The soils were collected (0–10 cm depth) from agriculturally used loamy Luvisol in northern Bavaria (49°54′ northern latitude; 11°08′ eastern longitude, 500 masl) in August 2014, sieved to 2 mm, and stored for 1 day at +5°C except for samples (n = 3) used to determine dry weight. Mean annual temperature and precipitation at the site are +7°C and 874 mm, respectively. The soil had a pH (KCl) of 4.88, a pH (H₂O) of 6.49, a total organic carbon (TOC) and total nitrogen content of 1.77 and 0.19%, respectively, and a cation exchange capacity of 13 cmol_C kg^-1.

Screw-cap glass microcosms (10 cm diameter and height of 12 cm) with a base layer of quartz sand were used for incubations. Eighty grams samples of soil were transferred to sample rings and installed on ceramic plates above the quartz sand. The soils were rewetted to field capacity by adding 10 ml of water to the underlying sand at +5°C for 2 days. Thereafter, the microcosms were preconditioned at +5, -5, and -20°C for 24 h. These temperatures were chosen to simulate (1) average annual soil temperature at the sampling site (control), (2) average winter temperature, and (3) lowest temperatures in some winters on this site and common deep freezer storage temperatures, respectively. Four position-specifically ¹³C-labeled isotopomers of glucose (¹³C-1, ¹³C-2, ¹³C-4, and ¹³C-6), uniformly ¹³C-labeled (U-¹³C) and unlabeled glucose (natural abundance background) were applied to the soils in separate microcosms with four replicates of each. 5 ml of 2.55 mM glucose solutions were applied on top of the soil. This C concentration was less than 5% of microbial biomass C and therefore, does not alter microbial community structure (Kuzyakov, 2010). Cups with 5 ml of 1 M NaOH were placed in each microcosm to trap CO₂. 20% (w/v) of NaCl was dissolved in NaOH traps at -20°C to depress the freezing point. The microcosms were sealed and incubated at the respective temperatures. NaOH in the vials was exchanged after 10 h, 1, 2, 3, 6, and 10 days at +5 and -5°C and in addition after 15, 20, and 30 days at -20°C. Soil samples were collected after 1, 3, and 10 days at +5 and -5°C and also after 30 days at -20°C, to account for the slow processes expected in deeply frozen conditions. Thirty grams of each sample was immediately subjected to chloroform fumigation-extraction, as described below.

A Nalimov outlier test with 95% significance level was performed on CO₂. ¹³C incorporation into bulk soil, microbial biomass, and CO₂ were tested for significant differences between the positions, time of incubation and effect of temperature with a factorial analysis of variance (ANOVA). If assumptions of normality and homogeneity of variances with groups were not met, outcomes were validated by a non-parametric Kruskal–Wallis ANOVA. Significant differences were determined with Tukey’s honest significance difference (Tukey’s HSD) post hoc test at a confidence level of 95%. Statistical tests were performed with Statistica (version 12.0, Statsoft GmbH, Hamburg, Germany).

CO₂ was released down to -20°C, although cumulatively, there were 3- and 10-fold reductions at -5 and -20°C, respectively, compared to +5°C at day 10. Similarly, recovery of glucose-derived ¹³C in CO₂ at -5 and -20°C was respectively 19 and 24% lower than +5°C (Figure 1A). At +5°C, ¹³C recovery in CO₂ showed preferential oxidation of the glucose C-1 position. In contrast, the glucose C-4 position was the most respired at subzero temperatures. Irrespective of temperature, two phases of glucose C mineralization were observed: (1) high glucose ¹³C respiration within the first 3 days at 5°C and 2 days at -5 and -20°C and (2) low respiration of glucose-derived ¹³C thereafter (Figures 1B–D).

At day 1, ¹³C recovery in bulk soil from uniformly labeled glucose was over 24% higher at -5°C than at +5 and -20°C. After the first day, recovery did not differ with temperature (Figure 2A). Similarly, position-specific patterns of ¹³C recovery in bulk soil were comparable between +5 and -20°C but differed at -5°C. At +5 and -20°C, ¹³C recoveries in bulk soil from C-2, C-4, and C-6 were significantly higher (p < 0.05) than C-1 (Figures 2B,D). In contrast, recovery from C-2 and C-6 was significantly higher (p < 0.05) than from C-1 and C-4 at -5°C (Figure 2C).

From uniformly labeled glucose, the highest ¹³C recovery from microbial biomass was obtained at -5°C and it was two and three times higher than at +5 and -20°C, respectively (Figure 2A). The position-specific patterns of ¹³C incorporation into microbial biomass were not affected by temperature. The ¹³C recovery in microbial biomass from C-1 was ≈1.7 times lower than C-2, C-4, and C-6 at each temperature (Figures 2B–D). Moreover, recovery from each position did not differ over time at any incubation temperature.

The ratio of extracellular ¹³C recovery in bulk soil to microbial biomass was highest at -20°C (factor 3), followed by +5°C (factor 1.5) and -5°C (factor 0.6) (Figure 2A).

Total extractable C in soil was highest at -20°C, being 1.4- and 2.3-fold higher than +5°C and -5°C, respectively (Figure 3A). Whereas DOC did not differ over time at -20 and +5°C, it increased by a factor of 1.5 between day 3 and 10 at -5°C, to reach a level similar to +5°C. ¹³C recovery in DOC at -20°C was also two times higher than both at -5 and +5°C (Figure 3A). Less than 0.4% of the applied ¹³C was recovered in DOC on day 1, irrespective of temperature (Figure 3A). The position-specific ¹³C patterns detected in CO₂ were different to DOC at each temperature. Whereas recoveries in DOC at -5°C did not differ between the four glucose positions, the recovery from the glucose C-6 position was twice as high as C-1 at +5 and -20°C (Figures 3B–D).

Microbial activity has been detected in soil down to -39°C (Panikov et al., 2006; Schaefer and Jafarov, 2016). This has evoked interest in the adaptive metabolic mechanisms of microorganisms at such extremely low temperatures. At subzero temperatures, low thermal energy is expected to slow down processes such as microbial substrate uptake (Hoehler and Jorgensen, 2013). However, less than 0.4% of applied ¹³C was recovered in DOC on the first day, irrespective of temperature (Figure 3A) suggesting that all applied glucose was taken up by microorganisms within the first day. These results agree with previous data showing that substrate uptake of highly available substances, such as glucose, is largely independent of temperature (Herbert and Bell, 1977; Schimel and Mikan, 2005). At the end of the incubation period, cumulative CO₂ was 3- and 10-fold lower at -5 and -20°C than +5°C, respectively (Figure 1A). The results agree with studies showing strong temperature sensitivity of CO₂ production in frozen soils (Monson et al., 2006; Panikov et al., 2006; Drotz et al., 2010). Irrespective of temperature, two phases of glucose C mineralization were observed: Phase 1 with high rates of ¹³C recovery in CO₂ is consistent with intense glucose mineralization. Phase 2 thereafter, with low ¹³C recovery rates (Figures 1B,D), reflects mineralization of glucose-derived metabolites (Blagodatskaya et al., 2011). These results contradict those of Drotz et al. (2010), who observed low ¹³C in CO₂ during phase 1 and high in phase 2. This discrepancy may reflect the soil types used for incubation. While our soils were agriculturally used loamy Luvisols, their soils were forest Podsols from the boreal region sampled from the organic horizon. Cold temperatures were shown to increase concentration of DOC in these soils (Haei et al., 2010), which implies a high concentration of low molecular weight organic substances and therefore low competition for applied substrate among microbes. In contrast, our agricultural soil was low in organic matter and the competition between microbes for applied glucose was highest immediately after application, leading to a fast uptake and release of glucose-derived CO₂.

The ¹³C recovery pattern in CO₂ showed high oxidation of the C-1 position in control soils (Figure 1B), and revealed that glucose was predominantly catabolized via the pentose phosphate pathway (Caspi et al., 2008; Dijkstra et al., 2011c; Apostel et al., 2015). Contrary to our hypothesis, metabolic behaviors completely switched over to a preferential respiration of the glucose C-4 position at -5°C and this behavior was still weakly visible at -20°C (Figures 1C,D). High ¹³C recovery from the C-4 position reflects glucose transformation via glycolysis (Dijkstra et al., 2011c; Apostel et al., 2013). Significant ¹³C recovery from the C-1 position, especially at -20°C, implies that, at such very low temperatures the pentose phosphate pathway again plays a greater role than at -5°C. This shift in metabolic pathways may reflect the need for antifreeze compounds. To confirm this interpretation, we assessed ¹³C incorporation into microbial biomass.

Liquid water in soil is a prerequisite for microbial activity (Oquist et al., 2009). At +5°C, microbial activity was not constrained by water availability or slow substrate diffusion. The enzymes responsible for organic matter decomposition also remain active at this temperature (Razavi et al., 2015, 2016). Consequently, this temperature was favorable for glucose use, as reflected by the 4- and 13-fold higher ¹³C recovery in CO₂ than at -5 and -20°C, respectively. Nonetheless, this high catabolic use of glucose-derived ¹³C with increasing temperature did not correspond to anabolic use. The highest ¹³C incorporation into microbial biomass was recorded at -5°C. The position-specific ¹³C recovery patterns in bulk soil and microbial biomass at +5°C were complementary to the metabolic fluxes observed in CO₂, showing that glucose was predominantly metabolized via pentose phosphate pathway. The dominance of this pathway at similar temperatures under moderate C supply reflects the classical metabolic C allocation observed in previous studies (Dijkstra et al., 2011c). This is triggered by the need for pentose and NADPH for biosynthesis (Fuhrer and Sauer, 2009).

Although water availability, water activity, and thermal energy are low at subzero temperatures (Davidson and Janssens, 2006; Oquist et al., 2009; Stevenson et al., 2015), recent studies demonstrated microbial growth down to -15°C in permafrost isolates (Mykytczuk et al., 2013), -20°C in permafrost cores (Tuorto et al., 2014), and the microbial metabolism limit was set at -33°C (Rummel et al., 2014). Therefore, the substantial amount of ¹³C incorporated into microbial biomass we observed at -5°C is not surprising. To survive and grow at such extremely low temperatures, microorganisms induce a suite of physiological processes which include both metabolic and biomass adjustments (Wouters et al., 2000; Drotz et al., 2010). A better understanding of these metabolic and biomass adjustments will be necessary in estimating the role of microorganisms in biogeochemical cycles of frozen soils.

¹³C incorporation into microbial biomass at -5°C was two and three times higher than at +5 and -20°C, respectively, and this incorporation made up more than two-thirds of the total glucose-derived ¹³C recovered in bulk soil (Figure 2A). In a recent study, ¹³C incorporation into microbial DNA was not detected after 1 month at subzero temperatures (Tuorto et al., 2014). This suggests that glucose utilization by microorganisms at -5°C were mainly intracellular metabolic responses to freezing and was rapidly activated. Intracellularly, for the vast majority of prokaryotic cells, contain no unbound water (Coombe, 2002). Hence, deleterious effects of freezing to these microbes, include, but are not limited to osmotic imbalance. Since water is not a limiting factor at -5°C, due to freezing point depression by increasing solute concentration in unfrozen volume (Brooks et al., 1997; Price, 2000), microorganisms invest energy in the synthesis of intracellular compounds that minimize effects of osmotic stress brought about by freeze-dehydration. This freeze-dehydration alters cell membrane ultrastructure and membrane bilayer fusion and causes organelle disruption (Mazur, 2004). Glycerol and ethanol were recently identified as abundantly synthesized compounds at -4°C (Drotz et al., 2010). The synthesis of these compounds is consistent with glucose transformation via glycolysis, which is confirmed by the high ¹³C recovery from C-4 in CO₂. These compounds act as cryoprotectants and are compatible solutes against freeze-dehydration (Block, 2003). Glycerol also interacts with water through hydrogen bonds depressing its freezing point (Lovelock, 1954). Glycerol therefore limits intracellular damage caused by freezing, hence maintaining microbial activity and CO₂ production at subzero temperatures (Drotz et al., 2010). We emphasize that conversion of glucose to alcohols (ethanol and glycerol) does not limit synthesis of other cellular compounds such as phospholipids and proteins. Nonetheless, conversion of glucose to alcohols explains the high glycolysis (C-4) and low pentose phosphate pathway (C-1) fingerprints (Figure 1C) observed at -5°C (Figure 4).

We obtained lowest ¹³C incorporation into microbial biomass at -20°C (1.3 and 3 times lower than at +5 and -5°C, respectively; Figure 2A. Nonetheless, the ratio of extracellular ¹³C to microbial biomass in soil was highest (ratio 5) at -20°C (Figure 2A). This high ratio could not be attributed to the non-metabolized glucose that remained in the soil because less than 0.4% of applied ¹³C was recovered in DOC at day 1 (Figure 3A); moreover, extracellular ¹³C recovery increased again at later time points. The high ¹³C recovery in bulk soil compared to microbial biomass at -20°C suggests that microorganisms utilized glucose for extracellular antifreeze compounds. To resist physical, environmental, and biological stress, and to survive in diverse ecological niches, microorganisms form biofilms. These biofilms are mainly composed of EPS secreted by microbes. Recently, culture-based studies revealed EPS secretion at -15°C by bacterial cells (Mykytczuk et al., 2013). In addition to cryoprotection, the high polyhydroxyl content of EPS lowers the freezing point and ice nucleation temperatures (Nichols et al., 2005; Qin et al., 2007; De Maayer et al., 2014). A key intermediate linking the anabolic pathway of EPS production and the catabolic pathway of glucose degradation is glucose-6-phosphate. In this step, the C flux bifurcates between the formation of glycolysis products and the biosynthesis of sugar monomers for EPS production via the pentose phosphate pathway (Welman and Maddox, 2003; Figure 4). Increasing glucose concentration was found to increase EPS production at low temperatures (Cayol et al., 2015). EPS production and secretion potentially explain the relevance of glycolysis and the pentose phosphate pathways, as reflected by preferential oxidation of glucose C-4 followed by C-1 positions (Figure 1D). This would account for the higher extracellular ¹³C in soil at -20°C compared to -5 and +5°C (Figure 2A). The poor extractability of EPS in water (Hughes, 1997) may explain why extracellular ¹³C was not entirely recovered in DOC at -20°C. Nevertheless, high ¹³C recovery in DOC coincided with high absolute DOC at -20°C (Figure 3A). This is consistent with the results of both Krembs et al. (2002), who found that EPS concentration in ice correlated positively with DOC, and with those of Hentschel et al. (2008), who observed the highest DOC content in forest soil at -13°C, followed by successively lower values at +5 and -3°C. This pattern, based on forest soils in a field experiment, agrees perfectly with our observation on agricultural soils in a laboratory experiment (Figure 3A), suggesting that underlying adaptation strategies are rather general. The spectroscopic properties of DOC at -13°C did not indicate lysis of microbial biomass induced by freezing (Hentschel et al., 2008), which confirm the maintenance of cellular integrity.

In addition to EPS, organisms commonly synthesize AFP to overcome freezing effects. These AFP have been identified in fungi, bacteria, plants, insects, and polar fishes (Holt, 2003). Some bacterial AFP are secreted as lipoglycoproteins or anchored on the outer cell membrane as lipoproteins, stabilizing membrane lipids at low temperatures (Kawahara, 2002). Based on adsorption inhibition, bound AFP inhibit rapid water movement between ice crystals, preventing destabilization of small ice crystal grains and thus minimizing ice recrystallization (Yu et al., 2010). For this reason, 3–8% of water remains unfrozen as thin films coating organo-mineral particles. This protects viable cells sorbed onto their surfaces from mechanical destruction due to intrusive ice crystals growing in frozen soil (Gilichinsky et al., 2003). The unfrozen water provides channels through which nutrients reach the cells and through which waste products are eliminated by diffusion (Rivkina et al., 2000). The synthesis of extracellular AFP further underlines the greater relevance of the pentose phosphate pathway (oxidation of C-1) at -20°C than at -5°C (Figure 1D). Increased pentose phosphate pathway activity signifies increased NADPH production, which is required as a source of reduction equivalents in protein and lipid synthesis (Nelson et al., 2008). Moreover, synthesis of hardly extractable extracellular AFP could also account for the fivefold higher extracellular ¹³C recovery in bulk soil than in microbial biomass at -20°C (Figure 2A).

The position-specific tracing of metabolic pathways in our study revealed that microorganisms induce a suite of temperature-dependent physiological processes to cope with freezing. This enables both catabolic and anabolic processes. These physiological changes reflect microbial resilience, enabling them to inhabit any ecological niche. Whether these responses are accompanied by a microbial community shift or solely reflect phenotypic plasticity remains to be determined.