Earth’s life-sustaining oceans harbor diverse bacterial communities that display varying composition across time and space. While particular patterns of variation have been linked to a range of factors, unifying rules are lacking, preventing the prediction of future changes. Here, analyzing the distribution of fast- and slow-growing bacteria in ocean datasets spanning seasons, latitude, and depth, we show that higher seawater temperatures universally favor slower-growing taxa, in agreement with theoretical predictions of how temperature-dependent growth rates differentially modulate the impact of mortality on species abundances. Changes in bacterial community structure promoted by temperature are independent of variations in nutrients along spatial and temporal gradients. Our results help explain why slow growers dominate at the ocean surface, during summer, and near the tropics and provide a framework to understand how bacterial communities will change in a warmer world.

To explore the distribution of bacterial life strategies along principal axes of temperature variation in the ocean, we gathered 16S rRNA gene amplicon sequencing datasets of marine bacterial communities encompassing wide seasonal, latitudinal, and depth gradients.

We focused on three particular datasets in the main text to represent the three principal axes of temperature variation in free-living bacterial communities (Fig. 1a). First, to analyze seasonal data, we used an eight-year pelagic microbial time series from the Linnaeus Microbial Observatory (LMO) in the Baltic Sea, 11 km off the northeast coast of Öland (N 56°55.8540′, E 17°3.6420′) (green dot in Fig. 1a). Seawater samples have been collected since 2011 on a monthly to weekly basis, together with environmental variables like temperature, inorganic nutrients (nitrate, phosphate and ammonium), dissolved organic carbon (DOC), and chlorophyll a concentrations (fig. S7). The LMO dataset includes free-living (<3 m and >0.2 m) as well as particle-attached (>3 m) and non-fractionated (>0.2 m) filter fractions. DNA was extracted from filters according to and modified after. We then amplified the V3V4 region of the 16S rRNA gene with the primers 341f-805r. Amplicon sequencing for LMO data was undertaken at the Science for Life Laboratory, Sweden on the Illumina MiSeq platform (2 × 300 bp paired-end reads). Subsequently, the Ampliseq pipeline (https://github.com/nf-core/ampliseq) was applied with DADA2 to infer amplicon sequence variants (ASVs). The used bioinformatic software versions were: nf-core/ampliseq = v1.2.0dev; Nextflow = v20.10.0; FastQC = v0.11.8; MultiQC = v1.9; Cutadapt = v2.8; QIIME2 = v2019.10.0. Taxonomic annotation of LMO ASVs derive from the SILVA database (version132).

To assess bacterial growth distributions across latitude and depth, we used datasets from two published cruise reports. In 2012, the five-week cruise ANT 28-5 collected seawater samples in the epipelagic zone (20–200m) of 27 stations in the Atlantic Ocean along a transect spanning ~100 degrees of latitude: from the polar regions of South America to the waters off the coast of England (yellow dots in Fig. 1a). The ANT dataset does not contain publicly available environmental data other than temperature measurements (fig. S8), but does include small (<8 m and >3 m) and large (>8 m) particle-attached in addition to free-living (<3 m and >0.2 m) filter fractions. The TARA Oceans project was an ambitious four-year expedition (2009–2013) conducted in a modified sailboat, taking samples from 210 globally distributed sites (purple dots in Fig. 1a) at depths from the surface down to 1,000 meters. All TARA samples are free-living filter fractions (<3 m or <1.6 m and >0.2 m), and metadata on phosphates, nitrates, and oxygen is included (fig. S9).

In addition to these three datasets, we analyzed three other time series: a three-year study at the Pivers Island Coastal Observatory (PICO) site (34.7181 °N 76.6707 °W) near the Beaufort Inlet (US East Coast), a seven-year survey at the Service d’Observation du Laboratoire Arago (SOLA) sampling station (42°31′N, 03°11′E) in the Bay of Banyuls-sur-Mer, North Western Mediterranean Sea, France, and a five-year study at the USC Microbial Observatory at the San Pedro Ocean Time-series (SPOT) station in the San Pedro Channel (33.55°N, 118.4°W). Finally, we analyzed the effects of depth and latitude in the latitudinal P15S GO-SHIP transect, a 7,000-km decadally repeated transect from the ice edge (∼66°S) to the equator (0°S) in the South Pacific Ocean (fig. S10). Both SOLA and SPOT samples are free-living filter fractions (<3 m and >0.2 m SOLA, <1 m and >0.2 m SPOT), while PICO and P15S GO-SHIP samples are non-fractionated (>0.2 m).

Available environmental variables for PICO include temperature, insolation, nitrate + nitrite, phosphate, ammonium, dissolved inorganic carbon (DIC), and chlorophyll a concentrations; for SOLA: temperature, length of the day in hours, nitrate, phosphate, ammonium, and chlorophyll a concentrations. Bacterial samples of SPOT were taken at five depths (5, 150, 500, 890 m and at the depth corresponding to the deep chlorophyll maximum, DCM), but only temperature and oxygen measurements were available for all the sampled depths. Particulate organic carbon (POC) and chlorophyll a concentration could be obtained only for the 5m samples, while phosphate and nitrate for 5m and DCM samples. Microbial communities along the P15S GO-SHIP transect have been sampled in 80 stations from the surface to 6000m. Temperature, phosphate, nitrate + nitrite, ammonium, oxygen, and chlorophyll a concentration measurements are available from the surface to the mixed layer depth (MLD, around 150m). For samples below the MLD, ammonium and chlorophyll a concentration were not available. We analyzed the data above and below the MLD separately: we used the data within the MLD to explore community growth relationship with environmental variables along the latitudinal gradient and data below the MLD to explore community growth relationship with environmental variables across depths.