Many precautions should be taken to limit the modification of the commensal communities studied and the increase of interindividual variation not attributable to the experimental variables. The following factors can influence the human microbiota and should be considered when designing studies targeting the lung microbiota: the administration of antibiotics or neoadjuvant25,26,27,28, the size of the lesion, the type of surgical procedure, the type of pulmonary pathology under study, and living habits of patients (e.g., smoking status, physical exercise, buccal hygiene, alcohol consumption)29,30,31,32,33,34.
A more exhaustive list of concomitant factors was pointed out by Carney et al.35. However, as the different fields of microbiota studies expand, it is likely that additional variables that can alter its composition will be uncovered. The molecular tools currently used to analyze the human microbiota do not have the power to discriminate the impact of that many factors over the microbial profiles. Whenever possible, patients selected for lung microbiota studies should be extensively screened so that they can be as similar as possible. Longitudinal studies could also minimize the impact of those variables, as the same patient, with similar concomitant factors through the study, would be compared to himself overtime.
Tissue management steps should consider the contamination possibilities. In addition to the selection of a less contamination-prone procedure, such as thoracoscopic lobectomy, the manipulations and the instrument used in subsampling the excised organ should be taken into account. A combination of bleach and humid heat was chosen to sterilize the instruments used to sample the cancerous and healthy tissue as it was considered the most easily accessible method. The use of humid heat itself (autoclave) lacks the power to completely neutralize bacterial genomic DNA in solutions and on surfaces36. On the other hand, the utilization of bleach, or a chlorinated detergent, leads to the complete degradation of contaminating DNA on surfaces, such as benches and instruments37,38, but requires rinsing to avoid corrosion. Hence, combining both methods, soaking the instruments in bleach 1.6% for 10 min before rinsing with distilled water and autoclaving in a sterilization pouches, ensures a minimal amount of DNA has to be degraded by moist heat. The rest of the single-use equipment used was commercially sterilized with ionizing radiation.
Healthy lung tissue was subsampled from the pulmonary lobe containing the tumor to ensure that the developed method could be used on a variety of lung tissue samples. It could also act as a control of non-pathologic microbiota to allow comparisons of cancerous and non-cancerous samples within the same subject, hence minimizing the impact on inter-individual microbiota variations. In fact, Riquelme et al. found that the gut microbiota has the capacity to specifically colonize pancreatic tissue8. Correspondingly, the use of adjacent pulmonary tissue to the tumor could help get better insights at a specific colonization of the tumor by lung bacteria. A 5 cm distance between the tumor and the healthy sample was ensured to minimize the potential effect of increased inflammation surrounding the tumor. Furthermore, the lung microbiota composition seems to vary dependently on the position and depth of the respiratory tract, even inside a same lobe39. The healthy tissue was collected in the same tierce of pulmonary depth (Supplementary Fig. 4) in an attempt to sample a microbial community that it would be as representative of non-pathologic microbiota in the tumoral region as possible.
The homogenization of frozen and thawed pulmonary tissues was attempted and was unsuccessful, both with the use of only a 2.8 mm tungsten bead in the Retsch – MM301 mixer mill (30 beats/s, 20 min) or of the Fisherbrand 150 homogenizer with plastic probes (Fisher scientific, Pittsburg). The elasticity of the tissue or its frozen state make the mass nearly unbreakable. The use of the Liberase™ TM enzymatic cocktail (collagenase I & II, thermolysin) prior to the mechanical homogenization proved successful and a homogeneous suspension was obtained using the two-step homogenization protocol (Supplementary Fig. 3). Multiple ratios of liquid to mass of tissue were tested and 3 mL/g was found optimal, as it facilitates the homogenization without overly diluting to sample. A similar ratio of liquid to tissue was used in breast tissue microbiota study40. The samples were first thawed at 4 °C to reduce potential growth or degradation of microorganisms. The digestion was performed directly in the 50 mL collection tube to limit the tissue manipulation and ensure possible contaminant tracking.
Our team was also unable to replicate the results obtain by Yu et al. on larger tissue samples using 0.2 mg/mL of Proteinase K for 24 h13. The samples remained firm and turned brown. Using the Liberase™ cocktail enabled a much faster digestion (75 min) and broke down specifically the lung component responsible for its elasticity, the collagen.
Three commercially available DNA extraction kits were tested. They were selected for their previous successful use in the study of pulmonary or gut microbiota and their intended application as described by the manufacturer. The extraction kits were first tested on homogenized lung tissue spiked with whole-cell bacterial community to assess the efficiency of DNA extraction and recuperation of the commercial kits. The three kits were able to recover more than 88% of the genera added to the samples. All the genera that were not detected by the Microbial and Powersoil (Cutibacterium acnes, Bacteroides vulgatus, Bifidobacterium adolescentis, D. radiodurans, Clostridium beijerinckii, L. gasseri), with the exception of H. pylori, were Gram-positive bacteria. This type of bacteria has been reported to require more aggressive extraction methods to break their tougher cell walls19. However, the bacterial community did not go through the enzymatic and physical homogenization that usually takes place before DNA extraction since we needed to obtain a homogenous tissue sample that could be processed with or without spiked bacteria. These hard to lyse Gram-positive bacteria could have been fragilized by these processes, rendering them easier to break down during the extraction protocol. Furthermore, the detection of the artificially incorporated bacteria does not account for the natural physical association that may occur between the human tissue and microbial cells. Nonetheless, these high percentages of recovery were promising and lead us to continue with the characterization of the extraction kits in a real-life context, meaning the analysis of the base-level microbiota in pulmonary samples collected and processed through the entire pipeline.
Every measurement of the efficiency of extraction, including DNA yield (Supplementary Fig. 5), DNA purity (Supplementary Figs. 6 and 7), and alpha diversity (Fig. 1), pointed in the same direction. In fact, they all showed that the Blood extraction kit was the best option out of the three kits. Therefore, using the Blood kit is recommended as one of the pieces of a complete study design. Additionally, the presence of a high concentration of host DNA in tissue samples might tend to saturate the purification column, which could reduce to amount of bacterial DNA recovered. The superior DNA binding capacity of the affinity column of the Blood kit compared to the two others could explain its better performance and its higher yield in most cases. The samples extracted with the Blood kit were also associated with higher alpha diversity (Shannon index). Therefore, this extraction method was able to recover a higher number of different bacterial organisms (richness) and proportionality in the different OTUs (evenness). The absence of PCR inhibitors and a higher recuperation rate of bacterial DNA in the Blood extracted samples could have led to a more proper amplification in the sequencing process and to the recuperation of very low abundance bacterial DNA in the extraction eluate. For further research, it is advised to take the additional precaution of working under a biosafety cabinet or in the sterile field when analyzing the microbiota of lung tissues to reduce the risk of incorporation of airborne contaminants.
The Illumina Miseq sequencing platform with the use of dual-index strategy has become the dominant technology used in microbial ecology studies for its cost efficiency, low error rate, and user-friendliness41,42,43. Most studies interested in the pulmonary microbiota have also used this technology11,13,14. The sequencing of the 16S rRNA gene amplicon was favored over a shotgun sequencing method because of the overwhelming quantity of human DNA joining bacterial genomes in the pulmonary tissue. The 16S rRNA gene is the most used marker of bacterial identification. No consensus has been reached on the selection of the 16S rRNA gene variable region (V) to sequence for human microbiota18,44. However, it should be kept consistent across studies to allow comparisons. Targeting the V3–V4 regions was suggested using the universal primers developed by Klindworth et al.45. Several microbiota studies, including lung microbiota, have also used these regions7,13,46,47,48.
In the context of this study, genomic mock-community was spiked in DNA extracted from the pulmonary tissue at a biological meaningful concentration. Every genus added to the samples was successfully detected. Consequently, the high ratio of human DNA to bacterial DNA did not interfere with the amplification and detection steps of the sequencing procedure. The sequencing method in place seems adequate for its application in the characterization of pulmonary microbiota.
Contaminating bacteria or DNA can have an important impact of the microbial profile observed in very low biomass samples such as pulmonary tissue23. Consequentially, in addition to proper protocol selection, methodological design that attempts to follow, detect, and account for contamination was proposed. Its main features include the incorporation of a single negative control that monitors the incorporation of contaminants at every step of the experimental method (Supplementary Fig. 3). Since every step of the protocol prior to the extraction is meant to be executed in a single tube and only by the addition of reagents, it is possible to carry and detect the contaminants introduced throughout the procedure. On the contrary, microbiota study methodologies usually dictate for the incorporation of multiple controls at every step of the procedure (e.g. DNA extraction kit, PCR controls, etc.)18. Although more informative as to which step leads to contamination, it makes data analysis harder since the presence of contamination in the multiple controls cannot by added.
No bioinformatics standard operating procedure is available and what should be done with controls sequencing data is still under debate18. Some research groups tried to use a neutral community model49, additional qPCR data50, amplicon DNA yield, or prevalence algorithms51 to assess the influence of methodological contaminants. The removal of every bacterial OTU found in controls from the samples is often not appropriate as these OTUs might also be naturally present in the samples22. We propose using relative abundance ratio between samples and controls to remove contaminating OTUs. Since controls have much lower richness than extracted lung samples and that the total number of reads (sequencing depth) is distributed across every OTU, the relative abundance of reads for each OTU tend to be much higher in the control than the same OTU in samples. Therefore, if the relative abundance of an OTU is greatly superior in the sample than in the control, it is reasonable to think that the same OTU was also in the sample in a substantial quantity. To ensure that OTUs that were present in very low absolute abundance (e.g., from only 1–2 reads) do not lead to the removal of the highly abundant corresponding OTU in samples, only the OTUs with a ratio of 1000 (relative abundance of sample/relative abundance of sample) were kept. The rest of the OTUs found in controls were completely removed from the related samples, since the influence of contaminating DNA could not be differentiated from the pulmonary microbiota. This method would theoretically tolerate no more than 20 reads (0.1%) before removing the entire OTU from the sample if only one OTU was present in the samples (20,000 reads, 100%). The use of relative abundance helps reduce the absolute abundance bias induced by the divergence in sequencing depth. The OTUs were removed from both tissues at the same time or not at all to avoid adding artificial intraindividual variation. The authors acknowledge that the proposed contaminant management method does not have the in-dept validation of other methods, such as described by Davis et al. with the decontam package51. However, it does not share its limitations regarding the lack of consideration for OTU abundance and need of high number of controls to ensure sensitivity while using prevalence-based detection. Further research focused on the development of statistical methods to detect contaminant OTUs in the cases of lung microbiota is needed. This work is to be a starting point toward methodological standardization and its modular nature makes the bioinformatic contaminant management method proposed here interchangeable once a more robust one is uncovered.
Pearson’s correlation tests were performed on the number of reads per OTU between the samples and their respective controls. Although these values were not normally distributed (Shapiro-Wilk, p < 0.05) and were zero inflated, the Pearson’s test was still used, as it was found serviceable in these conditions by Huson et al.52 and did not have an appropriate alternative. The correlation of the sample extracted with the Blood and Microbial kits to their controls was significantly lower than the one extracted with the Powersoil (p = 0.027, 0.0039), which could indicate a higher level of contaminants in this last kit. The outperforming results of the Blood kit in the terms of DNA yield, purity, and alpha diversity corroborate its low correlation of its controls. In fact, the more bacterial DNA recovered from the sample, the lower the impact of contaminant present on the bacterial profile.
The extraction kit had a significant impact on the composition of the bacterial profiles found in the methodological controls, taking into account reads abundance (p = 0.019, R2 = 0.14374) or not (p = 0.0184, R2 = 0.09582). It is therefore a major contributor in the incorporation of contaminants in the samples. Ten genera were identified as “core” in the controls of every variation of the pipeline. Other steps of the protocol, such as the enzymatic homogenization and the sequencing method, were shared by all the pipelines and could have led to these similarities. The experimental design does not allow us to identify the origin of the contaminants.
The method described here shows some limitations. For instance, it might not be suitable for the use of culture method for the characterization and identification of microorganisms. The presence of thermolysin in the enzymatic homogenization cocktail could inadvertently reduce the recuperation rate of live organisms. Furthermore, the use of a 16S rRNA gene amplicon sequencing approach rules out the possibility of identification of fungal and viral microorganisms. The modification of the sequencing techniques could allow more versatility in targeted organisms.
Although great care was put in reducing the incorporation of bacterial contaminants in lung tissue samples, the methodological biases were not specifically measured. This preferential recovering and detection of some bacterial member over others is still of concern. The resources available to correct these biases are still very scarce, but McLaren et al. offer great evidence of its importance and concrete attempts53.
This method allows the identification of the bacterial members of the lung community, but not their functionalities. Therefore, it could be interesting to develop a metatranscriptomics methodology for the pulmonary environment to get better insight at the transcribed microbial genes, as performed in the gut microbiota field of research54.
Here, a comprehensive methodological pipeline for the study of lung cancer microbiota is proposed (Supplementary Fig. 3). A protocol of tissue collection and sample treatment that minimizes the risk of contamination was validated, a DNA extraction technique was adapted, and a bioinformatics pipeline to account for detected contaminants was designed. It is a first step toward protocol standardization. Although the perfect method does not exist, taking the appropriate precautions and being aware of its limitations ensure that appropriate conclusions are drawn and that the pulmonary microbiota field grows in a conscientious and reliable way.
Source: Ecology - nature.com
