A comparison of sample processing and bacterial lysis methods for dairy product DNA extraction using automation friendly kits

The bacterial composition of milk and other dairy products are important determinants of product safety and quality. Therefore, accurate assessments of bacterial populations in milk are of great importance to the dairy industry. High-throughput 16S rRNA gene sequencing technologies are increasingly used in examining the dairy ecosystems. However, without standards methods, inconsistent and contradictory results have been reported. Therefore, in this study, by using both bacterial cell mock community and real milk samples, we evaluated a number of factors including milk collection volumes (200 L, 1 mL, 10 mL and 30 mL), storage conditions (freeze with PBS or 25% glycerol and 5 cycles of freeze-thaw), propidium monoazide treatment to enrich for viable cells, DNA extraction kits (MagMAX Total, CORE and Ultra 2.0) and cell lysis methods. The method of cell lysis conferred the largest variation on bacterial diversities, with effect sizes of 0.76 (Bray-Curtis) and 0.54 (weighted UniFrac). Specifically, gentle lysis methods (bead beat for 10 s at 4 m/s or vortex at 1800 rpm for 10 s) paired with MagMAX Total kit and proteinase K digestion provided the most accurate mock community representation, resulting in up to 12.3-fold reduction in Bray-Curtis distances from theoretical values compared to CORE and Ultra 2.0 kits using harsh lysis methods (bead beat twice for 1 min at 6.5 m/s with 1 min interval on ice or vortex at 1800 rpm for 15 min). This DNA extraction method (Total + proteinase K + gentle lysis) was also validated by raw and pasteurized milk samples. For the upstream milk collection, at least 10 mL should be sampled for consistent representation of the full community. Storage conditions caused minor variations on bacterial diversity, with effect sizes of 0.12 (Bray-Curtis) and 0.09 (weighted UniFrac). PMA inhibition of dead cells were variable between organisms. Moreover, contamination from PMA (51%) and RNase A treatment (7%) were prevalent, highlighting the importance of negative controls.