Characterization and Functional Prediction of Bacteria in Ovarian Tissues

Summary

Immunohistochemistry staining and 16S ribosomal RNA gene (16S rRNA gene) sequencing were performed in order to discover and distinguish bacteria in cancerous and noncancerous ovarian tissues in situ. The compositional and functional differences of the bacteria were predicted by using BugBase and Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt).

Abstract

The theory of a "sterile" female upper reproductive tract has been encountering increasing opposition due to advancements in bacterial detection. However, whether ovaries contain bacteria has not yet been confirmed yet. Herein, an experiment to detect bacteria in ovarian tissues was introduced. We chose ovarian cancer patients in the cancer group and noncancerous patients in the control group. 16S rRNA gene sequencing was used to differentiate bacteria in ovarian tissues from the cancer and control groups. Furthermore, we predicted the functional composition of the identified bacteria by using BugBase and PICRUSt. This method can also be used in other viscera and tissues since many organs have been proven to harbor bacteria in recent years. The presence of bacteria in viscera and tissues may help scientists evaluate cancerous and normal tissues and may be aid in the treatment of cancer.

Introduction

Recently, an increasing number of articles have been published that prove the existence of bacteria in abdominal solid viscera, such as the kidney, spleen, liver, and ovary^1,2. Geller et al. found bacteria in pancreatic tumors, and these bacteria were resistant to gemcitabine, a chemotherapeutic drug². S. Manfredo Vieira et al. concluded that Enterococcus gallinarum was portable to the lymph nodes, liver and spleen, and it could drive autoimmunity³.

Since the cervix plays a role as a defender, bacteria in the upper female reproductive tract, which contains the uterus, fallopian tubes, and ovaries, have been minimally researched. However, some new theories have been established in recent years. Bacteria may have access to the uterine cavity during the menstrual cycle due to changes in mucins^4,5. Additionally, Zervomanolakis et al. confirmed that the uterus, together with the fallopian tubes, is a peristaltic pump controlled by the endocrine system of the ovaries, and this arrangement enables bacteria to enter the endometrium, fallopian tubes, and ovaries⁶.

The upper reproductive tract is no longer a mystery anymore thanks to the development of bacterial detection methods. Verstraelen et al. used a barcoded paired-end sequencing method to discover uterine bacteria by targeting at the V1-2 hypervariable region of the 16S RNA gene⁷. Fang et al. employed barcoded sequencing in patients with endometrial polyps and revealed the presence of diverse intrauterine bacteria⁸. Additionally, by using the 16S RNA gene, Miles et al. and Chen et al. found bacteria in the genital system of women who had undergone salpingo-oophorectomy and hysterectomy, respectively^5,9.

Bacteria in tumor tissues have gained increasing attention in recent years. Banerjee et al. discovered that the microbiome signature differed between ovarian cancer patients and controls¹⁰. Anoxynatronum sibiricum was associated with tumor stage, and Methanosarcina vacuolata might be used to diagnose ovarian cancer¹¹. In addition to ovarian cancer, other cancers, such as stomach, lung, prostate, breast, cervix, and endometrium, have been proven to be associated with bacteria^{12,13,14,15,16,17,18}. Poore et al. proposed a new class of microbial-based oncology diagnostics, foreseeing early-stage cancer screening¹⁹. In this protocol, we investigated the differences between cancerous and normal ovarian tissues by comparing the composition and function of bacteria in these two tissues.

Immunohistochemistry staining and 16S rRNA gene sequencing were performed to confirm the presence of bacteria in the ovaries. The differences and predicted functions of the ovarian bacteria in cancerous and noncancerous ovarian tissues were studied. The results showed the existence of bacteria in ovarian tissues. Anoxynatronum sibiricum and Methanosarcina vacuolata were related to the stage and the diagnosis of ovarian cancer, respectively. Forty-six significantly different KEGG pathways that were present in both groups were compared.

Protocol

This study was approved by the Medical Institutional Ethics Committee of the First Affiliated Hospital of Xi'an Jiaotong University (No. XJTUIAF2018LSK-139). Informed consent was obtained from all enrolled patients.

1. Criteria for entering the cancer group and the control group

1. For the cancer group, enroll patients who are primarily diagnosed with ovarian cancer, and after laparotomy, they are proven to have serous ovarian cancer by pathological findings.
2. For the control group, enroll patients that are primarily diagnosed with uterine myoma or uterine adenomyosis, without presenting any ovarian condition, and who have undergone hysterectomy and salpingo-oophorectomy.
   NOTE: This standard is not definite. Patients with diseases not affecting the ovaries who undergo hysterectomy and salpingo-oophorectomy can also be enrolled.
3. Exclude patients with one or more of the following criteria:
   Pregnant or breast-feeding women.
   Taking antibiotics 2 months prior to the surgery.
   Having fever or elevated inflammatory markers.
   Having inflammation of any kind.
   Having undergone neoadjuvant chemotherapy.

2. Gather samples

1. During the surgery, place the resected ovaries into a sterile tube and place the tube in liquid nitrogen for transport. Avoid touching anything else throughout the whole procedure.
2. Separate the ovaries into approximately 1-cm thick tissue samples with a pair of new sterile tweezers under a laminar flow cabinet. After separation, preserve samples at -80 °C.
   ​NOTE: All the procedures for gathering samples are aseptic, including separating the ovaries.

3. Sequence the 16S rRNA gene

1. Extract DNA.
   1. Add 1.2 mL of inhibit EX buffer into a 2 mL centrifuge tube. Then, add 180-220 mg of samples into the tube. Let the sample fully mix (70 °C water bath for 5 min and then vortex for 15 s).
   2. Centrifuge the tube for 1 min at 600 x g.
   3. Place 550 µL of the supernatant into a new 1.5 mL tube, and centrifuge for 1 min at 600 x g.
   4. Transfer 400 µL of the supernatant with 30 µL of proteinase K into another 1.5 mL tube.
   5. Add 400 µL of buffer AL and use a vortex mixer for 15 s.
   6. Incubate at 70 °C for 10 min.
   7. Add 400 µL of 96-100% alcohol. Use a vortex mixer for 15 s.
   8. Transfer 600 µL of mixture into an absorption column and centrifuge for 1 min at 13700 g. Exchange the lower tube. Repeat this step 11 times.
   9. Add 500 µL of buffer AW1, centrifuge for 1 min at 13,700 x g, and change the lower tube.
   10. Add 500 µL of buffer AW2, centrifuge for 3 min at 13,700 x g, and change the lower tube.
   11. Centrifuge for 3 min at 13,700 x g.
   12. Transfer the mixture into a new 1.5 mL tube, add 200 µL of buffer ATE, incubate at room temperature for 5 min and centrifuge for 1 min at 13,700 x g.
2. Quality testing. Use 1% Sepharose gel electrophoresis to test the quality. Add 400 ng of sample, 120 V, 30 mins. Ideal result: DNA concentration: ≥ 10 ng/µL, DNA purity: A260/A280 = 1.8-2.0, gross DNA: ≥ 300 ng.
3. Prepare the libraries using a 16S metagenomic sequencing kit according to the manufacturer's protocol.
   1. Perform PCR. Briefly, each 25 µL PCR reaction contains 12.5 ng of sample DNA as input, 12.5 µL of 2x KAPA HiFi HotStart ReadyMix and 5 µL of each primer at 1 µM.
   2. Carry out PCR using the following protocol: an initial denaturation step performed at 95°C for 3 min followed by 25 cycles of denaturation (95°C, 30 s), annealing (55°C, 30 s) and extension (72°C, 30 s), and a final elongation of 5 min at 72°C.
   3. Clean up the PCR product from the reaction mix with magnetic beads using the manufacturer's instructions.
   4. Repeat steps 3.3.1 and 3.3.2.
   5. Quality testing. Please refer to step 3.2.
   6. Repeat step 3.3.3.
   7. Quality testing. Use 1% Sepharose gel electrophoresis to test impurity, a spectrophotometer to test purity, a fluorometer to test the concentration, and an RNA assay kit to test integrity. Follow the manufacturer's protocol. Normalize and pool the libraries; then sequence (2 x 300 bp paired-end read setting) using 600 cycle V3 standard flow cells, producing approximately 100,000 paired-end 2 x 300 base reads.
      ​NOTE: The full-length primer sequences: 16S Amplicon polymerase chain reaction (PCR) Forward primer: 5' TCGTCGGCAGCGTCAGATGTGTATAAGA GACAG-[CCTACGGGNGGCWGCAG] and 16S Amplicon PCR Reverse primer: 5' GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAG-[GACTACHVGGGTATCTAATCC].

4. Analyze 16S rRNA gene sequencing data

1. Filter the raw reads of every sample based on sequencing quality with the software package QIIME 2-201802²⁰.
   1. Copy three files into the directory: emp-paired-end-sequences_01
      one forward.fastq.gz file that contains the forward sequence reads,
      one reverse.fastq.gz file that contains the reverse sequence reads,
      one barcodes.fastq.gz file that contains the associated barcode reads
   2. Execute
      qiime tools import \
      --type EMPPairedEndSequences \
      --input-path emp-paired-end-sequences_01 \--output-path emp-paired-end-sequences_02.qza
2. Remove the primer and adaptor sequences.
   qiime cutadapt trim-paired \
   --i-demultiplexed-sequences demultiplexed-seqs_02.qza \
   --p-front-f GCTACGGGGGG \
   --p-front-r GCTACGGGGGG \
   --p-error-rate 0 \
   --quality-cutoff 25 \
   --o-trimmed-sequences trimmed-seqs_03.qza \
   --​verbose
3. Shorten sequence reads in which both paired-end qualities are lower than 25. See above --quality-cutoff 25
4. Analyze the sequencing data.
   1. Gather sequences to form operational taxonomic units (OTUs) with a similarity cutoff at 97%.
      qiime vsearch dereplicate-sequences \
      --i-sequences trimmed-seqs_03.qza \
      --o-dereplicated-table table_04.qza \
      --o-dereplicated-sequences rep-seqs_04.qza
      ​
      qiime vsearch cluster-features-closed-reference \
      --i-table table_04.qza \
      --i-sequences rep-seqs_04.qza \
      --i-reference-sequences 97_otus.qza \
      --p-perc-identity 0.97 \
      --o-clustered-table table-cr-97.qza \
      --o-clustered-sequences rep-seqs-cr-97.qza \
      --o-unmatched-sequences unmatched-cr-97.qza
   2. For the OTUs, calculate the relative abundance in each sample. Abundance information is in table-cr-97.qza
5. Employ a native Bayesian classifier, which aims at the RDP training set (version 9; http://sourceforge.net/projects/rdp-classifier/), to sort all of the sequences. Mapped taxon information is in table-cr-97.qza
6. Within the given OTU, assign a classification that reflects the major coherence of the sequences to OTUs. Then, align the OTUs. See table-cr-97.qza and rep-seqs-cr-97.qza
7. Based on the sample group information, perform alpha diversity (including the Chao 1, ACE, Shannon, Simpson and Evenness indexes) and the UniFrac-based principal coordinates analysis (PCoA).
   qiime tools export \
   --input-path table-cr-97.qza \
   --output-path exported-feature-table
   exported-feature-table
   
   qiime diversity alpha \
   --i-table table-cr-97.qza \
   --p-metric observed_otus \
   --o-alpha-diversity observed_otus_vector.qza
   
   qiime diversity beta \
   --i-table table-cr-97.qza \
   --p-metric braycurtis \
   --o-distance-matrix unweighted_unifrac_distance_matrix.qza

5. Predict bacterial function

1. To predict the related representation of the characteristics of the bacteria, use BugBase²¹. The input OTU table for BugBase is prepared using the following commands.
   biom convert -i otu_table.biom -o otu_table.txt --to-tsv
   biom convert -i otu_table.txt -o otu_table_json.biom --table-type="OTU table" --to-json
   NOTE: The prediction is based on six phenotype categories (Ward et al. unpublished) (https://bugbase.cs.umn.edu/): Gram staining, oxygen tolerance, ability to form biofilms, mobile element content, pathogenicity, and oxidative stress tolerance.
2. Predict the functional composition of a metagenome by PICRUSt with the usage of marker gene data and a database containing reference genomes²².
   make_otu_table.py -i microbiome_97/uclust_ref_picked_otus/test_paired_otus.txt -t /mnt/nas_bioinfo/ref/qiime2_ref/97_otu_taxonomy.txt -o otu_table.biom &
   normalize_by_copy_number.py -i otu_table.biom -o normalized_otus.biom
   predict_metagenomes.py -i normalized_otus.biom -o metagenome_predictions.biom
   categorize_by_function.py -i metagenome_predictions.biom -c "KEGG_Pathways" -l 1 -o picrust_L1.biom
   categorize_by_function.py -f -i metagenome_predictions.biom -c KEGG_Pathways -l 1 -o metagenome_predictions.L1.txt
3. Analyze the differences in functions among each group with the help of STAMP^23,24. Please refer to the citations to operate the software.

6. Data

1. Use statistical software to calculate the significance of the findings. The indication of statistical significance should be set as P < 0.05.
2. Assess differences in age and parity by Student's t-test. Assess differences in menopausal status, history of hypertension and diabetes by the chi-square test. Assess differences in the number of ovarian bacterial taxa by the Mann-Whitney U test.

Representative Results

Patients
A total of 16 qualified patients were included in the study. The control group included 10 women with a diagnosis of benign uterine tumor (among them, 3 patients were diagnosed with uterine myoma, and 7 patients were diagnosed with uterine adenomyosis). Meanwhile, the cancer group contained 6 women with a diagnosis of serous ovarian cancer (among them, 2 patients were diagnosed with stage II, and 2 of them were diagnosed with stage III). The following characteristics showed no differences between patients in the control group and the cancer group: age, menopausal status, parity, history of hypertension, and history of diabetes (Table 1).

Table 1: Patient statistics Please click here to download this Table.

The richness and variety of ovarian bacterial species in both groups
The presence of bacteria using immunohistochemistry staining was shown in Figure 1. The alpha diversity of the microbes was analyzed as a method to detect the richness and variety of ovarian bacterial species. The number of species observed in the ovarian cancer tissues was smaller than that of the control group, with no significant difference. The richness (represented by the Chao 1 and ACE index) and the diversity (represented by the Shannon, Simpson, and Evenness index) of the bacterial species were both not significantly different between the cancer group and the control group (Figure 2).

Figure 2: 16S rRNA gene sequencing shows differences between the cancer and control groups in bacterial richness and diversity. (A) Observed species index (P = 0.06, Mann-Whitney U test); (B) Chao 1 index (P = 0.06, Mann-Whitney U test); (C) ACE index (P = 0.06, Mann-Whitney U test); (D) Shannon index (P = 0.32, Mann-Whitney U test; E. Evenness index (P = 0.48, Mann-Whitney U test); (F) Simpson index (P = 0.46, Mann-Whitney U test). We obtained reprint permission from previous publishers. This figure has been modified from Wang et al.¹¹. Please click here to view a larger version of this figure.

Description of ovarian bacteria
Deep sequencing of the V3-V4 16S rRNA gene region was performed on all samples to obtain a better understanding of the ovarian bacteria. The results showed that Proteobacteria was the most abundant phylum (67.10% in the control group and 67.20% in the cancer group), Firmicutes was the second most abundant phylum (23.77% in the control group and 23.82% in the cancer group), and the third most abundant phylum was Bacteroidetes (3.26% in the control group and 3.41% in the cancer group). When analyzing species of the control group, the main composition was consisted of Halobacteroides halobios (14.53%), followed by Gemmata obscuriglobus (11.07%) and Methyloprofundus sedimenti (10.69%). For the cancer group, Gemmata obscuriglobus was the richest in the cluster (13.89%), followed by Halobacteroides halobius (11.99%) and Methyloprofundus sedimenti (11.12%) (Figure 3).

Figure 3: Relative abundance of phyla (> 1%) and of the top 12 species in ovarian samples. (A) The relative abundance of the phyla (> 1%) in the ovaries of the patients in the control group. (B) The relative abundance of the phyla (> 1%) in the ovaries of patients with ovarian cancer. (C) The relative abundances of the 12 most abundant bacterial species in the ovaries of the control patients. (D) The relative abundances of the 12 most abundant bacterial species in the ovaries of ovarian cancer patients. We obtained reprint permission from previous publishers. This figure has been modified from Wang et al.¹¹. Please click here to view a larger version of this figure.

Different compositions of ovarian bacteria between the two groups
A comparison of different bacterial communities was carried out by PCoA using PERMANOVA. The results showed that the bacteria in the control group differed from those in the cancer group, P < 0.05 (Figure 4).

Figure 4: PCoA detects clusters of communities and the relative abundances of Anoxynatronum sibiricum and Methanosarcina vacuolata. (A) Communities were clustered using PCoA. PC1 and PC2 are plotted on the x and y axes. The red block indicates a sample in the ovarian cancer group. The blue circle indicates a sample in the control group. The samples from the ovarian cancer group were separated from other samples in the control group. (B) Communities clustered using PCoA. PC1 and PC2 are plotted on the x and y axes. The red block indicates a sample in the ovarian cancer group. The blue solid circle indicates a sample from a patient with uterine myoma, and the blue hollow circle is equal to a sample of a patient with uterine adenomyosis. (C) The relative abundance of Anoxynatronum sibiricum (control group: n = 10, cancer group: n = 6, P = 0.034, Mann-Whitney U test). (D) The relative abundance of Methanosarcina vacuolata (control group: n = 10, cancer group: n = 6, P = 0.001, Mann-Whitney U test). We obtained reprint permission from previous publishers. This figure has been modified from Wang et al.¹¹. Please click here to view a larger version of this figure.

Ovarian bacterial composition in cancer and control groups from different perspectives
An analysis of the ovarian bacterial composition was performed from different perspectives to further detect the differences in the identified ovarian bacteria. In Table 2, phylum, class, order, family, genus, and species levels were considered, and statistics are provided in the chart. In particular, there was an association between the relative abundance of Anoxynatronum sibiricum and the stage of the tumor, and Methanosarcina vacuolata was a specific sign when diagnosing ovarian cancer (Table 2).

Phenotypic conservation of ovarian bacteria in the two groups based on predicted functions
In the cancer group, the expression of genes related to potentially pathogenic and oxidative stress-tolerant phenotypes was increased compared with that of the control group (Wilcoxon signed-rank test, P =0.02 and P =0.002). No significant difference was found between the ovarian cancer and control groups in the following aspects: the phenotypes of aerobic, anaerobic, facultatively anaerobic, gram-positive, and gram-negative bacteria; mobile elements; and biofilm formation of the ovarian bacteria (Figure 5). Forty-six variant KEGG pathways between the bacteria in ovaries in the cancer and control groups were determined. The ovaries in the cancer group showed 26 increased pathways. Among them, the most highly related pathways were transporters. On the other hand, the bacteria in ovarian cancer tissue showed 20 reduced pathways. The most relevant functions were as follows: secretion system, unknown functions, and two-component system. The rest of the pathways are shown in Figure 6.

Figure 1: LPS immunohistochemical expression in ovaries. (A) Control group (10 x). Scale bars, 200 µm. (B) Control group (40 x). Scale bars, 50 µm. (C) Cancer group (10x). Scale bars, 200 µm. (D) Cancer group (40 x). Scale bars, 50 µm. Arrows point to LPS staining in the ovarian tissue. We obtained reprint permission from previous publishers. This figure has been modified from Wang et al.¹¹. Please click here to view a larger version of this figure.

Figure 5: Predicted metagenomes analyzed by BugBase. The expression of some genes in the cancer group was increased compared with that in the control group. These genes were related to potentially pathogenic (Wilcoxon signed-rank test, P = 0.02) and oxidative stress-tolerant phenotypes of the ovaries. (Wilcoxon signed-rank test, P = 0.002). We obtained reprint permission from previous publishers. This figure has been modified from Wang et al.¹¹. Please click here to view a larger version of this figure.

Figure 6: PICRUSt analysis of different KEGG pathways between the cancer and control groups. We obtained reprint permission from previous publishers. This figure has been modified from Wang et al.¹¹. Please click here to view a larger version of this figure.

Table 2: Richness (represented by the Chao 1 and ACE index) and the diversity (represented by the Shannon, Simpson, and Evenness index) of the bacterial species Please click here to download this Table.

Discussion

Ovarian cancer has a notable influence on women's fertility²⁵. Most ovarian cancer patients are diagnosed at late stages, and the 5-year survival rate is less than 30%¹⁸. Confirmation of bacteria in the abdominal solid viscera, including the liver, pancreas and spleen, has been published. The existence of bacteria in the upper female reproductive tract occurs because the cervix is not enclosed^2,3,4,5. However, whether ovaries, which are abdominal solid viscera, are sterile or not has not yet been determined. Additionally, whether bacteria in the ovaries are related to ovarian cancer is also an important question.

The significant differences in the bacteria that we found were compared between different groups. All of the procedures mentioned above were strictly germfree, including instruments, reagents, equipment, and the operation of the whole protocol. More importantly, we used ovaries from patients with benign uterine disease as the control group to counteract possible contamination. However, in this protocol, contamination cannot be avoided. Thus, since the cancer group and control group were analyzed in the same experimental environment, merely by comparing the differences between these two groups, we could obtain primary evidence about the microbiological origin of ovarian cancer.

The findings of bacteria in ovarian tissue might start a new field investigating the bacteria influencing ovarian cancer. Additionally, the unique presence and composition of bacteria in cancerous ovarian tissues might direct the carcinogenesis of ovarian cancer, and the therapeutic and prognostic targets of bacteria. Among the 46 KEGG pathways, functions related to the biosynthesis of vancomycin group antibiotics drew particular attention. This may provide further treatment options for ovarian cancer.

However, the protocol had some limitations. First, the samples could not be collected from healthy people for ethical reasons. The control group was ovaries of patients with benign uterine disease (including uterine myoma and adenomyosis). Second, the number of samples should be larger. The limited sample size of the study might hamper the accuracy of the results. Third, although the cancer group and the control group were under the same conditions, there were no negative or positive controls. In addition, contamination could not be avoided. To date, the study of the ovarian bacteria in patients with ovarian cancer is still in an early stage. A larger-scale study with more samples is needed.

Materials

Name	Company	Catalog Number	Comments
2200 TapeStation Software	Agilgent United States
AmpliSeq for Illumina Library Prep, Indexes, and Accessories	Illumina
Image-pro plus 7	Media Cybernetics
Leica ASP 300S	Leica Biosystems Division of Leica Microsystems
Leica EG 1150	Leica Biosystems Division of Leica Microsystems
Leica RM2235	Leica Biosystems Division of Leica Microsystems
LPS Core monoclonal antibody, clone WN1 222-5	Hycult Biotech
Mag-Bind RxnPure Plus magnetic beads	Omega Biotek	M1386-00
Mag-Bind Universal Pathogen 96 Kit	Omega Biotek	M4029-01
MiSeq	Illumina	SY-410-1003
Silva database	Max Planck Institute for Marine Microbiology and Jacobs University
the QuantiFluor dsDNA System	Promega	E2670
Trimmomatic	Björn Usadel
ZytoChem Plus (HRP) Anti-Rabbit (DAB) Kit	Zytomed Systems	HRP008DAB-RB