The Prevalence of pks⁺ E. coli in Cystic Fibrosis

1. Introduction

The gastrointestinal microbiome in cystic fibrosis (CF) demonstrates marked differences compared to healthy subjects from an early age [1]. A significantly altered composition of the gut microbiome [2,3] influences functionality [1,3], delays maturation [4,5] and contributes to a persistent, inflammatory state [4]. Improved life expectancy in CF patients, resulting from therapeutic advancements, has contributed to rising diagnoses of concomitant disorders, including gastrointestinal malignancy [6]. Individuals with CF harbour around 10 times the risk of developing colon cancer, and earlier onset of cancer, when compared to the general population [6,7]. Crucially, gut dysbiosis and increased inflammation is both present in CF and implicated in the development of colorectal cancer (CRC), suggesting these factors may play a role in CF-related CRC.

The gut microbiome itself becomes established and is stable from age four into adulthood [8]. However, an altered intestinal milieu is present in CF from early childhood. Therefore, there is reason to consider that the physiological steps which prime CF patients to develop CRC may be identifiable in childhood. Children with CF (cwCF) demonstrate significantly increased abundance of Escherichia coli (E. coli) in the gut [1,9]. While E. coli is commensal, some strains may cause pathology, particularly in vulnerable gastrointestinal environments like that of people with CF [10]. E. coli derived from the B2 phylogenetic group can produce colibactin, a genotoxic chemical compound, through expression of a polyketide synthesis (pks) locus [11]. By inducing deoxy-ribonucleic acid (DNA) damage and cell cycle arrest, colibactin promotes colon tumour growth, and modifies the tumour microenvironment [12].

Colibactin is strongly implicated in CRC in non-CF populations and has demonstrated causative links to microbial dysbiosis and intestinal inflammation [12,13]. Together, this suggests pks⁺ E. coli may be overrepresented in CF. However, no studies have investigated a potential role of colibactin in CF-related gut changes. Isolation of colibactin for research purposes is difficult due to chemical instability and the fact it is produced in incredibly small quantities [14]. To counter this, previous research methods have focused on the pks island and its regulation of colibactin production, to characterise potential genotoxic ability and therapeutic potential [15].

In this study, we aimed to investigate differences in pks island expression, as a proxy for colibactin, between cwCF and HC, and within cwCF over time. We also investigated inflammation and microbial diversity within the CF gut, and potential associations between pks expression, gut inflammation and dysbiosis.

2. Materials and Methods

Study Design

CwCF were recruited from the CF clinic at Sydney Children’s Hospital (SCH) Randwick, an Australian tertiary paediatric hospital, as part of the ‘Evaluating the Alimentary and Respiratory Tracts in Health and Disease’ (EARTH) research programme [16]. HC children were recruited from advertisement, word-of-mouth, and outpatient clinics (e.g., healthy children from the fracture clinic). Inclusion criteria into the EARTH programme has been previously outlined in the EARTH protocol [16]. Three consecutive stool samples, collected at separate timepoints spaced approximately six months apart were required for inclusion in longitudinal analysis. The EARTH research programme was approved by the South-Eastern Sydney Area Health Service, Human Research Ethics Committee, Sydney, Australia. Ethics number: HREC/18/SCHN/26 and 2019/ETH05421.

DNA Extraction and Sequencing

The DNA was extracted following manufacturer instructions from homogenised stool samples using the QIAamp DNA Mini Kit (Qiagen, Germantown, Maryland, USA), as utilised by Nielsen et al. [17]. The 16s rRNA gene amplicon sequencing was performed at the Ramaciotti Centre for Genomics, University of New South Wales on the Illumina MiSeq platform (v3, 2 × 300 bp). Sequences were quality filtered, clustered into unique zero-distance operation taxonomic units (zOTUs), and classified as described by Coffey et al. [1].

Faecal Calprotectin

Faecal calprotectin was measured to assess presence of intestinal inflammation, using the BÜHLMANN fCAL ELISA kit (BÜHLMANN Laboratories AG, Schönenbuch, Switzerland), as validated by Garnett et al. [18].

Identification of E. coli and pks-Producing E. coli

Polymerase chain reaction (PCR) was performed to screen faecal DNA samples for pks⁺ E. coli. Detection of pks⁺ bacterial DNA involved the amplification of a 283 bp sequence in the gene clbB, part of the pks island, with the following primer set: forward primer

5’-GCGCATCCTCAAGAGTAAATA- 3’, reverse primer

5’ -GCGCTCTATGCTCATCAACC-3’ [19]. To confirm the presence of E. coli specific pks⁺, samples were also screened for E. coli, via amplification of a primer set universal for the E. coli specific gene UidA: forward primer 5’-AAAACGGCAAGAAAAAGCAG-3’, reverse primer 5’–ACGCGTGGTTACAGTCTTGCG-3’ [20]. Samples were considered as pks⁺ E. coli positive if they were positive for the clbB and UidA gene cluster on PCR.

Invitrogen® Super Mix (Invitrogen/Thermo Fisher Scientific, 10572014) and Invitrogen® Platinum™ II Hot-Start Green PCR Master Mix (2X) (Invitrogen/Thermo Fisher Scientific, 14001014) were used for the PCR reactions. DNA from pks⁺ E. coli, isolated from Mutaflor® Escherichia coli strain Nissle 1917 (Mutaflor, 123456), was used as a positive control, while a negative control was prepared by substituting the DNA with nuclease free water. The final volume of the PCR reactions was 25 μl, consisting of 22.5 μl of Invitrogen® Super Mix or Invitrogen® Platinum™ II Hot-Start Green PCR Master Mix (2X), 0.5 μl of each primer (forward and reverse) and 1.5 μl of the isolated stool sample DNA. Amplification was performed using a SimpliAmp™ Thermal Cycler (Applied Biosystems/Thermo Fisher Scientific, Waltham, Massachusetts, USA) (94°C for 2 minutes, followed by 35 cycles of 94°C for 15 seconds, 58.5°C for 15 seconds, 72°C for 60 seconds, then a final holding step at 4°C). The products were visualized and evaluated on a prepared 1% agarose gel containing SYBR® Safe DNA gel stain (Invitrogen/Thermo Fisher Scientific, S33102).

Statistical Analysis

Statistical analysis was performed with the use of RStudio v2023.03.0+386. Diagnostic differences in pks⁺ E. coli prevalence between cohorts were compared with Fisher’s Exact test and described as percentages. Continuous variables were described by mean and standard deviation (SD) and compared through paired Student’s t-tests. Statistical significance was reported at the 5% significance level. Alpha diversity was reported using two metrics: Richness - a measure of unique zOTUs - and Shannon diversity index - a measure of species abundance and evenness. Linear mixed models were constructed to control for age and gender when comparing continuous variables in longitudinal data. Permutational multivariate analysis of variance (PERMANOVA) tests (permutations = 1000) were utilised to test if beta diversity was significantly different between groups (‘CF versus HC’ and “Timepoint 1‘ versus ‘Timepoint 2’ versus ‘Timepoint 3’) and for age and gender using the vegan function adonis2 [21]. Graphs were generated using ggplot2 in R [22]. Differences in the proportion of pks prevalence between groups were determined using Pearson’s chi-squared test for independence. Microbial alpha diversity and phylogeny-based beta diversity were calculated with a dataset subsampled to 33839 and 13497 sequences per sample for baseline and longitudinal analysis respectively.

3. Results

Demographics

The study involved 55 children with CF and 55 HC [16]. Both the CF and HC group were comprised of 26 males (47%). The mean (SD) age of CF patients and HC was 7.7 years (5.1), and 7.8 years (5.0) respectively, p = 0.17. Among the CF patients, 48 patients (87%) were pancreatic insufficient. Anthropometric z scores for our CF cohort were also calculated based on WHO regression data for the general paediatric population and are presented in Table 1A. The CF subgroup (n = 23) included for longitudinal analysis was comprised of 10 males and 13 females. The mean (SD) age at each timepoint was 7.0 years (4.2), 7.8 years (4.2) and 8.8 years (3.9) for timepoints 1, 2 and 3 respectively. Of this subset, 19 patients (83%) were pancreatic insufficient (Table 1B).

Table 1. Demographics for cross-sectional comparison (1A) and longitudinal analysis (1B).

1A: Baseline				Cystic fibrosis patients			Healthy controls		P-value	95% CI
Number of participants				55			55
Male sex, n (%)				26 (47%)			26 (47%)		1
Mean age in years (SD)				7.66 (5.1)			7.78 (5.0)		0.17	(-0.29,0.05)
Exocrine pancreatic status
		Pancreatic sufficient (%)		7 (13%)
		Pancreatic insufficient (%)		48 (87%)
Anthropometrics
		Mean weight z-scores (SD)		-0.24 (1.23)
		Mean height z-scores (SD)		0.00 (1.01)
		Mean BMI z-scores (SD)		0.39 (0.71)
1B: Longitudinal			Cystic fibrosis patients
Number of participants			23
Male sex, n (%)			10 (38%)
Mean age in years (SD)			T1		T2	T3
			7.0 (4.2)		7.8 (4.2)	8.8 (3.9)
Exocrine pancreatic status
	Pancreatic sufficient (%)		4 (17%)
	Pancreatic insufficient (%)		19 (83%)

Cross-sectional analysis involved cystic fibrosis patients (n = 55) and healthy controls (n = 55) (Table 1A). Longitudinal analysis involved cystic fibrosis patients only (n = 23) (Table 1B). Fisher’s Exact test was utilised to determine gender differences between cystic fibrosis patients and healthy controls. Anthropometric z-scores were calculated using WHO regression data for the general paediatric population. Student’s t-test was utilised to determine if significant differences in age exist between CF and HC. CF = cystic fibrosis, HC = healthy controls, T1 = Timepoint 1, T2 = Timepoint 2, T3 = Timepoint 3, n = number of samples, CI = confidence interval, SD = standard deviation.

Table 1. Demographics for cross-sectional comparison (1A) and longitudinal analysis (1B).

1A: Baseline				Cystic fibrosis patients			Healthy controls		P-value	95% CI
Number of participants				55			55
Male sex, n (%)				26 (47%)			26 (47%)		1
Mean age in years (SD)				7.66 (5.1)			7.78 (5.0)		0.17	(-0.29,0.05)
Exocrine pancreatic status
		Pancreatic sufficient (%)		7 (13%)
		Pancreatic insufficient (%)		48 (87%)
Anthropometrics
		Mean weight z-scores (SD)		-0.24 (1.23)
		Mean height z-scores (SD)		0.00 (1.01)
		Mean BMI z-scores (SD)		0.39 (0.71)
1B: Longitudinal			Cystic fibrosis patients
Number of participants			23
Male sex, n (%)			10 (38%)
Mean age in years (SD)			T1		T2	T3
			7.0 (4.2)		7.8 (4.2)	8.8 (3.9)
Exocrine pancreatic status
	Pancreatic sufficient (%)		4 (17%)
	Pancreatic insufficient (%)		19 (83%)

Cross-sectional analysis involved cystic fibrosis patients (n = 55) and healthy controls (n = 55) (Table 1A). Longitudinal analysis involved cystic fibrosis patients only (n = 23) (Table 1B). Fisher’s Exact test was utilised to determine gender differences between cystic fibrosis patients and healthy controls. Anthropometric z-scores were calculated using WHO regression data for the general paediatric population. Student’s t-test was utilised to determine if significant differences in age exist between CF and HC. CF = cystic fibrosis, HC = healthy controls, T1 = Timepoint 1, T2 = Timepoint 2, T3 = Timepoint 3, n = number of samples, CI = confidence interval, SD = standard deviation.

Microbiome Diversity and Inflammation

Mean (SD) microbial richness was significantly decreased in CF patients compared to HC (108 (45) versus 207 (71); p < 0.001) (Figure 1A, 1D). Mean (SD) Shannon diversity (H) was also significantly decreased among CF patients compared to HC (2.2 (0.65) versus 3.2 (0.65); p < 0.001) (Figure 1B, 1E). When accounting for age, Shannon diversity remained significantly lower relative to HC (p = 0.001). No difference in either alpha diversity indices were observed between males and females. CF participants had significantly elevated mean (SD) calprotectin levels, indicative of increased intestinal inflammation relative to HC (232 (402) mg/kg vs 64 (73) mg/kg; p < 0.003). This disparity remained unchanged when controlling for age (Figure 1C, 1F).

Figure 1. Alpha diversity indices and faecal calprotectin for cross-sectional comparison of children with cystic fibrosis versus healthy controls. Boxplots depict microbial richness (number of zOTUs)) (1A), Shannon diversity (Shannon Index (H)) (1B), and the concentration of faecal calprotectin (mg/kg) (1C) in faecal samples of the cystic fibrosis and healthy control cohorts. Scatterplots illustrate sample microbial richness (1D), Shannon diversity (1E), and faecal calprotectin (1F) in the cystic fibrosis and healthy control cohorts with respect to age (years). Coloured lines indicate the cohort means, while shaded regions indicate 95% confidence intervals constructed from generalised linear models (1D, 1E, 1F). CF = cystic fibrosis, HC = healthy controls, zOTU =zero-distance operation taxonomic unit, calprotectin = a measure of intestinal inflammation.

Figure 1. Alpha diversity indices and faecal calprotectin for cross-sectional comparison of children with cystic fibrosis versus healthy controls. Boxplots depict microbial richness (number of zOTUs)) (1A), Shannon diversity (Shannon Index (H)) (1B), and the concentration of faecal calprotectin (mg/kg) (1C) in faecal samples of the cystic fibrosis and healthy control cohorts. Scatterplots illustrate sample microbial richness (1D), Shannon diversity (1E), and faecal calprotectin (1F) in the cystic fibrosis and healthy control cohorts with respect to age (years). Coloured lines indicate the cohort means, while shaded regions indicate 95% confidence intervals constructed from generalised linear models (1D, 1E, 1F). CF = cystic fibrosis, HC = healthy controls, zOTU =zero-distance operation taxonomic unit, calprotectin = a measure of intestinal inflammation.

Prevalence of pks⁺ E. coli at Baseline

The presence of clbB was identified in 21 CF patients and 26 HC (Table 2A). Of the 47 individuals positive for the clbB gene, four patients (three CF, one HC) were negative for the UidA gene cluster. There was no significant difference in proportions of clbB positive subjects between CF and HC using Pearson’s chi-squared test for independence (p = 0.32, df = 1). In our CF cohort, the presence of clbB was also not associated with any significant differences in mean calprotectin concentration, microbial richness, or Shannon diversity (Table 2B).

Table 2A depicts a two-way frequency table of clbB gene status (positive versus negative) against diagnosis (cystic fibrosis versus healthy controls) from cross-sectional baseline analysis of faecal DNA samples. The clbB gene was screened for using PCR to determine the prevalence of the pks island. Student’s t-test was utilised in Table 2B to determine if correlations between clbB prevalence and microbial diversity indices (Shannon diversity and microbial richness) or calprotectin exist. Pks = polyketide-synthase, Microbial richness = a measure of unique zOTUs, Shannon diversity = a measure of species abundance and evenness, Calprotectin = a measure of intestinal inflammation, SD = standard deviation.

Prevalence of pks⁺ E. coli in Longitudinal Analysis of the CF Cohort

Chi-squared test for independence revealed no significant difference in clbB prevalence across timepoints (p = 0.4, df = 2), even when accounting for age (p = 0.6) and antibiotics usage (p = 0.18). However, only 7/23 (30%) subjects were consistent in their pks⁺ E. coli status across all three timepoints (positive or negative). The fluctuation in pks⁺ E. coli prevalence across each timepoint is presented in full in Table 3. Furthermore, 12/69 (17%) of the total samples were negative for UidA. A linear mixed model was used to investigate the effects of age, gender, and sample timepoint on alpha diversity indices, with Patient ID included as a random effect to account for repeated measures. There was no significant difference in alpha diversity parameters (richness, Shannon diversity or chao1) between sample timepoints (variance = 412.9, 0.15, 344.9; p = 0.9, 0.2, 0.9 respectively) and no effect of gender. Alpha diversity parameters increased with age, but this was only a significant predictor for chao1 (β = 4.07, SE = 1.75, t = 2.32, p = 0.03 (Table 4).

The prevalence of pks⁺ E. coli was determined by being positive for both the clbB and UidA gene cluster on PCR. Here, ‘+’ indicates that pks⁺ E. coli is present in the sample, ‘-‘ indicates that pks⁺ E. coli is absent in the sample, and a grey space indicates that E. coli was undetectable in this sample following UidA gene PCR. Each row corresponds to one, de-identified cystic fibrosis patient involved in the longitudinal analysis. Samples were collected across three timepoints, spaced roughly six months apart. pks⁺ E. coli = polyketide synthase-positive Escherichia coli.

4. Discussion

This study is the first of its kind to investigate the colibactin-producing pks⁺ E. coli in the CF gut, and we identified no evidence suggestive of increased pks⁺ E. coli prevalence in cwCF. Furthermore, pks⁺ E. coli prevalence was not linked to differences in calprotectin concentration, gut microbial richness, or Shannon diversity index within cwCF. While many CF patients demonstrated fluctuations in pks⁺ E. coli status across timepoints, there was no significant difference overall. We also identified significant reductions in microbial diversity of the CF gut, and significant differences in bacterial composition when compared to HC. CF patients exhibited alterations in over 70 genera, including that of Fusobacterium, which has demonstrated previous links to CRC development [1]. Faecal calprotectin was significantly higher in CF compared to HC but within CF, there was no difference in calprotectin concentrations over time.

Our results suggest that increased abundance of pks⁺ E. coli is not a major contributor to alterations in gut richness or diversity for cwCF. Furthermore, we did not identify any significant relationship between inflammation or gut microbial diversity and the status of pks⁺ E. coli. While alpha and beta diversity provide an important characterisation of the gut microbiota, they do not characterise DNA changes which occur following exposure to genotoxic bacteria, such as pks⁺ E. coli. Future research should target a more definitive understanding of these DNA changes, and their implications, on CRC development.

The significantly lower gut microbial richness and diversity in cwCF and increased faecal calprotectin mirrors previous studies in patients with CF [3,23]. Within CF, we observed no significant temporal changes in microbial alpha diversity or gut inflammation.

While the implications of pks⁺ E. coli on CRC development have been previously reported in non-CF poopulations [19,25], when comparing the presence of pks⁺ E. coli between the HC cohorts of these studies, they are markedly different. Determinants of pks⁺ E. coli prevalence are evidently multifactorial, and while pks⁺ E. coli may be a viable pathogenic pathway towards CRC in CF, it could be one of many distinct pathways of carcinogenesis. Further, we identified pks in 4 samples (3 CF, 1 HC) which did not have E. coli, suggesting other, non-E. coli species may have the capacity to express pks, with or without pathogenic effect [26]. The consequences of increased levels of these non-E. coli pks⁺ positive species in colorectal carcinogenesis is not yet understood.

Lastly, our analysis identified no significant difference in the proportion of CF individuals who were pks⁺ E. coli gene positive over time. However, 16 individuals experienced fluctuations in pks⁺ E. coli status across successive timepoints, indicating a high degree of temporal instability. One previous study of an IBD cohort identified greater temporal microbiota instability between timepoints when compared to HC, particularly with changes in disease stability [27]. Disease instability could be applicable to our longitudinal observations in pks⁺ E. coli and explain the lack of diagnostic differences in our cross-sectional analysis. Furthermore, only three patients maintained positive pks⁺ E. coli status across all three timepoints. It remains unclear whether the persistence of pks⁺ E. coli is required for genotoxic effect, or if the cumulative influence of transient, repeat exposures is satisfactory to produce carcinogenesis. Given the high variability we identified in pks⁺ E. coli status, the concept of repeat, transient exposures could be relevant in CF-related CRC.

Limitations

Some limitations of our current study should be considered. While use of the clbB gene in PCR, as a proxy for the pks island, has been validated in previous studies, this gene represents only one of multiple within the pks gene cluster. Metagenomic analysis could also enable quantification of gene expression in future studies and avoid the primer specificity inherent to PCR-based methods.

Further, our study accounted for demographic co-variates such as age, gender, and clinical variables such as use of antibiotics, but other factors remain, such is the therapeutic complexity involved in management of CF. Furthermore, the high-energy, high-fat diet utilised in CF nutrition management influences GI microbial composition [29] and previous literature indicates some diets, such as the Western-style diet, are conductive to higher rates of CRC containing pks⁺ E. coli [30]. Evaluation of the role these variables play on the CF gut has already begun [29], but should further their analysis to analyse a potential role in CRC development.

5. Conclusions

Taken together, our findings suggest that pks⁺ E. coli is unlikely to directly contribute to the increased inflammatory environment and dysbiotic state that develops early in cwCF. While our findings do not support the utility of colibactin as a prognostic indicator for CRC risk in cwCF, future studies are needed within the adult CF population to understand the role pks may play in inflammation and CRC tumorigenesis later in life. Characterisation of pks in older CF patients with polyp formation or diagnosis of CRC would further this understanding. Based on the higher load of bacteria with carcinogenic links within the CF gut, along with the temporal variability we identified in pks⁺ E. coli, future research should also consider both the cumulative impact of multiple genotoxic species as well as effect of recurrent colonisation of these genotoxic species over time.