Microbial carbon use efficiency predicted from genome-scale metabolic models

Abstract

Respiration by soil bacteria and fungi is one of the largest fluxes of carbon (C) from the land surface. Although this flux is a direct product of microbial metabolism, controls over metabolism and their responses to global change are a major uncertainty in the global C cycle. Here, we explore an in silico approach to predict bacterial C-use efficiency (CUE) for over 200 species using genome-specific constraint-based metabolic modeling. We find that potential CUE averages 0.62 ± 0.17 with a range of 0.22 to 0.98 across taxa and phylogenetic structuring at the subphylum levels. Potential CUE is negatively correlated with genome size, while taxa with larger genomes are able to access a wider variety of C substrates. Incorporating the range of CUE values reported here into a next-generation model of soil biogeochemistry suggests that these differences in physiology across microbial taxa can feed back on soil-C cycling.

Fig. 1

Potential and substrate-limited CUE. Histogram of predicted potential CUE (purple) and predicted CUE under lysine-limitation (pink) across taxa

Fig. 2

Substrate-limited CUE. Boxplot of average CUE values across all taxa under potential and constrained scenarios. Boxplot width is proportional to number of models with a given constraining reaction. Dashed red line shows average for potential CUE. Shaded region shows range of values typically used in biogeochemical models. Solid lines within boxplots show median. Bottom and top edges of boxes represent 25th and 75th percentiles, respectively. Whiskers demarcate minimum and maximum datapoints within 1.5× of the interquartile range

Fig. 3

Phylogenetic heatmap of potential CUE. Phylogenetic heatmap of potential CUE values from draft metabolic models. Labels on tips correspond to kBase accession numbers

Fig. 4

Mean CUE versus GC content. Mean CUE versus GC content for manually curated metabolic models. Species in order of increasing GC content are Clostridium ljungdahlii DSM 13528, Staphylococcus aureus subsp. aureus N315, Saccharomyces cerevisiae S288c, Methanosarcina barkeri str. Fusaro, Bacillus subtilis subsp. subtilis str. 168, Thermotoga maritima MSB8, Synechocystis sp. PCC 6803, Escherichia coli str. K-12 substr. MG1655, Shigella boydii Sb227, Salmonella enterica subsp. enterica serovar Typhimurium str. LT2, Klebsiella pneumoniae subsp. pneumoniae MGH 78578, Geobacter metallireducens GS-15, Mycobacterium tuberculosis H37Rv. Mean CUE was calculated from CUE on growth on each of the following C-sources individually: D-Glucose, Fumarate, Acetate, Acetaldehyde, 2-Oxoglutarate, Ethanol, Formate, D-Fructose, L-Glutamine, L-Glutamate, D-lactate, L-Malate, Pyruvate, Succinate

Fig. 5

Potential CUE versus genome size. Potential CUE regressed against genome size (bp). Blue lines show GLS fit. Points are colored by phylum

Fig. 6

Ecosystem C stocks and fluxes with variable microbial communities. Annual totals for C cycle pools and respiration rates for models for high efficiency taxa (CUE = 0.9) relative to low efficiency taxa (CUE = 0.2) across 100 years. Dashed line represents no difference in model estimates at the two CUE values