The microbial fermentation behind sourdough bread is among our oldest technologies, yet there are many opportunities for sourdough science to learn from traditional bakers. We analyzed the microbiome structure and function of 40 starters (grown from 10 types of flour) over 14 days and identified 6 distinct stages of succession. At each stage, bacterial taxa correlate with key functions that determine bread quality, including pH, rise, and aromatic profile. Day 1 starter cultures were dominated by microorganisms commonly associated with plants and flour and by aromas similar to toasted grain/cereal. Bacterial diversity peaked from days 2-6 as taxa shifted from opportunistic/generalist bacteria associated with flour inputs, toward specialized climax bacterial communities (days 10–14) characterized by acid-tolerant taxa and fruity (p<3.03e-03), sour (p<1.60e-01), and fermented (p<1.47e-05) aromas. This collection of traits changes predictably through time, regardless of flour type, highlighting patterns of bacterial constraints and dynamics that are conserved across systems and scales. Yet, while sourdough climax communities exhibit similar markers of maturity (i.e., pH ≤4 and enriched in Lactobacillus [mean abundance 48.1%], Pediococcus [mean abundance 22.7%], and/or Gluconobacter [mean abundance 19.1%],), we also detected specific taxa and aromas associated with each type of flour. Our results address important ecological questions about the relationship between community structure and function and may enable bakers to deliberately select for specific sourdough starter and bread characteristics.

Five researchers grew a total of 40 starters from 10 different flour types: 4 replicates each of five flours containing gluten–unbleached all-purpose flour (processed from Triticum aestivum), red turkey wheat (Triticum aestivum), rye (Secale cereale), emmer (Triticum dicoccon), and einkorn (Triticum monococcum) – and five gluten-free flours – teff (Eragostis tef), millet (Eleusine coracana), sorghum (Sorghum spp.), buckwheat (Fagopyrum esculentum), and amaranth (Aramanthus caudatus). The all-purpose, red turkey wheat, emmer, einkorn, rye, and millet flours were milled at Boulted Bread Bakery (Raleigh, NC), while the teff, sorghum, buckwheat, and amaranth flours were purchased from commercial vendors online.

On day 0, two level tablespoons of flour and two tablespoons of distilled water were mixed in a half-pint wide-mouth glass jar with a sterilized spoon. All spoons were washed with soap and water and wiped with 70% ethanol immediately before mixing. A paper towel was fastened over the mouth of the jar with a rubber band to prevent large particles or insects from settling into the jar while still allowing environmental microbes to colonize the flour-water mixture. After 24 hours, we measured the maximum height of the starter in cm, described the smell(s) of the starter using free association, then mixed the starter and measured pH using short-range (0–6, 0.5 interval) Hydrion Brilliant Paper (MicroEssential Laboratory, Brooklyn, NY). Next, we and transferred 1 mL starter to a sterile, labeled 2-mL microcentrifuge tube, which was stored at -20°C for DNA sequencing. Then we removed 1 tablespoon of sourdough to discard. Finally, we added 1 tablespoon fresh flour and 1 tablespoon water to the remaining starter and mixed thoroughly with the spoon. These measuring and refreshing steps were repeated once a day (at 24-hour intervals) for 14 days. 

We collected three 1-mL samples of each flour type (n=30) and three 1-mL samples of distilled water to represent the resource inputs to each starter on day zero. We also collected a daily 1-mL aliquot from each starter on days 1, 2, 3, 6, 10, and 14 (n=240), as part of the discard. All aliquots were transferred to a sterile, labeled 2-mL microcentrifuge tube, which was stored at -20C for DNA sequencing. Together with the flour and water aliquots, a total of 273 samples were shipped on dry ice for DNA extraction, PCR amplification, and Illumina multiplexed sequencing of the bacterial 16S v4 region using the 515f/806r for bacteria at the Fierer Microbial Community Sequencing Lab (Boulder, CO).

 The Fierer Microbial Community Sequencing Lab demultiplexed raw sequences to produce fastq files with the DADA2 pipeline, as described at (https://github.com/fiererlab/dada2_fiererlab/tree/626a1aff9abe1fcdcac7fdf57ba51f3cf1db1e8e).