Introduction

A microbiota refers to a collection of microorganisms within a specific environment and encompasses bacteria, viruses, archaea, and fungi. The corresponding microbiome refers to this collection of microorganisms and their genomes1. The microbiota can be classified within localized regions, such as the gut and respiratory tract, and their composition is found to vary widely across different sites1. As bacteria are the most abundant component of the microbiota2, most studies are focused on bacterial microbiota at body sites that have high microbial biomass, such as the gut. The gut microbiota has thus been widely studied and shown to play important roles in maintaining human health3. More recently, there has been growing interest in the role of microbiota at low-microbial biomass body sites, such as the respiratory tract4. The microbial communities within the respiratory tract contain potential respiratory pathogens, but also act as a first-line defense against respiratory infections such as COVID-19, possibly modulating the outcomes of respiratory disease5. A growing number of studies6, including our own7, shows that the respiratory microbiota presents a significant therapeutic and prophylactic potential against respiratory infections and severe patient outcome. However, one of the challenges in analyzing low-microbial biomass respiratory microbiota samples is the lack of reliable, sensitive, fast, and cost-effective nucleic acid isolation techniques that can be used for large-scale research and clinical settings to advance our understanding of respiratory microbiota in relation to human health.

Numerous methods and commercial kits have been developed for nucleic acid extraction, for example the QIAamp DNA Microbiome kit, the Molzym Ultra-Deep Microbiome Prep kit, and the MasterPure Complete DNA and RNA Purification kit, which can be applied to various sample types including low-microbial biomass swabs and fluids. However, these commercial kits are often costly, and some use filter-based spin-column methods which can be difficult or impossible to automate using liquid handling robots to achieve higher throughput. Cost-effective, high-throughput alternatives exhibiting comparable or improved performance to the currently available commercial kits would make larger studies feasible and accelerate advancements in microbiota research. In Norway, during the COVID-19 pandemic, a test kit was initially developed for rapid isolation of viral RNA to detect SARS-CoV-2 infection with a sensitivity comparable to available commercial kits8. The extraction procedure, based on the use of NAxtra magnetic nanoparticles is low-cost, robust, and reliable, providing high-quality nucleic acids within a short time. When automated, nucleic acid extraction can be completed within 14 min for up to 96 samples on the robot systems KingFisher Flex and KingFisher Duo Prime9. It can also be done on higher-throughput liquid handling robots such as the Tecan Fluent® Automated Workstation, which could process 288 samples in one run within an hour. The NAxtra protocol was subsequently adapted and further validated for nucleic acid extraction from other types of viruses8, as well as mammalian DNA and RNA9. However, the suitability of this protocol for other biological applications, including bacterial microbiota characterization, remains to be assessed.

Here we report the outcome of a simple pilot study to establish whether further adaptation of the NAxtra protocol is possible for the broader application of profiling bacterial microbiota in low-microbial biomass samples. Here, rather than performing an expensive and time-consuming comprehensive study, we carried out a pilot study of a small number of (i) low biomass nasopharyngeal aspirates and nasal swabs, and (ii) saliva samples, which contain relatively higher microbial biomass. We evaluated these samples by characterizing their bacterial microbiota using 16 S rRNA gene sequencing, which we then compared to the same sample types from previous studies using other extraction protocols. In addition, we compared the bacterial microbiota profiles from different respiratory sample types in this study.

Materials and methods

Ethics

Samples used in this study were collected in the context of routine diagnostics, and the assays were performed on residual anonymized patient materials from a Norwegian-accredited diagnostic service laboratory supporting diagnostic processes, including nucleic acid extractions for standard RT-PCR assays, thereby not requiring informed consent and ethics committee approval, in compliance with IRB regulations and Norwegian National Legislation of Infection Prevention Act [Smittevernloven] (§ 3–7): Laboratories and institutions may carry out methodological testing using available sample material without the consent of those who have given the samples if the purpose of the testing is to develop new methods or improve existing methods for the identification and description of a communicable disease. https://lovdata.no/dokument/NL/lov/1994-08-05-55?q=Smittevernloven.

Sample collection and nucleic acid extraction

Nasopharyngeal aspirate and nasal swab samples (n = 8 each), and saliva samples (n = 8) were randomly selected from routine diagnostic sampling at Oslo University Hospital (OUH), Oslo, Norway, and treated anonymously. Total nucleic acid was extracted using the NAxtra™ nucleic acid extraction kit (Lybe Scientific) according to manufacturer’s instruction8 https://lybescientific.com/products/ on a Tecan Fluent® Automated Workstation at the laboratory in the Department of Microbiology at OUH, with minor modification, i.e., the elute buffer (water) was decreased from 100 µl in the original protocol to 80 µl to increase the DNA concentrations. The sample input volume was 100 µl. The DNA concentrations in the elutes were measured using a Qubit™ dsDNA HS kit on a Qubit 3.0 Fluorometer (Life Technologies, Darmstadt, Germany), and 2 µl DNA eluate was used as input volume.

16 S rRNA gene sequencing

The V3-V4 region of the 16 S rRNA gene was amplified based on the two-step PCR procedure (25 cycles were used for first PCR and 8 cycles for second PCR), as described in the Illumina application note10 using primer sequences derived from Klindworth et al.11. Positive control ZymoBIOMICS Microbial Community DNA Standard II (Log distribution) (Zymo Research, Irvine, CA, USA) was used to control for PCR amplification of 16 S rRNA gene and sequencing, and negative control (water) was used to identify and remove potential contaminations (i.e., taxa present in DNA elution buffer water). The final libraries were verified using the Tapestation 4200 with D1000 reagents (Agilent, Santa Clara, USA) and quantified on a SpectraMax M3 fluorometric plate reader (Molecular Devices, San Jose, CA, USA) using Quant-it HS Assay reagents (Thermo Fisher Scientific, Waltham, MA, USA). Libraries were pooled based on their concentrations and then sequenced on an Illumina MiSeq platform (Illumina, San Diego, CA, USA) with 300 bp paired end reads (v3 reagents). 20% PhiX control library was added to the libraries, and cluster density was reduced to 80% of regular levels. Base calling and production of demultiplexed FASTQ files were performed by running RTA v1.18.54.4 and bcl2fastq v2.20.0.422.

Bioinformatics and statistics

The demultiplexed FASTQ files were analyzed using QIIME2 version 2022.8 implemented in a SLURM script. Reads were demultiplexed and adapters were trimmed using cutadapt with forward and reverse sequences CCTACGGGNGGCWGCAG and GACTACHVGGGTATCTAATCC respectively. To remove low-quality calls at the ends of the reads, adapter-trimmed reads were truncated at 250 bp for the forward read and 190 bp for the reverse read. Denoising and amplicon sequence variant (ASV) inference was performed using DADA2. Taxonomic classification of the ASVs was then carried out with the feature-classifier classify-sklearn command with SILVA release 138 as the reference. A phylogenetic tree was estimated using the align-to-tree-mafft-fasttree command, which generates a MAFFT alignment from reads, and fasttree was used to estimate a phylogenetic tree from these alignments. To remove possible contaminations (taxa present in the negative control water sample), a vector was created with all ASVs except the ones present in the negative control. This vector was then applied to the ‘prune_taxa()’ function from the R phyloseq package on the ‘phyloseq’ object. Alpha diversity indices, including observed ASVs and Shannon index, were calculated using the phyloseq package with the function ‘estimate_richness ()’. Differences in alpha diversity indexes between groups were tested with the Kruskal-Wallis test, followed by a Wilcoxon rank sum test when statistically significant indexes were identified by Kruskal-Wallis. Benjamini–Hochberg correction was used to adjust p-values from the Wilcoxon rank sum test for multiple testing. Beta diversity analysis was based on Bray–Curtis distances determined using the phyloseq package. Samples were clustered according to bacterial composition using principal coordinates analysis (PCoA) based on the Bray–Curtis distances. Permutational multivariate analysis of variance (PERMANOVA) was performed to test the differences in bacterial composition between groups using the vegan package (with the Adonis function) with the Bray–Curtis dissimilarity method. The QIIME2 analysis SLURM script and downstream R analysis and visualization scripts are available on GitHub at (https://github.com/CBGOUS/baisi24).

Results

DNA recovery from respiratory samples using the NAxtra protocol

Saliva samples showed higher DNA yield ranging from 0.242 to 17.8 ng/µl among individual samples, as compared to nasopharyngeal aspirates ranging from 0.286 to 12.8 ng/µl, and nasal swabs ranging from 0.058 to 4.44 ng/µl (Table 1).

Table 1 DNA yield of respiratory samples using the NAxtra protocol.
Full size table

Microbiota diversity and composition in respiratory samples

One nasopharyngeal aspirate sample 2A failed in 16 S rRNA gene amplification and thus was excluded in the subsequent sequencing and analysis. After quality filtering, including removal of possible contaminations (i.e., taxa present in the negative control water), an average of 334,315 reads per sample was produced across 23 samples. Saliva samples showed a statistically significant higher bacterial alpha diversity measured by observed ASVs richness and Shannon diversity index, compared to the nasopharyngeal aspirates and nasal swabs (B-H adjusted p < 0.05, Wilcoxon rank sum test) (Fig. 1A and B). A PCoA plot showed that nasopharyngeal aspirates were separated from nasal swabs and saliva samples, while nasal swabs and saliva samples clustered more closely (p < 0.05, PERMANOVA) (Fig. 1C). A total of 104 ASVs were common to all three sample types, while saliva and nasal swabs shared additional 297 ASVs, saliva and nasopharyngeal aspirates shared additional 42 ASVs, nasal swabs and nasopharyngeal aspirates shared additional 26 ASVs (Fig. 1D).

Fig. 1
figure 1

Bacterial microbiota diversity and composition in different samples. Box plots showing the alpha diversity indices represented by (A) observed amplicon sequence variants (ASVs), and (B) Shannon diversity. Boxes represent the interquartile range between the 25th and 75th percentiles, and the horizontal line inside the box denotes the median value, and the diamond represents the mean value. Asterisks refer to the Benjamini–Hochberg corrected p value, where *p < 0.05, **p < 0.001. (C) Beta diversity illustrated by principal coordinates analysis (PCoA) based on Bray–Curtis distance. (D) Venn diagram showing the numbers of common and distinct ASVs among different sample types.

Full size image

At the genus level, the most abundant taxon in saliva samples was Streptococcus, followed by Rothia, Prevotella, Veillonella, Neisseria, and Haemophilus (Fig. 2A). In nasal swabs, Streptococcus was the most abundant, followed by Chryseobacterium, Prevotella, Veillonella, Gemella and Leptotrichia. The most abundant taxon in nasopharyngeal aspirates was Corynebacterium, followed by Moraxella, Streptococcus, Staphylococcus, and Dolosigranulum. We observed higher similarity of bacterial composition among individual saliva samples than individual nasopharyngeal aspirate and nasal swab (Fig. 2A). A similar tendency was observed at the family level (Fig. 2B). We observed high abundance of certain gram-positive bacteria in different respiratory samples, for example, Streptococcus, Rothia, Corynebacterium and Staphylococcus, in line with previous findings in such samples12. These data suggest that the NAxtra protocol works efficiently on bacterial microbiota profiling, without additional lysozyme treatment specifically for gram-positive bacteria.

Fig. 2
figure 2

Abundance of bacterial taxa in different samples. (A) Genus level. (B) Family level. The 20 most abundant genera and families are indicated in different colors, the remaining taxa are binned into the ‘Others’ category.

Full size image

Impact of sequencing depth on nasal and saliva microbiota characterization

Rarefaction curves are used as a qualitative method to estimate the bacteria richness as a function of sequencing depth. Our results showed that the rarefaction curves reached a plateau at a higher sequencing depth for saliva samples than for nasopharyngeal aspirates and nasal swabs (Fig. 3). For nasopharyngeal aspirates and nasal swabs, the rarefaction curves reached a plateau at a sequencing depth of 10,000 reads for most samples. While for saliva samples, approximately 50,000 reads per sample were required to capture the full set of ASVs.

Fig. 3
figure 3

Rarefaction curves. The curve is created for each sample to assess the saturation at different sequencing depths for recovery of bacterial taxa presented by amplicon sequence variants (ASVs). Nasopharyngeal aspirate and nasal swab samples reach a plateau more rapidly (~ 10,000 reads) compared to saliva samples (~ 50,000 reads).

Full size image

Discussion

In this study, we performed a preliminary assessment of the suitability of a fast, low-cost, robust and high-throughput NAxtra™ nucleic acid extraction kit8 for characterizing bacterial microbiota in low-microbial biomass respiratory samples using 16 S rRNA gene sequencing. DNA concentrations and total yields of nasal samples in this study were higher than those reported by Giske et al., where the same input volume of the original sample (100 µl) and a lower volume of elution buffer (50 µl) was used13. DNA yields isolated from saliva samples using the NAxtra protocol were comparable to those reported using other methods with the same saliva sample input volume (100 µl)13. Saliva samples showed higher DNA density than nasal samples, in line with the report from Giske et al.13. DNA yields within the same sample types were variable in this study, as reported elsewhere for nasal samples14. It should be noted that other factors may also impact the DNA yield, including variations in sampling populations, sampling methods, and storage buffers15. Thus, results from different studies should be compared with caution, and further comparisons between NAxtra and other methods are required to validate the results. Importantly, microbiota profiling showed a high abundance of certain gram-positive bacteria e.g., Streptococcus, Staphylococcus in our tested samples, suggesting that the NAxtra protocol and NAxtra lysis buffer works efficiently on gram-positive bacteria without additional use of lysozyme.

We compared the diversity and composition of the nasal, nasopharyngeal and saliva microbiota. The results showed that the bacterial microbiota of saliva samples was significantly richer than that of nasal swabs and nasopharyngeal aspirates, and saliva bacterial microbiota varied less among individual samples than did nasal and nasopharyngeal microbiota. This aligns with findings from studies comparing the oral and nasal bacterial microbiota from same participants16,17, although the saliva and nasal samples were from different participants in this study. It has been shown that distinct microbiota profiles exist at different sites along the respiratory tract, for instance, bacteria Prevotella, Streptococcus, Veillonella, Rothia are predominant in the oropharynx, while in the nasopharynx, the dominant bacteria are Moraxella, Corynebacterium and Neisseria12. Similarly, our study showed the most abundant bacteria in saliva samples were Streptococcus, Rothia, Prevotella and Veillonella; while in nasopharyngeal aspirates, Corynebacterium and Moraxella were most abundant. Interestingly, we observed that bacterial composition in nasal swabs was distinct from nasopharyngeal aspirates, while closer to saliva. This could be due to that the anterior nasal cavity is physiologically closer to oral cavity than nasopharynx, the niche-specific selective growth conditions may shape the microbial communities along the respiratory tract4,18. Our finding was supported by another study reporting a related but distinct bacterial microbiota composition between the anterior nares and nasopharynx19, yet few existing studies have so far simultaneously compared microbiota among nasal cavity, nasopharynx, and oral cavity. As the metadata was unavailable for the study participants, we could not make any further biological interpretation on the microbiota differences observed amongst different sample types.

Few studies have assessed the 16 S rRNA gene sequencing depth for characterization of low-microbial biomass samples. In this study, given the small number of samples, we obtained ultra-deep sequencing dataset with an average coverage of 334,315 clean reads per sample, allowing evaluation of the required sequencing depth for low-microbial biomass samples. Our results showed that saliva samples reached a plateau at a higher sequencing depth than nasal samples, which is reasonable as samples with higher microbial density and diversity would require greater sequencing depth. In general, our study showed that a 16 S rRNA gene sequencing depth of 50,000 reads per sample is sufficient to capture a full bacteria composition in nasal and saliva samples. This allows the high throughput characterization of large numbers of samples in a single 16 S rRNA gene sequencing run. Conversely, this suggests that the cost of DNA extraction could constitute a significant portion of the total expenses in studies with large sample sizes. Therefore, our extraction protocol, with its low cost, could make larger studies more feasible.

We acknowledge limitations of this pilot study. Firstly, the sample size was small and there are no technical replicates, thus the comparisons of bacterial community from different sample types lack statistical support. Secondly, we did not compare the NAxtra protocol with other DNA extraction methods in this study, thus we were unable to draw comparisons between extraction methods for bacterial microbiota profiling in this study, although its sensitivity and selectivity has been proven to be comparable to products on the market in detecting viral and mammalian nucleic acids8,9. Thirdly, due to the restricted metadata associated with the samples, the biological interpretation of this study is limited. Fourthly, since we were using low-microbial biomass samples which produced accordingly low DNA yields, we were unable to reliably measure the purity of extracted DNA and ascertain any effect this may have had on DNA amplification. Nevertheless, from the outset, our goal was not to perform a comprehensive study to fully evaluate the protocol, for example, in a high throughput clinical setting, but rather to establish whether such a comprehensive study was justified. As such, our results indicate that a larger scale study containing replicates and comparing different extraction protocols is indeed warranted and could be carried out by a larger group with access to more resources. This would further verify our preliminary findings that indicate the potential of NAxtra protocol for large-scale microbiota studies and clinical diagnostics of pathogens in respiratory and other low biomass samples.

In conclusion, we report an assessment of a new nucleic acid extraction method on bacterial microbiota profiling in low-microbial biomass respiratory samples. The bacterial composition identified using the NAxtra protocol is comparable to that in similar sample types reported elsewhere using different extraction protocols. Given the simplicity, reduced time, low-cost, and high throughput, the NAxtra method has the potential to serve as an attractive option for large-scale microbiota research and routine diagnostics of respiratory pathogens in clinical settings. However, a more comprehensive study is required to clearly establish its benefits for microbiota analysis as compared to other protocols.