This study analysed 47 hospitalised patients exposed to antibiotic treatment to observe effects of antibiotics on the gut microbiota and determine its clinical relevance by identifying specific changes in the bacterial community. Patient sampling was limited to an older adult/pensioner population (subject age: 77±9; mean±SD). It is widely considered that the gut microbiota develops in complexity with age (Fanaro et al., 2003; Palmer et al., 2007), therefore this study does not surmise the effect of antibiotics on children.
Following data acquisition, patient samples were collated into 4 groups: ‘BEFORE’, ‘DURING’, ‘ST_PE’ & ‘LT_PE’ (criteria shown in Methods). Despite the expected dissimilarities between these groups (Figure 1), the lack of pairwise statistical significance reduces their reliability. Due to the extent of variability in the gut microbiota composition between patients, it can be suggested that this ‘grouped’ analysis may not be the best method for analysing the impact of antibiotics over time. The individuality of a patient’s gut microbiota (Benson et al., 2010; Turnbaugh et al., 2009) within the groups may influence the results causing large outliers, wide ranges and a lack of statistical significance. For example, two out of three people undergoing ciprofloxacin therapy demonstrated a profound response in their gut microbial taxa to the antibiotic, and large individual variations were observed (Dethlefsen et al., 2008). If these participants were grouped, significant changes would be hidden within the average. Box plots, median calculations and appropriate statistical tests (Wilcox, 2005) have been used throughout this study to limit the influence of outliers and demonstrate the data’s large range. This however still does not account for the large patient-patient variation, which has been a challenge in showing the statistical significance of changes. Changes observed between the groups in this research is supported by previous studies (Hill et al., 2010; Panda et al., 2014; Dethlefsen, 2008; Schaumann et al., 2007; Cochetiere, 2005; Young and Schmidt, 2004), however the low patient numbers, overlapping of groups and ultimately the lack of statistical significance limits reliability. In future studies, rather than using groups, individuals could be assessed longitudinally over time. Several studies have used this method to observe gut microbiota changes as a result of antibiotic perturbation and were successful in demonstrating statistical significance (Engelbrektson et al., 2008; Jakobsson et al., 2010). This would require controlled study participants, as a number of factors such as sample taking, patient criteria and the antibiotic type administered are difficult to control when using the samples of randomised hospitalised patients, such as in this study. However, with the acquired patient data, the demonstrated trends of dissimilarity between groups provided a basis on which to analyse the effect of antibiotics on the gut microbiota.
Antibiotics were concluded to have an effect on the patient microbiota, as the number of antibiotics taken by an individual is a driver of variation between patient’s total gut microbiota (Figure 2). This was supported by regression analysis showing that as the number of antibiotics taken by a patient increases, there is greater dissimilarity between the community composition of samples (p<0.05). The use of regression analysis to provide information about data dependencies has been questioned (Armstrong et al., 2012), therefore a One-Way ANOSIM confirmed the significance of antibiotics’ role as a driver of sample dissimilarity. Few studies have analysed the effect of multiple antibiotics however those that have demonstrated their significant influence on the gut microbial community (Ladirat et al., 2014; Zhao et al., 2014). Different antibiotic agents influence different changes in the gut microbiome (Levy, 2000; Sullivan, Edlund and Nord., 2003; Nord and Edlund, 1990; Nord et al., 1993; Rashid et al., 2015), therefore it can be inferred that effects of multiple antibiotics are additive, resulting in a greater change in the gut (figure 2).
To determine how the antibiotics affect sample-sample variation, species diversity and richness were compared between the four groups (Figure 3). The diversity of bacteria within the human gut has long been studied using both cultivation and modern molecular techniques (Ferrer et al., 2014; Ley et al., 2006; Flanagan et al., 2007; Jernberg et al., 2007; Woodmansey et al., 2004; ). The microflora composition from one gut to the next differs remarkably, with diversity linked to a number of environmental variables such as diet (Flint et al., 2007; Xu and Knight, 2015) and genetics (Spor et al., 2011; Hansen et al., 2010; Goodrich et al., 2014). Our results (Figure 3) align with previous research, concluding that antibiotics reduce diversity in the gut (Woodmansey et al., 2004; Dethlefsen et al., 2008; Pérez-Cobas et al., 2012; Ferrer et al., 2014). Low diversity in the gut microbiota has been associated with a number of clinical conditions, such as increased risk of colorectal cancer (DeWeerdt, 2015), Crohn’s disease (Manichanh et al. 2006; Dicksved et al. 2008) obesity (Turnbaugh et al., 2009) and rheumatoid arthritis (Scher and Abramson, 2011). ChaO1 and Shannon diversity indices demonstrated similar patterns, however with a greater decrease in Shannon diversity post antibiotic therapy – showing a decrease in the number of species as well as their uneven distribution. This is known as competitive exclusion; antibiotics can cause the decrease in certain species, resulting in the proliferation of others, causing a negative impact on host health (Modi et al., 2014; Miller et al., 1957; Bohnhoff et al., 1962). For example, commensal Lactobacilli spp. induce the secretion of intestinal mucins which inhibit adhesion of pathogenic E. coli (Mack et al., 1999). Antibiotics significantly reduce Lactobacilli (Bartosch et al., 2004; Lode et al., 2001; Lidbeck et al., 1988) which can therefore result in the colonisation and infection of E.Coli. Diversity scores are an important measure of gut microbial community change as a result of antibiotic perturbation. The results also demonstrate gut resilience to antibiotics as there are no ‘long term’ diversity changes. Another study found similar results of diversity changes, with five out of six subjects returning to pre-treatment levels by day 30 after amoxicillin therapy (Cochetiere et al., 2005). Diversity scores in this study can be deemed as inconclusive as the results were not significant when compared between groups. This could be due to the lack of statistical dissimilarity between the four groups, as these groups are not experimentally strong and can be considered a profound limitation of this experiment. In future, experimental groups should be created prior to sample collection or an alternative method could be considered, such as the study of individuals previously mentioned. Diversity scores do not give a complete view of the gut bacterial changes, therefore it was important to observe the community composition at the taxonomic levels of each stage of antibiotic treatment. The analysis of phylum-level changes within the gut microbiota has found associated clinical disorders such as obesity (Backhed et al, 2004; Turnbaugh et al., 2009) and inflammatory bowel disease (Hansen et al. 2010; Frank et al. 2007; Sokol et al. 2006).
It is widely considered that Bacteroidetes and Firmicutes dominate the gut microflora in healthy subjects (O’Hara et al., 2006; Turnbaugh et al., 2006; Mariat et al., 2009). Our results supported this, as the dominant phyla found in the majority of samples were Bacteroidetes and Firmicutes, making up a median of over 95% of the total bacterial phyla across all stages of antibiotic treatment (Figure 4 A-D). The domination of these two bacterial phyla demonstrates resilience of these taxa, and thus by extension a large percentage of the gut microbiota, against antibiotic perturbation. Their predominance also shows the importance of their presence and role within the human gut. For example, Ley et al. found that obese persons possessed a greater proportion of Firmicutes compared to Bacteroidetes, whilst weight loss was associated with the opposite (Ley et al., 2006). This Firmicutes/Bacteroidetes ratio has also been linked to age, with significant differences between adults and elderly (Firmicutes/Bacteroidetes ratio of 10.9 and 0.6 respectively). The elderly subjects were identified to have different digestive physiology when compared to young adults (25-45 years old) which may be linked to microbiota composition (Mariat et al., 2009). These studies demonstrate the importance of Bacteroidetes and Firmicutes proportions within the gut microbiota and also their links to the functionality and health of the human host. It also highlights the potential influence of patient factors, which were not considered in this study.
The Bacteroidetes phylum decreased in number during antibiotics, this increased to almost pre-treatment levels short term post end of treatment, however still did not reach pre-treatment levels at the long term post end stage. Firmicutes can be observed to experience a steady decrease throughout the 4 stages, however this was not significant. Other studies have also found this problem in finding statistical significance, due to the large variation between individuals in Firmicutes and Bacteroidetes proportions (Walters et al., 2014; Moran and Shanahan, 2014). Also, different changes have been observed with regards to these dominant phyla. Some support our findings, with Bacteroidetes and Firmicutes decreasing as a result of antibiotics (Schaumann et al., 2007; Manichanh et al., 2010). Conversely, it has been observed that Bacteroidetes increase during treatment (Perez-Cobas et al., 2013; Panda et al., 2014). Differences in observations could be due to the usage of different antibiotics, dosage and research models. This demonstrates a downfall of this study, as changes as a result of specific antibiotics cannot be observed – therefore it is difficult to see significant changes. The detailed observation of the Bacteroidetes and Firmicutes at a refined taxonomic level was important to develop a more in depth analysis of the community composition and functionality of the gut.
The Bacteroidetes are a gram negative phylum found, in this study, to be composed of 4 abundant families. The relationship between Bacteroidetes and its human host is one of mutualism, as the fitness of both parties increase as a result of their partnership (Backhed et al., 2005). Bacteroidetes produce butyrate by fermentation which has been demonstrated to influence a healthy gut as a result of its antineoplastic properties (Kim and Milner, 2007). Bacteroidetes also have a role in the degradation of host derived carbohydrates (Salyers et al., 1977) and plant polymers. This polysaccharide digestion in the gut provides the host with short chain fatty acids available as an energy source which constitutes 7–10% of the daily calorie supply (Hooper et al., 2002; Thomas et al., 2011). The Bacteroidaceae family are obligate anaerobes and the most abundant of the 4 Bacteroidetes families. The proportions of the Bacteroidaceae family increased during antibiotic treatment increasing further short term post end, followed by a large decrease below pre-treatment levels. Bacteroides were the only genus identified within this family and is associated with Western diets – consisting of sugars and animal fats (Wu et al.,2011). Bacteroides are resistant to many antibiotics and are a reservoir for bacterial resistance (Salyers et al., 2004; Löfmark et al., 2006). This explains their increase during treatment as a result of competitive exclusion; other bacteria will decrease as a result of antimicrobial therapy, allowing the Bacteroides the space and resources to proliferate. Bacteroides have complex systems enabling them to dominate within the gut. For example Bacteroides thetaiotaomicron possess small regulatory proteins known as extracytoplasmic function sigma factors (Helmann, 2002) as well as two component systems that allow sensing and adaptation to environmental stresses such as antibiotic perturbation. Increases in Bacteroides spp. have been associated with obesity (Bervoets et al., 2013; Vael et al., 2011), and therefore may be a problem for patients experiencing a long term increase in this genera as a result of antibiotics. Bacteroides returned to pre-treatment levels in the majority of patients, however two possessed a gut microbiota made up of 60% Bacteroides post therapy. Their ‘before’ samples had relatively low proportions, therefore this large change means that they are now more at risk of obesity as a result of antibiotics. These two samples fortunately had corresponding ‘before’ samples in order to create this comparison, this shows the importance of studying individuals as patients will experience different changes and thus will have varying levels of susceptibility to other clinical defects.
Rikenellaceae maintained similar levels during antibiotic treatment, decreasing during the short term and long term recovery (Figure 5F). Cox supported these findings, showing reductions in Rikenellaceae and suggesting an association of this bacteria with adult metabolism (Cox et al., 2014).
Porphyromonadaceae increased throughout treatment and short term recovery, however the gut demonstrates its resilience to antibiotic perturbation as its levels decrease during the long term phase. Our findings aligned with previous research, showing the proliferation of the Parabacteroides spp. (the most dominant species of the Porphyromonadaceae family) during antibiotic therapy. This is due to its possession of resistance genes, for example β-lactamase production was identified in 92% of Parabacteroides distasonis isolates (Nakano et al., 2011).
Prevotellaceae remained at near negligible levels before, during and short term post end of antibiotic therapy, but increased largely in the long term. This family is involved in the digestion of complex polysaccharides and is therefore strongly associated with diet (De Filippo et al., 2010) and the lack of diet consideration within this study may influence the data trends. For example, in two patients who possessed a high abundance of Prevotellaceae, the increase at the long term stage was still significant (data not shown), once again highlighting the importance of observing individuals. Stool consistency has been recently associated with the gut microbiota and in particular, the Prevotella enterotype is linked with loose stools. It is uncertain whether the diet, subsequent ‘transit-time’ or ‘water content’ causes the increase in Prevotella however this research can be applied to the effect of antibiotics. A common side effect of antibiotic therapy is ‘loose stools’ and therefore the findings of Vandeputte et al. may help to explain the observed increase in Prevotellaceae in this study (Vandeputte et al., 2015). From this analysis, it can be suggested that knowledge of the gut bacterial enterotype of an antibiotic taking patient may assist the prevention or treatment of antibiotic-associated health defects. It has been observed therefore that changes in Bacteroidetes as a result of antibiotic therapy can have a profound impact on patient wellbeing.
The Firmicutes are a gram positive phylum possessing two dominant families: Ruminococcaceae and Lachnospiraceae. Proportions of the Ruminococcaceae family increased during treatment quickly followed by ‘normal’ levels at the short term recovery period and this was maintained long term. Interestingly, Dethlefsen et al. found the opposite, observing depletion of the Ruminococcaceae family as a result of Ciprofloxacin therapy – a conserved response across the majority of subjects within the study (Dethlefsen et al., 2008). These opposed findings highlight the possible influence of confounding patient factors on this research. Diet, age and health status were also found to influence alterations in the gut microbiota of elderly long-stay hospitalised patients (Claesson et al., 2012). The study found an increased abundance of Oscillibacter (genus of the Ruminococcaceae family) along with an increased incidence of frailty. These findings support the idea that patient factors should be considered when analysing gut microbial data.
Lachnospiraceae was the second most abundant family of the Firmicutes phyla. Proportions decreased during antibiotic treatment by almost half, continuing to decrease short term post end of therapy and the proportion did not return to the pre-treatment composition. This aligned with previous research showing similar patterns of loss and restoration of the Lachnospiraceae family as a result of antibiotics (Ferrer et al., 2012). Other studies demonstrated the potential of this family to protect the patient against pathogenic bacteria such as C.difficile (Reeves et al., 2012). Lachnospiraceae produces short chain fatty acids from complex polysaccharides helping to maintain gut homeostasis (Cotta et al., 2006) and low concentrations of this short chain fatty acid end product have been associated with C.difficile infections (Hove et al., 1996; Rolfe et al.,1984).
Enterococcaceae varied largely between samples; during antibiotic therapy it became the most abundant family for two patients, increasing dramatically in two more. Enterococcaceae are gram positive bacteria, and are of particular clinical interest. Microorganisms of this family (Enterococcus faecalis and Enterococcus faecium for example) are important multi-drug resistant pathogens involved in hospital acquired infections (HAIs) (Vincent, 2003; Olawale et al., 2011; Hunt, 1998). Enterococci possess intrinsic resistance to beta-lactams as well as aminoglycosides and acquired resistance to more antibiotics, not only this but vancomycin resistant strains have also been discovered (Cetinkaya et al, 2000; Fraser, 2015). Therefore most antimicrobials will not reduce Enterococci and instead promote their growth and facilitate the transmission of resistance genes to other bacteria. This can render the patient susceptible to acquiring more infections. However this increase in Enterococcaceae only occurs in 4 patients. Variation between patients’ gut microbiota demonstrates the influence of individuality, further acknowledging that there are patient predispositions that can affect the changes observed.
The results have demonstrated the ability of antibiotics to affect the bacterial community composition and diversity. A limitation of this analysis is that patients were receiving different types of antibiotic, and in different quantities e.g. patient 047 had taken 7 antibiotics in 35 days. Our results demonstrate that the number of antibiotics taken has an effect on the gut microbiota (Figure 2) and also that the type of antibiotic is a driver of dissimilarity between samples (Figure 7A). Most research to date focuses on the comparison between effects of one type or class of antibiotics on the gut microbiota (Panda et al., 2014; Dethlefsen et al., 2008; Jernberg et al., 2007; Jernberg et al., 2010; Jernberg et al., 2005; Jakobsson et al., 2010). This is an effective means of research as it distinguishes the specific changes that may occur as a result of a class of antibiotics, enabling the application of this knowledge clinically for example in the restoration of the disturbed gut flora. This is an ideal approach for this type of research.
In this study, from all samples collected during antibiotic treatment, just over half were associated with one antibiotic – with augmentin and tazocin being the only two therapies that possess more than two samples. Therefore a ‘single-drug’ focus was used to analyse Chao1 species richness and Shannon diversity between 3 groups: ‘Before’, augmentin-only and tazocin-only (Figure 8 A&B). Both antibiotics decreased the species richness and diversity, however tazocin acting more so than augmentin. This dip in diversity implicates the eviction of ‘normal’ and antimicrobial susceptible bacteria, resulting in an increased growth and dominance of resistant strains (Modi et al., 2014; Miller et al., 1957; Bohnhoff et al., 1962). The shift in population could also allow for the colonisation of pathogenic bacteria and fungi (Mack et al., 1999; VanScoy et al., 2013; Gerding et al ,2008; Sullivan et al., 2001). It has been observed that different antibiotics result in different changes in the gut ecosystem, with the route of administration, luminal content concentration and spectrum of antibiotics having an influence (Nord et al 1984a, 1984b; Edlund & Nord, 1993; Sullivan et al., 2001; Janczyk et al., 2007). Augmentin is a broad spectrum, oral penicillin antibiotic that contains amoxicillin and a beta-lactamase inhibitor (clavulanate potassium). Both amoxicillin and clavulanate potassium are effectively absorbed from the gastrointestinal tract, (White et al., 2004). This effective uptake of augmentin limits the destruction of the gut bacterial flora compared to other poorly absorbed agents. Antibiotics that are poorly absorbed from the gut, such as rifaximin, are used to treat gastrointestinal infections (Scarpignato and Pelosini, 2005). Due to its almost non-absorption from the gut, rifaximin has been shown to ‘reset’ the gut microbiota of patients with functional bowel disorders (Mishkin, 2001; Sharara et al., 2006; Pimentel et al., 2011). However there are concerns with this treatment about the potential overgrowth of pathogenic bacteria, infections with C.difficile and antibiotic resistance (Valentin et al., 2011; Curry et al., 2009; Khardori, 2010; Chang et al.,2008). Tazocin is also a penicillin class antibiotic containing piperacillin and tazobactam but is administered intravenously. As well as absorption, the route of antibiotic administration has been discovered to influence gut dynamics; findings suggest that oral antibiotics have a larger effect on antibiotic resistance development when compared to I.V injections (Zhang et al., 2013). Therefore, it would be expected that augmentin would have a larger effect on diversity, however this was not the case. Other studies have found the opposite to Zhang et al., with results showing a greater change in the gut microflora diversity as a result of IV antibiotics compared to other routes of administration (Zhao, 2014). Therefore other factors must influence the difference in diversity observed between the two antimicrobial agents. Tazocin was found to have a pronounced effect on the lower intestinal microbiota due to its high concentration in the bile (Nord and Edlund 1990). Research has shown that patients treated with a course of tazocin experienced a decrease in most bacterial flora, however Bacteroides and anaerobic gram negative cocci were unaffected (Rafii et al., 2008). Augmentin increases Enterococci, Bacteroidetes, and aerobic gram positive bacteria whilst decreases have been observed in Bifidobacteria, Lactobacilli, and Clostridia (Lode et al., 2001). It can therefore be inferred that there are different global effects in the gut microbial flora community depending on the type of antibiotic and in future, studies could potentially focus on single-drug research (discussed further below). This research is clinically relevant when looking at hospital acquired infections, such as the overgrowth and colonisation of Clostridium difficile as a result of a perturbed gut flora (4.9% of patients taking Tazocin acquired a C.difficle infection; VanScoy et al., 2013).
The lack of statistical significance between these antibiotics encouraged further analysis on the effect of antibiotic type. Regression analysis estimated the significance of the effect of different antibiotics on the total bacterial OTU composition (Figure 7B). Amoxicillin, meropenem and fluoroquinolone were found statistically significant. The links between variables estimated by regression only demonstrate correlation, not causation. Also, when looking at the application of these three antibiotics, they were often prescribed to patients as multi-drug combinations. This therefore limits the reliability of the results as their influence cannot be inextricably separated from the action of the other prescribed drugs. This poses a definitive challenge for this area of research. Whilst single-drug research may be the best method for studying the effect of antibiotics on the gut, the unpredictability of patient illness and requirements for multiple antibiotic prescriptions limits its clinical application. For example, patients admitted into intensive care units (ICUs) are prescribed multiple antibiotics due to the severity of their illness. A study found that a mean of 6.2 antibiotics (range from 5.1 to 12) were prescribed per patient (Williams et al., 2011). The effect of antibiotics on the gut microbiota therefore cannot be predicted or anticipated when such a cocktail of drugs has been administered. This is not only limited to ICU patients as a number of antibiotics are usually prescribed in combinations, as this can be key in combatting resistant bacteria (Holm et al., 1986; Fitzgeral et al., 2006; Cokol et al., 2011). A common combination of antibiotics in this study, for example, was benzylpenicillin and fluoroquinolones. As a future perspective, it could be useful to identify the common combinations of antibiotics and analyse their affect in unison rather than looking at individual classes of drug (Ubeda et al., 2010). This research could be fundamental in reducing hospital acquired infections such as C.difficile, as the use of multiple antimicrobials has been shown to increase the risk of infection (Bartlett and Gerding, 2008).
In conclusion, the results of this research have aligned with previous studies on the trends of change in community structure and diversity of the gut microbiome as a result of antibiotic perturbation. Antibiotics have an effect on the community composition and reduce the diversity of the gut bacteria during and short term post end of antibiotic treatment; some changes can still be seen long term post end of therapy. The increase and overgrowth of resistant bacterial groups, such as Bacteroidaceae and Enterococcaceae, along with the loss of predominant groups, such as Lachnospiraceae, can result in the loss of colonisation resistance rendering the patient more susceptible to infections such as C.difficile. However resilience of the gut was demonstrated by some microflora families, such as Ruminococcaceae, as a result of their restoration to pre-treatment levels. This demonstrates the importance of observing specific bacterial groups, as not all microorganisms will have the same response to antibiotic perturbation and more importantly, this may have a range of subsequent clinical effects on the host. Inter-sample variation, the abundance of contributing patient factors and the groupings of samples provided a limitation for seeing the significance and confirming the reliability of the results. Given larger cohorts and samples, it would be better to analyse this data in individuals longitudinally in order to observe the true dynamics of change of the gut microbiota. Also, testing common antimicrobial drug combinations rather than the usual ‘single-drug’ approach could be a method for the future in analysing the effect of antibiotics on the gut microbiome.