ddPCR allows 16S rRNA gene amplicon sequencing of very small DNA amounts from low-biomass samples

Abstract

Background: One limiting factor of short amplicon 16S rRNA gene sequencing approaches is the use of low DNA amounts in the amplicon generation step. Especially for low-biomass samples, insufficient or even commonly undetectable DNA amounts can limit or prohibit further analysis in standard protocols.

Results: Using a newly established protocol, very low DNA input amounts were found sufficient for reliable detection of bacteria using 16S rRNA gene sequencing compared to standard protocols. The improved protocol includes an optimized amplification strategy by using a digital droplet PCR. We demonstrate how PCR products are generated even when using very low concentrated DNA, unable to be detected by using a Qubit. Importantly, the use of different 16S rRNA gene primers had a greater effect on the resulting taxonomical profiles compared to using high or very low initial DNA amounts.

Conclusion: Our improved protocol takes advantage of ddPCR and allows faithful amplification of very low amounts of template. With this, samples of low bacterial biomass become comparable to those with high amounts of bacteria, since the first and most biasing steps are the same. Besides, it is imperative to state DNA concentrations and volumes used and to include negative controls indicating possible shifts in taxonomical profiles. Despite this, results produced by using different primer pairs cannot be easily compared.

Keywords: 16S rRNA gene sequencing; Low-biomass samples; Very small DNA amounts; ddPCR.

Fig. 1

Overview of experimental procedure of this study. Experiments are divided into two parts. Left, control experiments (shaded green) were used for checking if the additional ddPCR step did not introduce bias in the resulting taxonomical profiles. Right, experimental procedures (shaded red) were described to detect the minimum input amount of DNA input necessary to produce reliable 16S rRNA gene sequencing results

Fig. 2

Control experiment to test for bias possibly introduced due to the extra ddPCR step after 2nd-step PCR. Samples processed with ddPCR (Samples marked “D” for ddPCR processed, 12 ng DNA used) are compared to standard short amplicon controls which were not ddPCR processed (Samples marked “C” for Control, 12 ng DNA used). Four human samples: T1 (red), T28 (orange), T29 (green), T30 (turquoise) and two mock communities: Zymo (pink), ZIEL2 (blue) were sequenced using primer pairs amplifying different V-regions. A Meta Multi-Dimensional Scaling (MDS) shows that samples cluster significantly differently due to their origin and not by preparation method. B The dendrogram shows that clustering is dependent on sample origin even though clustering within a sample is effected by the V-region targeted. C Taxonomic profiles at genus-level of Sample-C and Sample-D for human samples. D As before, for mock samples from Zymo and (E) and ZIEL2. Note, the taxonomic profiles at the genus-level show only minor differences when the same V-region is targeted

Fig. 3

Taxonomical profiles at the genus level for two human samples (T1, T30) and two mock communities of known composition (Zymo, ZIEL2). For every sample, different initial DNA amounts were used for 1st-step PCRs and, further, different V-regions were targeted

Fig. 4

Proof of concept study for low biomass samples using water of ponds and rivers. A Ten different samples were selected from water bodies in and around the city of Freising, Germany. The map was provided by GoogleMaps. B The rarefaction curve shows that cleaned water samples (MilliQ and desalted water samples) have significantly lower number of reads and species compared to all other water body samples. C The multi-dimensional scaling (MDS) plot shows that cleaned water samples (‘control’) cluster significantly apart from the water body samples. D Significant differences were found when comparing the water body samples to the controls for richness or, e.g., the abundance of the genus Limnohabitans