Optimization of high-throughput 16S rRNA gene amplicon sequencing: an assessment of PCR pooling, mastermix use and contamination

16S rRNA gene sequencing is widely used to characterize human and environmental microbiomes. Sequencing at scale facilitates better powered studies but is limited by cost and time. We identified two areas in our 16S rRNA gene library preparation protocol where modifications could provide efficiency gains, including (1) pooling of multiple PCR amplifications per sample to reduce PCR drift and (2) manual preparation of mastermix to reduce liquid handling. Using nasal samples from healthy human participants and a serially diluted mock microbial community, we compared alpha and beta diversity, and compositional abundance where the PCR amplification was conducted in triplicate, duplicate or as a single reaction, and where manually prepared or premixed mastermix was used. One hundred and fifty-eight 16S rRNA gene sequencing libraries were prepared, including a replicate experiment. Comparing PCR pooling strategies, we found no significant difference in high-quality read counts and alpha diversity, and beta diversity by Bray–Curtis index clustered by replicate on principal coordinate analysis (PCoA) and non-metric dimensional scaling (NMDS) analysis. Choice of mastermix had no significant impact on high-quality read and alpha diversity, and beta diversity by Bray–Curtis index clustered by replicate in PCoA and NMDS analysis. Importantly, we observed contamination and variability of rare species (<0.01 %) across replicate experiments; the majority of contaminants were accounted for by removal of species present at <0.1 %, or were linked to reagents (including a primer stock). We demonstrate no requirement for pooling of PCR amplifications or manual preparation of PCR mastermix, resulting in a more efficient 16S rRNA gene PCR protocol.


Group
difference by PERMANOVA analysis, p=0.99), whereas the mock community samples are clearly distinct from nasal samples (significant difference by PERMANOVA analysis, p=0.001).
In the PCoA plot, replicates cluster very closely, such that the red 'group' line is not visible.- Supplementary Figure 8 Alpha diversity by multiple indices of mock community replicates.
Plot comparing PCR reaction conducted in triplicate (25µl), duplicate (40µl), and as a single reaction (75µl).Replicates by type of mastermix preparation -manual mix (red) and premixed (green) are linked by a blue-line.Alpha diversity is calculated after rarefication of high-quality reads.Note, y-axis is adjusted to reflect alpha indices variation seen amongst nasal samples (Supplementary Figures 4,5,and 15).Data presented are from Experiment 2.

Original Protocol
The manually prepared master mix contained 5ul 5x Q5 Buffer (NEB, USA), 0.5ul 10mM dNTPs (NEB, USA), 0.25ul Taq polymerase (NEB, USA), 14.25ul nuclease free water (ThermoFisher Scientific, USA), 1.25ul of both the forward and reverse indexed cartridge purified primers (ThermoFisher Scientific, USA) diluted to10uM in nuclease free water as before.Primer sequences are included below.Each 2.5ul aliquot of nucleic acid extraction was run with 22.5ul of master mix in triplicate.PCR was run for 32 cycles, initially at 98ºC for 2minutes, with 30 cycles of 98ºC for 30 seconds, 50ºC for 30 seconds, 72ºC for 1 minute and 30 seconds and finished with 72ºC for 5mins.Triplicate PCR products were pooled into single reactions per sample and sample pools were purified using an AMPure XP (Beckman Coulter) workflow at a ratio of 1X.Libraries were quantified using the Qubit High Sensitivity dsDNA kit (ThermoFisher) and equimolar library pools created.The equimolar pools were purified by gel electrophoresis and the Wizard SV Gel and PCR Clean Up Kit (Promega) before submission for sequencing.

Process used in Operations to generate the data
PCR was performed to amplify bacterial 16S ribosomal gene regions using V1V2 specific primers with attached adaptors and indexes.Manually prepared PCR reaction mastermixes were made using the Q5 High-Fidelity Polymerase Kit (New England Biolabs), according to the original protocol as described above.Pre-mixed mastermixes contained 12.5ul Q5 Hot Start High-Fidelity 2X Master Mix (New England Biolabs) and 7.5ul nuclease free water (ThermoFisher Scientific).2.5ul nucleic acid extract was used per 25ul triplicate reaction, 4ul per duplicate 40ul reaction and 7.5ul per single 75ul reaction.1.25ul of each forward and reverse primers (ThermoFisher Scientific) diluted to 10uM with nuclease free water were used per 25ul reaction and scaled accordingly for 40ul and 75ul reactions.Mastermix reagent volumes were scaled accordingly.PCR was run for 30 cycles, initially at 98ºC for 2minutes, with 30 cycles of 98ºC for 30 seconds, 50ºC for 30 seconds, 72ºC for 1 minute and 30 seconds and finished with 72ºC for 5mins.Triplicate and duplicate PCR products were pooled into single reactions per sample respectively and all samples were purified using an AMPure XP (Beckman Coulter) workflow at a ratio of 0.8X.Libraries were quantified using the AccuClear Ultra High Sensitivity dsDNA Quantitation kit (Biotium) and equimolar pools were subsequently created using a Biomek NX-8 liquid handler (Beckman Coulter).Samples were sequenced using the Illumina MiSeq (300bp paired-end reads, v3 Reagent Kit).Process controls included an extraction control, a negative PCR water control, an aliquot of the glycerol used for storage and an aliquot of the water used to dilute the mock community.

Proposed optimisations
Based on our findings, the optimised process would utilise single PCR reactions (75ul), removing the need to set up multiple reactions per sample, significantly saving time and effort.Additionally, this eliminates the requirement for subsequent pooling per sample post-PCR, removing the risk of pooling incorrect samples together, together with the increased level of sample tracking that would otherwise be necessary to support this step at scale.A pre-mixed mastermix would be used in place of a manually prepared mastermix, for speed, convenience and to reduce the risk of manual handling error.

Mis-assigned taxa by arb
A few operational taxonomic units were evaluated with BLAST where the species was disconcordant with the taxa genus and/or not a prokaryote, as assigned by arb, or where an uncultured bacterium was seen across the mock microbial community suggesting it was either an expected mock microbial community member or a significant contaminant:

Results from replication study (Experiment 2) PCR Pooling
After quality filtering of samples used to assess the requirement for pooling of PCR (Figure 1), median read counts were 126552, 110890, and 128873 from PCR reactions in triplicate, duplicate or as a single reaction, respectively from experiment 2. Pairwise Mann-Whitney-U test comparisons showed no significant difference in high-quality read counts generated from reactions in triplicate vs duplicate (p = 0.31), triplicate vs single (p=0.42), or single vs duplicate (p=0.15).We then investigated variation in alpha diversity (measures of the within sample diversity).We did not observe any significant difference between PCR pooling strategies using Kruskall-Wallis tests by Shannon, Simpson, Fisher, Chao1, and Observed indices, and replicates from pair-wise PCR pool conditions showed a strong correlation by Kendall's rank correlation coefficient (Supplementary Table 1, Supplementary Figure 6).
Beta diversity (measure of the similarity or dissimilarity between two samples) by Bray-Curtis index clustered by replicate on examination of the PCoA and NMDS ordination plots, and did not significantly differ between PCR pooling strategies by PERMANOVA analysis (F(2)= 0.47, p = 0.94).The groups did differ by PERMANOVA analysis when compared by sample type i.e. mock vs healthy nasal sample (F(2)= 35.2, p = <0.001)(Supplementary Figure 9).
Further, relative abundance of samples by all technical replicates (including various types of mastermix used) appeared to remain similar (Figure 4 and Figure 5).

Mastermix preparation
After quality filtering of samples used to assess mastermixes for Experiment 2 (Figure 1), difference in read counts from samples with manually prepared mastermix (median = 125222) or premixed mastermix (median = 120012) by Mann-Whitney-U test comparison did not reach statistical significance (p=0.91).Alpha diversity of replicates from manually prepared or premixed mastermix methods by Shannon, Simpson, Fisher, Chao1, and Observed indices, did not significantly differ between by Mann-Whitney-U comparison (Supplementary Table 3).Beta diversity by Bray-Curtis index clustered by mastermix preparation replicate on examination of the PCoA and NMDS ordination plots (Supplementary Figure 11).Further, relative abundance of samples by all technical replicates (including various types of mastermix used) appeared to remain similar (Figure 3).Replicate numbers in the repeat experiment (Experiment 2) were low and examined with the mock community serially diluted samples alone.

STORMS
Examples of the population of interest could be: adults with no chronic health conditions, adults with type II diabetes, newborns, etc.This is the total population to whom the study is hoped to be generalizable to.The sampling method describes how potential participants were selected from that population.
If the participants are from a substudy of a larger study, provide a brief description of that study and cite that study.
Clearly state how cases and controls are defined.
An example of relevant physiological state might be pre/post menopausal for a vaginal microbiome study; examples of stage in the life history of disease could be whether specimens were collected during active or dormant disease, or before or after treatment.

Geographic location
State the geographic region(s) where participants were sampled from.
MIxS: geographic location (country and/or sea,region) Geographic coordinates can be reported to prevent potential ambiguities if necessary.

Relevant Dates
State the start and end dates for recruitment, followup, and data collection.

STORMS
Recruitment is the period in which participants are recruited for the study.
In longitudinal studies, follow-up is the date range in which participants are asked to complete a specific assessment.Finally, data collection is the total period in which data is being collected from participants including during initial recruitment through all follow-ups.

Eligibility criteria List any criteria for inclusion and exclusion of recruited participants.
Modified STROBE Among potential recruited participants, how were some chosen and others not?This could include criteria such as sex, diet, age, health status, or BMI.
If there is a primary and validation sample, describe inclusion/exclusion criteria for each.

Antibiotics Usage
List what is known about antibiotics usage before or during sample collection.

STORMS
If participants were excluded due to current or recent antibiotics usage, state this here.
Other factors (e.g.proton pump inhibitors, probiotics, etc.) that may influence the microbiome should also be described as well. NA

Analytic sample size
Explain how the final analytic sample size was calculated, including the number of cases and controls if relevant, and reasons for dropout at each stage of the study.This should include the number of individuals in whom microbiome sequencing was attempted and the number in whom microbiome sequencing was successful.

STORMS
Consider use of a flow diagram (see template at https://stormsmicrobiome.org/figures).Also state sample size in abstract.
If power analysis was used to calculate sample size, describe those calculations.
Yes Methods

Longitudinal Studies
For longitudinal studies, state how many follow-ups were conducted, describe sample size at follow-up by group or condition, and discuss any loss to follow-up.

STORMS
If there is loss to follow-up, discuss the likelihood that drop-out is associated with exposures, treatments, or outcomes of interest.When interpreting the findings, consider how the interpretation of the findings may be summarized or quoted for the general public such as in press releases or news articles.
If causal language is used in the interpretation (such as "alters," "affects," "results in," "causes," or "impacts"), assumptions made for causal inference should be explicitly stated as part of 6.0 and 13.0.

4 :
Ordination plots of Bray-Curtis dissimilarity indices between replicate samples from different PCR pools.Principle Component Analysis (PCoA) (A) and Non-metric multidimensional scaling (NMDS) (B) of Bray-Curtis dissimilarity indices.Nasal samples obtained from healthy participants at the Wellcome Sanger Institute (green) and mock community isolates (blue) are represented.Replicates from different PCR pools are linked by a red-line.Replicates by various PCR pooling strategies cluster (no significant

6 :
Data presented are from Experiment 1. Supplementary Figure 5: Principle Component Analysis of Bray-Curtis dissimilarity indices between replicate samples from PCR pooling strategies (linked by green line) from nasal samples of healthy volunteers via the Wellcome Sanger Institute (blue).Replicates from libraries with different PCR pooling strategies are nearly indistinguishable.High-quality reads from the Operational Taxonomic Unit Matrix are rarefied and then converted to percentage abundance for each sample.Data presented are from Experiment 1. Principle Component Analysis of Bray-Curtis dissimilarity indices between replicate samples from PCR pooling strategies (linked by red line) from mock community serially diluted preparations.(A) Axis of PCoA is consistent with the PCoA plot in Supplementary Figure 8 and the mock community samples are observed to closely cluster.(B) Axis of PCoA is exaggerated to allow closer examination of the mock community cluster.High-quality reads from the Operational Taxonomic Unit Matrix in A and B are rarefied and then converted to percentage abundance in each sample.Data presented are from Experiment 1.

9 :
Alpha diversity by multiple indices post mothur qc of healthy nasal swabs obtained from the Healthy Volunteers at the Wellcome Sanger Institute.Plot comparing PCR reaction replicates conducted in triplicate at 25µl (red), duplicate at 40µl (green), and as a single 75µl reaction (blue), replicates linked by purple-line.Alpha diversity is calculated after rarefication of high-quality reads.Data presented are from Experiment 2.
Alpha diversity by multiple indices post mothur qc of healthy nasal swabs from the CARRIAGE study, comparing manual (green) and premixed (blue) mastermix replicates (linked by red-line).Alpha diversity is calculated after rarefication of high-quality reads.Data presented are from Experiment 1. Supplementary Figure 16: Ordination plots of Bray-Curtis dissimilarity indices between replicate samples from different mastermix preparations.Principle Component Analysis (PCoA) (A) and Non-metric multidimensional scaling (NMDS) (B) of Bray-Curtis dissimilarity indices.Nasal samples from healthy participants (CARRIAGE) (green) and mock community

PCoA NMDS Supplementary Figure 10:
Ordination plots of Bray-Curtis dissimilarity indices between replicate samples from different PCR pools.Principle Component Analysis (PCoA) (A) and Non-metric multidimensional scaling (NMDS) (B) of Bray-Curtis dissimilarity indices.Nasal samples from healthy volunteer obtained via the Wellcome Sanger Intitute(green) and mock community isolates (blue) are represented.Replicates from different PCR pools are linked by a red-line.Replicates from different PCR pooling strategies cluster (no significant difference Axis of PCoA is exaggerated to allow closer examination of the mock community cluster.High-quality reads from the Operational Taxonomic Unit Matrix in A and B are rarefied and then converted to percentage abundance in each sample.Data presented are from Experiment 2. Supplementary Figure 11: Principle Component Analysis of Bray-Curtis dissimilarity indices between replicate samples from PCR pooling strategies (linked by green line) from nasal samples of healthy volunteers obtained via the Wellcome Sanger Institute (blue).Replicates from libraries with different PCR pooling strategies are nearly indistinguishable.High-quality reads from the Operational Taxonomic Unit Matrix are rarefied and then converted to percentage abundance in each sample.Data presented are from Experiment 2. B Supplementary Figure 12: Principle Component Analysis of Bray-Curtis dissimilarity indices between replicate samples from PCR pooling strategies (linked by red line) from mock community serially diluted preparations.(A) Axis of PCoA is consistent with the PCoA plot in Supplementary Figure 11and the mock community samples are observed to closely cluster.(B) Supplementary Figure 13: Boxplot of high-quality reads post mothur qc.(A) Comparing manual (red) and premixed (green) mastermix replicates from healthy nasal swabs (CARRIAGE study).(B) Comparing healthy nasal swabs (CARRIAGE) (beige) with mock community experiments (blue).Data presented are from Experiment 1.

.71 Sample Listeria monocytogenes Pseudomonas aeruginosa Bacillus subtilis Salmonella enterica Escherichia coli Lactobacillus fermentum Staphylococcus aureus
Replicates from different mastermix preparations are linked by a red-line.Replicates from individuals with different mastermix preparations cluster (no significant difference by PERMANOVA analysis, p=1), whereas the mock community samples are clearly distinct from nasal samples (significant difference by PERMANOVA analysis, p=0.001).In the PCoA plot, replicates cluster very close, such that the red line connecting them is not visible.Data presented are from Experiment 1.

Table 3 :
Differences in Alpha diversity between mastermix choice using Mann-Whitney-U tests.No significant difference is seen by Shannon, Simpson, Fisher, Chao1, and Observed indices between manually prepared and premixed mastermix.Alpha diversity is

STORMS checklist. An editable version for adaptation and inclusion in publications is available from https://stormsmicrobiome.org
Other data types Report other data analyzed--e.g.metabolic function, functional potential, MAG assembly, and RNAseq.