Warming effects on soil physicochemical and biological properties
We analyzed subarctic grassland soil samples from eight stable natural soil temperature gradients located in the same landscape at two adjacent sites being subject to sustained soil warming for years (8 y; MTW, medium-term warming site; 4 gradients) and decades (>50 y; LTW, long-term warming site; 4 gradients) (Fig. 1A), respectively. The analyzed samples comprised two distinct soil temperature regimes (AT, ambient soil temperatures and ET, +6 °C above ambient soil temperatures). To evaluate sample selection and characterize the individual sample groups we analyzed soil physicochemical and biological properties (Table S1). On average, total C, N and P (Ctot, Ntot, Ptot), dissolved organic C and P (DOC, DOP), and microbial C, N and P (CMO, NMO, PMO), as well as RNA and water content were higher in AT than in ET (Fig. 1B). Dissolved organic N (DON) and DNA content varied considerably between MTW and LTW, resulting in a lack of consistent differences between AT and ET for these variables. The pH was slightly higher in ET. These properties were individually tested for significant differences between defined groups (Fig. 1B, right panel, Table S2). Significant differences were mainly observed between the two temperature groups AT and ET. While Ctot, Ntot, and Ptot, did not differ significantly between AT and ET (Padj < 0.097), pH, DOC and DOP, as well as most biological properties (CMO, NMO, PMO, and RNA content, respectively) did. Besides DON, no significant differences between the two non-warmed control groups (MTW-AT and LTW-AT) were observed (Fig. 1B). These results largely mirror previous studies conducted at the same sites that also report lower substrate concentrations and microbial biomass contents in the warmed soils (15–18), indicating high reproducibility and suitability of the experimental sites for a comparative analysis of medium- and long-term warming effects.
In addition to the lower RNA contents in the warmed plots (Fig. 1B) and a strong linear correlation between microbial biomass (using microbial C as proxy) and RNA per g dry weight soil (Fig. 1C, scatterplot, Table S3), we observed a trend towards lower RNA content per unit of microbial biomass in the warmed soils (Padj = 0.083). This was contrasted by a significantly higher DNA content per unit of microbial biomass (Fig. 1C, boxplots).
Taken together, our observations are in line with previous studies which characterized these warmed plots as significantly divergent from their non-warmed counterparts (15–19). The new observations also support the proposition that an initial acceleration of biotic activity due to warming leads to decreases in C, N, and P pools within the first years after the onset of warming, followed by a decrease of microbial biomass (15). Warming also affected the ratio between nucleic acids and microbial biomass, an observation that was not made previously.
Warming effects on microbial gene expression
Illumina paired-end sequencing of total RNA from 16 soil samples produced an average of 6.69 ´ 106 mRNA reads per sample (Table S4). Bacteria dominated the mRNA read pools (93.36–98.52%), followed by Eukaryota (1.28–6.45%), Archaea (0.5–1.6‰), and viruses (0.2–1.9‰). An average of 2.66 ´ 106 mRNA reads per sample was assigned to a molecular function (KO number) defined in the KEGG Orthology database (21) (Table S4). These reads were also dominated by Bacteria, accounting for 90.82–98.47% of the functionally annotated mRNA reads (see Table S5 for a detailed comparison). We further analysed all functionally annotated mRNA reads assigned to a KEGG metabolic pathway or functional complex. Rarefied abundance counts (Table S6) indicate no significant effect of warming on functional richness (Fig. S1, Table S7) while a PERMANOVA analysis revealed a significant effect of warming on the composition of expressed KOs assigned to a KEGG metabolic pathway or functional complex (Fig. S2, Table S8). To identify the nature of this functional response we explored in more detail the functional assignments to KEGG categories and performed a differential gene expression analysis. KEGG offers a hierarchical structure with four layers, KEGG1 (broad functional categories), KEGG2 (functional sub-categories), KEGG3 (functional pathways and complexes), KO (molecular functions, i.e. orthologous genes encoding proteins, enzymes, and enzyme subunits). In total, 3164 unique KOs assigned to a KEGG metabolic pathway or functional complex were detected in the metatranscriptomes. Of all functionally annotated mRNA reads, 78.9%, represented by 2732 KOs, were assigned to only one KEGG1 category (Fig. S3), while the remaining mRNA reads were assigned to two or more KEGG1 categories. To avoid redundancy, we next explored the reads assigned to a single KEGG1 category and corresponding KEGG2 sub-categories. Their distribution across the soil temperature groups led to four major observations (Fig. S3): (i) Higher relative gene expression levels for major sub-categories such as Carbohydrate metabolism, Amino acid metabolism and Lipid metabolism, in the warmed soils (KEGG1: Metabolism). (ii) Lower relative gene expression levels for Energy metabolism in the warmed soils (KEGG1: Metabolism). (iii) Lower relative gene expression levels for sub-categories associated with protein biosynthesis, i.e. Translation, Folding, sorting & degradation, and Transcription in the warmed soils (KEGG1: Genetic information processing). (iv) Higher relative gene expression levels for Replication & repair in the warmed soils, albeit most pronounced in MTW-ET (KEGG1: Genetic information processing). Only within the KEGG1 category, Metabolism, a substantial fraction of the reads was assigned to two or more KEGG2 categories. Within the other KEGG1 categories, the fractions of ambiguous assignments were <1% of all KEGG2 assignments (Table S9). The very same observations (i.e. i–iv) were made if all ambiguous assignments were included (Fig. S4).
These initial findings pointed towards substantial overall changes in the most basic cellular functions such as protein biosynthesis, replication, energy conservation, and central metabolisms. Next, we employed an abundance-pattern filter (see Materials and Methods) to identify KEGG3 pathways and functional complexes exhibiting putative warming-induced differential expression. More than 40% of all mRNA reads assigned to a KEGG metabolic pathway or functional complex were associated with KEGG3 categories that exhibited warming-induced differential abundance patterns (Fig. S6, Table S10). Out of 56 KEGG3 categories with a visually distinct pattern, 13 showed lower relative gene expression levels in ET than AT, while 43 KEGG3 categories showed higher relative gene expression levels in ET than AT. The comparatively high relative abundances of eukaryotic mRNA reads in the LTW-AT samples (4.7% ± 1.8 SD) compared to MTW-AT, MTW-ET, and LTW-ET (1.5% ± 0.6 SD) may skew or mask biologically and ecologically relevant changes. Thus, we complemented the abundance-pattern analysis with a differential gene expression (DGE) analysis on the KO level (orthologous genes), excluding eukaryotic assignments. A total of 787 KOs were included in the DGE analysis (see Materials and Methods), out of which 66 KOs (i.e. 8.4%) were differentially expressed (false discovery rate (FDR) < 0.05) when comparing all AT with all ET soils (Table S11). Comparing gene expression profiles of MTW-AT and LTW-AT revealed no significant differences between the two non-warmed control groups (Table S11). We then assigned the differentially expressed KOs to the hierarchical KEGG structure (Table S12). The majority of the differentially expressed KOs fell within KEGG3 categories associated with protein biosynthesis, i.e. Ribosome, Protein export, RNA polymerase, and RNA degradation. In addition, several KOs assigned to these functional categories with an FDR between 0.05 and 0.1 shared the same expression pattern. The summarized expression of genes for all of these functions was lower in the warmed soils (Fig. S7) indicating a down-regulation of the protein biosynthesis machinery in warmed soils. Furthermore, several differentially expressed KOs were assigned to the Energy metabolism KEGG3 category Oxidative phosphorylation, the lower summarized expression of these genes in the warmed soils (Fig. S7) indicating a down-regulation of membrane-bound energy harvesting complexes. The majority of the remaining differentially expressed KOs were associated with one or more central carbohydrate metabolisms and nucleotide and amino acid metabolisms. Lists of all differentially expressed KOs and KOs with an FDR between 0.05 and 0.1 are provided as Supplementary Tables (Table S13–S15).
Down-regulated enzymes and enzyme complexes
The protein biosynthesis machinery. Due to the high number of differentially expressed KOs encoding enzymes involved in protein biosynthesis (Fig. S7) we decided to analyze this aspect of cellular metabolism in more detail. Our analysis included (i) the bacterial DNA-directed RNA polymerase, responsible for the transcription of DNA into rRNA (ribosomal RNA), mRNA (messenger RNA), and small non-coding RNAs, (ii) the bacterial ribosome, consisting of ribosomal proteins and rRNA and the site of protein synthesis (translation of mRNA into peptides), (iii) bacterial RNA degradosomes, responsible for the degradation and recycling of RNAs, (iv) bacterial protein folding complexes, responsible for the proper folding of peptides into mature proteins, and (v) the bacterial Sec dependent pathway of post-translational translocation of proteins across membranes (Fig. 2A, drawings). The relative abundances of mRNA reads encoding RNA polymerase subunits, ribosomal proteins, GroEL and DnaK (protein folding), and Sec translocase subunits (protein export) were consistently lower in the warmed soils (Fig. 2A, boxplots). For all these enzyme complexes we found a substantial fraction of KOs with significantly lower relative expression levels in the warmed soils, the pattern being largely consistent between MTW and LTW (Fig. 2A, table, Table S16). KOs encoding ribosomal proteins accounted for most of the differentially expressed KOs, with significantly lower relative expression levels in the warmed soils (overall 19 KOs). Within MTW, 22 ribosomal proteins showed significantly lower relative KO expression levels in the warmed soils, supported by eight ribosomal proteins with lower but not significant relative KO expression levels in the warmed soils (FDR 0.05–0.1). In LTW, we observed significantly lower relative KO expression levels in the warmed soils for seven ribosomal proteins. Supporting this observation, were 10 ribosomal protein encoding genes with lower, but not significantly different, relative KO expression levels in the warmed soils (FDR 0.05–0.1) (Table S16). Two ribosomal proteins showed significantly higher relative expression levels in the warmed soils of LTW, indicating that not all ribosomal proteins are regulated by the same feedback mechanisms (22). The expression of Sec subunits (protein export) showed highly consistent temperature responses in MTW (Table S16). While less clear in LTW, the summarized mean relative mRNA abundances being overall lower in the warmed soils (Fig. 2A, Protein export-boxplot). A contrasting temperature response was shown for RNA degradosome sub-units, with overall higher, but not significantly different, expression in the warmed soils relative to the ambient soils (Fig. 2A). Taken together our results suggest that the entire protein biosynthesis machinery, including enzyme complexes involved in transcription, translation, and protein processing, may be downregulated in the microbiomes subject to soil warming of +6 °C above ambient temperatures, irrespectively of the duration of warming (8 y vs. >50 y). Notably, the lower relative abundances of mRNA reads encoding ribosomal proteins suggests a downregulation of ribosome synthesis in the warmed soils. This was supported by the lower total RNA content per unit of soil (Fig. 2A, RNA-boxplots) and total RNA content per unit of microbial C (Fig. 1C) in both medium- and long-term warmed soils.
Energy metabolism. We observed lower expression of KOs associated with the Energy metabolism sub-category Oxidative phosphorylation in the warmed soils (Fig. S7). Oxidative phosphorylation is the final step in aerobic respiration (Fig. 2B, drawings). Electrons harvested from the oxidation of carbohydrates are transferred via a membrane-bound electron transport chain, O2 acts as terminal electron acceptor, and the generated proton-motive force is used by the ATP synthase to phosphorylate ADP and produce ATP, the major energy currency of all cells. Even though respiratory electron transport chains vary among Bacteria (23), we observed lower mean relative abundances of multiple common complexes including ATP synthase in the warmed soils (Fig. 2B, boxplots). Of all investigated membrane bound complexes, only succinate dehydrogenase, and only in MTW, did not exhibit a warming-dependent expression pattern. However, succinate dehydrogenase represents a non-pathway-specific enzyme (21), offering a potential explanation. Despite few genes being significantly differentially expressed (Fig. 2B, table), the large number of genes encoding metabolically connected membrane-bound enzyme complexes displaying a lower mean relative expression level in the warmed soils suggests that downregulation of the oxidative phosphorylation complexes may be a microbial response to soil warming.
Up-regulated enzymes and enzyme complexes
Central (carbohydrate) metabolisms. KEGG2 and KEGG3 categories with overall higher relative expression levels in the warmed soils were to a large extent observed within the KEGG1 category Metabolism (Fig. S3, S6, and Fig. S7), prompting us to analyze community transcript profiles focusing on central (carbohydrate) metabolisms (Fig. 3A). We observed only few differentially expressed KOs but all had higher relative expression levels in the warmed soils. Individual KEGG metabolic pathway maps (21) reflecting KEGG3 categories (e.g. Pyruvate metabolism, Glycolysis / Gluconeogenesis, and Methane metabolism) overlap and the few differentially expressed KOs mainly represent KOs assigned to several KEGG3 categories. Nevertheless, a few noteworthy observations were made (Fig. 3A).
Enzymatic reaction steps leading from pyruvate to Fatty acid biosynthesis via acetyl-CoA showed higher relative expression levels in the warmed soils while steps leading from pyruvate to Oxidative phosphorylation showed the opposite trend, irrespectively of the duration of warming (8 y vs. >50 y) (Fig. 3A). This matches our detailed analysis on Fatty acid biosynthesis (Fig. S8) and Oxidative phosphorylation (Fig. 2B), as an upregulation of fatty acid biosynthesis would require an increased flow of C from the central intermediates, acetyl-CoA and pyruvate, while a downregulation of Oxidative phosphorylation might be linked to a downregulation of succinate supply. Some integral pathways exhibited inconsistent temperature responses. For example, the first steps in glycolysis (from α-D- and β-D-glucose to β-D-fructose-6P) showed higher relative expression levels in the warmed soils and include differentially expressed KOs (Fig. 3A). However, most subsequent steps did not reveal a clear temperature trend. A previous in situ study on samples from the same LTW plots, taken one summer earlier, reported significantly higher biomass-specific organic C uptake rates (mg C g-1 CMO h-1) and higher biomass-specific respiration rates (mg C g-1 CMO h-1) in the warmed soils (ET, +6 °C) compared to their ambient counterparts (16). However, unlike the very consistent patterns within protein biosynthesis described above, our more detailed investigation of central (carbohydrate) metabolisms did not reveal sufficiently convincing patterns to conclude that a common microbial upregulation of central (carbohydrate) metabolisms is occurring due to soil warming and ambiguous KOs complicate interpretation.
Cell replication. We observed higher relative expression levels for amino acid and nucleotide metabolisms (Fig. S7), Glycerolipid metabolism (Fig. S7), Fatty acid biosynthesis (Fig. S8), and Peptidoglycan biosynthesis (Fig. S6), important pathways in the production of major biomolecule building-blocks and cell membrane and cell wall biosynthesis. Taken together these results prompted us to have a closer look on cell replication (Fig. 3BC). Bacterial cell replication can be divided into two closely interacting cycles, (i) the DNA cycle, i.e. DNA replication and chromosome segregation, and (ii) the division cycle, i.e. cytokinesis and cell separation (24). FtsZ is a key enzyme and regulator of bacterial cell division and an upregulation of its synthesis has been shown to correlate with the formation of the Z-ring, a ring-like structure at the division site constricting during cell division (24, 25). We saw indications for higher average expression levels of DNA polymerase III subunits and DNA replication initiation factors (Fig. 3B) as well as higher average expression levels of FtsZ in the warmed soils (Fig. 3C), but these patterns were not significant. Cellular DNA content depends on the stage of growth, usually being higher in actively replicating cells (26). The total DNA content per unit of microbial C was significantly higher in the warmed soils (Fig. 1C), further indicating a higher relative abundance of cells being replicated in the warmed soils. This interpretation is strongly supported by a previous study reporting significantly higher biomass-specific microbial growth rates (mg C g-1 CMO h-1; measured via DNA replication rates) in samples from the same LTW plots (ET, +6 °C), taken one summer earlier, compared to their ambient counterparts (16). However, extracellular or relic DNA stemming from dead microorganisms, which can represent up to 40% of prokaryotic and fungal DNA in soil (27), could also have contributed to the higher DNA:biomass ratios in the warmed soils.
Decoupling of microbial community structure and functional temperature response
A simple explanation for changes in gene expression between ambient and warmed microbiomes could be a compositional change in the active microbial communities as a response to warming. Therefore, we analyzed the taxonomic composition of mRNA read pools in detail, combining diversity and variance analysis with abundance-pattern and DGE analyses. An average of 6.02 ´ 106 mRNA reads per sample was taxonomically classified (Table S4). As described above, mRNA read pools were largely dominated by Bacteria (93.36–98.52%; Table S5, Fig. S9). Overall, more than 1,000 different prokaryotic, eukaryotic, and viral families were detected in the metatranscriptomes (Table S6). Mean family richness did not differ significantly between warmed soils and their ambient counterparts (Table S7), albeit showing a trend towards a reduced richness in the warmed soils (Fig. S1B). Furthermore, a PERMANOVA analysis revealed no significant effect of warming on the taxonomic composition of mRNA reads assigned on family level (Fig. S2B, Table S8), contrasting the observation of a significant effect of warming on the composition of expressed KOs (Fig. S2A). Correspondingly, a differential expression analysis of all mRNA reads assigned on family level did not reveal any overall significant differences in the taxonomic composition between ambient and warmed soils (AT vs. ET, Table S19). Repeating the differential expression analysis on subsets representing bacterial families and/or considering only functionally and taxonomically classified reads identified up to three families with differential abundances between AT and ET (Table S19). However, these taxa accounted for less than 0.05% of the whole communities, affirming that these minor changes in the taxonomic composition were not responsible for the gene expression changes. Furthermore, no significant differences in the community composition of the mRNA read pools between the ambient sites (MTW-AT vs. LTW-AT) that could potentially explain the lack of taxonomic changes were observed (Table S20). In contrast, significant differences between the warmed sites were observed (MTW-ET vs. LTW-ET, Table S20) prompting us to investigate taxonomic composition of mRNA read pools, abundance pattern, and changes in relative abundances of both sites individually (Fig. S9). While the taxonomic composition of mRNA reads assigned on family level did not change in response to medium-term warming (MTW-AT vs. MTW-ET), it did in response to long-term warming (LTW-AT vs. LTW-ET) (Table S21). Up to 44 families, 5 orders, 3 classes, and 6 phyla showed a differential abundance between LTW-AT and LTW-ET. These taxa accounted for less than 2% of the whole communities (Table S21) and up to 75% of the families displaying a differential abundance between LTW-AT and LTW-ET represented fungal families with lower relative mRNA read abundance in the warmed soils (Table S21). Thus, the earlier-mentioned functional changes observed within the bacterial community in response to warming (Fig. 2, Fig. 3) were not related to a shift in the taxonomic composition (of mRNA reads assigned on family level). We attempted to increase the taxonomic resolution by assembling the metatranscriptome reads (see Supplementary Materials). Our approach using rnaSPAdes (28), however, did not provide a sufficient number of long mRNA contigs (<10% of functionally annotated mRNA contigs were longer than the unassembled reads). Also, previous studies employing amplicon sequencing of rRNA genes did not indicate warming-induced changes of the microbial community composition at genus and operational taxonomic unit (OTU) level (15, 16, 29). Taken together these results provide evidence for a decoupling of microbial community structure and functions, as recently indicated in studies of different ecosystems including short-term warmed peat soil (30–34). In line with this, an extensive screening of all phyla, classes, orders, and families with a mean relative abundance ≥1‰ in MTW-AT, MTW-ET, LTW-AT, or LTW-ET revealed that on average nearly two thirds of all taxa (63.7%) showed taxon-specific expression patterns reflecting the overall temperature-dependent expression patterns of central carbohydrate metabolisms, energy metabolism, and protein biosynthesis (Fig. 3D, Fig. S11). However, the percentage of taxa exhibiting the pattern varied between 43 and 95% depending on the KEGG3 category and the warming duration and was clearly highest for the KEGG3 categories related to protein biosynthesis and energy metabolism.
Downregulation of the protein biosynthesis machinery as common physiological response of microorganisms to soil warming?
Downregulation of the protein biosynthesis machinery across the bacterial community (Fig. 3D), especially of ribosomal proteins, was the most pronounced response of the microbial communities to soil warming of +6 °C observed in this study (Fig. 2A). Combined with the reduced total RNA content per unit of microbial C in the warmed soils (Fig. 1C), these data suggested that the ribosome content per bacterial cell is lower in the warmed soils than in their ambient counterparts, as rRNA can account for >90% of the total cellular RNA content (35). Besides higher temperatures, the warmed ForHot soils are also characterised by a lower C, N, and P content (16–18). Starving Escherichia coli and Salmonella spp. cells reduce their ribosomal content (36, 37), suggesting that a downregulation of the translation machinery, is metabolically favorable in nutrient-limited ecosystems (38). Since the translation machinery accounts for up to 40% of total cellular proteins (39) and protein biosynthesis is the costliest type of macromolecular synthesis (40–42) a reduced number of ribosomes would furthermore reduce the energy costs (ATP) related to cell maintenance and replication. The indicated warming-induced downregulation of membrane-bound energy harvesting complexes including ATP synthase (Fig. 2B) might therefore be directly linked to the downregulation of the protein biosynthesis machinery. The described downregulation of the protein biosynthesis machinery and energy harvesting complexes do not necessarily reflect a downregulation of the processes itself (i.e. protein biosynthesis and energy generation). Protein and ATP synthesis rates (per ribosome and per ATP synthase complex, respectively) are also highly affected by temperature, with higher temperatures resulting in higher rates. This fits to the higher relative expression levels of genes for nucleotide, amino acid, and inorganic phosphate (Pi) supply, to the protein and ATP synthesis machineries (Fig. S7 and Fig. 2B). Consequently, the warming-induced downregulation of the protein biosynthesis machinery and energy harvesting complexes is not contradictory to a warming-induced upregulation of cell replication and increased biomass-specific growth rates (16). It is long known that increased temperatures accelerate mRNA synthesis and protein synthesis rate per ribosome (peptide chain elongation rate) in the model organism E. coli (43). Conversely, it was indicated more recently that lower growth rates induced by lower protein synthesis rates are associated with an increased content of ribosomal proteins (39, 44).
Protein biosynthesis is a central cellular process and related complexes, especially ribosomes, are conserved in all organisms. Our observation of a warming induced downregulation of the protein biosynthesis machinery across the soil bacterial community led to the following hypothesis: Reduction of cellular ribosome content is a physiological key response of soil microorganisms exposed to warming (Fig. 4A).
To test this hypothesis, we conducted two supplementary experiments. We extracted RNA from soil samples collected from the same AT and ET plots of the geothermal soil temperature gradients in a different season (autumn) four years later (October 2020) and conducted a short-term warming experiment using LTW-AT soil sampled at the same time (Fig. 4BCD). Our rationale for the first test was that an observation of the same pattern of RNA content relative to soil dry weight, during a different time-point and season, would be a first indication that ribosome reduction is a long-term warming response. The rationale for the experiment was that if this response is also triggered by short-term temperature change, it is directly influenced by temperature, not just indirectly due to altered nutrient concentrations (Fig. 1B).
The in situ RNA content per unit of soil extracted in autumn 2020 reflected the pattern observed in summer 2016, with a lower RNA content in the warmed soils (Fig. 4C). This indicates ribosome reduction throughout the seasons as a response of microorganisms to medium- and long-term soil warming (i.e. years and decades of warming, respectively). Likewise, we observed a massive reduction to approximately half the RNA content after one week of incubation of ambient soils (AT) at higher temperature (+6°C), remaining at the lower level throughout the experiment (Fig. 4D, upper bar plots). The decrease of RNA content per unit of microbial C (a proxy for microbial biomass) matched the decrease per gram dry weight soil (Fig. 4D lower bar plots). Thus, cellular ribosome reduction appears to be a fast but stable response to soil warming. We think that higher reaction rates, including higher protein synthesis rates per ribosome (43), caused by warming might allow microorganisms to reduce their cellular ribosome content and increase serial protein synthesis. A resulting ratio change between ribosomal and metabolic proteins was supported by the observed transcriptional patterns, although patterns for most central (carbohydrate) metabolisms were not significant. Such re-allocation of energy and matter could potentially accelerate overall metabolism and growth. This presumption is in line with theoretical and experimental work pointing at the importance of covariation of microbial protein composition and growth rate and suggesting that a careful tuning of highly abundant proteins, such as ribosomal proteins, liberates most resources for acceleration of other reactions ((39) and references therein). Very recently, optimal cellular resource allocation was described to be rather constant across non-extreme temperatures, assuming the same temperature-dependency of metabolic and translational rates, but resource allocation and ribosome content will vary with the nutrient status (predictive modelling approach calibrated with E. coil quantitative proteome data) (45). As such, cellular ribosome reduction in medium- and long-term warmed ForHot soils could be mainly caused by the lower C, N, and P content in the warmed soils. We did not observe a reduction in total C and N content during the short-term warming experiment (Table S24), indicating that the increased temperature is the main driver of cellular ribosome reduction on the short-term, but the quality of substrate may have changed. However, studying and quantifying substrate availability and accessibility in heterogeneous soil ecosystems and the nutrient status of individual members of complex soil microbial communities is very challenging. Thus, determining the relative effect of environmental variables affected by warming, such as soil temperature and substrate availability (and quality), on metabolic and physiological changes of microbial cells remains. Furthermore, cultivation-independent in situ studies need to be complemented by pure culture studies on ecologically relevant soil microorganisms, to assess how widespread the observed changes are among individual community members and ecologically relevant functional guilds. Additionally, microbial cells reducing their cellular ribosome content might subsequently reduce their cell sizes. It has been shown that E. coli cell sizes decrease slightly at higher temperatures (46). Thus, the carbon turnover effects of changing cellular density, cell composition, and cell size, potential key microbial responses to global warming, should be targeted in future studies.