16S amplicon sequencing gives more reliable data for human milk microbiota studies than shotgun metagenomics sequencing Johanne Spreckels1, Sanzhima Garmaeva1, Trishla Sinha1, Ranko Gacesa2, Alexander Kurilshikov1, Marloes Kruk1, Hiren Ghosh2, Hermie Harmsen3, Jingyuan Fu1,4, Alexandra Zhernakova1 1Department of Genetics, University Medical Center Groningen, Groningen, the Netherlands; 2Department of Gastroenterology and Hepatology, University Medical Center Groningen, Groningen, the Netherlands; 3Department of Medical Microbiology, University Medical Center Groningen, Groningen, the Netherlands; 4Department of Pediatrics, University Medical Center Groningen, Groningen, the Netherlands Background: Human milk microbes are important for the development of the infant gut microbiota and immune system. Many studies have investigated the milk microbiota using 16S amplicon (16S) sequencing, while only a few studies tried investigating the milk microbiome with whole metagenomics shotgun sequencing (MGS). Studying the human milk microbiome with MGS is challenging since milk contains a large number of human cells and only few microbes. Objective: We aimed to optimize milk sample preparation protocols to increase the amount and quality of microbial MGS data for human milk microbiome studies. Methods: DNA was isolated from a mock community, three human milk samples and negative controls with and without two different bacterial enrichment strategies (E1, E2) prior to DNA extraction, and subjected to 16S and MGS sequencing. Additionally, DNA was isolated from the same samples using four DNA extraction kits (A: MagMAX TNA Isolation Kit, B: DNeasy PowerSoil Pro Kit, C: Milk DNA Extraction Kit, D: QIAamp Fast DNA Stool Mini Kit) without enrichment and used the DNA for 16S sequencing. All isolations were performed in duplicates or triplicates. Results: All expected genera were detected in MGS and 16S sequencing data of mock communities isolated with the bacterial enrichment method E1. For mock samples prepared using the enrichment method E2, MGS library preparation was not successful, and only three of eight expected genera were detected when using 16S sequencing. The bacterial enrichment method E1 decreased both the percentage of microbial reads of the total number of MGS reads and the absolute number of microbial MGS reads in milk samples compared to samples prepared without enrichment. Even though similar bacterial genera were detected in most duplicates or triplicates of each milk sample, the detected genera were often found in negative controls as well, rendering the milk MGS data untrustworthy. Mock communities isolated without bacterial enrichment with the DNA isolation kits A-D and analyzed with 16S sequencing contained all expected bacterial genera, and kits A and B represented the true mock composition the best. The use of kits A and B also led to significantly lower 16S read numbers in negative controls than kits C and D. Both kits A and B identified bacteria commonly reported in human milk, and microbial profiles of the tested milk samples were similar with the two methods. Conclusion: Sample preparation and DNA isolation methods affect the results from milk microbiome studies. The tested bacterial enrichment methods do not improve milk MGS data and, also without bacterial enrichment, we cannot recommend the use of MGS to investigate the human milk microbiota. Instead, to study the human milk microbiota composition, we suggest DNA isolation with the MagMAX (A) or the PowerSoil Pro (B) kit and the use of 16S sequencing.