Comparison of two commercial kits for faecal microbiome extraction
Five individuals participated in each stage of the experiment. Their stool was collected at the same conditions and time, vigorously homogenised, and analysed simultaneously. To minimise the variance caused by technical personnel, the same laboratory staff extracted gDNA for all experiments.
For the first validation stage, two available commercial kits were chosen. Both kits, PureLink™ Microbiome DNA Purification Kit (PL) and ZymoBIOMICS™ DNA Miniprep Kit (ZR) were validated for bacterial genomic DNA extraction for microbiome analysis. In addition to the chemical bacterial disruption, microbes were mechanically lysed by bead-beating, as recommended for such studies. Moreover, the PL kit suggests even one more additional heat-lysis step. The application of the multistage lysis approach ensures recovery of microbial DNA from the most complicated samples such as Gram-positive bacteria. Although standard PL protocol suggests gDNA extraction from pelleted stool samples while the ZR kit provides protocol from suspension material, we tested these kits from all possible starting materials, specifically from the pellet (P), suspension mix (M), and supernatant (S). To minimise the possible inhibitory effect of the DNA/RNA Shield™ solution (Zymo Research) for PL extraction protocol, the same amount of cooled sterile phosphate-buffered saline (PBS) was added to the faecal suspension. Samples were prepared similarly for both kits and then treated according to the manufacturer’s recommended procedure. We tested the supernatant remained after centrifugation of the faecal suspension in the DNA/RNA Shield™ solution as it may contain gDNA from dead lysed cells in the PBS-Shield solution, which might have lysed during the sample collection, storage, or inside the human gut. We included the supernatant as one of the starting materials for our analysis to better understand the amount and impact of gDNA from lysed bacteria on microbiome test interpretation.
To assess the quality control of the DNA extracted kits and approve the absence of the bacterial contamination from them, additional negative kit-ome controls were done (Suppl. Figure 1), where sterile PBS taken for sample preparation was used instead of gut microbiome samples. Both kit-ome controls were negative with only 0.0011% or 0.0014% of the total reads for PL and ZR correspondently. Thereby, all negative controls, including no-template library preparation and both kit-omes, performed almost similarly and confirmed cleanliness of the work and low impact of extraction kit impurities on the final sequencing results.
As seen in Figure 2A, the total concentration of gDNA obtained by using the ZR kit was higher than for the PL protocol for both pellet and suspension material. Additionally, using only the ZR kit it was possible to extract a substantial gDNA amount from suspension material. Both kits showed a negligible amount of gDNA extracted from supernatants, suggesting an insignificant proportion of dead bacteria for the current collection method used. Although the amount of DNA from lysed bacteria in the supernatant was insignificant, the extraction of total gDNA from suspension material has its advantages. Suspension material gives a comprehensive overview of the whole microbial composition of the gut, which in some cases, such as long storage at a higher temperature, could be more representative than the pelleted sample. Additionally, the quality of extracted DNA checked by the OD 260/230 ratio (Figure 2B) was higher in the case of the ZR kit. gDNA integrity checked by agarose gel phoresis (Suppl. Figure 2) indicated high-quality extraction by both kits. Altogether, despite the suitability of both kits for gDNA isolation for microbiome analysis, the ZR kit exhibited better quality characteristics of isolated DNAs.
The next step of microbiome analysis was performed by sequencing the V4 region of bacterial 16S rRNA on iSeq 100 platform (Illumina). For suspended or pelleted material, the amplicon library was prepared with 12.5 ng of total extracted gDNA, which is different for supernatant gDNA, which concentration was much lower. As the amplicon sequencing technology assumes PCR amplification stages that often increase the impact and ratio of low-represented bacterial species, special attention for comparison studies should be put for the equal DNA load during library preparation, which is not always possible. As a result, the quantitative comparative analysis of supernatant was not equivalent to the pellet and suspension gDNA bacterial composition, representing only a set of lysed bacteria but not their real ratio in the probe.
The non-metric multidimensional scaling (NMDS) of sequenced samples (Figure 2C) showed clear grouping by the individual, indicating similarity independent of the isolation kit, or starting material taken for gDNA extraction. However, the bacterial composition of the supernatant samples considerably differed from the pelleted and suspension material. Variation among the samples of one individual was most likely connected to the heterogeneity of the initial material than to the isolation method used.
Despite the better characteristics of gDNA isolated by the ZR kit, microbiome profiles obtained by both kits were quite similar and it was difficult to prefer one or another extraction protocol. Considering the equal amount of gDNA for library preparation, the alpha-diversity of the gut microbiome assessed by Shannon’s index was similar between suspension and pellet samples for both kits (Figure 3A). Supernatant samples were different due to the smaller amount of gDNA taken for library preparation and consequently showed lower counts of detected bacterial species. Moreover, the number of sequence reads was more reproducible for the ZR isolation method regardless of the starting material (Figure 3B). Surprisingly, this number was even higher for three out of five supernatant samples. This was different in the case of the PL kit, which had the lowest amount of reads from supernatant samples. Such a phenomenon can affect the interpretation of the results of microbiome studies. In general, both methods did not result in considerable differences according to Shannon’s diversity index while the ZR isolation kit provided more reproducible read counts across the various starting materials. The only observed preference between the two kits in favour of ZR was noticed by the inclusion of the supernatant samples, which were not incorporated in further microbiome profile analysis. Most likely the difference between these two kits is more significant only in the case of low DNA concentration.
A more comprehensive analysis of individual bacteria species might uncover further discrepancies between the kits used, but to make a reliable conclusion, it is necessary to analyse a substantial number of participants and perform systematic statistical analysis. Besides, our personal experience with the PL kit usage in different laboratories revealed its low lab-to-lab reproducibility of gDNA extraction from the same sample. Overall, despite the higher gDNA amount and the better quality parameters for the ZR kit, the biological variability between individuals was much more significant than the observed differences in gDNA isolation technology.
The effect of material storage conditions on the gut microbiome
ZR DNA extraction kit was chosen for this study stage as it showed slightly more stable and qualitative results and is compatible with the Shield collection solution produced by the same company. gDNA was extracted from the suspension mixture of a faecal sample in the preservation solution. Faecal samples were stored for up to three weeks at room temperature (RT), +4 °C, and -20 °C. The latter was either taken for analysis as a new aliquot (-20 °C new) or repeatedly frozen and thawed (-20 °C).
NMDS of sequenced samples kept at various temperature conditions revealed their uniform distribution among the individuals independent of storage parameters, indicating stable results for all tested temperatures (Figure 4A). Analysis of the top ten most abundant bacterial strains of each person showed patterns very similar to the controls and uniform across the temperatures and time points (Figure 6). This supports the efficiency of the Shield solution to conserve faecal samples at different temperatures during the three weeks of storage. Shannon’s index registered a decline in diversity for samples stored at RT (Figure 4B) but considering the magnitude, it was not significant. The storage of samples for one week had almost no impact on the number of detected bacteria for all other temperature conditions. Sørensen dissimilarity index (Figure 4C), which was close to zero, also corroborated the stability of the microbial communities during the tested period and confirmed that the repeated freezing and thawing had no significant impact on microbiome identification.
Partial least squares discriminant analysis was applied to investigate the differences between the samples in more detail. This algorithm identifies features that vary between the treatments, in this case, the storage conditions. Overall, there were no common differences between the tested storage conditions, nevertheless, a model built using the data of three out of five most similar individuals (S1, S3, S5) presented some clear trends (Figure 5). Only the most important species are shown (variable importance in projection > 1) on the plot. Considering the divergence between biological samples, results revealed that storage at -20 °C had the smallest impact on the microbiome composition compared to the control. The higher the temperature of the storage, the further away the cluster was from the control, indicating changes in the microbial profile. Nevertheless, these changes were minor in absolute terms, as confirmed also by the low explained variance of the two components (13.6%).
Next, a heatmap analysis was performed to visualise the sequence data (Figure 6). As the gut microbiomes had more than 440 different bacterial species, only 80 of the most abundant species were included in the heatmap. The most represented bacterial species in all individuals were F. prausnitzii followed by Lachnospiraceae and Ruminococcacaeae spp. Again, the abundance of species was stable among biological individuals independent of the storage time and conditions. As seen in Figure 5, some samples, such as S4, had a more constant distribution of species compared to the control, whereas others (S2 and S5) were more divergent at different storage conditions. However, in general, we can claim that keeping faecal samples at +4 °C or -20 °C for one week in the Shield solution had minimal impact on the microbiome composition. Additionally, repeated freezing and thawing procedure did not change dramatically the whole bacterial distribution, indicating that the same sample can be repeatedly used in case a re-test is needed.
Gram-positive bacteria have a thick peptidoglycan layer in their cell wall that makes them more difficult to lyse than most Gram-negative bacteria, which can influence the whole microbiome analysis. To understand the specific effect of storage conditions on Gram-positive and Gram-negative bacteria, several most abundant bacterial species from these categories were selected and visualised as heatmaps (Figure 7A, 7B). This analysis also did not show any evident negative impacts of sample storage conditions or time. A slight reduction of Gram-negative and an increase of Gram-positive bacteria during the storage period were observed for individuals S4 and S5. At the same time, S2 and S3 showed the opposite tendency, whereas the distribution between Gram-positive and Gram-negative bacteria in S1 fluctuated in response to the storage conditions. Altogether, this data indicates a low impact of temperature and up to three weeks of storage time on the detection shift between various bacteria groups.