroidetes vulgatus and Prevotella sp. showed steep increases during intervention and retreated after the intervention had ended. Alistipes shahii increased in abundance during intervention and further increased after intervention had ended (Figure 2e).
Ruminococcaceae UCG-014 decreased after the intervention and regained baseline levels six months after the intervention had ended. However, a drop in abundance of this taxon was also observed in the placebo group. A slight decrease in abundance of Anaerostipes hadrus was observed in the probiotic group, but this OTU was absent in the placebo group before and after intervention. Dorea longicatena was also identified as significantly altered; however, it was only present in four samples (two samples in the probiotic group before intervention and two samples in the placebo group after intervention). Intervention-independent longitudinal changes were observed that mainly included the increase or decrease of various sequences identified as Bacteroides vulgatus, a starch-utilizing bacterium.

Associations with Clinical Findings
To assess clinical significance of the observed changes in the microbiome at the end of intervention, their associations with biomarkers of gut permeability (faecal zonulin, lactulose-mannitol ratio, and diamine oxidase), bacterial translocation (lipopolysaccharide, sCD14, and lipopolysaccharide binding protein) and innate immune response (neopterin, and resting burst by neutrophils) were examined. Correlation analysis associated changes in neopterin to changes in Alistipes shahii abundance (r s = 0.354; p = 0.006), and changes in zonulin to changes in Syntrophococcus sucromutans and Prevotella sp. (r s = −0.311; p = 0.018 and r s = −0.285; p = 0.030, respectively). Accordingly, patients with an increase in Alistipes shahii after probiotic intervention show a significant increase in neopterin compared to placebo-treated patients and patients without that bacterial increase. Moreover, patients with an increase in Syntrophococcus sucromutans and/or Prevotella sp. showed a significant decrease in zonulin compared to placebo-treated patients (Figure 3). Other biomarkers showed no significant correlation to the identified OTUs. Nutrients 2020, 12, x FOR PEER REVIEW 7 of 11

Discussion
Our data show that after six months of probiotic intervention the probiotic strains Lactobacillus brevis, Lactobacillus salivarius and Lactococcus lactis were enriched in the stool of cirrhotic patients and the abundance of Faecalibacterium prausnitzii, Syntrophococcus sucromutans and Alistipes shahii (all short-chain acid producers), as well as Bacteroides vulgatus, and a Prevotella species were increased in the probiotic group at the end of the intervention. Changes in the microbiome correlated with changes in serum neopterin levels and gut permeability.
When efficacy of probiotic therapies is discussed, it is necessary to know whether the probiotic preparation in question can reach the intended target in the intestine. Probiotic colonization shows a very high inter-individual variability and the resident microbiome can oppose the introduction of exogenous bacteria [32]. Cirrhotic patients show a significantly reduced microbial richness and therefore a reduced colonization resistance [10,33,34]. This could lead to a more efficient integration of probiotic bacteria into the resident microbiome [8]. In liver cirrhosis, an increase in the abundance of probiotic bacteria after ingestion has so far only been shown in patients with decompensated disease, where probiotic bacteria may find more favourable conditions for colonization due to severe dysbiosis [27,28]. In the present study, we demonstrated the transient enrichment of probiotic bacteria in compensated cirrhosis. Previous reports have shown that the enrichment of the microbiome with probiotic strains is dependent on ongoing administration [35][36][37] and our data confirm this, since we show that the abundance of probiotic bacteria regresses after probiotic therapy has been completed.
Probiotic intervention in our study was not associated with monumental changes in the composition of the resident microbiome whereas in trials with decompensated patients, changes in major taxa of the human microbiome, such as an increase in Clostridiales or a reduction in Enterobacteriaceae with probiotic therapy were observed [27,28]. Changes observed in the present study were only found on OTU level and no major taxon significantly shifted in abundance. A possible reason for the different results of these studies might be found in the underlying health condition of the study cohorts. Compared to compensated cirrhosis, the microbiome of patients with decompensated cirrhosis is less stable due to various factors, such as medical treatment and hospitalization [9]. Therefore, probiotics might be able to induce larger changes in decompensated cirrhosis; however, in relatively stable conditions, big shifts in composition might not be achieved as easily.
Whether small changes in the microbiome are sufficient to induce clinical benefits remains unclear. We found correlations between the modulation of Alistipes shahii and the increase in neopterin, a macrophage-derived antimicrobial molecule that can induce reactive oxygen species

Discussion
Our data show that after six months of probiotic intervention the probiotic strains Lactobacillus brevis, Lactobacillus salivarius and Lactococcus lactis were enriched in the stool of cirrhotic patients and the abundance of Faecalibacterium prausnitzii, Syntrophococcus sucromutans and Alistipes shahii (all short-chain acid producers), as well as Bacteroides vulgatus, and a Prevotella species were increased in the probiotic group at the end of the intervention. Changes in the microbiome correlated with changes in serum neopterin levels and gut permeability.
When efficacy of probiotic therapies is discussed, it is necessary to know whether the probiotic preparation in question can reach the intended target in the intestine. Probiotic colonization shows a very high inter-individual variability and the resident microbiome can oppose the introduction of exogenous bacteria [32]. Cirrhotic patients show a significantly reduced microbial richness and therefore a reduced colonization resistance [10,33,34]. This could lead to a more efficient integration of probiotic bacteria into the resident microbiome [8]. In liver cirrhosis, an increase in the abundance of probiotic bacteria after ingestion has so far only been shown in patients with decompensated disease, where probiotic bacteria may find more favourable conditions for colonization due to severe dysbiosis [27,28]. In the present study, we demonstrated the transient enrichment of probiotic bacteria in compensated cirrhosis. Previous reports have shown that the enrichment of the microbiome with probiotic strains is dependent on ongoing administration [35][36][37] and our data confirm this, since we show that the abundance of probiotic bacteria regresses after probiotic therapy has been completed.
Probiotic intervention in our study was not associated with monumental changes in the composition of the resident microbiome whereas in trials with decompensated patients, changes in major taxa of the human microbiome, such as an increase in Clostridiales or a reduction in Enterobacteriaceae with probiotic therapy were observed [27,28]. Changes observed in the present study were only found on OTU level and no major taxon significantly shifted in abundance. A possible reason for the different results of these studies might be found in the underlying health condition of the study cohorts. Compared to compensated cirrhosis, the microbiome of patients with decompensated cirrhosis is less stable due to various factors, such as medical treatment and hospitalization [9]. Therefore, probiotics might be able to induce larger changes in decompensated cirrhosis; however, in relatively stable conditions, big shifts in composition might not be achieved as easily.
Whether small changes in the microbiome are sufficient to induce clinical benefits remains unclear. We found correlations between the modulation of Alistipes shahii and the increase in neopterin, a macrophage-derived antimicrobial molecule that can induce reactive oxygen species production by neutrophils [38,39]. This increase in antimicrobial strategies is consistent with previous reports that attest antimicrobial properties to probiotic strains or a boost of host derived immune responses to their ingestion [8,25,[40][41][42][43][44]. Moreover, Alistipes species were found to be depleted in patients with liver cirrhosis and non-alcoholic fatty liver disease [10,45]. An increase of Alistipes might signify an opposition to liver disease specific dysbiosis.
The increase of OTUs attributed to Synthrophococcus and Prevotella were correlated to a decrease in zonulin, an endogenous tight-junction regulating protein and is used as a biomarker for gut permeability, where high levels in stool or serum are indicative of a barrier dysfunction [46,47]. Decrease in zonulin levels during probiotic interventions has been shown before in obese patients with type 2 diabetes with the same product as used in the present study, and in healthy trained men with a comparable multispecies product [23,48]. Furthermore, the same product as in the presented trial showed a reduction of bacterial translocation in type 2 diabetic patients after a six month intervention and in postmenopausal women after 12 weeks [49,50]. Our results suggest that the improvement of the gut barrier is dependent on intervention-related modulation of the microbiome. Further, improvements in clinical outcomes in other diseases might be linked to successful modulation of the microbiome [32,51].
Besides the intervention-specific changes in the microbiome, OTUs that were identified as Bacteroides vulgatus showed fluctuations throughout the study in both groups. Bacteroides vulgatus utilizes starch as a primary substrate [52], and since these changes in abundance are prevalent in both groups, it is possible that this reflects a reaction of the microbiome to the product matrix, which consisted primarily of maize starch and was the sole ingredient in the placebo. Reactions to a product matrix should be a focus in future studies and considered in the product development.
Limitations: Although the analysed samples were from a randomized, double blind, placebo-controlled study, the inclusion in this subgroup analysis was based on liver disease severity. Accounting for the significant differences in the probiotic modulation of the microbiome between compensated and decompensated cirrhosis and the instability of the microbiome during decompensation at the same time would require a robust control group. Due to uneven dropout rates, such a control group was not available and patients with decompensated cirrhosis could not be selected for this analysis.
Furthermore, it should be noted that the differences in microbiome composition obtained by one method could not be reproduced by other statistical tools. Although the methods are not identical, they were both optimized for the identification of differentially abundant OTUs in compositional data. The controversial nature of their results might simply indicate that they are based on different algorithms and weigh specific characteristics of the data set differently. However, they might also indicate that the results are not robust enough to be identified by more than one method and need to be reproduced in future studies. Therefore, we encourage the careful deliberation of the results before interpretation.

Conclusions
After a six-month intervention with probiotics, an increase in probiotic bacteria and other beneficial taxa in stool of compensated cirrhotic patients was evident. Most of these changes were not permanent, suggesting that ongoing administration might be necessary to achieve long-term effects in this patient cohort. Further studies are necessary to elucidate the clinical relevance of microbiome changes.