Background
First proposed in the early 1950s, an infectious aetiology for prostate cancer has since been widely investigated using conventional and serology-based case–control designs and some cohort studies but the evidence from these has been generally weak and inconsistent. A causal association is yet to be established.
Recent support for a role of infection in prostate cancer risk came from the detection of a novel candidate,
Propionibacterium acnes, within prostate cancer tissues [
1,
2]. There is also evidence of association between prostate cancer risk and gene variants of COX-2 [
3], RNASEL [
4] and TLR4 [
5], identified in cases of hereditary prostate cancer, indicating that infection and the host response to infection may be involved in the development of prostate cancer.
Studies that have investigated the role of infectious agents in the aetiology of prostate cancer have adopted single organism targeted approaches or have identified microbial constituents based on amplification of various hypervariable regions of the 16S rRNA gene in concert with traditional cloning and sequencing methods [
6‐
9]. Single organism targeted approaches are limited by their specificity while traditional broad-range 16S rRNA gene amplification, cloning and Sanger sequencing can be laborious and costly, depending on the scale of the study, number and complexity of samples. When compared with conventional sequencing methods, cyclic array-based massively parallel sequencing (MPS) methods, albeit with shorter read length capability and less accuracy in base calling, offer efficiencies in terms of cost, time and scalability.
The principal hypothesis that guided the direction of the work presented in this study was that persistent, rather than transient, infection of the prostate gland by a sexually transmitted or other infectious agent would be associated with risk. Thus, evidence of infection at the tissue level was sought by utilising two different molecular approaches, targeted partial 16S rRNA gene sequencing and total RNA sequencing using MPS. The overall objective of this study was to investigate the presence of infectious agent(s) in histopathologically determined aggressive prostate cancer cases (Gleason score ≥ 8).
Discussion
We used broad-range methods (one targeted and one agnostic) to explore and characterise microbial constituents within the prostate tissue of men with aggressive prostate cancer.
Previous studies have investigated the presence of bacterial, viral and prokaryotic organisms and their association with prostate cancer [
9,
23,
24] using other methodologies including traditional bacterial culture, specific, targeted PCR and bacterial 16S rRNA amplification, traditional cloning and capillary sequencing methods. The advantage of MPS, in this context, is the capacity to sequence the entire genomic/transcriptomic content of samples without
a priori knowledge of specific genes and targets [
25], in addition to its sensitivity and high-throughput capability. However, despite the advantages of applying new technology to a decades-old question, the data generated and the methods used for data analysis were still in early development. As this field evolves, the methods, data, analytical tools and strategies will become more refined and enable further elucidation of these study questions.
To date, five studies [
8,
9,
26‐
28] have investigated and characterised bacterial 16S rRNA sequences in prostate tissue collected from prostate cancer patients. Only one of these studies [
28] found no evidence of 16S rRNA sequences in prostate cancer tissues. Four studies [
8,
9,
26,
27] demonstrated the presence of bacterial sequences in 88.9, 85.7, 19.6 and 87% of patients, respectively. The most common organisms identified in these studies were members of the family
Enterobacteriaceae and specifically species related to
Escherichia coli. These findings are consistent with the results of the present study. In addition, analysis of the 16S rRNA V4 region sequencing data identified
Actinobacter spp.,
Pseudomonas spp. and
Streptococcus spp. as being present in 95% of all prostate samples therefore members of the ‘core’ community, in accordance with Sfanos et al. (2008). Analysis of the V2-V3 region also identified
Enterobacteriaceae,
Escherichia spp. as the predominant taxa within this sample of prostate tissues in addition to
Staphylococcus spp,
Streptococcus spp,
Moraxella spp., and
Propionibacterium acnes as members of the ‘core’ community.
Distinguishing between contamination of tissue and ‘true’ prostatic microbial constituents is one of the main challenges of bacterial community studies. Studies [
8,
27] have suggested that the presence of bacterial sequences in prostate cancer tissues reflects bacterial contamination of the prostate via transrectal prostate biopsy of prostate which is routinely performed to confirm a diagnosis of prostate cancer. This could explain the presence of bacterial 16S rRNA sequences in prostate tissue samples from prostate cancer patients and the range of organisms detected in our dataset also supports this hypothesis.
Catheterization of patients has also been suggested as a way in which the prostate may be contaminated with bacteria. Hochrieter et al. (2000) detected 16S rRNA sequences in all four prostate tissue samples taken from a benign prostatic hyperplasia (BPH) patient that had an indwelling catheter for several weeks before radical prostatectomy [
27]. Gorelick et al. (1988) performed quantitative bacterial culture of prostate tissues from prostatectomy patients to determine the prevalence of prostate bacterial infection or colonization [
29]. They reported that 34% of patients with a pre-operative indwelling catheter returned a positive prostatic culture. Organisms were identified as common urinary tract pathogens including
E. coli and
Streptococcus fecalis. The pre-operative status with respect to catheterization of patients included in this study is unknown, however, it is a possibility that bacterial sequences identified in our samples could have been introduced in this way.
Sequences representing
Propionibacterium acnes were detected in the V2-V3 16S rRNA dataset in 95% of samples albeit at low abundance. This study reports a 95% prevalence of
P. acnes in prostate tissue samples which is consistent with the 100% prevelance of
P. acnes detected in prostatic intraepithelial neoplasia (PIN) lesions and 78% of prostate cancer tissues reported by Fehri et al. (2011) but approximately two-fold higher than the prevalence of
P. acnes reported by other studies [
1,
2,
9,
30]. The present study could not determine whether the
P. acnes sequences detected in the V2-V3 dataset represented either urogenital or cutaneous strains. Therefore, it is difficult to ascertain if the
P. acnes detected in these samples represent contamination through laboratory handling and reagents or if they have biological significance.
The study design and methods employed in this study had several limitations that may have diminished the ability to detect infectious organisms in prostate tissues that were of clinical significance. The study design employed to identify potential infectious agents associated with prostate cancer was limited by study sample collection methods, the sampling of prostate tissue, small sample size and sensitivity of detection (total RNA sequencing). In addition, there were inherent limitations to our study design including the presence of multiple 16S rRNA gene copies, extraction methods, library preparation, experimental controls and bioinformatics approaches.
The 16S rRNA gene occurs in at least one copy of every bacterial genome, however can also occur as multiple and heterogeneous copies with copy number ranging from 1 to 15 [
31]. The
E. coli genome contains seven copies of the 16S rRNA gene and the
P. acnes genome three copies [
32]. Most 16S rRNA gene surveys assume that the relative abundance of 16S sequences are an accurate surrogate measure of the relative abundance of microorganisms in studies of community composition [
31]. However, differences in the copy number/heterogeneity of the target 16S rRNA gene may result in overestimation of diversity and abundance [
33,
34]. Therefore, inferences made on the basis of relative abundance of 16S rRNA genes may not be an accurate representation of actual community composition [
31,
35] and variation in 16S rRNA gene copies can be a source of significant systemic bias within 16S rRNA gene surveys [
33]. This study did not normalize for variation in 16S rRNA copy number and therefore it is unlikely that the reported relative abundances of taxa identified reflected
actual taxa abundance. However, there are software tools [
31] and a publicly available curated database (ribosomal RNA operon copy number database or rrnDB [
35]) that could be applied to estimate actual organism abundance from 16S rRNA gene abundance data in future work.
There is considerable scope to extend and improve upon the experimental design of this study in investigating a persistent infectious aetiology for prostate cancer. Incorporating a prospective study design that collected tissues specifically for PCR- and sequencing-based analyses may reduce the prevalence of contaminating sequences. Inclusion of (a) control group(s) that included samples from lower grade and less aggressive prostate cancer cases and cancer-unaffected prostates such as those from organ donors, cystoprostatectomy and/or BPH cases would allow comparison between the microbial constituents of different prostate pathologies (if any) and normal prostate tissue. In addition, a greater number of cases would ensure that the study is sufficiently powered to detect differences in microbial communities (if any) between groups. Sampling a greater proportion of the prostate gland at several anatomical sites would provide comprehensive coverage of the prostate gland as a whole. With regard to 16S rRNA amplicon sequencing, the inclusion of extraction, PCR and water controls in sequencing runs would also provide a profile of laboratory contaminants so that ‘true’ microbial constituents (if any) could be distinguished from contaminating sequences. Normalization of 16S rRNA datasets to account for heterogeneity of 16S rRNA gene copies would also provide more accuracy with respect to relative organismal abundance. In terms of RNA sequencing, depletion of host RNA and enrichment of microbial rRNA and mRNA may increase detection sensitivity. If microorganisms of interest were detected, follow-up studies including verification of specific infectious agents in original nucleic acid samples via PCR and tissue localization studies would be warranted.
Acknowledgements
Data analyses of the V2-V3 and V4 16S rRNA datasets were carried out by Gayle Philip, Life Sciences Computation Centre, Victorian Life Sciences Computation Initiative, University of Melbourne. Data analyses of the total RNA sequencing datasets were carried out by Dieter Bulach, Life Sciences Computation Centre, Victorian Life Sciences Computation Initiative, University of Melbourne.
Amplification primers that targeted the V2-V3 hypervariable region of the 16S rRNA gene were adapted/designed by Josef Wagner (JW), Murdoch Children’s Research Institute.
The authors would like to express their appreciation to the study participants who kindly donated tissue to the Australian Prostate Cancer BioResource. The Australian Prostate Cancer BioResource is supported by the National Health and Medical Research Council of Australia Enabling Grant (no. 614296) and by a grant from the Prostate Cancer Foundation Australia.