Lactulose drives a reversible reduction and qualitative modulation of the faecal microbiota diversity in healthy dogs

Hepatic encephalopathy is a frequent and debilitating complication of liver disorders. Lactulose is an established and reasonably effective treatment, yet with incompletely understood mechanisms of action. The aims of this study were to examine how the faecal microbiota composition changed before, during and after lactulose treatment in a large animal model. Healthy, privately owned dogs (n = 18) completed a prospective cohort study. Faecal samples were collected weekly, while the subjects were either on their usual diet (week 1), or a standardised diet (weeks 2–9), with added oral lactulose in weeks 6–7. DNA extraction and 16S rRNA gene sequencing were undertaken. Faecal samples from week 7 had a significantly lower microbiota richness/diversity, based on observed operational taxonomic units, Shannon/Chao1 indexes and Pielou’s Evenness. Beta diversity based on UniFrac distances was significantly different in week 7 compared to weeks 1, 5 and 9. At the phylum level, week 7 was associated with a significant increase of Firmicutes and Actinobacteria, and a decrease of Bacteroidetes and Fusobacteria, when compared to weeks 5 and 9. In summary, we have shown that lactulose induces a reversible qualitative and quantitative change of the faecal microbiota, which may explain its clinical efficacy in the management of hepatic encephalopathy.

Faecal dysbiosis is known to occur in covert and overt HE. Cirrhotic patients with minimal HE (MHE) harbour a higher proportion of urease-producing Streptococcus salivarius in stool samples, positively correlating with serum ammonia accumulation 13 . Additionally, the cirrhosis dysbiosis ratio (CDR), a previously validated ratio of autochthonous to non-autochthonous taxa in stool samples of cirrhotic patients, is lowered after development of severe HE, indicating worsened dysbiosis, and associated with 30-day mortality and organ failure 14 . Moreover, the presence of specific bacterial families (Alcaligenaceae, Porphyromonadaceae, Enterobacteriaceae) is strongly associated with poor cognition and inflammation in HE patients 15 . Considering the ongoing evidence regarding microbiome disruption in cirrhosis and HE 16 , it is likely that manipulation of the microbiota may contribute to improved outcomes. Interventions with proposed positive impact have so far included probiotics 17 , diet 18 and faecal microbiota transplantation 19 .
The impact of lactulose in ameliorating cirrhosis and HE-associated faecal dysbiosis is controversial. A direct impact has been supported by studies based on culture-dependent methodologies, namely: increased Bifidobacterium, Lactobacillus and Bacteroidaceae colonies, and reduced Enterobacteriaceae, Enterococcus and yeasts in patients with MHE, alongside with improved blood ammonia, psychometric tests and reduced risk of developing overt HE 20 ; and increased total aerobic and anaerobic bacterial counts, and lactobacilli in cirrhotic patients without clinical HE, alongside with decreased faecal pH 21 . Conversely, studies based on culture-independent techniques have not substantiated an effect of lactulose in the microbiome of cirrhotic patients without HE 22 and have reported only a minimal change in cirrhotic patients with HE 14 , including after lactulose withdrawal 23 .
However, no study using next generation sequencing techniques has assessed quantitative and qualitative changes of the intestinal or faecal microbiome, i.e. changes based on both the abundance and the presence or absence of microbial communities, after lactulose treatment in patients entirely naïve to lactulose. The effects of oral lactulose on the faecal microbiome of healthy humans have only been evaluated through either culture-based [24][25][26] or culture-independent methods targeting predominant bacterial groups [27][28][29][30] . As diet was only standardised in two of those studies 24,30 , it seems likely that a variable diet could have impacted results 31 .
The human faecal microbiome is closer to the canine faecal microbiome when compared to pigs or mice 32 . Dogs evolved to cohabit with people, and hence adapted to a similar diet 33 . As they are frequently kept as companion animals, they are also exposed to the same environment. In addition, dogs can equally suffer from HE and lactulose is frequently used as supportive treatment in this condition 34 . Consequently, companion dogs can represent a useful comparative model to explore the faecal microbiome in HE as well as the effects of certain interventions on its composition, richness and function.
Therefore, the aims of this study were to investigate the magnitude and duration of quantitative and qualitative changes of the faecal microbiota by lactulose in healthy privately-owned dogs fed a standardised commercial diet. It was hypothesised that oral lactulose administration would significantly and transiently change the faecal microbiota in healthy dogs.

Methods
Prospective cohort study design. Dogs owned by members of staff at the Hospital for Small Animals, the Royal (Dick) School of Veterinary Studies, University of Edinburgh, were recruited with the following inclusion criteria: no current history of any disease; up to date vaccination and deworming records; and no current or recent administration of medications. A faecal sample was requested to be collected from each subject weekly, pertaining to the interventions schematised in Fig. 1.
The standardised diet was a commercial maintenance diet for adult dogs (Hill's ™ Science Plan ™ Canine Adult Advanced Fitness ™ Large Breed with Chicken, Hill's Pet Nutrition Ltd., Guildford, UK) and the lactulose was a 3.5 g/5 ml oral solution (Sandoz Ltd, Hampshire, UK). The dose of lactulose was calculated at 0.5 ml/kg and given every 12 hours, unless excessively soft or unformed faeces were noticed, at which point the subjects' owners would notify one of the authors (MFF) and subsequently decrease the dose by 25% each time, aiming to achieve a soft faecal consistency that would be still amenable for manual collection.
Faecal samples were collected into plain bijoux tubes, kept frozen at −20 °C for a maximum of 24 hours and transferred afterwards to a −80 °C archiving freezer.
Informed consent was obtained from each subject's owner. The study was approved by the University of Edinburgh's Veterinary Ethical Review Committee (reference number 112-14) and carried out in accordance with the institution's relevant guidelines and regulations.
Faecal DNA extraction, amplification and sequencing. Each sample was defrosted, manually homogenised and DNA extraction performed with a commercial kit (PowerSoil ® DNA Isolation Kit, MO BIO Laboratories, Inc., Carlsbad, CA, USA) according to the manufacturer's instructions 35 . Amplification of DNA was undertaken with PCR of the hypervariable V4 region of the 16S ribosomal RNA (rRNA) gene, using dual-indexing primers (515F/806R), followed by amplicon quantification (Quant-iT ™ PicoGreen ® , Invitrogen, Briefly, sequence data with sequence quality information was imported through the "fastq manifest" format PairedEndFastqManifestPhred33. Demultiplexed sequence counts were summarised showing a median of 58,478 sequences per sample, with a minimum of 27,297 and maximum of 134,220. Sequence quality trimming, chimera filtering and feature table construction was performed through the DADA2 pipeline 38 . Mapping of feature identifiers to sequences was created using links for the Basic Local Alignment Search Tool (BLAST) 39 , and multiple sequence alignment of the sequences was undertaken by the Fast Fourier Transform (MAFFT) program 40 . This was followed by filtering of the alignment with the mask plugin 41 , generation of a phylogenetic tree with the FastTree program 42 and application of midpoint rooting. The QIIME2 ™ q2-diversity plugin was used for rarefaction analysis and computation of alpha diversity metrics (observed operational taxonomic units [OTUs], Shannon diversity index, Chao1 and Pielou's Evenness), as well as beta diversity metrics (weighted and unweighted UniFrac distances). Finally, assigning taxonomy to sequences was performed with a pre-trained Naïve Bayes classifier (Greengenes 13_8 99% OTUs) through the QIIME2 ™ q2-feature-classifier plugin 43 .
A human-extrapolated CDR: ratio of commensal autochthonous taxa (Lachnospiraceae, Ruminococcaceae, Veillonellaceae, and Clostridiales Insertae Sedis XIV) to potentially pathogenic non-autochthonous taxa (Enterobacteriaceae and Bacteroidaceae) 44 , was calculated without the inclusion of Clostridiales Insertae Sedis XIV, as this taxon was not detected in the studied population.
Statistical analyses used included descriptive statistics, as well as the following inferential statistical tests: Kruskal-Wallis test to compare differences between subjects regarding alpha diversity metrics; Wilcoxon matched-pairs signed rank test to compare differences between weeks regarding alpha diversity metrics, taxonomy frequencies and the CDR; paired t-test to compare differences between weeks regarding the presence of Lachnospiraceae; and Permutational Multivariate Analysis of Variance (PERMANOVA) test to compare differences between subjects and weeks regarding beta diversity metrics 45 . Normality was assessed with the Shapiro-Wilk test. Statistical significance level was set at p < 0.05.

Results
Population's baseline characteristics. A total of 21 dogs were enrolled, with a median age of 5 years (range of 2-10 years). Sex distribution included 12 females (11 neutered) and 9 males (7 neutered). Just over half (n = 11) of the cohort was represented by crossbred dogs, with the remaining 10 dogs distributed as follows: two each of English Cocker Spaniel, English Springer Spaniel, Border Collie and Podenco; and one each of Boston Terrier and Labrador Retriever.
Adherence to study protocol and side effects related to lactulose administration. Three dogs did not complete the study for the following reasons: acute diarrhoea following dietary indiscretion (n = 1); requirement of a NSAID (meloxicam) for pain management secondary to presumptive degenerative joint disease (n = 1); and lip fold dermatitis after being licked by another dog in the household once starting lactulose (n = 1). In addition, two faecal samples (one each from weeks 4 and 8) were accidentally not collected. In total, 172 faecal samples were analysed, one of which failed sequencing (week 6). Side effects associated with lactulose administration included excessively soft faeces (n = 7) and unformed faeces (n = 1), all resolving after a dose reduction of 25% and 50%, respectively, therefore not requiring a drop out from the study.
Alpha diversity. Community richness and diversity values were different between subjects regarding the following metrics: observed OTUs (p < 0.0001), Shannon diversity index (p = 0.0025), Chao1 index (p < 0.0001) and Pielou's Evenness (p < 0.0001). Assessment of community richness and diversity across time is shown in Fig. 2. There was a reduction of all the above metrics at week 7 (standardised diet and oral lactulose), when compared to weeks 1 (usual diet), 5 (standardised diet) and 9 (standardised diet after having stopped lactulose). Conversely, values from weeks 1, 5 and 9 didn't differ from each other apart from Shannon diversity index values between week 5 and 9 (Table 1).
Beta diversity. Bacterial community presence/absence (qualitative) and abundance (quantitative) assessments with unweighted and weighted UniFrac distances, respectively, are illustrated in Fig. 3. These were overall different between subjects (p < 0.001). When analysed across time, the distances at week 7 were different from weeks 1, 5 and 9. However, values at weeks 1, 5 and 9 didn't differ from each other ( Table 2).
taxonomy. The relative and absolute distribution of phyla abundance across time is depicted in Fig. 4.
Irrespective of week number (5, 7 or 9), the most abundant phylum was Firmicutes, followed by Bacteroidetes. A shift of phyla was observed at week 7, with the third most abundant phylum being Actinobacteria, followed by Fusobacteria and Proteobacteria. In contrast, for both weeks 5 and 9, Fusobacteria was the third most abundant phylum, followed by Proteobacteria and Actinobacteria. Moreover, when compared to weeks 5 and 9, week 7 was associated with a higher abundance of both Firmicutes (p = 0.0056 and 0.0047, respectively) and Actinobacteria (p = 0.0090 and 0.0104, respectively), and lower abundance of both Bacteroidetes (p = 0.0304 and 0.0120, respectively) and Fusobacteria (p = 0.0040 and 0.0008, respectively). The abundance of these phyla was similar between www.nature.com/scientificreports www.nature.com/scientificreports/ A total of 20 families were found to be present in the microbiota of at least half of the subjects at one or more time points (Fig. 5a). Different abundances across time were found in eight families (Fig. 5b), six of which showed a different abundance in week 7 when compared to weeks 5 and 9 (Fig. 5c), while both these time points www.nature.com/scientificreports www.nature.com/scientificreports/ had similar abundances. Significant changes at week 7 included an increased representation of Veillonellaceae and Bifidobacteriaceae, and a decreased abundance of Fusobacteriaceae, Bacteroidaceae, Ruminococcaceae, Alcaligenaceae, Lachnospiraceae and Peptococcaceae. cirrhosis dysbiosis ratio. When analysing the CDR across time (Fig. 6), the highest values were obtained at week 7 with a median of 25.07 (range of 0.08-460.04). These were different from week 1 (median 3.19, range 0.18-30.98, p = 0.0079), week 5 (median 2.54, range 0.18-29.78, p = 0.0003) and week 9 (median 2.10, range 0.24-15.75, p < 0.0001). Conversely, the CDRs calculated for weeks 1, 5 and 9 were similar to each other (p = 0.5678 for week 1 vs 5; p = 0.2288 for week 1 vs 9; and p = 0.5509 for week 5 vs 9).

Discussion
To date, there are no publications employing untargeted or global culture-independent methodologies to assess the faecal microbiome in healthy people receiving lactulose. In people with HE, further limitations often apply, given the common use of standard of care treatment by the time of study enrolment (lactulose, rifaximin, antacids), precluding an evaluation of the microbiome in non-treated HE 14,18 .  , or a standard diet after having stopped lactulose (week 9). Samples from week 7 were significantly different (p < 0.01, PERMANOVA test) from weeks 1, 5 and 9, while none of the latter differed from each other. www.nature.com/scientificreports www.nature.com/scientificreports/ To the authors' knowledge, this study is the first to investigate the effect of lactulose on the canine microbiota. Collectively, lactulose induced a reversible reduction and qualitative modulation of the faecal microbiota diversity in this population of healthy dogs, while on a commercial standardised diet. Both alpha and beta diversity metrics were affected and specific taxa were implicated in this change.

Metric
The impact of lactulose has been studied in healthy mice and pigs through culture-independent methods, most showing an increased alpha diversity, which is in contrast to the present study [46][47][48][49] . However, experimental animals are kept in laboratory-controlled conditions and both species are known to have a more distant www.nature.com/scientificreports www.nature.com/scientificreports/ microbiome from humans relative to dogs 32 . Assessing companion dogs, alongside maintaining the advantage of dietary control, may be therefore valuable given their natural shared environment with humans 31 .
No significant changes of the microbiota were observed in this study due to diet change alone. This is likely explained by the relative similarity of composition of commercial diets (most of the original diets consisted of www.nature.com/scientificreports www.nature.com/scientificreports/ various brands of dry dog food), in comparison to homemade or raw food 50,51 . There were marked inter-individual differences concerning overall alpha and beta diversity, which was not surprising, given the non-relatedness of the subjects and the diversity of breeds evaluated 52,53 .
Significant changes in the abundance of main phyla were observed with the introduction of lactulose and matched by changes noted at family taxa. Within the Firmicutes and Actinobacteria, Veillonellaceae and Bifidobacteriaceae increased; and decreases in Bacteroidaceae and Fusobacteriaceae likely reflect lower abundances of Bacteroidetes and Fusobacteria.
Veillonellaceae was the most abundant family after the administration of lactulose. Its significance in liver disease has been controversial, as previous studies have reported both increases and decreases in human cirrhosis with and without HE 13,14 . Bacteria of this family convert lactate to acetate and butyrate 54 . Previously, the latter has been positively correlated with the presence of Veillonellaceae in canine faeces 55 . Increases in microorganisms producing these SCFA could be beneficial, as in people, acetate was negatively correlated with pro-inflammatory cytokines in cirrhosis, and butyrate was protective against the development of HE 16 . The administration of lactulose has been associated with increased acetate and butyrate production, together with decreased branch-chained fatty acids isobutyrate and isovalerate 46,47,56 . Similarly to the present study, several others have reported increased numbers of faecal Bifidobacteriaceae [24][25][26][27][28][29][30] , which produce lactate and acetate 57 , as well as contribute to metabolic cross-feeding, stimulating other butyrate-producing bacteria 58 .
Reductions of both Fusobacteriaceae and Alcaligenaceae induced by oral lactulose have not been previously reported. This finding is potentially significant, as their presence in the stools of patients with cirrhosis and HE has been associated with worsened inflammation/endotoxaemia and poor performance on cognitive tests 15 . On the contrary, decreases in Bacteroidaceae 24,26 and Ruminococcaceae 30 have been previously reported with lactulose administration. Bacteroides spp., are known to produce isovalerate 59 , as well as pro-inflammatory and barrier-disruptive neurotoxins/enterotoxins 60 , and β-glucuronidase 61 , a potential carcinogenic. Reduced faecal activity levels of this enzyme have been demonstrated previously with lactulose administration 26,62 . Besides producing butyrate 63 , Ruminococcaceae and Lachnospiraceae are also known to produce β-glucuronidase 64 . The lack of a significant increase in Lactobacillaceae with lactulose administration has been demonstrated before; hence, this study confirms that it is not a major hallmark of lactulose use 24,25,29,30 .
CDR is a measure of dysbiosis in humans: healthy people are reported to have a higher ratio than patients with cirrhosis, and the presence of HE was associated with an even lower CDR 14 . Although extrapolated from studies in people, the increase of CDR during lactulose administration could represent improvement of dysbiosis, despite an overall lower microbiota diversity and richness.
Limitations of this study include the small number of dogs, the collection of voided faecal samples rather than mucosal or luminal colonic samples, the lack of storage buffer/cryoprotectant and the fact that 16S rRNA gene sequencing allows assessment of taxonomy and abundance, but no species level resolution, nor extrapolation of functional data from the microbiome. However, given the longitudinal study design, each subject could serve as their own control, allowing observation of general trends of microbiota changes. For ethical reasons, invasive collection methods were not employed in these privately owned animals 47,65 . Short-term storage of faecal samples without cryopreservative or buffer is likely to not impact on major phyla distribution 66,67 . To investigate functional changes of the microbiome, high-throughput techniques (whole metagenome shotgun sequencing), metabolomics or metatranscriptomics could have been performed, but this was not within the scope of the current study.
In conclusion, lactulose can drive a reversible quantitative and qualitative modulation of the faecal microbiota in this dog model. Future research is warranted to focus on the investigation of microbiome dynamics in canine models of spontaneous naturally occurring HE (e.g. congenital portosystemic shunts) 68 . This could Figure 6. Cirrhosis dysbiosis ratio (CDR), calculated from the canine faecal microbiota, extrapolated from humans. Cohort of healthy owned dogs while either on their usual diet (week 1), a standard diet (week 5), a standard diet and oral lactulose (week 7), or a standard diet after having stopped lactulose (week 9). **p < 0.01; ***p < 0.001; ****p < 0.0001 (Wilcoxon matched-pairs signed rank test).