Effect of radiotherapy on the gut microbiome in pediatric cancer patients: a pilot study

Study design and participants’ demography

Ten fecal samples were collected from three cancer patients and two healthy controls. The study included two sets of controls: self-control (same patients before starting radiotherapy sessions) and healthy individuals. Healthy is defined as not consuming any chemotherapeutic drugs or antibiotics (antibiotic-free period was defined by at least one month before sample collection) and never being exposed to radiotherapy before. All the participants in the study (patients and healthy individuals) were males with age range between 3.5 and 7 years old.

Participants in the current study were treated in Children Cancer Hospital, Egypt (CCHE 57357) and were diagnosed with rhabdomyosarcoma in the pelvic region. The treatment protocol entailed 50.4 Gy (180 cGy/fraction) that is the highest dose of radiation prescribed to pediatric cancer patients. The sample collection points were defined as follows: pre-point, before starting radiotherapy sessions; mid-point, ranging from day 12 to day 16; and the last collection point, day 26–28. All patients were treated with a complex treatment protocol including chemotherapy (Cyclophosphamide, Vincristine and Dactinomycin) and radiotherapy dose of 50.4 Gy on 28 fractions (180 cGy/fraction). The patients started the radiotherapy sessions either after 12 weeks of chemotherapy (Patient-1 and patient-3) or after four weeks of chemotherapy (patient-2). During the chemotherapy, a set of antibiotics were prescribed according to the patients’ case and needs (Table 1).

The institutional review board approvals were granted before sample collection, from the American University in Cairo and Children Cancer Hospital Egypt (CCHE 57357). The participants have signed a child assent form and the guardians have approved the participation in study and signed a parental permission form. Fecal samples were collected from the participants with the help of their parents in sterile falcon tubes, then transferred on ice before storing in −80 °C freezer until DNA extraction.

DNA extraction

Microbial DNA was extracted from the collected fecal samples by DNA extraction kit from stool, QIAamp DNA stool Mini Kit (Qiagen, USA), according to the manufacturer’s instruction. The protocol used enhances the non-human DNA over the human DNA extracted from the sample, through optimization of the lysis conditions as described in the manufacturer’s protocol. The DNA was eluted in 50 µl elution buffer provided by the kit and stored in −20 °C until sequencing.

16S rRNA sequencing

The extracted DNA was sent to Eurofins Genomics in Germany for sequencing of 16S rRNA. The sequence was performed on Illumina MiSeq platform, targeting V3–V5 variable regions of the 16S rRNA. The run was performed on 2 × 300 paired-end reads. The target region (V3–V5) length is approximately 700 bp. The obtained raw data are available on NCBI BioProject number: PRJNA545788.

Data analysis

The 16S rRNA sequencing data received from Eurofins Genomics were already demultiplexed. The analysis was completed on QIIME 2 pipeline (version 2017.12) (Bolyen et al., 2018), q2cli command line interface. The denoise command was used followed by length trimming to 230 bp. Feature table was constructed on the same program using Deblur approach. The Deblur approach uses an error profile that operates on per-sample bases and depends on the read length and diversity in amplicon sequences. Thus, it offers a higher sensitivity and requires lower computational power compared to other OTU clustering algorithms (Amir et al., 2017). Taxonomy classes were assigned using the constructed feature table in comparison with SILVA database (Silva-119 99% OTUs full-length sequences). Unrooted phylogenetic tree was also constructed on QIIME 2 (qiime phylogeny fasttree). The taxonomic classification was appended to the feature table and exported as a biom format file. The phylogenetic tree was exported as Newick tree format and the sequences corresponding to the classified OTUs were exported into a fasta file. All exported files from Qiime2 analysis were imported to phyloseq package (McMurdie & Holmes, 2013) on R-CRAN for figure plotting and further analysis. It is important to mention that reads were not rarefied to an even sampling depth (McMurdie & Holmes, 2014). Shannon Entropy was calculated from Fasta sequences of Streptococcus and Escherichia-Shigella bacterial genera using Oligotyping software (Eren et al., 2014). The threshold was selected as 0.2 (Eren et al., 2014) above which the variation in considered as mutation rather than sequencing error.

Sample nature and reads quality

High-throughput sequencing generated a sum of 904,685 reads in both directions (mean reads per sample per direction = 42,407.9 and median reads per sample per direction = 41,645.5). The average read length obtained (Forwards reads ∼280 bp, Reverse reads ∼250 bp). All reads were trimmed to 230 bp after denoising on QIIME 2 (qiime deblur denoise-16S).

Alpha diversity analysis

Different indices of alpha diversity measures (Observed OTU, Chao1, Shannon, Simpson, InvSimpson, and Fisher) all showed a higher diversity in healthy controls compared to cancer patients in the three-time points collected (Fig. 1). After completing the radiotherapy treatment along with the antibiotics courses (Last), the alpha diversity has generally declined when compared to the mid-point after 12–15 fractions and before radiation. The outliers at each time point were obvious, which could be attributed to the personal variations between the patients, along with the variation due to the set of antibiotics prescribed (antibiotic course duration and antibiotic course timing relative to the radiotherapy and sample collection time).

Alpha diversity per sample

Alpha diversity measures (Chao1 and Shannon) per sample did not indicate a direct relationship between exposure to radiation, intensive use of antibiotics, and abundance of all bacterial species per sample (Fig. 2). Patients-1 and 2 experienced an increase in the alpha diversity after the two aforementioned exposures, unlike patient-3 who experienced a massive drop in alpha diversity. On the other hand, a pattern was inferred between high alpha diversity and response to treatment (Table 2). The low bacterial abundance was associated with a positive response to radiotherapy and vice versa. It is important to mention that the Chao1 index reflect the richness only (number of bacterial species per sample), while Shannon index reflects both richness and evenness (relative abundance of species that make up the richness).

Variation in the bacterial abundance at different taxonomic levels

At the phylum level, the two healthy controls showed normal variation between the four most abundant bacterial phyla (Firmicutes, Bacteroidetes, Proteobacteria and Actinobacteria) (Fig. 3), with domination of Firmicutes and relatively high abundance of Proteobacteria (Odamaki et al., 2016).

After the exposure to different doses of radiation that ranged between (21.6 Gy and 50.4 Gy), along with the antibiotic courses (Table 1), the relative abundance of Firmicutes decreased while the Proteobacteria increased in the three patients. The phylum abundance (Fig. 3) showed disturbance in microbial phyla when compared to controls. However, when comparing each patient to himself (at the different time points collected), it was found that Actinobacteria, Bacteroidetes and Proteobacteria phyla increased after antibiotics and radiation while Firmicutes decreased, which may explain the increase in alpha diversity (Fig. 2).

A higher resolution or an intuition of more specificity to the bacterial taxa was needed. The frequency table at the genus level was obtained for all samples. The search criteria for the specific increasing genera was set as follows: (1) the frequency per patient was higher at each time point (fluctuating taxa were excluded), (2) frequencies per taxa were elevated in the three patients or completely absent in one patient. According to these criteria, eight different genera were identified (Fig. 4), six of which belong to the Firmicutes phylum (that had overall decreased). One belongs to Proteobacteria and the last is a member of Bacteroidetes phylum (Table S1).

Variation at different nucleotide positions before and after radiotherapy

To determine the effect of exposure to intensive radiation on the variation in bacterial DNA sequences, Shannon entropy was determined. The reads used in the analysis were trimmed to 200 bases to exclude low quality base calling. From the bacterial genera identified (Table S1), two were randomly selected to determine Shannon entropy at a single nucleotide position (comparing before radiation to after radiation at the same nucleotide position). The entitled analysis was performed on Streptococcus and Escherichia-Shigella. Shannon entropy increased (Figs. 5 and 6) at definite nucleotide positions in both genera and post exposure to radiation. Values below 0.2 were considered as sequencing errors and accordingly excluded. Escherichia-Shigella, gap insertions were clear (Fig. 5) to reach the best sequence alignment, which indicates a high probability of single point mutations in the bacterial DNA post radiotherapy. It is important to note that the Shannon entropy was performed only on a definite sequence length (approx. 200 bp from V3–V4 region of the 16S rRNA of the bacterial genome).

Beta diversity analysis

To determine the effect of increasing the exposure of radiation on the bacterial diversity, beta diversity analysis was performed based on Bray-Curtis dissimilarity metric. Samples were grouped according to the stage of sample collection (Fig. 7). The bacterial diversity appears to decline with the stage (by increasing the exposure to radiation). The control samples share a high extent of similarity in terms of the bacterial taxa presence in each sample, while the pre samples had the highest recorded level of variation among the bacterial taxa per sample. Finally, for the mid and last sample groups, both the beta diversity, across the groups and across the samples in the same group, decreased as the exposure to radiation increased.