Study on Effects of Probiotics on Gut Microbiome and Clinical Course in Patients with Critical Care Illnesses

Ventilator-associated pneumonia (VAP) is a nosocomial infection contracted by ventilator patients in which bacteria colonize the upper digestive tract and contaminated secretions are released into the lower airway. This nosocomial infection increases the morbidity and mortality of the patients as well as the cost of treatment. Probiotic formulations have recently been proposed to prevent the colonization of these pathogenic bacteria. In this prospective observational study, we aimed to investigate the effects of probiotics on gut microbiota and their relation to clinical outcomes in mechanically ventilated patients. For this study, 35 patients were recruited (22 probiotic-treated and 13 without probiotic treatment) from a cohort of 169 patients. Patients in the probiotic group were given a dose of 6 capsules of a commercially available probiotic (VSL#3®:112.5 billion CFU/cap) in three divided doses for 10 days. Sampling was carried out after each dose to monitor the temporal change in the gut microbiota composition. To profile the microbiota, we used a 16S rRNA metagenomic approach, and differences among the groups were computed using multivariate statistical analyses. Differences in gut microbial diversity (Bray Curtis and Jaccard distance, p-value > 0.05) between the probiotic-treated group and the control group were not observed. Furthermore, treatment with probiotics resulted in the enrichment of Lactobacillus and Streptococcus in the gut microbiota of the probiotic-treated groups. Our results demonstrated that probiotics might lead to favorable alterations in gut microbiome characteristics. Future studies should focus on the appropriate dosages and frequency of probiotics, which can lead to improved clinical outcomes.


Introduction
Hospital -acquired infections are the major cause of morbidity and mortality in critically ill patients worldwide. Ventilator-associated pneumonia (VAP) is reported as the most common hospital-acquired infection among critically ill patients admitted to intensive care unit (ICU) with incidence as high as 2 to 16 episodes per 1000 ventilator days [1]. One of the recent measures being extensively studied is the role of probiotics in preventing the incidence of nosocomial infections. However, the beneficial effect has not been consistently demonstrated across all studies and the effect of probiotics has varied depending on the probiotic strain used and the patient population studied. The human gastrointestinal tract consists of a diverse variety of microorganisms to the tune of 1014 representing over 1000 different species of bacteria from more than 150 different genera [2]. They are known to have an important role in human health, especially in metabolism and homeostasis. The gut is an important organ for stress after sepsis and other acute stressors [3]. Several illnesses like auto-immune, chronic liver disease, inflammatory bowel disease, and diabetes have been linked to alterations in the gut microbiome [4,5]. Dysbiosis is a common term used to indicate disorders of gut microbiota [6]. Dysbiosis has been studied in ICU populations and Shimizu et al. had shown that critical illness leads to a decimation of the gut microbiota population and the number of total obligate anaerobes was directly linked to mortality [7]. Moreover, dysfunction of the intestinal epithelium, immune system, and resident bacteria can lead to multiple organ dysfunction [8]. Several factors like antibiotics, vasoactive agents, antacids, and sedatives may work to aggravate the acute changes in gut microbiota [9]. However, it was previously reported that in critical illnesses, genera Enterococcus, Staphylococcus, and Enterobacteriaceae comprised the majority of bacteria in the gut during critical illnesses along with the reduction in the diversity of the gut microbiome [10]. It was earlier reported that the ratio in the gut of Bacteroidetes and Firmicutes (B/F ratio) changed during critical illnesses and extreme ratios of B/F were associated with poor prognosis [9]. Extreme changes in the ratio were thus thought to portend poor outcomes but it was not clear whether it had an associative or causative role. Thus, critical illness has been thought to be associated with decrease in commensal bacteria, which might lead to an increased chance of hospital-acquired infections. Most microbes in the body cannot be isolated by culture methods because of their anaerobic nature [9]. Conventional culture methods cannot thus capture the changes in the microbiota [8]. A metagenomic study using the 16S rRNA gene, which relies on amplification of the bacterial 16S rRNA gene with parallel processing, has been devised as an alternative to determine microorganisms present in the gut [11]. The use of probiotics like Feacalibacerium prausnitzii in some conditions like nonalcoholic fatty liver disease (NAFLD) improves liver health and reduces inflammation. [12]. It is believed that certain bacteria secrete anti-inflammatory peptides, which reduce local inflammation and strengthen gut epithelial barrier function through the agency of certain short-chain fatty acids [11][12][13]. Probiotic use in critical care has been an area of focus recently. Several studies have tried to evaluate its role in such a population. However, the role of probiotics in preventing infection has not been substantiated and has been a matter of controversy. The beneficial effects of probiotics have varied depending on the probiotic strain used and the patient population studied. In this study, we planned to study the effect of probiotics on the dynamics of gut microbiota, and also correlate it clinically with occurrences of nosocomial infections like VAP.

Patient Information
We conducted a prospective non-randomized interventional study in a single medical ICU at a tertiary center in India, with 2:1 allocation of patients in probiotic and control groups respectively. Being a pilot study, a sample size of convenience consisting of 22 in probiotic group patients and 13 in control group patients was planned. The study was conducted over 2 years, during the study period patients who were critically ill on mechanical ventilation admitted to medical ICU were screened for eligibility. We included patients with (a) age more than 18 and below 80 years, (b) on mechanical ventilation for < 72 h at the time of screening, (c) expected need for mechanical ventilation for at least 72 h. We excluded patients with (a) refusal to give consent, (b) failure of enteral feeding, (c) pregnancy or lactation, (d) severe multiple organ dysfunction defined by APACHE-II score >25 at screening, (e) previously received mechanical ventilation in the last 14 days. Written informed consent was taken from the first-degree family members or nextof-kin and patients and were sequentially allocated to probiotic or control groups. The study protocol (study no: IEC PG-624/28.11.2019) was approved by the AIIMS institutional ethics.

Trial Interventions
All patients were subjected to the preventive strategies of VAP including daily screening for weaning, protocolized weaning, strict hand hygiene, daily ET cuff pressure monitoring, closed circuit suctioning, pantoprazole for stress ulcer prophylaxis, oral care, and a semi-recumbent position with a head elevation of 30°, if not contraindicated. The first fecal sample was collected as soon as possible, at least within 72 h of initiation of mechanical ventilation. Laxatives and rectal enemas were prescribed as required to induce bowel movement. The stool samples were collected in specially designed vials with integrated spatula. After the collection of the first sample, the patients in the probiotic group were initiated on probiotics (VSL#3®: 112.5 billion CFU per capsule) at a dose of 6 capsules per day divided into 3 doses for 10 days. Each VSL#3® capsule contained the following bacterial strains Streptococcus thermophilus, Bifidobacterium breve, Bifidobacterium longum (reclassified as B. lactis), Bifidobacterium infantis (reclassified as B. lactis), Lactobacillus acidophilus, Lactobacillus plantarum, Lactobacillus paracasei, Lactobacillus delbrueckii subsp. bulgaricus (reclassified as L. helveticus). The missed capsules were given on successive days to a total of 60 capsules. In case capsules could not be given due to nothing per oral (NPO) status for more than three successive days, the patients were excluded from the study. The second and third fecal samples were collected on days 3-5 and days 7-10 from recruitment.

Clinical Outcome Measures
Clinical outcome measures used in this study were changes in gut microbiome at different time-points which was the primary outcome; secondary measures included ICU acquired 1 3 infections, namely ventilator-associated pneumonia (VAP) and bloodstream infection (BSI) were defined according to 2016 CDC/ATS guidelines. Diarrhea was defined as either ≥ 3 loose stools/day. The other outcomes measured were all-cause mortality, serial APACHE-II, SOFA score, blood lactate, days of antibiotic consumption, duration of mechanical ventilation, duration of ICU, and hospital stay.

Antimicrobial Administration
The dose, duration, choice, and route of administration of antimicrobial therapy of each patient was prospectively recorded daily and entered into a per-decided proforma. The antibiotic-usage days and antibiotic-free days were calculated from the collected data.

Follow-up
The baseline data (e.g., demographics, comorbidities, diet, source of sepsis, indication of intubation, severity of illness, smoking habit, heavy alcohol use, peri-intubation complications) and follow-up data (e.g., probiotic administration and reasons for not administering, relevant medications including antimicrobials, adherence to VAP prevention bundle, culture results, clinical diagnoses, length of stay, mortality) were recorded in the proforma. The presence of VAP or BSI was confirmed by consensus of three independent physicians considering relevant clinical, radiologic, and microbiologic data from patients with clinical suspicion of VAP/BSI. Any reasons for protocol non-compliance were also recorded.

Serious Adverse Events
Isolation of any Lactobacillus, Bifidobacterium, Streptococcus species in a sterile site or cultured as the sole or predominant organism in a non-sterile site prompted discontinuation of the study product. Serious adverse events were documented as mentioned in protocol.

Sample Collection
As the patients were on mechanical ventilation with sedation, the diapers were checked at regular intervals for passage of stools. Laxatives and enema were administered as necessary. The samples were collected in a sterile container with integrated spatula to avoid contamination and immediately transported to the refrigerator maintaining the cold chain. The samples were stored at −80°C until DNA extraction.

Metagenomic DNA Extraction
The fecal samples were stored at −80°C prior to DNA extraction. Of frozen samples, 200 mg were used for DNA extraction using the THSTI method [14]. Briefly, bacterial cell walls were lysed using three enzymes, lysozyme, lysostaphin, and mutanolysin which resulted in the spheroplast formation. Spheroplast was treated with guanidinium thiocyanate (GITC) followed by bead beating and heating enabled final lysis. The isolated DNA was precipitated by isopropanol and contaminants were removed by adding RNAse, sodium acetate, and 75% ethanol. The quantity and quality of the isolated DNAs were confirmed by agarose gel electrophoresis and spectrophotometry, respectively. Extracted DNA from all the samples was quantified using Nanodrop spectrophotometer 2000 (Thermo Scientific, USA).

Library Preparation for Nanopore and Illumina Sequencing
The Amplicon libraries for Nanopore were prepared using a Ligation sequencing kit (SQL-LSK109, Oxford Nanopore technologies) and PCR barcoding kit (EXP-PCR096, Oxford Nanopore technologies). Briefly, the protocol includes end-reparation, barcoding, and sequencing adapter ligation. A total of 200 ng of purified amplicon DNA from each sample was end-repaired using NEBnext ultra II end repair kit (New England Biolabs, MA, USA), cleaned up with 1X AmPure beads (Beckmann-Coulter, USA). Barcode adapter ligation (BCA) was performed with NEB blunt/TA ligase (New England Biolabs, MA, USA) and cleaned with 1× AmPure beads. Barcode adapter ligated products were quantified using a Qubit fluorometer barcoded using PCR reaction with LongAmp Taq 2× Master mix (New England Biolabs, MA, USA) and cleaned up with 1.6× AmPure beads (Beckmann-Coulter, USA). Barcode sequences are detailed in the table below. Barcoded samples were quantified and pooled at equimolar concentration. Pooled barcoded samples were end-prepared using NEBNext Ultra II End Repair/dA-Tailing Module (New England Biolabs, MA, USA). End-repaired DNA was cleaned up with 1× AmPure beads. Adapter ligation (AMX) was performed for 15 min using NEB blunt/TA ligase (New England Biolabs, MA, USA). Library mix was cleaned up using Ampure beads and finally eluted in 15 μl of elution buffer. Sequencing was performed on GridION X5 (Oxford Nanopore Technologies, Oxford, UK) using SpotON flow cell R9.4 (FLO-MIN106) in a 48-h-sequencing protocol. Nanopore raw reads ("fast5" format) were base called ("fast" format) and demultiplexed using Guppy v2.3.4.
For metagenomic sequencing through MiSeq illumina, extracted DNA was further quantified with Qubit fluorometer to get an exact estimate of the dsDNA. Samples with concentration of 5ng/μl were considered for 16S metagenomic library preparation following the Illumina 16S Metagenomic Sequencing Library Preparation. For amplification PCR, each 25μl reaction contained 12.5μl of 2× KAPA HiFi HotStart ReadyMix, 1μM of both the forward and reverse primer pairs and 2.5μl of the template DNA. PCR amplification using the following conditions: 3 min at 95°C, followed by 25 cycles of 95°C for 30s, 55°C for 30s, and 72°C for 30s followed by a final extension at 72°C for 5 min. Upon completion of the amplification PCR, products were cleaned using AMPure XP beads. The cleaned up products were subjected to index PCR that attaches dual indices and Illumina sequencing adapters using the Nextera XT Index Kit. For index PCR, a 50μl of reaction mixture contained 25μl of 2× KAPA HiFi HotStart ReadyMix, 5μl of both the Nextera XT Index 1 and 2 primers, 10μl of PCR grade water, and 5μl of template. Indexing was carried out by following conditions: 3 min at 95°C, followed by 8 cycles of 95°C for 30s, 55°C for 30s, and 72°C for 30s followed by a final extension at 72°C for 5 min. The final library is further cleaned using AMPure XP beads and validated in an Agilent 2100 Bioanalyser instrument [15].

16S rRNA Gene Sequencing Data Analysis
Both the Nanopore and Illumina raw reads quality were evaluated by the program FastQC [16]. Trimming on the Nanopore raw reads was carried out using Porechop [17] to retain reads with quality above Q12. Karaken2 [18] tool was used to align the processed reads against the SILVA 138.1 [19] database for bacterial classification. Estimation of abundances in each of the samples as performed through the Bracken pipeline [20]. Similarly, the Miseq generated paired end reads were processed with dada2 (v. 1.20.1) pipeline [21] in the R environment [22]. Forward and reverse sequences were trimmed at 270 and 230 bp, respectively to retain sequences with phred score above Q30. Successfully merged sequences were used to cluster into amplicon sequence variants (ASVs). Merged sequence clusters were then curated and refined. ASVs were aligned with Lambda [23] and taxonomic assignments were made based on the SILVA database 138.1 [19].
Sample read information tables and taxonomy tables were merged onto a phyloseq object for downstream diversity analyses. For estimation and calculation of alpha diversity indices, the data were normalized as described on the phyloseq protocol for preprocessing of samples (https:// joey7 11. github. io/). The estimate_richness function from the phyloseq package was used for estimating the alpha diversity. Similarly, beta diversity distances (Bray Curtis and Jaccard) were calculated using functions of phyloseq [24] and vegan package [25]. However, for taxonomic analyses, the data was filtered to retain ASVs that occurred greater than 10 −5 times across all the samples.

Statistical Analysis
All the statistical tests were performed in the R platform (version 4.1.1) by using base function and calling specialized packages. Comparison among the groups and timepoints were computed using the Kruskal-Wallis H test. However, for pairwise comparison, pairwise Wilcoxon rank sum test was used. Differential abundances of the taxonomic ranks among the groups were computed by Kruskal-Wallis H test. The difference between proportions was explained with false discovery rate (FDR) corrections using Benjamini-Hochberg (BH) method. The significance testing for beta diversity metrics was carried out using anosim and permanova from the vegan package [25]. To describe a patient's demographic, clinical, and investigation parameters, the data were summarized and analyzed using statistical packages for social sciences (SPSS version 24.0) or STATA (version 14) software [26]. Qualitative data were expressed as numbers and percentages as appropriate. For comparison of categorical data, the Chi-square/Fisher's exact test was performed to see the association. For continuous data, mean (standard deviation) or median (first quartile to the third quartile) was used depending on data distribution. Kaplan-Meier probability estimates were calculated for development of VAP and BSI. Hazard ratios of VAP and BSI were calculated by Cox-regression analysis adjusted to account for differences in baseline parameters. In general, a p-value < 0.05 was considered statistically significant.

Baseline Characteristics
We assessed 169 patients for eligibility of which 86 patients were included in the study, and 83 were excluded (Fig. 1). The reasons for ineligibility include 33 patients received MV for >72h, 19 patients had severe illness (APACHE-II score>25), 23 patients were expected to receive mechanical ventilation for less than 72 h, four patients were antenatal or lactating mothers, two refused to give consent, and two patients received other probiotics. Of the 86 patients included in the study, 60 patients received probiotics plus standard care, and 26 patients received standard care only. Of these, 22 patients in the probiotic group and 13 patients in the control group completed the study. Baseline characteristics were compared between the probiotic group (n=22) and control group (n=13) as listed in Table 1. Probiotics capsules were administered to patients in the probiotics group according to the study protocol with adherence of ~95%; the missed doses were administered on successive days. The two groups did not differ significantly except the proportion of patients with chronic liver disease (CLD), which was significantly more in the probiotic group than the control group (probiotic group-49%, control group-0%). The most common source of sepsis was pneumonia followed by soft tissue infection, meningitis or encephalitis, and genitourinary infections, and they were distributed similarly between groups. During DNA extraction and library preparation for NGS, a total of 19 samples were excluded from 105 samples obtained from 22 probiotic-treated and 13 control subjects due to low quality and quantity of tagged DNA fragments. Therefore, 86 samples (22 probiotic-treated and 10 control) could be retained for downstream analyses.

Gut Microbiome Composition
In case of nanopore sequencing, a total of 6521377 reads were generated with a minimum read of 24167 and maximum reads of 115949. Similarly, for MiSeq, we obtained a total of 2386014 reads with a minimum read of 156572 and a maximum of 216621 reads (Supplementary Table 1).

Gut Bacterial Diversity
For diversity analysis, reads from Nanopore and MiSeq were merged at the ASV level. The alpha diversity indices viz. Chao1, Shannon, and Simpson of the gut microbiome of the cohort were calculated and significance was tested among the groups and time-points of probiotic treatment. Chao1 index ranged from 20-249, Shannon diversity was in the range of 0.7-3.8, whereas Simpson diversity ranged from 0.25-0.96 (Supplementary Table 3). In comparison among the groups (probiotic-treated and without probiotic), no statistical significance was observed in Chao1 (p = 0.42), Shannon (p = 0.3), and Simpson (p = 0.53) diversity (Fig. 5A), B), and C)). Similarly, alpha diversity between various time-points of probiotic treatment also did not result in any difference in Chao1 (Kruskal-Wallis, p = 0.28), Shannon (Kruskal-Wallis, p = 0.13), and Simpson (Kruskal-Wallis, p = 0.29) diversity (Fig. 6A), B), and C)). However, in case of beta diversity, we observed that the Bray Curtis and Jaccard distance matrices (Supplementary Table 4) showed subtle differences between probiotic-treated and without probiotic groups. Both the group, represented on the ordination plot with a variance of 24.1% and 15.1%, respectively (Fig. 7A)). The statistical significance of this variance was computed using multi factor Permutational multivariate analysis of variance (PERMANOVA) through ADONIS, which reflected that the probiotic-treated and without probiotic group had variance among them (R 2 = 0.0187; p = 0.171), whereas variance for time-points were (R 2 = 0.0845; p = 0.121 ) (Fig. 7B)).

Compositional Variations Among the Group and Time-points
The serial changes among the identified commensal and pathogenic taxa (genus level) were tested at three different time-points viz T 0 , T 1 , and T 2 of both the probiotic-treated and without probiotic group (Table 2). In the first time-point and the second time-point, i.e. day 0 and between days 3 and 5, respectively, no significant differences among the taxa were observed. However, on the third time-point days 7-10, we observed a subtle increase of Streptococcus (p = 0.03) in the probiotic-treated group. Notably, the probiotic-treated group was administered with commercially available VSL3 probiotic, which contains live strains of Streptococcus. We further extended our analysis to the species level, in order to understand the effect of probiotic administration. For this significance, testing was done among each of the VSL3 taxa. Among the Lactobacillus genus, only L. paracasei was significantly higher among the probiotic-treated group (p = 0.0032) (Fig. 8A)). Although other strains of Lactobacillus such as L. acidophillus, L. delbruckii, and L. plantarum were detected in both the groups, no statistical significance was observed (Fig. 8B)-D)). While L. rhamnosus was detected only in the probiotic-treated group, L. casei could not be detected in our analysis. Among the Bifidobacterium strains, B. longum, and B. breve were detected barring only B. infantis. However, we did not observe any significant differences in the detected Fig. 2 Taxa tree depicting the organization of identified taxa from phylum through genus level for Nanopore reads. On the lower right-hand side is the color scale, the color intensity of the nodes represents the abundances of taxa Bifidobacterium strains among the probiotic-treated and without probiotic cohorts (Fig. 8E), F)). The most interesting observation was made in the case of S. thermophilus, which is only strain of Streptococcus present in VSL#3 probiotics. S. thermophilus was two-folds higher (p = 0.00072) in the probiotic-treated as compared to the without probiotic group, which concurred with our earlier finding in the genus level (Fig. 8G)). The fold changes of the differentially abundant species were then analyzed in the three time-points for both probiotic and without probiotictreated groups. We observed that in case of L. paracasei T1P (days 3-5, probiotic-treated) was significantly higher than T0WP (day 0, without probiotic) and T1WP (days 3-5), without probiotic-treated) (Fig. 9A). Whereas in the case of S. thermophilus, we observed a sequential increase in abundance with the progression of probiotic treatment (Fig. 9B)). Interestingly, we found that the abundance of S. thermophilus was significantly higher (p = 0.001) as compared to the abundances of T2WP subjects.

Clinical Outcomes
One episode of VAP occurred in 14 (63%) patients of the probiotic group and 8 (61%) patients of the control group with incidence of VAP estimated to be 138 and 170 per 1000 days of mechanical ventilation; ventilator-free days at end of 30 days of study period was found to be 19.9 days (±5.7), 18.2 days (±3.0) in the probiotic and control group respectively; with no significant statistical difference(p-0.51 and 0.85 respectively). The onset day of VAP was found to be similar in both groups (probiotic-6, IQR: 4.2-7; control-8, IQR: 5.7-8.7; p-value 0.30). The mini-BAL samples showed culture positivity in 10 (45%) patients of the probiotic group and 8 (61%) patients of the control group. The most common isolate was found to be Acinetobacter baumani in both the groups; probiotic-8 (36%) and control-5 (38%); with a p-value of 0.08. Kaplan-Meier curves estimating VAP occurrence rates are shown in (Fig. 10); the  The blood stream infection (BSI) was defined as bacteremia occurring 3 days after recruitment. The incidence of BSI was found to be 4 (18%) in probiotic group and 3 (27%) in control group. BSI occurrence estimates were calculated by Kaplan-Meier curves (Fig. 11). The onset day of BSI was found to be similar in both groups (probiotic-12, IQR: 5-14; control-11, IQR: 7.5-15.5; p-value 0.30). The common isolates on blood culture were K. pneumoniae (p 4%, c 23%; p-value = 0.49), Pseudomonas aeruginosa (p 4.5%, c 0) and Methicillin-resistant coagulase negative Staphylococcus aureus MR-CONS (p 4%, c 7%; p-value: 0.62). The adjusted hazards ratio was found to be 0.32 (95% CI, 0.33-3.08). The incidence of diarrhea (13% vs. 15%, p-value: 0.62) and days of diarrhea (probiotic: 6.5, control: 6; p-value = 1.0) was similar in the probiotic group and control group. The overall mortality was similar in both groups (probiotic group 12 (55%); control group 10 (71%), p 0.31), sepsis with refractory shock was most common cause (8 (36%) probiotic group and nine (69%) control group). The severity scores assessed by mean APACHE-II

Antimicrobial Administration
The carbapenem group of antibiotics was used in 17 (73%) probiotic patients and 11 (84%) control group patients. The total days of the antibiotic requirement until discharge was 20.8 days ( ± 9.3 days) in the probiotic group and 22.3 daysn (± 10 days) in the control group (p 0.67). Frequently prescribed antibiotics were penicillin, cephalosporins, carbapenems, polymyxins for gram-negative sepsis; cephalosporins, linezolid, teicoplanin for gram-positive infections tailored according to culture sensitivity results, and   anti-tubercular agents, their usage was found to be similar between the groups.

Serious Adverse Events
None of the patients had lactobacillus bacteremia or isolation of lactobacillus in sterile sites.

Discussion
Our study was aimed to determine whether probiotics preserved the composition of gut microbiota and prevented the occurrence of ventilator-associated pneumonia (VAP) and blood stream infections (BSI) in mechanically ventilated patients. Human gut is dominated by obligate anaerobes in health, whose diversity reduced in patients with sepsis [1,7]. Dysbiosis, characterized by the alteration in composition and diversity of gut microbiome, is a key factor in critical illness and leads to the progression of sepsis [27]. VAP is a common nosocomial infection with high mortality and morbidity. Myriad of factors in mechanically ventilated patients (antibiotic usage, delayed gastric emptying, sedation use, stress hormones) facilitate colonization of gut by potentially pathogenic microorganisms (PPMO) and its direct aspiration or through its effects on gut-lung axis is thought to be precursor event of VAP. Zeng and colleagues in their randomized controlled trial demonstrated there is decreased acquisition of PPMO's with administration of probiotics (Bacillus subtilis and Enterococcus faecalis) to critically ill patients on mechanical ventilation [28]. Similarly, a few trials employed probiotic formulations consisting of bacterial genera Lactobacillus (casei, paracasei, rhamnosus, plantarum) and Bifidobacterium (breve, bifidum) to study role of probiotics in prevention of gut colonization by pathogenic bacteria and showed mixed results [4,27,29,30]. Majority of these studies administered probiotics at lower doses of 1-50 billion CFU per day in contrast to our study (600-700 billion CFU per day) [4,26,28]. The duration of therapy was largely variable between studies, either until successful weaning from mechanical ventilation or for 2-4 weeks. In our study, we administered probiotics for a fixed duration of 10 days, with the missed doses administered on the following days. Earlier findings reported that gastric colonization of potentially pathogenic microorganisms was lowered with administration of probiotics in patients with critical illnesses [28,31]. However, the above studies, and majority of previous studies determining the role of gut microbiome in critically ill, employed semi-quantitative and quantitative bacterial culture methods determine the changes in gut microbiome [26,[28][29][30]. The major drawback of these conventional methods is inability to accurately determine changes in bacterial composition, as more than 1500 strains of the gut bacteria cannot be grown on culture media due to their anaerobic nature [32]. Only a few studies employed next-generation sequencing (NGS) method, which is a novel sequencing technique with ability to delineate the bacterial taxa more accurately than the conventional culture methods.
In the present study, we found that administration of probiotics had no significant impact in the alpha diversity in comparison to the control group. In contrast, in a striking study, it was previously reported that total bacterial number had a lower tendency to decline in patients receiving symbiotic therapy [7]. However, in the case of beta diversity in our study, statistical significance was observed both among the probiotic and control group, and also among the time-points. Though we could demonstrate that probiotic administration for a short duration has lowered the abundances of common ICU pathogens such as Acinetobacter, Enterococcus, Pseudomonas, Salmonella, and Staphylococcus in the probiotic group, the difference was not statistically significant. In agreement with our study, Shimizu et. al., in his recent RCT, reported an increasing trend in numbers of pathogenic bacteria, Enterococcus and Pseudomonas, with duration of critical illness; this was partly prevented with administration of synbiotics [28]. A group of subjects recruited for this study was administered probiotic VSL-3 orally for a period of 10 days. Three fecal samples were taken for gut microbiome analysis from each patient in probiotic and control groups. First sample was collected on day zero, immediately following it probiotics were initiated. Second was collected between day three to five and third between day seven and ten of initiation of the probiotic capsule. No samples were taken after stopping the probiotic capsules. Majority of the strains present in this given probiotic were detected, demonstrating the strength and resolution of the sequencing techniques used. Apart from S. thermophilus, none of the VSL#3 bacteria were significantly abundant in the probiotic-treated group. Members of the Lactobacillus and Bifidobacterium group are transient and might have failed to establish in the gut owing to their transient nature. The diseased condition of the subjects might also be a contributing factor, which might have resulted in non-adhesion. Furthermore, we speculate that prolonged exposure to probiotics, i.e., increasing the time-points, the probiotic might be successful in seeding beneficial bacteria into the gut microbiome. Moreover, the differences in geographical origin, lifestyle, and food habits of our study patients, which significantly impact the composition of an individual's microbiome, could have blunted the beneficial effects of probiotics on gut microbiome.
The incidence of diarrhea defined as >3 loose stools on at least 1 day (13% vs. 15%), and median days of diarrhea (6.5 days vs. 6 days) were similar between probiotic and control groups. However, in a contrasting study, it was demonstrated that administration of probiotics containing Saccharomyces boulardii was associated with a decrease in the incidence of diarrhea and percentage of diarrheal days [33]; this difference could be probably due to different probiotic strains used in both studies. We could not demonstrate significant impact of probiotics on diarrhea, due to lower number of diarrheal episodes in either group, probably attributed to higher severity of critical illness contributing to paralytic ileus. In line with previous studies, there were no significant beneficial effects of probiotic administration on the incidence of diarrhea in critically ill patients [27,30].
In our study, the VAP rates were high in both the groups (138 and 170 for 1000 ventilator days in the probiotic and control group), and probiotics failed to demonstrate any beneficial effect on the incidence of VAP. The pathogens cultured on mini-BAL fluid were similar between the groups; most common being Acinetobacter baumannii (57% vs. 62%) and K. pneumoniae(14% vs. 25%), resembling the common isolates of nosocomial infections [21]. Higher VAP rate observed in our study could be attributed to longer duration of mechanical ventilation, high incidence of VAP in our ICU settings, and higher severity of critical illness in our study population. Probiotics did not significantly impact occurrence of VAP, which could be due to heterogeneity of population at baseline, smaller sample size underpowered to detect a difference in incidence of VAP, a relatively shorter course of probiotic administration in comparison. In agreement with our study, administration of probiotics had no impact on VAP as demonstrated by several previous studies [27,34,35]. However, there are also several reports where there was a significant decrease in the incidence of VAP in patients receiving probiotics, which could be probably due to differences in definition of VAP used in study [4,28]. These claims were further claimed by systemic reviews and meta-analysis, which showed beneficial effects of probiotics in prevention of VAP [36,37]. The studies considered for the systemic reviews were based on surgical ICU patients, who have lesser likelihood of development of VAP along with a large heterogeneity in studies selected for the analyses.
The incidence of BSI was similar (18% and 27%) in the probiotic and control group respectively; the commonly isolated organisms were P. aeruginosa and MR-CONS. Based on our observations of previous studies, a decrease in incidence of bacteremia was observed with administration of probiotics [30,38]. In addition to reduced incidence of VAP and BSI, it was also reported that the probiotic administration was associated with a decreasing trend of other septic complications, urinary tract infections, and soft tissues infections [30]. Bacteremia is partly determined by the gut mucosal colonization of bacteria rather than the luminal bacteria. Fecal microbiome, which represents the luminal microbiome, may not be a sensitive indicator of changes in mucosal colonization, thus the bacteremia [39].
As our study was a single center, unblinded, non-randomized trial without a matched placebo, which could lead to potential bias, the results of our study should be interpreted with caution and may not be generalizable.
To the best of our knowledge, our study is the first in India to evaluate the effects of probiotics on the gut microbiome in critically ill patients and correlate it with clinical outcomes. The choice of probiotics, VSL#3 used in our study has been commonly used both in practice and prior research. Though above 500 different bacterial species were profiled through 16S metagenomics, we failed to establish the association of all these bacteria with clinical outcomes. Other limitations of our study include small sample size, single center, non-homogeneous study population, unblinded, and non-randomized nature of study, which could have led to significant heterogeneity, making the comparison between the groups questionable. Our study was limited to patients admitted in medical ICU predominantly with diagnosis of sepsis, so the results of this study may not be extrapolated to patients admitted to surgical ICU. Fecal samples were collected from the diapers of patients, which has inherent risk of contamination by perineal flora. The effect of probiotics on gut microbiome beyond 10 days was not studied. Multicentric studies are required for a better understanding of the effects of probiotics on the gut microbiome and clinical course in critically ill patients as well as confirming the findings of our study.

Conclusion
Probiotics in the commonly used dosages did not significantly alter the diversity of the gut microbiome. There was a decrease in the relative abundance of certain pathogenic bacteria in the probiotic group in comparison to the control group. Probiotic administration did not have a significant benefit in decreasing incidence of ventilator-acquired pneumonia, bloodstream infections or diarrhea, length of hospital or ICU stay, and all-cause mortality. Future studies should focus on appropriate dosages and frequency of probiotics, which can prevent the change in diversity of the gut microbiome. Data Availability Metadata and 16S rRNA gene sequences are submitted to the Sequence read archive (SRA) (NCBI) database https:// www. ncbi. nlm. nih. gov/ sra/. The submission IDs of the sequences are SUB11977043 (BioProject ID PRJNA874880) and SUB11991923 (PRJNA875248). Sequences will be available to the public immediately after acceptance of the article.