Overview of trial enrolment
Thirty-two patients were recruited from outpatient clinics at King’s College Hospital or from suitable inpatients on the wards. From the 2nd March 2018 to 31st July 2019, 318 patients aged 18 to 75 years were pre-screened for eligibility; 88 fulfilled the inclusion criteria. 32 consented to be enrolled and of these, 23 patients were randomised, with 21 receiving treatment (15 FMT; 6 placebo) [Figure 1]. Randomisation was undertaken using a validated online randomisation programme in a 3:1 ratio of FMT: placebo. Treatment allocation was by block randomisation with randomly varying block sizes. All 15 patients receiving FMT completed the trial and attended all 4 follow-up visits. Five of the six patients receiving a placebo completed the trial, with the sixth patient attending all but the day 90 follow-up visit.
Characteristics of the study population
The randomised participants had a mean age of 57.1 +/- 11 years, 26% (6/23) were female and 78% (18/23) were Caucasian. 52% had alcohol-related cirrhosis. Overall, treatment arms were balanced with half of the participants in each arm having ascites and one-third having a history of overt HE with similar baseline levels of blood ammonia. [Table 1]
Table 1
Baseline characteristics of the treated study cohort
|
Total
|
FMT
|
Placebo
|
p-value
|
n
|
Mean (SD) or %
|
n
|
Mean (SD) or %
|
n
|
Mean (SD) or %
|
Age
|
21
|
57.38 (10.73)
|
15
|
57.80 (10.77)
|
6
|
56.33 (11.57)
|
0.70
|
Sex: female
|
5
|
23.8%
|
3
|
20.0%
|
2
|
33.3%
|
0.60
|
Ethnicity
|
|
0.032
|
White
|
16
|
76.2%
|
13
|
86.7%
|
3
|
50.0%
|
-
|
Black
|
1
|
4.8%
|
0
|
0.0%
|
1
|
16.7%
|
-
|
Asian
|
2
|
9.5%
|
0
|
0.0%
|
2
|
33.3%
|
-
|
Other
|
2
|
9.5%
|
2
|
13.3%
|
0
|
0.0%
|
-
|
Weight (kg)
|
21
|
83.32 (20.23)
|
15
|
87.16 (19.31)
|
6
|
73.72 (20.95)
|
0.15
|
Body Mass Index
|
21
|
29.15 (7.38)
|
15
|
30.10 (7.69)
|
6
|
26.79 (6.55)
|
0.23
|
Smoker
|
5
|
23.8%
|
4
|
26.7%
|
1
|
16.7%
|
1.00
|
Aetiology
|
|
0.51
|
Alcohol
|
12
|
57.1%
|
8
|
53.3%
|
4
|
66.7%
|
-
|
Metabolic/NAFLD
|
4
|
19.0%
|
3
|
20.0%
|
1
|
16.7%
|
-
|
Alcohol/Metabolic
|
1
|
4.8%
|
1
|
6.7%
|
0
|
0.0%
|
-
|
Hepatitis C (RNA negative)
|
3
|
14.3%
|
3
|
20.0%
|
0
|
0.0%
|
-
|
Primary biliary cholangitis
|
1
|
4.8%
|
0
|
0.0%
|
1
|
16.7%
|
-
|
Diabetes
|
4
|
19.0%
|
3
|
20.0%
|
1
|
16.7%
|
1.00
|
Prior Ascites
|
10
|
47.6%
|
7
|
46.7%
|
5
|
50.0%
|
1.00
|
Prior Spontaneous Bacterial Peritonitis
|
1
|
4.8%
|
1
|
6.7%
|
0
|
0.0%
|
1.00
|
Prior Hepatic Encephalopathy
|
13
|
61.9%
|
9
|
60.0%
|
4
|
66.7%
|
1.00
|
Prior Variceal Bleeding
|
8
|
38.1%
|
5
|
33.3%
|
3
|
50.0%
|
0.75
|
TIPS
|
1
|
4.8%
|
1
|
6.7%
|
0
|
0.0%
|
1.00
|
Encephalopathy grade at randomisation
|
|
0.62
|
0
|
15
|
71.4%
|
10
|
66.7%
|
5
|
83.3%
|
-
|
1
|
6
|
28.6%
|
5
|
33.3%
|
1
|
16.7%
|
-
|
2
|
0
|
0.0%
|
0
|
0.0%
|
0
|
0.0%
|
-
|
Ascites at randomisation
|
2
|
9.5%
|
2
|
13.3%
|
0
|
0.0%
|
1.00
|
White Blood Cell Count (x109/L)
|
21
|
4.35 (1.66)
|
15
|
4.22 (1.33)
|
6
|
4.66 (2.43)
|
0.82
|
Neutrophils (x109/L)
|
21
|
2.71 (1.08)
|
15
|
2.68 (0.88)
|
6
|
2.79 (1.58)
|
1.00
|
C-reactive protein (mg/L)
|
21
|
5.07 (6.12)
|
15
|
4.53 (3.67)
|
6
|
6.42 (10.43)
|
0.23
|
Haemoglobin (g/L)
|
21
|
123.3 (20.0)
|
15
|
123.5 (20.0)
|
6
|
123.0 (21.9)
|
0.91
|
Platelet count (x109/L)
|
20
|
109.0 (61.1)
|
15
|
105.5 (50.6)
|
5
|
119.6 (92.5)
|
1.00
|
INR
|
20
|
1.46 (0.34)
|
14
|
1.48 (0.38)
|
6
|
1.40 (0.22)
|
0.97
|
Sodium (mmol/L)
|
21
|
137.0 (4.5)
|
15
|
137.1 (5.1)
|
6
|
136.5 (3.1)
|
0.41
|
Urea (mmol/L)
|
20
|
5.47 (2.38)
|
15
|
5.61 (2.70)
|
5
|
5.04 (1.03)
|
0.90
|
Creatinine (µmol/L)
|
21
|
73.57 (30.31)
|
15
|
75.33 (34.25)
|
6
|
69.17 (18.93)
|
0.67
|
ALT (IU/L)
|
21
|
33.00 (19.52)
|
15
|
29.27 (15.28)
|
6
|
42.33 (26.90)
|
0.15
|
AST (IU/L)
|
20
|
50.65 (27.16)
|
14
|
45.86 (14.75)
|
6
|
61.83 (44.97)
|
0.71
|
Total bilirubin (µmol/L)
|
20
|
32.30 (22.38)
|
14
|
28.64 (14.37)
|
6
|
40.83 (35.23)
|
0.71
|
Albumin (g/L)
|
20
|
36.85 (4.65)
|
14
|
37.93 (4.63)
|
6
|
34.33 (3.93)
|
0.11
|
Ammonia (µmol/L)
|
21
|
69.38 (27.55)
|
15
|
71.27 (28.22)
|
6
|
64.67 (27.73)
|
0.56
|
Lactate (mmol/L)
|
21
|
1.59 (0.53)
|
15
|
1.55 (0.52)
|
6
|
1.68 (0.58)
|
0.75
|
MELD Score
|
19
|
13.11 (2.16)
|
13
|
13.38 (2.10)
|
6
|
12.50 (2.35)
|
0.45
|
Primary Outcome:
(i) Feasibility
83% (95%CI 64–97%) of those that consented and screened were eligible. 87% (95%CI 66–97%) of all patients randomised were treated and followed up to completion at 90-days.
(ii) Tolerability
Overall, FMT was well tolerated with only minor transient gastrointestinal side effects reported within the following 7 days. One individual vomited within 2-hours of receiving the FMT (bile not faeces) and 3/15 (20%) passed a type 6/7 bowel motion within 2-hours of receiving the FMT. No patients receiving FMT reported a foul taste or smell, flatulence or reflux.
(iii) Safety
There were 9 serious adverse events (SAEs) occurring in 4/15 patients in the FMT arm and one SAE in 1/6 patients in the placebo arm. None of the SAEs were deemed related to the FMT by an independent data monitoring committee. Two patients treated with FMT developed suspected infections (no organisms cultured); one patient developed pedal cellulitis on day 15 post-FMT following a podiatry intervention and one patient developed a urinary tract infection on day 87 post-FMT. There were no infections in the placebo group [Supplementary Table 1].
Secondary clinical and mechanistic outcomes:
(i) FMT restored faecal microbial richness with temporal donor engraftment
Gut microbial metagenomic species (MGS) richness plays an essential role in regulating health and immunity and is reduced in cirrhosis. (4, 5) We confirmed patients with cirrhosis had reduced faecal MGS richness compared to healthy donors [p = 1.73e-5 Fig. 2a]. The faecal microbial composition was distinct from healthy donors [Figure 2d-e]. FMT treatment restored recipient MGS species richness [p = 0.038 Fig. 2b] with significant donor engraftment on day 7 [Figure 2f] recently linked to enhanced clinical success. (20) In line with previous studies, (21) we observed antibiotic treatment adversely impacted engraftment [subjects 1 and 7; Fig. 2f]. FMT modified the microbiome composition [Figure 2g] increasing the abundance of beneficial Alistipes putredinis (decreased A. putredinis correlates with liver decompensation), (22) Firmicutes bacterium and Faecalibacterium prausnitzii, a strict anaerobe that produces butyrate (short-chain fatty acid, SCFA) associated with gut health [Figure 2h]. (23, 24) Following separation into Cluster-1 enriched for beneficial obligate anaerobes and Cluster-2 containing pathobionts, FMT increased cluster-1 and reduced cluster-2 at day-7 [Figure 2j-k]. The changes were temporal as by day-30 post-FMT a resurgence (outgrowth) of pathobiont cluster 2 was observed which was further enhanced at day-90 post-FMT [Figure 2f, j and k]. However, there are some residual effects of FMT at D90 compared to baseline.
(ii) FMT ameliorates cirrhosis-associated immune dysfunction (CAID) and augments anti-bacterial cytokine responses
BT is a significant driver of CAID characterised by compromised immune responses to bacterial pathogens with lymphocyte anergy and dysregulated cytokine production including circulating interleukin (IL)-17A and increased susceptibility to developing infection. (8) FMT increased plasma IL-17A (day-7; p = 0.002) [Figure 3a] involved in antibacterial function, systemic immune modulation, and gut barrier repair. (25, 26) Conversely, plasma IL-1β concentrations were higher in placebo-treated patients than in FMT [Figure 3b]. IL-1β is integral to the host-response and resistance to pathogens, however, it also exacerbates acute tissue injury during chronic disease. (26) Indeed, FMT augmented systemic immune cell responses following bacterial stimulation (Heat-killed Escherichia coli 0111:B4 [HKEB] and Escherichia coli lipopolysaccharide [LPS]). Whole blood peripheral blood mononuclear cells incubated with HKEB or LPS 7-days post-FMT, increased tumour necrosis factor alpha (TNF-α) production (HKEB p = 0.0025; LPS p = 0.032) and IFN-γ-induced protein 10 (IP-10) (LPS p = 0.0002) production [Figure 3c-d]. IL-1β production was increased in response to HKEB (p = 0.0085) in the placebo arm compared to FMT on day-90 [Figure 3e]. Collectively FMT contributed to a reduction in immune paralysis and a drive towards the restoration of immune antibacterial function [Figure 3a-e].
(iii) FMT promotes microbial hyodeoxycholic acid and butyrate production promoting gut barrier repair
Gut microbial-host co-metabolites are understood to contribute to gut barrier function via impact upon host immune responses and direct impact upon epithelial cell homeostasis. Patients with high faecal butyrate levels post-FMT had lower plasma IFNγ and had higher levels of the faecal secondary bile acid hyodeoxycholic acid (HDCA) (p = 0.048; R2 = 0.57) [Figure 3f]. Interferons are important in host defence against infection. (27) Gut microbiota-derived HDCA attenuates inflammatory responses to LPS via TGR5, AKT, and NF-κB suppression, decreasing TNF-α, IL-6, and IL-1. (28) HDCA also modifies gut barrier integrity by suppressing intestinal epithelial cell proliferation through FXR-PI3K/AKT, alongside increasing the abundance of the gut bacteria associated with bile acid metabolism. (29)
Fatty Acid Binding Protein 2 (FABP2) is expressed in small intestinal epithelial cells. It is a marker of epithelial shedding and when elevated in the blood is a marker of gut barrier damage. (24) Plasma FABP2 levels were higher in the placebo group compared to the FMT group at 7, 30 and 90 days (p = 0.014, p = 0.0062, p = 0.0315, respectively) [Figure 3g]. PCoA showed that FMT had a significant effect on components associated with intestinal barrier integrity. Albumin exerts several homeostatic functions as a potent scavenger, antioxidant, and immunomodulatory molecule, including the transport of fatty acids as well as modulating intestinal barrier function. (30) Plasma albumin rose across the 90-day follow-up period and faecal FABP2 levels fell (p = 0.014) following FMT but not placebo [Figure 3h]. Faecal FABP2 levels began to rise again in the FMT group compared to placebo at 30 days (p = 0.03) [Figure 3i] mirroring the resurgence of pathobiont cluster 2. Undigested carbohydrate is fermented to D-lactate by lactic acid–producing bacteria [LAB] present in the colon. In dysbiosis LAB species are either lost or, with excessive carbohydrate consumption, overgrown resulting in excessive D-lactate passing into the systemic circulation (a biomarker of BT). Plasma D-lactate was reduced at day-7 post-FMT compared to placebo (p = 0.0077) [Figure 3j]. Faecal D-lactate levels increased post-FMT treatment compared to placebo [p = 0.043 Extended data Fig. 1]. Indeed, enhanced proportions of the LAB-containing phylum Bacillota are observed in cluster 1 in these patients [Figure 2j-k]. Faecal proteomic analysis demonstrated FMT increased the abundance of human proteins [Supplementary Table 3] involved in antimicrobial mucosal barrier defence and suppression of microbiome and pathogen entry into the host including human zymogen granule proteins, secretory IgA and several neutrophil-derived molecules including calprotectin, neutrophil defensin 1 and cathepsin G [Figure 3k-n]. Over the duration of the study markers of barrier disruption (alpha-amylase 2B, lithstathine-1-α, intestinal-type alkaline phosphatase and neprilysin) were enhanced in the placebo stool but not following FMT [Supplementary Table 3, Fig. 3k-n].
FMT also initiated microbial and human metabolic reprogramming, particularly of glutamine. Intestinal epithelial cells and immune cells largely depend on glutamine availability to survive, proliferate, function, and ultimately defend against pathogen invasion [Supplementary Tables 3 & 8, Fig. 3k-n]. (31) Taken together, these data indicate that FMT is associated with a change in bacterial species resulting in microbial and human metabolic reprogramming that augments intestinal barrier integrity and immune defence.
(iv) FMT modulates the translocating microbiota towards health
In healthy individuals, the circulating human blood microbiota is mostly composed of members of the phyla, Pseudomonadota (synonym Proteobacteria) and to a lesser extent of Actinomycetota (synonym Actinobacteria) and Bacteroidota (synonym Bacteroidetes) phyla. (32, 33) Intestinal epithelial and immune cells limit the translocation of gut microbiota across the intestinal barrier. In patients with cirrhosis and portal hypertension, however, there is exaggerated BT of pathogenic organisms. Increased alpha-diversity of the blood microbiome (at genus and species level) was observed in the placebo group [Figure 4a, 4c and Extended Data Fig. 2a] over time but not in the FMT-treated cohort [Figure 4b and extended data Fig. 2b]. FMT modified the profile of the species translocating across the gut barrier [Figure 4e-f, Extended Data Fig. 2c-d and Supplementary Tables 5–6]. The relative abundance of Bacillota (synonym Firmicutes) such as Clostridia, including the Ruminococcaceae family, was increased in patients receiving FMT [Figure 4e-f; Supplementary Table 5]. Ruminococcaceae convert primary bile acids into secondary bile acids and are major butyrate producers such as butyrate and are depleted in cirrhosis. (34)
Blood 16S rDNA gene-targeted sequencing was further analysed with PICRUSt 2.0 to predict altered bacterial functions altered. Enhanced levels of microbial enzymes involved in butyrate metabolism, with an increase in butyrate kinase [EC.2.7.2.7] following FMT (day 30; p = 0.0013) and aldehyde dehydrogenase [EC.1.2.1.10] (day 90; p = 0.003) were observed [Figure 4e-g]. This enrichment was accompanied by a rise in faecal butyrate (day 90; p = 0.0258, Extended Data Fig. 2e) pointing to FMT promoting engraftment of butyrate-producing species. The local anti-inflammatory effects of butyrate render intestinal epithelial cells tolerant of commensal bacteria and additionally confer a protective role against unchecked inflammation. (35)
(v) FMT attenuated microbial-associated ammonia production and upregulated ammonia utilisation by augmenting the anaerobic metabolism of L-aspartate
Plasma ammonia reduced following FMT (day 30; p = 0.0006) [Figure 5a] and faecal ammonia rose (days 7, 30 and 90; p = 0.019, p = 0.011 and p = 0.025, respectively) [Figure 5b]. This was associated with amelioration in biomarkers of systemic inflammation [Figure 5c]. Patients with higher ammonia levels at baseline had higher levels of faecal calprotectin (R2 = 0.79; p = 0.002). Calprotectin is a surrogate marker of intestinal inflammation and a consequence of neutrophil migration into an inflamed intestinal mucosa [Figure 5c]. The reduction in plasma ammonia was associated with an improvement in HE grade (R2= -0.601; p = 0.02) [Figure 5c] and the rise in faecal ammonia was associated with lower faecal LPS levels (R2= -0.71; p = 0.003) [Figure 5d]. In contrast, plasma TNF-α and IL-1β increased with placebo.
Metagenome prediction from blood 16s rDNA sequencing using PICRUSt 2.0 revealed increases in the abundance of microbial enzymes involved in nitrogen assimilation and a reduction in those required for denitrification and ammonification [Figure 5f]. Nitrite reductase [EC.1.7.2.1], which reduces nitrite to nitric oxide (NO), was increased from day-7 (p = 0.0043), day-30 (p = 0.0125) and day-90 (p = 0.0004) post-FMT [Figure 5g, Extended Data Fig. 3a.]. Nitrite reductase [EC.1.7.2.1] is usually expressed in obligate anaerobes (36) and coincided with an increase in Cluster-1 containing obligate anaerobes on day-7 post-FMT. The respiration and growth of E. coli are reversibly suppressed by NO. (37) NO produced by facultative or anaerobic bacteria diffuses into the mucosal microaerobic zones facilitating colonic bacterial production of SCFAs which provide respiratory fuels for intestinal epithelia and modulating mucosal immunity. (38) Nitrite reductase further reduces NO and hydroxylamine to ammonia, and sulphite to sulphide (39), was enriched in patients with higher plasma ammonia [Figure 5j] and increased post-FMT from day 30 to 90 coinciding with the resurgence of the pathogenic Cluster-2 [Figure 5h]. Nitrite reductase [EC.1.7.2.2] confers a selective advantage for pathogenic bacteria to colonise mucosa. (39) A decrease in the abundance of the ammonia-generating enzyme L-seryl-tRNASec selenium transferase [EC.2.9.1.1] which metabolises L-serine from dietary protein into glycine and ammonia was observed post-FMT (day-90; p = 0.0027) whilst an increase was observed with placebo (p = 0.025) [Figure 5k and m].
Aspartate ammonia-lyase [EC.4.3.1.1] was significantly enriched in the blood microbiome post-FMT (day-90; p = 0.0004) [Figure 5l; Extended Data Fig. 3c]. In anaerobic conditions, EC.4.3.1.1 combines ammonia with fumarate to produce L-aspartate which requires the transport of succinate but in aerobic growth conditions, L-aspartate is metabolised by aspA-encoded aspartase to produce ammonia. (40) The increase in Cluster-1 rich in obligate anaerobes post-FMT supports FMT-promoting growth conditions that drive the anaerobic catabolism of L-aspartate consuming ammonia.
Faecal proteomic analysis revealed that FMT increased the abundance of faecal microbial enzymes involved in denitrification and ammonification with a reduction in enzymes required for nitrogen assimilation and excretion via the urea cycle (Fig. 5n). Glutamine synthetase [EC.6.3.1.2] involved in the incorporation of ammonia into glutamate was increased in the FMT-treated group over the course of the trial, being 17.4-fold higher in the FMT group at day 90. Enzymes involved in ammonia excretion by the urea cycle were modified following FMT. Bacterial ornithine carbamoyltransferase [EC2.1.3.3, Q328S8, P75473, B2G653, Q93JF1, P0C0N1] which converts ammonia and L-aspartate via L-ornithine to urea were enhanced in the stool (p = 0.006, Q328S8). FMT increased urinary hippurate, a marker of gut diversity (41) and metabolic health (42) (day-30; p = 0.0299) [Figure 6a] and decreased urinary succinate (day-7; p = 0.0005, day-30; p = 0.0162) [Figure 6b]. Metagenome prediction from blood 16s rDNA sequencing using PICRUSt 2.0 revealed changes in the abundance of microbial enzymes involved in benzoate metabolism [Figure 6c-g] and glycine [Figure 6h] post-FMT culminating in increased hippurate production [Figure 6n] and consumption of ammonia indicative of urinary excretion.