Abstract
Farmed soils contribute substantially to global warming by emitting N2O (ref. 1), and mitigation has proved difficult2. Several microbial nitrogen transformations produce N2O, but the only biological sink for N2O is the enzyme NosZ, catalysing the reduction of N2O to N2 (ref. 3). Although strengthening the NosZ activity in soils would reduce N2O emissions, such bioengineering of the soil microbiota is considered challenging4,5. However, we have developed a technology to achieve this, using organic waste as a substrate and vector for N2O-respiring bacteria selected for their capacity to thrive in soil6,7,8. Here we have analysed the biokinetics of N2O reduction by our most promising N2O-respiring bacterium, Cloacibacterium sp. CB-01, its survival in soil and its effect on N2O emissions in field experiments. Fertilization with waste from biogas production, in which CB-01 had grown aerobically to about 6 × 109 cells per millilitre, reduced N2O emissions by 50–95%, depending on soil type. The strong and long-lasting effect of CB-01 is ascribed to its tenacity in soil, rather than its biokinetic parameters, which were inferior to those of other strains of N2O-respiring bacteria. Scaling our data up to the European level, we find that national anthropogenic N2O emissions could be reduced by 5–20%, and more if including other organic wastes. This opens an avenue for cost-effective reduction of N2O emissions for which other mitigation options are lacking at present.
Similar content being viewed by others
Main
Until the mid-twentieth century, crop production was severely limited by nitrogen, requiring farmers to recycle this element in a reactive form within their agroecosystems. This constraint is reflected in the agricultural treatise by Marcus Porcius Cato (234–143 bc) De Agri Cultura, which recommends to “save carefully goat, sheep, cattle, and all other dung”9. The invention of the Haber–Bosch process in 1908 eliminated the nitrogen constraint by producing ammonium from atmospheric nitrogen. The Haber–Bosch process was a breakthrough, saving the world from starvation10, but has also become a problem because it allowed farmers to use nitrogen in excess, with marginal economic penalties for losing nitrogen to the environment. As a result, most agroecosystems have become nitrogen-enriched and leaky, releasing ammonia to the atmosphere and nitrate to the groundwater and surface water, at scales that induce eutrophication and threaten the quality and resilience of both terrestrial and aquatic ecosystems worldwide2,11,12,13. The global scale of the problem becomes apparent when considering that the flux of reactive nitrogen into the biosphere has practically doubled since the industrial revolution, primarily owing to nitrogen produced through the Haber–Bosch process14.
Nitrogen fertilization causes emissions of the greenhouse gas N2O, both from agricultural soils themselves (direct emissions) and from the natural environments owing to the input of reactive nitrogen lost from the farms (indirect emissions). These farming-induced emissions account for substantial shares of the escalating concentration of N2O in the atmosphere since the industrial revolution1,15,16. A comprehensive analysis of global N2O emissions for 2007–201617 estimated that total direct and indirect emissions were 2.3–5.2 and 0.6–2.1 Tg N2O-N yr−1, respectively, in total accounting for >50% of the total anthropogenic N2O emissions (4.1–10.3 Tg N2O-N yr−1).
Mitigation
Reducing the anthropogenic impacts on nitrogen cycling and N2O emissions has become a major environmental challenge for the twenty-first century owing to the severity of these issues. An obvious place to start is to improve the nitrogen-use efficiency of agroecosystems by reducing their losses of ammonia and nitrate12. This can be achieved by policy instruments to induce shifts in existing farming technologies and implementation of emerging ones13,18,19,20.
Although improving nitrogen-use efficiency can reduce emissions, deliberately manipulating the soil microorganisms holds even greater potential for achieving substantial reductions. N2O emitted from soils is produced by denitrifying bacteria, denitrifying fungi, ammonia-oxidizing archaea, ammonia-oxidizing bacteria5 and abiotic chemical reactions21. Whereas ammonia-oxidizing archaea, ammonia-oxidizing bacteria and denitrifying fungi are net sources of N2O because they lack the enzyme N2O reductase, denitrifying bacteria can be either sinks, sources or both: N2O is a free intermediate in their stepwise reduction of nitrate to molecular nitrogen, NO3− to NO2− to NO to N2O to N2, catalysed by enzymes encoded by the genes nar and nap; nirS and nirK; cNor and qNor; and nosZ, respectively3. The organisms use this pathway to sustain their respiratory metabolism under hypoxic and anoxic conditions. Denitrifying bacteria are extremely diverse regarding their catabolic potential, their regulation of denitrification22,23 and their denitrification gene sets: a substantial share of denitrifying bacteria in soils have truncated denitrification pathways, lacking one to three of the four genes coding for the complete pathway23,24. This has been taken to suggest that denitrification is essentially ‘modular’ (that is, that each step of the pathway is catalysed by a separate group of organisms, rather than by organisms carrying out all of the steps of the pathway)25. The truth is probably a bit of both4,26. Of note, an organism with a truncated denitrification pathway lacking nirS and nirK is not a denitrifying bacterium sensu stricto.
Being the only sink for N2O in soils, the enzyme N2O reductase (NosZ) has been the target for recent attempts to mitigate N2O emissions from soils. An intervention that strengthens this sink will lower the N2O/N2 product ratio of denitrification and hence reduce the propensity of the soil to emit N2O into the atmosphere5,27. This can be achieved by liming to increase the soil pH: the synthesis of functional NosZ is enhanced by pushing the soil pH towards the upper end of the normal pH range of farmed soils (pH 5–7)28. As a result, liming acidified soils will reduce their N2O emissions by 10–20%, albeit with a next-to-neutral climate effect owing to the CO2 emission induced by lime application29,30.
N2O-respiring bacteria
Increasing the abundance of N2O-respiring bacteria (NRB; Box 1) could decrease the emission of N2O (ref. 31). NRB with a complete denitrification pathway can be net sinks of N2O if their denitrification regulatory networks secure earlier and/or stronger expression of NosZ than of the other denitrification enzymes6,32, or if their electron flow is channelled preferentially to NosZ (ref. 33). Their effect as N2O sinks is plausibly conditional, however, as regulation of their anaerobic respiratory pathway can be influenced by environmental conditions. By contrast, bacteria that are equipped with nosZ, but lack nirS and nirK, are more likely to be effective sinks for N2O (ref. 34). In the following, we will call them non-denitrifying NRB (NNRB) because they are unable to denitrify, sensu stricto (Box 1). NNRB are sinks for N2O in hypoxia and anoxia, unless equipped with enzymes catalysing nitrate ammonification (that is, reduction of NO3− to NH4+ via NO2−). Such NNRB organisms catalysing nitrate ammonification have been found to produce significant amounts of N2O if provided with high nitrate concentrations35; or when using Fe3+ as electron acceptor, thus inducing abiotic N2O formation by chemical reaction of Fe2+ with NO2− (ref. 21).
We know too little about the ecology and physiology of NNRB to selectively enhance their growth in situ4, but their potential as agents to reduce N2O emissions from soils is indisputable, as demonstrated by laboratory incubations of soils amended with NNRB grown ex situ36. Recently, it was suggested6 that such soil amendment can be carried out inexpensively on a large scale, by using waste from biogas reactors (digestates), destined for soils as organic fertilizers, both as a substrate and vector for NRB or NNRB. By anoxic enrichment culturing with N2O as the sole electron acceptor, these authors successfully enriched and isolated NRB with a strong preference for N2O, which could grow aerobically to high cell densities in digestates, and showed that amending soils with NRB-enriched digestates lowered the N2O/N2 product ratio of denitrification. The isolates obtained were not ideal, however, because they had genes for the entire denitrification pathway, and their catabolic capacities were streamlined for growth in digestate, not soil. In a follow-up study7, the authors designed a dual substrate enrichment strategy, switching between sterilized digestate and soil as substrates, to deliberately select for NRB and NNRB with a broader catabolic capacity and physiochemical tolerance. The enrichments became dominated by strains classified as Cloacibacterium (based on 16S rRNA gene amplicon sequencing), and the isolated strain Cloacibacterium sp. CB-01 was deemed promising: it carries the genes for reduction of NO and N2O but lacks the genes for reduction of NO3– and NO2−, thus qualifying as an NNRB (Box 1). A subsequent meta-omics analysis of the enrichments and the genome of CB-01 suggested that surface attachment and utilization of complex polysaccharides contributed to its fitness in soil8.
Here we have evaluated the ability of CB-01 to reduce N2O emission from soil, when vectored by digestate. We examined several regulatory and enzyme kinetic traits to assess its inherent strength as an N2O sink. We then tested its capacity in ‘real life’ by conducting field experiments in which soils were fertilized with digestate in which CB-01 had been grown to a high cell density. Last, we assessed the potential of this technology for reducing N2O emissions across the European Union.
The respiratory phenotype
The genome of CB-01 contains nosZII but lacks any genes coding for dissimilatory reduction of NO3− and NO2−, predicting a phenotype able to respire N2O (but neither NO3− nor NO2−), which was confirmed experimentally. In response to oxygen depletion, CB-01 reduced N2O to N2, but was unable to produce N2O from NO2−(ref. 7). The fact that it has cNor, coding for NO reductase, means that it could produce N2O from NO, but the NO kinetics indicates minor NO reductase activity7. This qualifies CB-01 as an NNRB (Box 1), and the laboratory incubation of soils fertilized with digestates containing CB-01 produced marginal amounts of N2O (ref. 7).
The capacity of a strain to reduce N2O emissions is commonly judged by a set of biokinetic parameters31, and we investigated these for CB-01, for comparison with other strains. In all experiments (unless otherwise stated), CB-01 was grown as batch cultures in GranuCult nutrient broth (Merck) containing meat peptone and meat extract, at pH 7.3 and 23 °C.
Growth yield
Based on the bioenergetics and charge separation for aerobic and anaerobic respiration of canonical denitrifying organisms, having NosZI (Box 1), the growth yield in terms of grams of cell dry weight per mole of electrons (\({Y}_{e \mbox{-} {N}_{2}O}\)) is about 60% of that for aerobic growth (\({Y}_{e \mbox{-} {O}_{2}}\))37. For CB-01, which has NosZII, \({Y}_{e \mbox{-} {N}_{2}O}\) was 85% of \({Y}_{e \mbox{-} {O}_{2}}\) (Extended Data Fig. 1a,b), which lends support to the claim that electron flow to NosZII conserves more energy (by charge separation) than that to NosZI (refs. 38,39).
Cell-specific respiration and growth rates
Measured aerobic and anaerobic respiration rates during unrestricted growth were used to estimate maximum growth rates, µmax, by nonlinear regression (Extended Data Fig. 1c,d), and the maximum rate of electron flow per cell to O2 and N2O was calculated on the basis of the measured growth yields (Vmax = µmax/Y). The estimates are \({\mu }_{max{{\rm{O}}}_{2}}\) = 0.29 h−1 (s.d. = 0.006), \({\mu }_{max{{\rm{N}}}_{2}{\rm{O}}}\) = 0.11 h−1 (s.d. = 0.001), \({V}_{{{\rm{maxO}}}_{2}}\) = 0.72 fmol O2 per cell per hour, \({V}_{{{\rm{maxN}}}_{2}{\rm{O}}}\) = 0.66 fmol N2O per cell per hour. In terms of electron flow rates per cell, we get \({V}_{{{\rm{m}}{\rm{a}}{\rm{x}}{\rm{e}}-{\rm{O}}}_{2}}\) = 2.9 fmol of electrons to O2 per cell per hour, \({V}_{{{\rm{m}}{\rm{a}}{\rm{x}}{\rm{e}}-{\rm{N}}}_{2}{\rm{O}}}\) = 1.3 fmol of electrons to N2O per cell per hour. This shows that CB-01 slows down its respiratory metabolism by about 50% when switching from aerobic to anaerobic respiration.
Oxygen repression of N2O respiration
N2O respiration under oxic conditions has been reported for several organisms31. Such aerobic N2O respiration would be desirable for an organism to effectively scavenge N2O in soil, but we found no evidence for this in CB-01: aerobically raised cells monitored as they depleted oxygen did not initiate N2O respiration before the oxygen concentration reached below 1–2 µM, whereas cells previously exposed to anoxia (hence with intact NosZ enzymes) initiated N2O respiration at 4–6 µM O2 (Extended Data Fig. 2).
Affinity for O2 and N2O
It is commonly assumed that an organism’s ability to effectively mitigate N2O emissions depends on its affinity for N2O. We determined the apparent half-saturation constant for O2 and N2O reduction in CB-01 by nonlinear regression of rates per cell versus concentrations of the two gases in the liquid, and found \({K}_{{{\rm{mO}}}_{2}}\) = 0.9 µM O2 (s.e. = 0.27) and \({K}_{{{\rm{mN}}}_{2}{\rm{O}}}\) = 12.9 µM N2O (s.e. = 1.2; Extended Data Fig. 3). The relatively low \({K}_{{{\rm{mO}}}_{2}}\) was expected as the genome of CB-01 contains genes coding for cbb3-type high-affinity cytochrome c oxidases8.
Comparing the N2O sink strength
To compare CB-01 with other organisms as a sink for N2O in soil, we have summarized the biokinetic parameters for various N2O-respiring organisms by plotting their ‘catalytic efficiency’ (Vmax/Km) against their Vmax on a cell dry weight basis (Fig. 1b). This suggests that CB-01 is far from being the best among N2O-respiring organisms: it is on par with the average of others with respect to Vmax, which is a measure of the N2O sink strength at high N2O concentrations (»Km = 12.9 µM N2O ≈ 389 ppmv in the gas phase at 15 °C), but it scores poorly at low N2O concentrations (Vmax/Km for CB-01 is only 3% of the average for the others). The apparent bet-hedging (Fig. 1a), explored in more detail in several experiments (Extended Data Fig. 4) would clearly add to its inferiority as an N2O sink. However, the bet-hedging was clearly depending on the growth medium: when growing in digestate, all cells switch to anaerobic respiration in response to oxygen depletion (Extended Data Fig. 5d–g).
As judged by kinetics of N2O respiration in pure culture, CB-01 scores strikingly low compared to other N2O-respiring organisms as a sink for N2O: the kinetics of N2O respiration in response to O2 depletion indicate bet-hedging (that is, that only a fraction (FNosZ) of the cells express NosZ and start growing by N2O respiration after O2 depletion). a, The phenomenon for a single vial. Measured O2 and N2O (triangles), and simulated values (solid lines), using a simplified version of the bet-hedging model of ref. 51, with FNosZ = 0.03. Of note, the decline of N2O concentrations before about 18 h is due to sampling loss. The yellow line shows the simulated cell density, and the dashed black line shows simulated N2O for FNosZ = 1. The inset shows measured and simulated total electron flow in the vial. Two replicate vials showed very similar kinetics, and their FNosZ, estimated by model fitting, were 0.032 and 0.039. b, A condensed comparison of CB-01 with other N2O-respiring organisms regarding its capacity to scavenge N2O. Here we have plotted Vmax/Km against Vmax (mmol N2O per gram of cell dry weight per hour) for CB-01 and a range of other organisms with NosZI and NosZII, as measured by others (see Extended Data Table 1 for details and citations). The comparison shows that CB-01 is close to the average with respect to Vmax, but its Vmax/Km ratio is very low owing to the low apparent affinity for N2O (Km = 12.9 µM N2O).
Effects of CB-01 on N2O emissions
CB-01 was found to grow exponentially by aerobic respiration in autoclaved digestate, reaching a cell density of about 109 cells per millilitre after 20 h. At this point, about 1% of the organic C in the digestate had been consumed, and the growth rate declined gradually, plausibly owing to depletion of the most easily available substrate components reaching a final density of about 6 × 109 cells per millilitre after 2 days, as judged by oxygen consumption (Extended Data Fig. 5a–c), and growth yield based on quantitative PCR (qPCR) quantification of CB-01 cells (Extended Data Fig. 1b).
We conducted three outdoor experiments in which the soils were fertilized with digestates in which CB-01 had been grown to about 6 × 109 cells per millilitre. Control treatments were fertilized with the same digestate, in which the CB-01 cells had been killed by heat (70 °C), thus securing practically identical N and C availability in the soils with and without metabolically active CB-01 cells. This type of control treatment is crucial for correctly assessing the effect of CB-01 metabolism, as the incorporation of any organic material will induce transient peaks of N2O emissions. Experimental details are provided in the Methods.
The first field experiment demonstrated that the initial peak of N2O flux induced by the fertilization with digestate was practically eliminated by CB-01 (Fig. 2), and that CB-01 continued to have a strong effect throughout; a second peak in N2O emission induced by precipitation (day 12) was reduced by 51%; and the later emission peaks induced by re-fertilization with digestate without CB-01 (indicated by arrows) were reduced by 31, 67 and 46%.
N2O flux from buckets with soil throughout 90 days after fertilization (14 July 2021) with digestate (11 l m−2) in which the NNRB strain Cloacibacterium sp. CB-01 had been grown to about 6 × 109 cells per millilitre, quantified by qPCR with primers specific for CB-01. Control buckets were fertilized with the same digestate in which CB-01 had been killed by heat (70 °C for 2 h). The buckets were sown with ryegrass (Lolium perenne), and the soil moisture content was sustained by daily water additions during the first 10 days. Buckets were re-fertilized with a lower dose of autoclaved and pH-adjusted digestate without CB-01 (4.6 l m−2) after 19, 33 and 89 days. The top panel shows N2O flux measured by the dynamic chamber method52 with 3 min enclosure time, operated by a field robot (Supplementary Fig. 1). The insert is a rescaled plot for day 89–93. The emissions are shown as single dots for each enclosure, and with a floating average for each treatment (solid lines, n = 8 replicate buckets for each treatment, calculated by a Gaussian kernel smoother). The lower panels show the average soil temperature (at 0–5.5 cm depth) and water-filled pore space (WFPS) from n = 4 loggers (s.d. of the mean is shown as lighter coloured ribbons). The fluxes show clear diurnal fluctuations, driven by temperature, and transient peaks in response to a rain event (day 12) and in response to re-fertilization (marked by arrows). The percentage reduction of N2O emissions (cumulated flux) by CB-01 was calculated for selected periods, shown by the green arrows with 95% confidential intervals (Methods). The additional control buckets receiving water instead of digestates emitted negligible amounts of N2O (result not shown).
Given the number of CB-01 cells added with the digestate (6.6 × 1013 cells per square metre of soil surface), and the Vmax = 0.6 fmol N2O per cell per hour (Extended Data Fig. 1), the potential N2O consumption rate, if all the added CB-01 cells were respiring N2O at maximum rate, is 1.1 g N2O-N m−2 h−1. The peak N2O flux 1–2 days after fertilization was reduced by about 85 mg N2O-N m−2 h−1, which is about 8% of the estimated potential. For the subsequent peaks of N2O flux, the apparent N2O respiration by CB-01 (that is, the reduction of the flux) was ≤4 mg N2O-N m−2 h−1, which is ≤0.36% of the initial potential. This decline in apparent N2O respiration by CB-01 was plausibly a result of two factors: a gradually declining rate of N2O provision by the indigenous microbiome, and a gradually declining number of CB-01 cells.
One would expect that the effect of CB-01 as an N2O sink would be marginal in periods with low emissions: low emissions are due to low water-filled pore space (that is, drained soil), low respiration rate (limited by available organic C substrates) or both, resulting in marginal hypoxic and anoxic volumes within the soil matrix40. Under such conditions, the primary source of N2O emission could be nitrification41, and CB-01 as an N2O sink would be confined to the remaining hypoxic microsites. Inspections of the relationship between the effect of CB-01 and the N2O emissions in the control soil (that is, with dead CB-01) lend some support to this: although CB-01 reduced the emissions even for periods with modest emissions, the effect was clearly strongest in periods with high emissions (Fig. 3).
Emissions from soil fertilized with digestate containing live Cloacibacterium sp. CB-01 plotted against the emissions from soil fertilized with digestate containing dead CB-01 cells (same data as in Fig. 2). a–c, The results for the low-emission (<400 μg N2O-N m−2 h−1; part a), intermediate-emission (<4,000 μg N2O-N m−2 h−1; part b) and high-emission (>4,000 μg N2O-N m−2 h−1; part c) ranges. d, A log-scaled plot of the ratio between emissions from soil with live and dead CB-01 plotted against the emission from soil fertilized with digestate containing dead CB-01.
We reasoned that the capacity of CB-01 to reduce N2O emissions could be influenced by soil type. Soil pH is plausibly crucial because the synthesis of functional N2O reductase is increasingly impeded by declining pH within the range 4–7, both in CB-01 (ref. 7) and most other NRB5. Soil organic carbon content (SOC) could also have an impact. This is because the abundance of CB-01 relative to the abundance of indigenous N2O-producing bacteria would be inversely related to SOC, as the abundance of indigenous bacteria in soil is directly related to SOC42. To explore this, we replicated the bucket experiment (Fig. 2), but with four different soils spanning a range of pH levels and including a soil with very high organic carbon content (Fig. 4 and Extended Data Fig. 6).
The measured emission after application of digestates with and without CB-01 to four different soils (17 September 2021). The organic carbon contents of the soils were 15.8% (organic-rich clay loam of pH 5.26), 3.21% (neutral-pH clay loam of pH 6.70), 0.75% (sandy silt soil of pH 4.15) and 3.23% (low-pH clay loam of pH 4.50) of dry weight. The pH(CaCl2) before fertilization with digestate is given in the panels. The emissions are shown as single dots for each enclosure, and with a floating average for each treatment (solid lines, n = 6 replicate buckets for each treatment) as in Fig. 2. The percentage reduction of N2O emissions (cumulated flux) by CB-01 is shown by the green arrows with 95% confidential intervals (Methods).
The emissions were low compared to those in the first experiment, plausibly owing to lower temperatures (September versus July), but CB-01 significantly reduced the emissions from all four soils. The strong effect in the acidic sandy silt soil (pH 4.15) was unexpected, as CB-01 proved unable to reduce N2O at such low pH (ref. 7). However, the incorporation of digestate in this soil increased the pH(CaCl2) of the sandy silt soil by more than one pH unit (Extended Data Fig. 6), reflecting its weak buffer capacity. Most probably, the CB-01 embedded in the digestate experienced an even higher local pH (pH of the digestate was 7.3). The results for the three clay loam soils show a stronger effect of soil pH: CB-01 had a clearly stronger effect in the neutral-pH clay loam (pH 6.7) than in the two more acidic clay loams (low-pH clay loam of pH 4.5; organic-rich clay loam of pH 5.26).
Finally, we scaled up to a field plot experiment, fertilizing 0.5-m2 plots with digestate with live and dead CB-01, mixed into the upper 10-cm layer of the soil as in the bucket experiments. The experiment was conducted on field plots that had been limed with 2.3 kg m−2 of dolomite in 2014, with an average pH(CaCl2) = 6.13 (s.d. = 0.10). The high emissions during the first 4 days (Fig. 5) show diurnal variations, peaking when the soil temperatures reach their maximum, and a substantial effect of CB-01. Subsequent emissions, measured at low frequency throughout 280 days, were much lower and the effect of CB-01 was not statistically significant, albeit with a wide confidence interval. The very low soil temperature could be the reason for the meagre effect.
The 0.5-m2 field plots with clay loam of pH 6.13 were fertilized by mixing digestate into the upper 10 cm (20 August 2022), with live and dead CB-01 as in previous experiments (n = 6 replicate plots for each treatment). The top panel shows emissions throughout 290 days, and the insert shows emissions during the first 10 days. The percentage reduction of N2O emissions (cumulated flux) by CB-01 for the periods 0−10 and 10–290 days is shown by the green arrows with 95% confidential intervals (Methods). The lower panels show soil temperature and WFPS for n = 4 loggers, with ribbons representing the s.d. of the mean.
Survival in soil
Soil microbiome engineering by inoculation is an emerging field, promising new possibilities in enhancing agricultural efficiency and sustainability43. It is challenging, however, because inoculants are invariably found to die out rapidly, plausibly due to a multitude of abiotic and biotic barriers impeding establishment44. CB-01 was obtained through a dual substrate enrichment technique aimed at isolating organisms capable of withstanding the abiotic challenges of soil7. However, this selection process did not account for the biotic barriers that organisms may encounter in soils, such as competition for resources, antagonism and predation, as highlighted previously45.
To assess the ability of CB-01 to survive in soil, we used qPCR with specific primers to measure the abundance of CB-01 genomes in soil (Methods) throughout the long-term field bucket experiment (Fig. 2), and throughout a laboratory incubation of soil amended with digestate with CB-01 (Methods); the results are shown in Fig. 6. During the laboratory incubation, there was a fast first-order reduction in abundance during days 3–7, and a much slower first-order reduction thereafter. By contrast, the abundance was sustained at a high level throughout 90 days in the field buckets, albeit gradually declining. The sustained CB-01 population in the bucket experiment explains why the effect on the N2O emission was sustained (Fig. 2).
The abundance of CB-01 was assessed by qPCR (Methods). The panel shows the genome abundance in the long-term field bucket experiment (Fig. 2) and in the laboratory incubation experiment (Methods). In the field bucket experiment, additional digestate (without CB-01) was incorporated 2 days before each soil sampling for qPCR. A single dot represents an individual soil sample (n = 8), and the line is the fitted exponential function Nt = N0e−d×t, in which Nt is the abundance at time t, and d is the apparent first-order death rate (estimated half-life T1/2 = ln(2)/d). For the laboratory incubation, three phases can be recognized: an initial apparent growth during the first 2–3 days, followed by a rapid first-order decline during the subsequent 4–5 days, and a slow first-order decline thereafter. Of note, the measured CB-01 genome abundance in the field plots after 280 days indicated similar average first-order death rates (0.02 per day, T1/2 = 34 days; Extended Data Table 2).
The discrepancy between the field and the laboratory experiments demands a scrutiny. In the field bucket experiment, digestate (not inoculated with CB-01) was applied three times during the course of the experiment, with soil sampling for quantification of CB-01 abundance conducted 2 days after each application. As digestate is a suitable substrate for CB-01, growth of CB-01 in response to each dose could contribute to the sustained population.
Another factor could be protozoal grazing, which was plausibly more intense in the laboratory incubations than in the field experiment, owing to the higher soil moisture content at the time of CB-01 incorporation. In the laboratory experiment, the digestate with CB-01 was dripped onto soil that was already very wet (0.53 ml per gram of soil dry weight) and retained this high soil moisture throughout. In the field bucket experiment, CB-01–digestate was harrowed into relatively dry soil (0.34 ml per gram of soil dry weight), and the soil remained modestly moist throughout (Fig. 2). There is ample evidence that low soil moisture protects a bacterial inoculum against protozoal grazing, ascribed to increasing tortuosity, and localization of bacteria in small pores that are inaccessible to the protozoa46. Although we recognize that this is a speculative explanation, it warrants further experimental investigation owing to the potential practical implications.
A legitimate concern would be that the heavy inoculation with CB-01 could affect the indigenous microbiota47. We investigated this by analysis of 16S rRNA gene amplicons, excluding the operational taxonomic unit that circumscribed CB-01 (Methods), and found that the digestate itself had a transient impact (with or without live CB-01), but we were unable to discern any consistent difference between the treatments with live versus dead CB-01, which both converged towards the composition of pristine soil (Extended Data Fig. 7).
Laws and regulations for the use of inoculants vary from country to country, but all are likely to forbid the use of NNRB if they carry genes for antibiotic resistance or pathogenicity. We were unable to identify such genes in CB-01 (Methods).
Extrapolating to national emissions
To assess the potential emission reductions by NNRB compared with other available techniques such as optimized N fertilization and nitrification inhibitors, we estimated emissions for Europe 2030 with the greenhouse gas and air pollution interactions and synergies (GAINS) model48,49 (Methods).
Consistent with using a uniform emission factor in GAINS (from the Intergovernmental Panel on Climate Change (IPCC)50) of 1% of N applied to be emitted as N2O, a uniform factor for emission reductions was also assumed. From the experiments, we conclude that 60% of emission reductions due to NNRB may be considered a conservative estimate. In Extended Data Table 3, emission reductions are shown by European country for a 2030 scenario if emissions from the application of liquid manure alone are reduced by 60%. All other anthropogenic emissions have been left unchanged. Under these assumptions, the total anthropogenic N2O emissions from Europe decrease by 2.7% owing to NNRB being introduced and applied to all liquid manure systems. This figure is higher in countries that have a high share of liquid manure systems in their agriculture; hence, it increases to 4.0% for EU27 (27 EU member countries).
Ongoing work explores the possibility to extend the technology by growing NNRB in all types of organic waste used to fertilize soils, and by combining the application of mineral N fertilizers with incorporation of NNRB-amended organic wastes. This requires new strains, technologies and investments, but with a great potential, reducing EU27 agricultural emissions by a third (31%; Extended Data Table 3).
It needs to be pointed out that an emission reduction of 60% as derived here for NNRB is much larger than emission reductions typically reported for N2O abatement measures. GAINS, for example, assumes nitrification inhibitors to be able to reduce emissions by as much as 38%, and high-tech mechanical fertilizer-saving technologies (‘variable rate application’) to be able to save only 24% of the emissions48.
Future development
This study presents a proof of concept demonstrating a feasible utilization of NNRB to curb N2O emissions from farmland. By using organic waste as substrates and vectors, massive soil inoculation is achieved, which can secure reduced N2O emissions throughout an entire growth season, despite a gradually declining NNRB abundance. To ensure the robustness and versatility of this biotechnology, we will need an ensemble of new NNRB strains, capable of thriving in waste materials beyond digestates. New NNRB strains will probably vary regarding their ability to tolerate abiotic and biotic stress factors present in the soil. The dual substrate enrichment technique7 selects for strains tolerant of abiotic, but not biotic, stress. Consequently, innovative techniques are necessary for selecting strains that tolerate the biotic stress.
Methods
Robotized batch cultivations for respiratory phenotype
NNRB have attracted much interest recently as net sinks for N2O in soils, potentially curbing N2O emissions4,31. NNRB strains vary grossly in their apparent capacity to act as N2O sinks, assessed by determining their biokinetic parameters: NNRB strains are commonly assumed to be strong N2O sinks if they have strong affinity (low apparent Km) for N2O and a high maximal rate of N2O reduction (Vmax), or simply a high catalytic efficiency (that is, a high Vmax/Km)38. Another desirable, albeit speculative, feature would be to reduce N2O under oxic or at least hypoxic conditions53.
To assess Cloacibacterium sp. CB-01 along these criteria, we conducted in-depth investigations of its respiratory phenotype by batch culturing in the robotized incubation system designed and described previously54,55, with the OpenLAB CDS 2.3 software for GC data acquisition (Agilent). The system hosts up to 30 parallel stirred batch cultures (normally 50 ml) in 120-ml gas-tight serum vials (crimp-sealed with butyl rubber septa) with a He atmosphere (with or without N2O and O2), which are sampled frequently for measuring the concentrations of O2, N2, N2O, NO and CO2 in the headspace. Robust routines are established for calculating the rates of production and consumption of all the gases (taking sampling loss and leakage into account), and for calculating gas concentrations in the liquid as a function of measured gas concentrations in the headspace and the rate of transport between liquid and headspace. These routines are included in a spreadsheet that is publicly available, including a set of instruction videos56. The system has been used in numerous investigations of the respiratory phenotypes of denitrifying bacteria6,7,33,57,58,59,60,61,62.
To enable refined analyses of the respiratory phenotype of CB-01, we initially determined the cell dry weight (femtograms per cell), and the growth yields for aerobic (\({Y}_{{{\rm{O}}}_{2}}\), cells per mole of O2) and anaerobic (\({Y}_{{{\rm{N}}}_{2}{\rm{O}}}\), cells per mole of N2O) respiration by measuring the cell yields in batches provided with various amounts of O2 and N2O. This enabled inspection of the cell-specific respiration rates (fmoles per cell per hour) throughout subsequent batch incubations, based on measured rates (moles of O2 and N2O per vial per hour) for each time interval between two gas samplings, and the estimated cell number in the vial for the same time interval (=Nini + \({Y}_{{{\rm{O}}}_{2}}\) × cumO2 + \({Y}_{{{\rm{N}}}_{2}{\rm{O}}}\) × cumN2O, in which Nini is the initial number of cells at time 0, and cumO2 and cumN2O are the cumulated consumption of the two gases). The cell-specific rates calculated this way allowed an analysis of the affinity for O2 and N2O by plotting cell-specific rates of O2 and N2O against the concentrations of the two gases in the liquid as the cultures depleted the gases, and fitting the Michaelis–Menton function to these data (least squares). Batch cultures provided with both N2O and O2 in the headspace were monitored as they depleted O2 and switched to respiring N2O, thus determining the critical concentration of O2 (in the liquid) at which the cells started to respire N2O. The kinetics of electron flow throughout such transitions from aerobic to anaerobic respiration were used to assess the fraction of cells expressing N2O reductase in response to O2 depletion, using a simplified version of the model developed previously60.
All phenotype experiments were conducted at 23 °C. The medium used was GranuCult nutrient broth (product number 1.05443, Merck): 8 g l−1, containing meat peptone and meat extract, pH-adjusted to 7.3 with NaOH. Additional experiments were conducted with autoclaved digestate (aerated and pH-adjusted to 7.3, as described below).
Culturing CB-01 in digestate for field experiments
For each field experiment, fresh digestate was collected from a wastewater treatment plant close to Oslo (VEAS), described in ref. 6. Averaged values of the quality parameters for the period of digestate collection were: dry matter content = 3.97 wt% (s.d. = 0.16), ignition loss of dry matter = 55.6% (s.d. = 2), pH = 7.72 (s.d. = 0.07) and NH3 + NH4+ = 1.71 g N l−1 (s.d. = 0.12).
Before cultivation of CB-01, the digestate was heat-treated, aerated and pH-adjusted. For the field bucket experiments, the digestate was autoclaved (121 °C for 20 min), and then sparged with air (while stirred) for 48 h to secure chemical oxidation of Fe2+ to Fe3+, and then autoclaved again. Oxidation of Fe2+ by air sparging was considered necessary to avoid abiotic oxygen consumption, as the digestate had high concentrations of Fe2+ originating from the Fe3+ used as precipitation chemicals in the primary wastewater treatment, and reduced to Fe2+ in the anaerobic digesters6. The sparging caused the pH to increase to 9.4 owing to the removal of CO2, requiring a final pH adjustment to 7.3 (with HCl). The same procedure was used for the field plot experiment, except that autoclaving was replaced by heat treatment: 70 °C for 4 h.
CB-01 was then grown aerobically in the pretreated digestates, inoculated to an initial cell density of about 5 × 107 cells per millilitre, which were stirred and sparged with sterile air (filtered) at 23 °C. To monitor the growth of CB-01, we transferred subsamples of each batch (after inoculation) to 120-ml vials (50 ml per vial) with Teflon-coated magnetic stirring bars, which were placed in the incubation robot system for monitoring the O2 consumption (Extended Data Fig. 5a–c).
Field experiments
Emissions of N2O in all outdoor experiments were monitored by the ‘dynamic chamber’ technique52,63, operated by an autonomous field flux robot described previously64, and shown in detail in Supplementary Fig. 1.
Field bucket experiments
Soils for the bucket experiments were collected from agricultural fields in southern Norway, spanning a range of soil characteristics. The acid sandy silt soil (S) was taken from an agricultural field in Solør, Norway, dominated by fluvial sandy silt soils. The clay loam soils L, I and N were from different plots within a liming experiment near the Norwegian University of Life Sciences (59° 39′ 48.2″ N 10° 45′ 44.8″ E), limed in 2014 (ref. 41): the low-pH clay loam (L) received no lime, the intermediate-pH clay loam (I) was limed with 2.3 kg m−2 of dolomite, and the neutral-pH clay loam (N) was limed with 3 kg m−2 of finely ground calcite. Soil O was a clay loam soil from the same area as L, I and N (hence, with similar mineral components), but with a much higher content of organic C because it had been a wetland before cultivation. The soil characteristics are listed in Extended Data Fig. 6.
The soils used in the bucket experiments (S, L, N and O) were sieved (10 mm) in moist conditions and mixed thoroughly before filling into the buckets. The conically shaped buckets (height = 21.5 cm, top diameter = 23.5 cm, bottom diameter = 21.5 cm) had a total volume of 8.6 l. An approximately 1-cm layer of gravel (4–8 mm diameter) was placed at the bottom, covered with a nylon fibre cloth to prevent eluviation of the soil by drainage. For soils S, L and N, 8 kg soil dry weight was filled into each bucket, packed by thumping the bucket on the ground until the soil had reached a bulk density of 1 kg l−1. For the organic-rich clay loam soil, each bucket was filled with only 5.92 kg soil dry weight, reaching a bulk density of 0.74 kg l−1 after being packed to 8 l. The soil surface area of the buckets was 0.043 m2.
To secure equal initial amounts of NO3 m−2 for all soils, we mixed an amount of KNO3 to each soil to reach a level of 12 g N m−2 soil surface = 516 mg NO3-N per bucket (soil surface area = 0.043 m2). Digestate (480-ml per bucket = 11 l m−2 soil surface area) was mixed into the top ≈10 cm of the soil by ‘harrowing’, using a small hand-held rake. We used autoclaved digestates in which CB-01 had been grown to about 6 × 109 cells per millilitre, and as the control treatment we heat-treated this digestate (70 °C, 2 h), which effectively killed the CB-01 cells (tested by measuring respiration, results not shown). As an additional control treatment, buckets received water alone. The density of CB-01 cells per soil surface area immediately after application was 6.6 × 1013 cells m−2. The cell density in the upper 10 cm of the soil was about 6 × 108 cells per gram of soil dry weight for the soils S, L and N (bulk density = 1 kg l−1), and about 8 × 108 g−1 for soil O.
The buckets were placed on 1-m2 Plexiglass plates (1.5 mm), to avoid gas exchange with the soil below. The soil moisture (volumetric water content, m3 m−3) and temperature (°C) in the upper 5.5 cm of the soil were monitored by four Teros 11 sensors, connected to an EM50 logger (Meter Group). Emissions were measured by field flux robot, lowering the chambers over the buckets (Supplementary Fig. 1g).
In the first bucket experiment, using only soil N (Extended Data Fig. 6), starting on 14 July 2021, ryegrass (L. perenne) was sown the day after the incorporation of the digestate, and the emissions were monitored for 90 days. Within this time span, we added 200 ml autoclaved and pH-adjusted digestate (4.6 l m−2) without CB-01 three times (after 19, 33 and 89 days), to induce transient bursts of N2O emission. By the end of each burst of N2O emission induced by applying digestates, the upper 10 cm of the soil was sampled with an auger (diameter 1 cm) and stored in the freezer (−4 °C) until DNA extraction and subsequent molecular work. The auger was washed and sterilized with 70% ethanol between each sampling.
In a follow-up bucket experiment, all soils were included and monitored for 10 days, with no re-fertilization. Soil sampling was carried out after the first peak of N2O emissions, as described for the 90-day bucket experiment.
The digestate application’s influence on soil pH was tested in the laboratory by mixing soil with the same type and amount of digestate as applied to the 0–10-cm soil layers of the field buckets (0.11 ml per gram of soil) ±50% to show the potential pH in pockets with higher or lower than average concentration of digestate. Water was added (if needed) together with digestate to reach the same water-filled pore space (%) as in the field bucket experiment. The most prominent increase in soil pH was seen in the sandy silt soil (Extended Data Fig. 6), reflecting its low buffer capacity due to low content of clay and organic material (Extended Data Fig. 6), both known to be crucial factors determining the buffer capacity of soil65.
Field plot experiment
We established small (0.5 m2) test plots within larger field plots (8 m × 3 m) of a soil liming experiment (limed in 2014) on clay loam soil41,66 and re-limed with 174 g dolomite per square metre in 2019. We used the plots with soil I (Extended Data Fig. 6) that were previously limed with dolomite to pH(CaCl2) = 6.13 (s.d. = 0.10), and within each of the six replicate plots, we established two 0.7 m × 0.7 m test plots side by side (distance = 30 cm), fertilized with autoclaved digestate in which CB-01 had been grown to a cell density of about 6 × 109 cells per millilitre. We applied 4.5 l digestate per plot (= 9 l m−2), which was mixed into the upper ≈10 cm of the soil by a hand-held cultivator. The initial density of CB-01 was 5.4 × 1013 cells per square metre. If distributed throughout the soil layer that was sampled for analyses (0–10 cm depth = 125 kg soil dry weight per square metre, assuming a bulk density of 1.25 kg l−1), the initial cell density in the soil would be 4.3 × 108 cells per gram of soil. Soil samples for determining CB-01 abundance were taken from each plot (three replicate samples) before incorporation of digestate with CB-01, 9 days later, and after 10 months. The soil samples were stored in the freezer (−20 °C) until DNA extraction and following quantification by PCR.
The 0.5-m2 test plots were situated along the boardwalk for the autonomous field flux robot, which was used to monitor the N2O emissions (Supplementary Fig. 1f).
Calculations of emissions and statistical analyses
From the slope of the N2O regression lines (Supplementary Fig. 1e), the flux of N2O is calculated by the equation
in which \({q}_{{{\rm{N}}}_{2}{\rm{O}}}\) is the flux of N2O (mol m−2 s−1), a is the slope of the regression line (ppm s−1), h is the height (that is, the volume divided by the ground surface area) of the chamber (m), p is the pressure (Pa), R is the universal gas constant (J mol−1 K−1) and T is the temperature (K).
For graphic presentation of the emissions, we used the Gaussian kernel smoother67 to plot floating averages for each treatment (solid curves) together with individual measurements (as dots; Figs. 2, 4 and 5).
Cumulated N2O emissions over a period of time are approximated by using the trapezoidal rule on the estimated fluxes \((\int {q}_{{{\rm{N}}}_{2}{\rm{O}}}(t){\rm{d}}t\approx \sum ({q}_{{{\rm{N}}}_{2}{\rm{O}}}\left({t}_{i}\right)+{q}_{{{\rm{N}}}_{2}{\rm{O}}}\left({t}_{i+1}\right))({t}_{i+1}-{t}_{i})/2)\). This was carried out for each individual bucket and field plot.
The field plot experiment yielded paired data—six pairs (Xi, Yi), i = 1 ... 6, in which Xi are cumulated emissions from plots treated with NNRB, and Yi are cumulated emissions for control plots. This gives six ratios Ri = Xi/Yi. Confidence intervals for the mean of the ratios, 1/6 ΣRi, for two time periods were made with a Student’s t distribution (assuming that the ratios were normally distributed). These confidence intervals were similar to confidence intervals found by the Fieller method for ratios of paired data and also by simple nonparametric bootstrapping68.
As the field bucket experiments did not yield paired data, flux reduction statistics are calculated as ratios of means, rather than means of ratios, of cumulated fluxes. Confidence intervals of these ratios were made by the Fieller method for unpaired data69 and by simple nonparametric bootstrapping (the results were similar). The 95% coverage of the Fieller confidence intervals was tested by numerical simulations and a bootstrap-calibration of the confidence level was made, with negligible effects on the confidence intervals.
The plots in Figs. 2, 4 and 5 were prepared using the packages Tidyverse (v2.0.0)70, Pracma (v2.4.2)71, ggbreak (v0.1.2)72, patchwork (v1.1.3)73 and scico (v1.5.0)74, in the R Studio software (v4.3.2)75. Colours used in the figures are, in general, from the scientific colour maps as described in ref. 76. The Fieller and bootstrap confidence intervals were calculated using Python (v3.11.5)77 with Scipy (v1.11.2)78 and Pandas (v2.1.1)79, and Julia (v1.9.3)80.
Tracing CB-01 in digestate and soil
To quantify CB-01 cells in digestate and soil, we used qPCR with primers specific to members of the genus Cloacibacterium developed previously81. The primers 5′-TATTGTTTCTTCGGAAATGA-3′ (Cloac-001f) and 5′-ATGGCAGTTCTATCGTTAAGC-3′ (Cloac-001r) target a region of the 16S rRNA gene.
DNA was extracted with the DNeasy PowerSoil Pro Kit (Qiagen) according to the manufacturer’s protocol, except for the first step: bead beating of the cells was carried out at 4.5 m s−1 for 45 s in a FastPrep-24 (MP Biomedicals), instead of a vortex. To measure the concentration of DNA in the extract, we used a broad-range or high-sensitivity Qubit dsDNA Assay Kit (Thermo Fisher Scientific), depending on the expected concentration. The number of CB-01 16S rRNA gene copies in extracted DNA was quantified using a CFX96 Touch Real-Time PCR Detection System (Bio-Rad), running for 15 min at 95 °C followed by 40 cycles of denaturation (30 s at 95 °C), annealing (30 s at 55 °C) and elongation (45 s at 72 °C). The final concentration of the master mix contained 0.2 µM of each primer (Cloac-001f and Cloac-001r), and 1× HOT FIREPol EvaGreen qPCR Supermix (Solis BioDyne).
For calibration, we used DNA-extracted suspensions of washed cells containing 103, 104, 105, 106, 107 and 108 cells per millilitre, resulting in 2.4 × 101–2.4 × 106 16S templates per PCR tube (taking dilution into account, and the fact that each genome of CB-01 contains three 16S rRNA genes). Results from the qPCR were analysed using the CFX Maestro 1.1 software (v4.1.2433.1219 from Bio-Rad). To enable the use of the Cq values to estimate copy numbers, we used the generalized reduced gradient solver in Excel to fit the model (equation (1)) to the data:
in which N is the initial number of 16S rRNA gene templates in the PCR tube, NT is the number of amplicons per tube needed for signal detection (above background), e is the efficiency of the PCR amplification and Cq is the number of cycles needed for detection of a signal. The fitted parameters were NT = 7.68 × 1010 copies per tube and e = 0.85 (85% efficiency).
An independent dataset was provided by running qPCR with the same primers on extracted DNA from suspensions of unwashed CB-01 cells (in nutrient broth) with densities 104, 105, 106, 107 and 108 cells per millilitre. The log10 values of cell densities estimated by the Cq values were on average 104% of the expected value, with a standard deviation of 6%.
When using qPCR to estimate the CB-01 abundance in soil and digestate, inhibition of the polymerase can result in too high Cq numbers, hence resulting in underestimation of the gene abundance82. To investigate this, we spiked the different soils and the digestate with 109 CB-01 cells per gram of soil dry weight and per millilitre of digestate, respectively, extracted DNA from 0.2 g soil and 0.2 ml digestate, and eluted to a 50-µl DNA solution for each material, which was then diluted in tenfold steps from 0 (undiluted) down to 1/107. The results show a reasonable fit between model (predicted) and measured Cq values for all materials if diluting the extracted DNA to ≤1/10, except for the intermediate-pH clay loam (pH(CaCl2) = 6.13), which required dilution to ≤1/100 to eliminate inhibition (for further details, see Supplementary Fig. 2).
The result was used to approximate the lower limit for detection of CB-01 in soils and digestate: a cautious upper limit for Cq values to be trusted is 40 (that is, 34 templates per PCR tube; equation (1)). The polymerases were evidently inhibited by using undiluted DNA in the reaction (Supplementary Fig. 2); hence, a 1/10 dilution of the extracted DNA is needed for all soils except soil I, for which 1/100 dilution is required. This means that the PCR tube can maximally be loaded with DNA from 0.8 mg soil (0.08 mg for soil I) and 0.8 µl digestate. This implies a limit of detection around 4.3 × 104 templates per gram of soil (4.3 × 105 for soil I owing to dilution to 1/100) and per millilitre of digestate, or 1.4 × 104 CB-01 genomes per gram of soil and per millilitre of digestate (as the genome contains three copies of the 16S rRNA gene).
The real limit of detection for a CB-01 inoculum in soil and digestate could be higher than this, if indigenous genes are amplified with the primers. This was tested by running PCR on soil and digestates that had not been spiked with CB-01, along with analysing spiked samples in various experiments. The results are summarized in Supplementary Fig. 2. As there were several tubes with a negative result (Cq > 40), average values cannot be calculated. A cautious judgement would be that the ‘background’ PCR signal of the soil is Cq = 39–38, which is equivalent to 67–107 templates per PCR tube, or 21–36 CB-01 genomes per tube. For all soils except I, we used the Cq values for the PCR tubes loaded with 1/10 dilutions, which were thus loaded with DNA from 0.8 mg soil. For these, the background PCR signal is equivalent to 2.6–4.8 × 104 CB-01 genomes per gram, and 10 times higher for soil I (owing to 1/100 dilution of the DNA from this soil). For digestate, the average Cq was 31.98 (Fig. 2), which means that the untreated digestate contains 3.2 × 106 CB-01 16S templates per millilitre, or 1.1 × 106 CB-01 genomes per millilitre.
Survival of CB-01 in soil
Laboratory experiment
A soil incubation experiment was designed to assess the survival of CB-01 in soil, vectored by digestate, under constant temperature and moisture conditions, and without any subsequent incorporation of digestate (thus contrasting with the field bucket experiment, Fig. 2). CB-01 was first grown to about 6 × 109 cells per millilitre in autoclaved, aerated and pH-adjusted digestate (as for the field experiments). Neutral-pH clay loam soil (soil N, see Extended Data Fig. 6) was portioned into a set of 50-ml Falcon tubes (9.4 g soil dry weight, moisture content = 0.5 ml g−1 soil dry weight). To each tube, 4.2 ml sterile water and 0.85 ml digestate (with CB-01) were dripped onto the soil. The tubes were stored in a dark moist chamber at 15 °C, with loose lids to allow exchange of air. Control tubes received only sterile water. At intervals, two replicate tubes were frozen (−20 °C) for quantification of CB-01 16S rRNA gene abundance by qPCR as described above.
Field plot experiment
From each individual plot (Fig. 5) we took three replicate soil samples, 9 and 280 days after fertilization, for quantification of CB-01 abundance by qPCR.
Extrapolating to national emission reductions
We use the emissions quantified with the GAINS model48,49 for 2030 in Europe to estimate the possible reductions of the measure.
The experiments described in this paper demonstrate marked emission reductions on all soils tested, over extended periods. The strongest reductions have been seen for the initial N2O peak immediately after fertilization, but NNRB has shown to remain active over a period of 90 days. Cumulated emissions over the whole period have been reduced by at least 41% (for clay loam soils), up to 95% reduction. We may disregard the case of the smallest reduction as the emissions from these soils are also rather small, but the organic loam soils (55% reductions) need to be considered. Consistent with the uniform emission factor used in GAINS (from IPCC50) of 1% of N applied to be emitted as N2O for all conditions of crops, soil or type of fertilizer added, a uniform reduction factor of 60% of emission reductions due to NNRB, which we consider a conservative estimate, was also applied. In Extended Data Table 3, emission reductions are shown by European country for 2030 if emissions from application of liquid manure alone are reduced by 60%. This assumption is based on the understanding that liquid manure can easily be treated in biodigesters. The authors of ref. 83 assume, for the purpose of methane abatement, that anaerobic digestion becomes profitable only for large agricultural entities of at least 100 livestock units. According to GAINS numbers, this concerns 70% of all farms in Europe, which more probably reflect liquid rather than solid manure systems, so the above estimate remains valid for the main fraction of liquid manure available. Indirect emissions as well as other soil emissions due to grazing, mineral fertilizer additions or application of farmyard manure (solid manure systems) have been left unchanged. Note that the GAINS model (in agreement with IPCC50) does not account for potentially increased emissions due to dry periods or freeze–thaw cycles (the latter considered to potentially contribute as much as 17–28% to global soil emissions84) but it covers increased emissions from cropping histosols.
Under these assumptions, total N2O emissions from Europe decrease by 2.7% owing to NNRB introduced. This figure is higher in countries that have a high share of liquid manure systems in their agriculture; hence, for EU27 (27 EU member countries) the corresponding figure is 4.0%, if NNRB were used for all manure nitrogen applied from liquid manure systems.
If it were possible to extend the NNRB technology, using solid manure and plant residues as substrates and vectors, we speculate emission reductions could be achieved for all mineral and natural fertilizer actively applied on fields. Ongoing work has shown that although Cloacibacterium sp. CB-01 grows to high cell densities in plant residues, new strains that grow in manure have been enriched and isolated (K. R. Jonassen and S. H. W. Vick, unpublished results). Although further development will be needed to implement this, it is relevant to estimate their impacts. Applying NNRB also to these other substrates at the same reduction efficiency could decrease European emissions as well as EU27 emissions by about a quarter (24% and 23%, respectively). For agricultural emissions alone, this means that roughly a third (31%) could be eliminated. For this calculation, we assume that indirect emissions from agriculture (due to re-deposition of ammonia released from fertilizers, or due to nitrate leaching), manure-management-related emissions and emissions from histosols remain unaffected.
It needs to be pointed out that an emission reduction of 60% as derived here for NNRB is much larger than emission reductions typically reported for N2O abatement measures. For example, GAINS assumes nitrification inhibitors to be able to reduce emissions by as much as 38%, and high-tech mechanical fertilizer-saving technologies (‘variable rate application’) to be able to save only 24% of the emissions48. Of note, the percentage reduction of N2O emission by the NNRB technology is plausibly unaffected by ‘variable rate application’ and nitrification inhibitor, as the target for NNRB is to reduce the N2O/N2 product ratio of denitrification, whereas the two others target the concentration of NO3− and nitrification, respectively.
Effect of CB-01 on the soil microbiome
Microbial community composition was examined by amplicon sequencing of the 16S rRNA gene V3–V4 region. Purified DNA from soil samples was sent to Novogene Europe for amplification, library preparation and sequencing to generate 250-base-pair paired-end reads using the Illumina Novoseq platform. Reads, after primer removal, were processed using GHAP (v2.4)85, an in-house amplicon clustering and classification pipeline built around Usearch (v11.0.66)86, the RDP classifier (v2.13)87 and locally written tools for generating operational taxonomic units (OTU) tables. Reads were processed using default quality control and trimming parameters. Clustering was carried out at both 97% and 100% similarity to generate OTUs and zero-radius OTUs (zOTUs), respectively. The 16S rRNA gene sequence of Cloacibacterium sp. CB-01 (GCA_907163125) was then matched against the OTU and zOTU representative sequences using the Usearch usearch_global command at 97% similarity and 99% similarity, respectively, to determine which OTU and zOTUs circumscribe the Cloacibacterium sp. CB-01 inoculant. From visual inspection it appeared that two zOTUs (zotu45 and zotu611) may circumscribe Cloacibacterium sp. CB-01 owing to shared abundance profiles and taxonomic classifications. To confirm that these two zOTUs both matched to Cloacibacterium sp. CB-01, the two representative sequences were BLAST-searched88 against the Cloacibacterium sp. CB-01 genome, and it was observed that both zOTU sequences matched closely to two separate regions of the genome, presumably harbouring multiple slightly divergent copies of the 16S rRNA gene. To confirm this, the two 16S rRNA genes from the Cloacibacterium sp. CB-01 genome were matched back against the zOTU representative sequences using the usearch_global command at 99% similarity, at which they matched to both zotu45 and zotu611, separately. Owing to this, zotu45 and zotu611 were combined for downstream analyses.
To assess the impact of the various treatments on the soil microbial communities, α- and β-diversity measures were calculated for microbial communities from all samples using the OTU tables generated above. OTU tables were first modified by removing the OTU circumscribing Cloacibacterium sp. CB-01 (OTU_27) before rarifying the tables to 72,846 reads per sample using the Usearch otutab_rare command. Shannon’s89 and Simpson’s90 diversity indices were calculated using the Usearch -alpha_div command and β-diversity measures were calculated using the Usearch -beta_div command. Jaccard’s dissimilarity measures91 were then used to generate multidimensional scaling plots using the Scikitlearn MDS module92.
The β-diversity as shown by Jaccard’s dissimilarity measures indicated that early during the soil incubation period there is greater between-sample variation both within treatments and between soils treated with live CB-01 and those treated with water or dead CB-01, indicating an effect of CB-01 on the soil microbial communities (Extended Data Fig. 7a). This effect, however, disappears by the final time point, at which samples from live-CB-01-, dead-CB-01- and water-treated soils cluster together, suggesting that the effect of live CB-01 on native soil microbial communities is transient and microbial soil communities are not affected in the longer term by the addition of live CB-01. It should be noted that the effect over time throughout the experiment is also a much larger source of microbial community variation than the addition of live CB-01 cells, presumably owing to disturbances to the soil from digging, sieving and packing of pots. Similarly, no systematic effects are observed on the α-diversity of soil microbial communities throughout the experiment indicating that the CB-01 treatment does not reduce the complexity or evenness of soil microbial communities when added to soils with digestate organic matter as can be seen in the Shannon and Simpson diversity measures of samples taken throughout the experiment (Extended Data Fig. 7b,c).
Search for antibiotics resistance genes and pathogenicity in CB-01
Microorganisms produce secondary metabolites crucial for diverse microorganism–microorganism interactions, enhancing survivability and competitive fitness through antagonistic effects on competitors under limited growth conditions. This array of metabolites, including antibiotics, toxins, pigments, growth hormones and anti-tumour agents, can also contribute to virulence and human pathogenicity. Such traits, if encoded in the inoculant’s genome, would restrict the use of such organisms as inoculants in agricultural soil. Likewise, the use of an inoculant would be restricted if its genome contains antibiotic resistance genes.
We checked CB-01 for such traits, scrutinizing its assembled draft genome7 in Pathogenfinder (v1.1)93 and ResFinderFG (v2.0)94, using standard settings. This revealed no evidence of human pathogenicity or antimicrobial resistance genes.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Data availability
Data that support the findings reported in this study are available at Figshare (https://doi.org/10.6084/m9.figshare.25130507)95. The assembled draft genome of CB-01 was downloaded from the European Nucleotide Archive (accession number GCA_907163125). 16S rRNA sequence data were deposited in the National Center for Biotechnology Information Sequence Read Archive database under accession number PRJNA878624.
Code availability
Code for calculating confidence intervals of ratios of time-integrated fluxes are available at https://github.com/larsmolstad/cloacipaper_stats.
References
Davidson, E. A. The contribution of manure and fertilizer nitrogen to atmospheric nitrous oxide since 1860. Nat. Geosci. 2, 659–662 (2009).
Seitzinger, S. P. & Phillips, L. Nitrogen stewardship in the Anthropocene. Science 357, 350–351 (2017).
Kuypers, M. M., Marchant, H. K. & Kartal, B. The microbial nitrogen-cycling network. Nat. Rev. Microbiol. 16, 263–276 (2018).
Hallin, S., Philippot, L., Löffler, F. E., Sanford, R. A. & Jones, C. M. Genomics and ecology of novel N2O-reducing microorganisms. Trends Microbiol. 26, 43–55 (2018).
Bakken, L. R. & Frostegård, Å. Emerging options for mitigating N2O emissions from food production by manipulating the soil microbiota. Curr. Opin. Environ. Sustain. 47, 89–94 (2020).
Jonassen, K. R. et al. Nitrous oxide respiring bacteria in biogas digestates for reduced agricultural emissions. ISME J. 16, 580–590 (2022).
Jonassen, K. R. et al. A dual enrichment strategy provides soil- and digestate-competent nitrous oxide-respiring bacteria for mitigating climate forcing in agriculture. mBio 13, e00788–22 (2022).
Vick, S. H. W., Jonassen, K. R., Arntzen, M. Ø., Lycus, P. & Bakken, L. R. Meta-omics analyses of dual substrate enrichment culturing of nitrous oxide respiring bacteria suggest that attachment and complex polysaccharide utilisation contributed to the ability of Cloacibacterium strains to reach dominance. Preprint at bioRxiv https://doi.org/10.1101/2023.06.04.543644 (2023).
Cato, M. P. & Varro, M. T. On Agriculture Vol. 283 (Harvard Univ. Press, 1934).
Erisman, J. W., Sutton, M. A., Galloway, J., Klimont, Z. & Winiwarter, W. How a century of ammonia synthesis changed the world. Nat. Geosci. 1, 636–639 (2008).
Rockström, J. et al. Planetary boundaries: exploring the safe operating space for humanity. Ecol. Soc. 14, 32 (2009).
Sutton, M. A. et al. Too much of a good thing. Nature 472, 159–161 (2011).
Gu, B. et al. Cost-effective mitigation of nitrogen pollution from global croplands. Nature 613, 77–84 (2023).
Canfield, D. E., Glazer, A. N. & Falkowski, P. G. The evolution and future of Earth’s nitrogen cycle. Science 330, 192–196 (2010).
Reay, D. S. et al. Global agriculture and nitrous oxide emissions. Nat. Clim. Change 2, 410–416 (2012).
Kanter, D. R. & Brownlie, W. J. Joint nitrogen and phosphorus management for sustainable development and climate goals. Environ. Sci. Policy 92, 1–8 (2019).
Tian, H. et al. A comprehensive quantification of global nitrous oxide sources and sinks. Nature 586, 248–256 (2020).
Vatn, A. Environmental Governance: Institutions, Policies and Actions (Edward Elgar, 2015).
Davidson, E., Galloway, J., Millar, N. & Leach, A. N-related greenhouse gases in North America: innovations for a sustainable future. Curr. Opin. Environ. Sustain. 9–10, 1–8 (2014).
Zhang, X. et al. Managing nitrogen for sustainable development. Nature 528, 51–59 (2015).
Onley, J. R., Ahsan, S., Sanford, R. A. & Löffler, F. E. Denitrification by Anaeromyxobacter dehalogenans, a common soil bacterium lacking the nitrite reductase genes nirS and nirK. Appl. Environ. Microbiol. 84, e01985-17 (2018).
Bakken, L. R., Bergaust, L., Liu, B. & Frostegård, Å. Regulation of denitrification at the cellular level: a clue to the understanding of N2O emissions from soils. Phil. Trans. R. Soc. B 367, 1226–1234 (2012).
Lycus, P. et al. Phenotypic and genotypic richness of denitrifiers revealed by a novel isolation strategy. ISME J. 11, 2219–2232 (2017).
Pessi, I. S. et al. In-depth characterization of denitrifier communities across different soil ecosystems in the tundra. Environ. Microbiome 17, 30 (2022).
Roco, C. A., Bergaust, L. L., Bakken, L. R., Yavitt, J. B. & Shapleigh, J. P. Modularity of nitrogen‐oxide reducing soil bacteria: linking phenotype to genotype. Environ. Microbiol. 19, 2507–2519 (2017).
Gowda, K., Ping, D., Mani, M. & Kuehn, S. Genomic structure predicts metabolite dynamics in microbial communities. Cell 185, 530–546 (2022).
Klimasmith, I. M. & Kent, A. D. Micromanaging the nitrogen cycle in agroecosystems. Trends Microbiol. 30, 1045–1055 (2022).
Liu, B., Frostegård, Å. & Bakken, L. R. Impaired reduction of N2O to N2 in acid soils is due to a posttranscriptional interference with the expression of nosZ. mBio 5, e01383-14 (2014).
Hénault, C. et al. Management of soil pH promotes nitrous oxide reduction and thus mitigates soil emissions of this greenhouse gas. Sci. Rep. 9, 20182 (2019).
Wang, Y. et al. Potential benefits of liming to acid soils on climate change mitigation and food security. Glob. Change Biol. 27, 2807–2821 (2021).
Simon, J. in Enzymes for Solving Humankind’s Problems: Natural and Artificial Systems in Health, Agriculture, Environment and Energy (eds Moura, J. J. et al.) 185–211 (Springer, 2021).
Lycus, P. et al. A bet-hedging strategy for denitrifying bacteria curtails their release of N2O. Proc. Natl Acad. Sci. USA 115, 11820–11825 (2018).
Gao, Y. et al. Competition for electrons favours N2O reduction in denitrifying Bradyrhizobium isolates. Environ. Microbiol. 23, 2244–2259 (2021).
Shan, J. et al. Beyond denitrification: the role of microbial diversity in controlling nitrous oxide reduction and soil nitrous oxide emissions. Glob. Change Biol. 27, 2669–2683 (2021).
Mania, D., Heylen, K., van Spanning, R. J. & Frostegård, Å. The nitrate‐ammonifying and nosZ‐carrying bacterium Bacillus vireti is a potent source and sink for nitric and nitrous oxide under high nitrate conditions. Environ. Microbiol. 16, 3196–3210 (2014).
Domeignoz-Horta, L. et al. Non-denitrifying nitrous oxide-reducing bacteria - an effective N2O sink in soil. Soil Biol. Biochem. 103, 376–379 (2016).
van Spanning, R. J., Richardson, D. J. & Ferguson, S. J. in Biology of the Nitrogen Cycle (eds Bothe, H. et al.) 3–20 (Elsevier, 2007).
Yoon, S., Nissen, S., Park, D., Sanford, R. A. & Löffler, F. E. Nitrous oxide reduction kinetics distinguish bacteria harboring clade I NosZ from those harboring clade II NosZ. Appl. Environ. Microbiol. 82, 3793–3800 (2016).
Hein, S., Witt, S. & Simon, J. Clade II nitrous oxide respiration of Wolinella succinogenes depends on the NosG,‐C1,‐C2,‐H electron transport module, NosB and a Rieske/cytochrome bc complex. Environ. Microbiol. 19, 4913–4925 (2017).
Ball, B. Soil structure and greenhouse gas emissions: a synthesis of 20 years of experimentation. Eur. J. Soil Sci. 64, 357–373 (2013).
Nadeem, S., Bakken, L. R., Frostegård, Å., Gaby, J. C. & Dörsch, P. Contingent effects of liming on N2O-emissions driven by autotrophic nitrification. Front. Environ. Sci. 8, 8–16 (2020).
Taylor, J., Wilson, B., Mills, M. & Burns, R. Comparison of microbial numbers and enzymatic activities in surface soils and subsoils using various techniques. Soil Biol. Biochem. 34, 387–401 (2002).
Lawson, C. E. et al. Common principles and best practices for engineering microbiomes. Nat. Rev. Microbiol. 17, 725–741 (2019).
Kaminsky, L. M., Trexler, R. V., Malik, R. J., Hockett, K. L. & Bell, T. H. The inherent conflicts in developing soil microbial inoculants. Trends Biotechnol. 37, 140–151 (2019).
Albright, M. B. et al. Solutions in microbiome engineering: prioritizing barriers to organism establishment. ISME J. 16, 331–338 (2022).
van Veen, J. A., van Overbeek, L. S. & van Elsas, J. D. Fate and activity of microorganisms introduced into soil. Microbiol. Mol. Biol. Rev. 61, 121–135 (1997).
Mallon, C. A. et al. The impact of failure: unsuccessful bacterial invasions steer the soil microbial community away from the invader’s niche. ISME J. 12, 728–741 (2018).
Winiwarter, W., Höglund-Isaksson, L., Klimont, Z., Schöpp, W. & Amann, M. Technical opportunities to reduce global anthropogenic emissions of nitrous oxide. Environ. Res. Lett. 13, 014011 (2018).
Amann, M. et al. Cost-effective control of air quality and greenhouse gases in Europe: modeling and policy applications. Environ. Model. Softw. 26, 1489–1501 (2011).
IPCC. 2006 IPCC Guidelines for National Greenhouse Gas Inventories (Institute for Global Environmental Strategies, 2006).
Hassan, J., Bergaust, L. L., Wheat, I. D. & Bakken, L. R. Low probability of initiating nirS transcription explains observed gas kinetics and growth of bacteria switching from aerobic respiration to denitrification. PLoS Comput. Biol. 10, e1003933 (2014).
Cowan, N. et al. An improved method for measuring soil N2O fluxes using a quantum cascade laser with a dynamic chamber. Eur. J. Soil Sci. 65, 643–652 (2014).
Kim, D. D. et al. Identification of nosZ-expressing microorganisms consuming trace N2O in microaerobic chemostat consortia dominated by an uncultured Burkholderiales. ISME J. 16, 2087–2098 (2022).
Molstad, L., Dörsch, P. & Bakken, L. R. Robotized incubation system for monitoring gases (O2, NO, N2O N2) in denitrifying cultures. J. Microbiol. Methods 71, 202–211 (2007).
Molstad, L., Dörsch, P. & Bakken, L. R. Improved robotized incubation system for gas kinetics in batch cultures. ResearchGate https://doi.org/10.13140/RG.2.2.30688.07680 (2016).
Bakken, L. R. Spreadsheet for gas kinetics in batch cultures: KINCALC. https://www.researchgate.net/publication/348383243_spreadsheet_for_gas_kinetics_in_batch_cultures_KINCALC (2024).
Bergaust, L., Shapleigh, J. P., Frostegård, Å. & Bakken, L. R. Transcription and activities of NOx reductases in Agrobacterium tumefaciens: the influence of nitrate, nitrite and oxygen availability. Environ. Microbiol. 10, 3070–3081 (2008).
Bergaust, L., Mao, Y., Bakken, L. R. & Frostegård, Å. Denitrification response patterns during the transition to anoxic respiration and posttranscriptional effects of suboptimal pH on nitrous oxide reductase in Paracoccus denitrificans. Appl. Environ. Microbiol. 76, 6387–6396 (2010).
Hassan, J., Bergaust, L. L., Molstad, L., de Vries, S. & Bakken, L. R. Homeostatic control of nitric oxide (NO) at nanomolar concentrations in denitrifying bacteria – modelling and experimental determination of NO reductase kinetics in vivo in Paracoccus denitrificans. Environ. Microbiol. 18, 2964–2978 (2016).
Hassan, J., Qu, Z., Bergaust, L. L. & Bakken, L. R. Transient accumulation of NO2-and N2O during denitrification explained by assuming cell diversification by stochastic transcription of denitrification genes. PLoS Comput. Biol. 12, e1004621 (2016).
Kellermann, R. et al. Preparation for denitrification and phenotypic diversification at the cusp of anoxia: a purpose for N2O reductase vis-à-vis multiple roles of O2. Appl. Environ. Microbiol. 88, e01053-22 (2022).
Qu, Z., Bakken, L. R., Molstad, L., Frostegård, Å. & Bergaust, L. L. Transcriptional and metabolic regulation of denitrification in Paracoccus denitrificans allows low but significant activity of nitrous oxide reductase under oxic conditions. Environ. Microbiol. 18, 2951–2963 (2016).
Hensen, A. et al. Dairy farm CH4 and N2O emissions, from one square metre to the full farm scale. Agric. Ecosyst. Environ. 112, 146–152 (2006).
Molstad, L. et al. A field robot for autonomous laser-based N2O flux measurements. Geophys. Res. Abstr. 16, 16109 (2014).
Curtin, D., Campbell, C. & Messer, D. Prediction of titratable acidity and soil sensitivity to pH change. J Environ. Qual. 25, 1280–1284 (1996).
Byers, E., Bleken, M. A. & Dörsch, P. Winter N2O accumulation and emission in sub-boreal grassland soil depend on clover proportion and soil pH. Environ. Res. Commun. 3, 015001 (2021).
Hastie, T., Friedman, J. H. & Tibshirani, R. The Elements of Statistical Learning: Data Mining, Inference, and Prediction Vol. 2 (Springer, 2009).
Efron, B. & Tibshirani, R. J. An Introduction to the Bootstrap (CRC, 1994).
Motulsky, H. Intuitive Biostatistics: A Nonmathematical Guide to Statistical Thinking (Oxford Univ. Press, 2014).
Wickham, H. et al. Welcome to the Tidyverse. J. Open Source Softw. 4, 1686 (2019).
Borchers, H. W. pracma: Practical Numerical Math Functions. R package v.2.4.2 https://cran.r-project.org/package=pracma (2022).
Xu, S. et al. Use ggbreak to effectively utilize plotting space to deal with large datasets and outliers. Front. Genet. 12, 2122 (2021).
Pedersen, T. L. patchwork: The Composer of Plots. R package v.1.1.3 https://CRAN.R-project.org/package=patchwork (2023).
Pedersen, T. L. & Crameri, F. scico: Colour Palettes Based on the Scientific Colour-Maps. R package v.1.5.0 https://CRAN.R-project.org/package=scico (2023).
R Core Team. R: A Language and Environment for Statistical Computing (R Foundation for Statistical Computing, 2023).
Crameri, F., Shephard, G. E. & Heron, P. J. The misuse of colour in science communication. Nat. Commun. 11, 5444 (2020).
Van Rossum, G. & Drake, F. L. Python 3 Reference Manual (CreateSpace, 2009).
Virtanen, P. et al. SciPy 1.0: Fundamental algorithms for scientific computing in Python. Nat. Methods 17, 261–272 (2020).
The pandas development team. pandas-dev/pandas: Pandas. Zenodo https://doi.org/10.5281/zenodo.3509134 (2020).
Bezanson, J., Edelman, A., Karpinski, S. & Shah, V. B. Julia: A fresh approach to numerical computing. SIAM Rev. 59, 65–98 (2017).
Allen, T. D., Lawson, P. A., Collins, M. D., Falsen, E. & Tanner, R. S. Cloacibacterium normanense gen. nov., sp. nov., a novel bacterium in the family Flavobacteriaceae isolated from municipal wastewater. Int. J. Syst. Evol. Microbiol. 56, 1311–1316 (2006).
Lim, N. Y., Roco, C. A. & Frostegård, Å. Transparent DNA/RNA co-extraction workflow protocol suitable for inhibitor-rich environmental samples that focuses on complete DNA removal for transcriptomic analyses. Front. Microbiol. 7, https://doi.org/10.3389/fmicb.2016.01588 (2016).
Höglund-Isaksson, L., Gómez-Sanabria, A., Klimont, Z., Rafaj, P. & Schöpp, W. Technical potentials and costs for reducing global anthropogenic methane emissions in the 2050 timeframe – results from the GAINS model. Environ. Res. Commun. 2, 025004 (2020).
Wagner-Riddle, C. et al. Globally important nitrous oxide emissions from croplands induced by freeze–thaw cycles. Nat. Geosci. 10, 279–283 (2017).
Greenfield, P. Greenfield Hybrid Analysis Pipeline (GHAP). Commonwealth Scientific and Industrial Research Organisation https://doi.org/10.4225/08/59f98560eba25 (2017).
Edgar, R. C. Search and clustering orders of magnitude faster than BLAST. Bioinformatics 26, 2460–2461 (2010).
Wang, Q., Garrity, G. M., Tiedje, J. M. & Cole, J. R. Naive Bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy. Appl. Environ. Microbiol. 73, 5261–5267 (2007).
Altschul, S. F., Gish, W., Miller, W., Myers, E. W. & Lipman, D. Basic local alignment search tool. J. Mol. Biol. 215, 403–410 (1990).
Shannon, C. E. A mathematical theory of communication. Bell Syst. Tech. J. 27, 379–423 (1948).
Simpson, E. H. Measurement of diversity. Nature 163, 688–688 (1949).
Jaccard, P. The distribution of the flora in the alpine zone. 1. New Phytol. 11, 37–50 (1912).
Pedregosa, F. et al. Scikit-learn: Machine learning in Python. J. Mach. Learn. Res. 12, 2825–2830 (2011).
Cosentino, S., Voldby Larsen, M., Aarestrup, F. M. & Lund, O. PathogenFinder - distinguishing friend from foe using bacterial whole genome sequence data. PLoS ONE 8, e77302 (2013).
Gschwind, R. et al. ResFinderFG v2.0: a database of antibiotic resistance genes obtained by functional metagenomics. Nucleic Acids Res. 51, W493–W500 (2023).
Hiis, E. G. et al. “Unlocking the bacterial potential...” source data and calculus for all figures. Figshare https://doi.org/10.6084/m9.figshare.25130507 (2024).
Hein, S. & Simon, J. Bacterial nitrous oxide respiration: electron transport chains and copper transfer reactions. Adv. Microb. Physiol. 75, 137–175 (2019).
Bakken, L. R. & Olsen, R. A. Buoyant densities and dry-matter contents of microorganisms: conversion of a measured biovolume into biomass. Appl. Environ. Microbiol. 45, 1188–1195 (1983).
Suenaga, T., Riya, S., Hosomi, M. & Terada, A. Biokinetic characterization and activities of N2O-reducing bacteria in response to various oxygen levels. Front. Microbiol. 9, 697 (2018).
Qi, C. et al. Organic carbon determines nitrous oxide consumption activity of clade I and II nosZ bacteria: genomic and biokinetic insights. Water Res. 209, 117910 (2022).
Suenaga, T. et al. Enrichment, isolation, and characterization of high-affinity N2O-reducing bacteria in a gas-permeable membrane reactor. Environ. Sci. Technol. 53, 12101–12112 (2019).
Acknowledgements
We thank T. Fredriksen for assistance and supervision in the field, and the Centre for Plant Research in Controlled Climate (SKP) at the Norwegian University of Life Sciences for facilitating the field experiments. This work was supported by the projects NENIM, Research Council of Norway No. 286888, and NOX2N, Research Council of Norway No. 331811.
Author information
Authors and Affiliations
Contributions
L.R.B., E.G.H., L.M., S.H.W.V., K.R. and K.R.J. designed experiments and analysed data. E.G.H., L.R.B., S.H.W.V., L.M., K.R. and K.R.J. designed and conducted phenotyping experiments. E.G.H., L.R.B., S.H.W.V. and L.M. designed and conducted the field experiments. E.G.H., K.R.J. and K.R. traced CB-01 in soils. W.W. extrapolated national emission reductions. All authors contributed to writing.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature thanks Frank Löffler, Robert Sanford, Lisa Stein and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data figures and tables
Extended Data Fig. 1 Growth yield and growth rates by aerobic and anaerobic respiration.
Panel a: The growth yield of CB-01 by aerobic and anaerobic respiration assessed by batch cultivation in 50 mL nutrient broth (meat-peptone and meat extract) in 120 mL vials (crimp-sealed with butyl-rubber septa) with He-atmosphere, provided with N2O and O2. Vials were placed in the thermostatic water-bath (23 °C) of the robotized incubation system54,55. After temperature equilibration and subsequent release of overpressure due to N2O- and O2-injection, the vials were inoculated (3.5 × 1010 cells vial−1). Based on measured O2, N2O and N2 in the headspace, the cumulated reduction of O2 and N2O was estimated. The inserted panel shows an example of the gas kinetics in a single vial. When O2 and N2O had been depleted, the cell density was measured by OD600. The relationship between cell density and OD600 was determined in a separate experiment comparing OD600 with microscopic counts. A linear relationship was found for OD600 ≤ 0.5 (cell density = 3.34 × 109 mL−1 OD−1). The cell dry weight, determined by weighing (cells washed three times in distilled water by dispersion and centrifugation, then dried at 105 °C), was 108 fg cell−1 ± s.e. = 7.5 (n = 9). The measured yield per mol of N2O and O2 was found by using the Generalized Reduced Gradient Solver in Excel (Microsoft Office 365, v2309) for the entire dataset. The panel shows the result for individual vials, as a plot of the predicted cell density (based on the yields given below) against measured cell density. The estimated yields were YN2O = 1.7 × 1014 cells mol−1 N2O and YO2 = 4 × 1014 cells mol−1 O2. The yields per mol electrons are Ye-N2O = 0.85 × 1014 mol−1 e− to N2O and Ye-O2 = 1.0 × 1014 cells mol−1 e− to O2. The yields in terms of dry weight g are Ye-N2O = 9.2 ± 0.6 g mol−1 e− to N2O and Ye-O2 = 11 ± 0.8 g mol−1 e− to O2. In comparison, Bergaust et al. 58 found Paracoccus denitrificans to have Ye-O2 = 3.75 × 1013 cells mol−1 e− = 11.2 g cell dry weight mol−1 e− to O2 (cell dry weight = 298 fg), which is practically identical to Ye-O2 for CB-01. Ye-N2O was 85% of Ye-O2 for CB-01, which is high compared to that measured for P. denitrificans (53%), and compared to the expectations ( ~ 60%) based on the charge separation per electron for aerobic and anaerobic respiration for NosZ clade I37. However, there is mounting evidence that the electron pathway to NosZ Clade II generates more charge separations than the pathway to NosZ Clade I, which is thermodynamically possible31,38,96. Panel b: The growth yield is plausibly declining as a culture reach stationary phase by depleting the C-sources. To assess the growth yields under these conditions, we quantified the cell densities by real-time quantitative PCR (qPCR), and compared this with the estimated cell densities based on the oxygen consumption and YO2 = 4 × 1014 cells mol−1 O2. The panel shows this comparison for aerobic growth to high cell densities in nutrient broth and digestate, confirming lower YO2 when the cultures approach stationary phase, more so in digestate than in nutrient broth. We find that YO2 for growth in digestate is 2 × 1014 cells mol−1 O2, which has been used to calculate cell densities in the digestate for fertilization experiments. The cell densities are based on O2 consumption from a single vial, and technical replicates from that same vial were used for qPCR analysis (all points plotted). Panel c and d: The aerobic and anaerobic growth rates, culturing as explained for panel a. Panel c: O2-consumption rates, 6 vol% O2 in the headspace, inoculated with ~7 × 108 cells mL−1, and estimated growth rate (µ, h−1) for each vial (average = 0.29 h−1, s.d. = 0.006). Panel d: rates of N2O reduction, anoxic vials with 1.1 vol% N2O in headspace, inoculated with ~3.4 × 108 cells mL−1, and estimated µ (average = 0.11 h−1, s.d. = 0.001). Given these growth rates, and the growth yields (panel a) we calculate Vmax = µ/Y): VmaxN2O = 0.65 fmol N2O cell−1 h−1, VmaxO2 = 0.73 fmol O2 cell−1 h−1. The maximal electron transport rates are Ve maxO2 = 2.9 fmol e− to O2 cell−1 h−1, Ve max e N2O = 1.3 fmol e− to N2O cell−1 h−1.
Extended Data Fig. 2 Onset of N2O-reduction during O2 depletion.
Denitrifying bacteria vary as to how early they initiate anaerobic respiration during O2 depletion. To explore this for Cloacibacterium sp. CB-01, we ran several experiments, both in nutrient broth (panel a-c) and in digestate (panel d). Panel a and b show the results for experiments where the inoculum was raised through > 10 generations under strict oxic conditions, thus diluting out any N2O reductase that might be present in the cells. Panel a shows the gas measurements, the O2 concentration in the liquid as calculated from the O2 transport rate (see Molstad et al.54), and the rate of N2O-reduction (VN2O). Panel b shows the ratio VeN2O/Vetot, i.e. the fraction of total electron flow that goes to N2O (for each time increment), plotted against the O2 concentration in the liquid. Error bars show s.d. of the mean for n = 3 replicate vials. Panel c shows VeN2O/Vetot (plotted against [O2]) for an experiment where the inoculum had been grown in hypoxia, thus with NosZ expressed already. Panel d shows the result for growth in digestate, inoculated with cells raised aerobically.
Extended Data Fig. 3 Apparent affinity for O2 and N2O.
To assess the affinity for O2 and N2O, we measured the rates of O2- and N2O reduction as the batch cultures depleted the two electron acceptors. The rates as measured (mol vial−1 h−1) were converted to rates per actively respiring cell (VO2 and VN2O, fmol cell−1 h−1) based on the numbers of active cells in the vial at each time point (= the midpoint between two samplings). The number of active cells were the numbers of cells in the inoculum + new-grown cells as calculated from the cumulated consumption of O2 and N2O (and the growth yield per mol, Extended Data Fig. 1), as done previously for determining the affinity for NO59. For VN2O, the number of actively N2O-respiring cells was only a fraction of the total (as shown in Extended Data Fig. 4). The maximum rates Vmax and the apparent Km values were found by fitting the Michaelis Menten model V = Vmax × S / (Km + S) to the data (S is the concentration of O2 and N2O in the liquid) by least square, using the Generalized Reduced Gradient Solver in Excel. This was done for each individual vial, and for the collective datasets. Panels a and b show the results for O2, for a single vial (a) and for the entire dataset (b). Embedded in the panel is the estimated Km for each individual vial. Panels c and d show the results for N2O, for a single vial (c) and for the entire dataset (d). Embedded in the panel is the estimated Km for each individual vial. These results show a relatively strong affinity for O2 (Km ~ 1 µM O2), and a rather weak affinity for N2O (Km ~ 13 µM N2O).
Extended Data Fig. 4 Bet-hedging during transition from aerobic to anaerobic respiration in CB-01.
To investigate the characteristic denitrification regulatory phenotype of CB-01, 120 mL vials with 50 mL nutrient broth and O2 + N2O in He-atmosphere were inoculated with 2.7 × 108 cells per vial, which had been raised under strict oxic conditions. The vials were monitored for gas kinetics as the cultures grew by oxygen initially, and then switched to respiring N2O in response to O2-depletion. The experiment included four treatments, all with 1 mL N2O (50 µmol N2O vial−1), but five different amounts of O2 (0, 0.8, 1.4, 3 and 4.6 mL O2) and 3 replicate vials for each O2 level. The panel shows the result for the vials with 0.8 mL initial O2. Error bars represent s.d. of the mean for n = 3 replicate vials, in all panels (4a-d). Panel a shows the measured amounts of O2 and N2O per vial. Panel b shows the electron flow rates to O2 and N2O (and total electron flow rate as a dashed line) as calculated from the measured O2 and N2O. Two phenomena stand out: 1) as oxygen was depleted, the electron flow rate declined to very low values and the subsequent electron flow rate to N2O increased exponentially, and 2) the apparent growth rate is 0.12 h−1, which is slightly higher than the anaerobic growth rate of CB-01 determined previously. This is the typical pattern for a bet-hedging denitrifying organism, i.e. an organism which expresses denitrification enzymes only in a fraction of the cells32,51. Assuming this, we investigated the possible fraction of cells that express NosZ and engaged in anaerobic respiration and growth: Panel c shows the estimated total number of cells (based on the cumulated O2- and N2O-consumption and the yields per mol O2 and N2O, Extended Data Fig. 1), and the number of N2O-respiring cells (Nos-active) calculated from the measured N2O-reduction rate (VN2O, mol N2O vial−1 h−1) and the assumption that VmaxN2O = 0.65 fmol N2O cell−1 h−1 (as determined previously): NosZ-active cells vial−1 = VN2O/VmaxN2O. The blue dashed line is the estimated number of cells without Nos (Nos inactive), assumed to be cells entrapped in anoxia without NosZ, hence unable to synthesize NosZ. Panel d shows the number of Nos-active cells as numbers per vial, and as fraction of the total number of cells in the vial. This fraction increases with time due to growth by N2O-respiration, and the fraction at the time of O2 depletion is a crude estimate of the fraction of cells which were able to express NosZ before O2 is completely exhausted. We coin this fraction FnosZ, analogous to Fden for Paracoccus denitfrificans, which is the fraction of cells that express NirS before O2 is depleted51, thus avoiding entrapment in anoxia61. In P. denitrificans, Fden was proportional to the time length of the hypoxic phase preceding complete anoxia, ascribed to a stochastic initiation of transcription of nirS once the cells experience hypoxia. To investigate if the apparent bet-hedging in CB-01 shows the same pattern, we estimated FnosZ for 15 batch cultures, all provided with 1 mL N2O but different amounts of O2 (0, 0.8, 1.4, 3 and 4.6 mL O2, n = 3 replicate vials for each O2-level). The cell density at the time of O2-depletion increased with increasing initial O2 (panel e), whereas the time length of exposure to hypoxia (arbitrarily defined as 0.2–4 µM O2) declined (panel f). A simplified version of the bet-hedging model60, assuming instantaneous expression of NosZ in a fraction of the cells (FNosZ) as O2 reached below 0.5 µM, was fitted to observed gas kinetics (O2 and N2O) for each vial to estimate FNosZ (all other parameters were as determined previously (Extended Data Figs. 1–2). Contrary to our expectations, the estimated FNosZ decreased with increasing time length of exposure to hypoxia (panel g) and increased with cell density at the time of O2 depletion (panel h). Interestingly, practically all cells appeared to become entrapped in anoxia (FNosZ = 0.002-0.006) in the vials without any O2 injected. The results warrant further investigations to provide direct evidence for the cell differentiation (bet-hedging), and the mechanism causing FNosZ to increase with cell density. A tantalizing hypothesis is that quorum sensing induction of NosZ expression is involved. Interestingly, all cells expressed NosZ in response to oxygen depletion when growing in digestate (Extended Data Fig. 5d–g).
Extended Data Fig. 5 Aerobic respiration, growth and and transition anaerobic respiration of CB-01 in digestate.
Panels a-c show the kinetics of O2-consumption during cultivation of CB-01 in digestate for the field experiments, measured in 50 mL subsamples placed in the incubation robot. The error bars (seen in all panels except c), represent the s.d. of the mean for n = 3 replicate vials. Panel a shows the rates of O2 consumption in vials with CB-01 and in the sterile controls, and the net consumption by CB-01 (CB-01 minus sterile control). This increased exponentially during the first 24 h, with apparent growth rate 0.12 h−1, which is much slower than in nutrient broth (0.29 h−1, Extended Data Fig. 1). Panel b shows the cumulated O2 consumption by CB-01, and the estimated cell density assuming YO2 = 4.06 × 1014 cells mol−1 O2 (Extended Data Fig. 1), reaching 1.1 × 1010 mL−1. The cell density quantified by qPCR for a similar experiment only 47% of the density based on O2 (Extended Data Fig. 1), suggesting that YO2 for growth to high cell densities in digestate is ~50% of YO2 for optimal growth in nutrient broth (Extended Data Fig. 1b). Panel c shows estimated cell specific O2 consumption (VO2, fmol O2 cell−1 h−1), estimated growth rate, µ (h−1) = VO2 × Y, where Y = the measured growth yield by aerobic respiration (4.06 × 1014 cells mol−1 O2), and the estimated fraction of organic C in the digestate (10 mg C mL−1) consumed by CB-01 (sum of CO2 and assimilated C). This suggests that ~1 % of the organic C in the digestate was easily available monomers, supporting rapid growth of CB-01 to a cell density of ~1 × 109 cells mL−1 after 20 h, while subsequent growth was gradually declining as the organism utilized increasingly recalcitrant substrates, plausibly with a lower growth yield (YO2). Panels d-g show results of experiments designed to investigate if CB-01 is bet-hedging when growing in digestate: Bet-hedging of CB-01, as observed when cultured in nutrient broth (Extended Data Fig. 4) would reduce its capacity to scavenge N2O when vectored by digestate to soil. In theory, however, the bet-hedging could depend on the growth medium, since the fraction of cells expressing NosZ (FnosZ) increased with the cell density (Extended Data Fig. 4), plausibly due to accumulation of compounds stimulating the expression of NosZ. On this background, we conducted experiments similar to that shown in Extended Data Fig. 4, but using digestate instead of nutrient broth as a growth medium. The panels show the results of one treatment, in which 120 mL serum vials with 50 mL autoclaved and aerated digestate (and stirring magnets), He-atmosphere + 1.4 mL O2 and 1 mL N2O, were inoculated with 2.5 × 108 CB-01-cells (5 × 106 cells mL−1), raised under strict oxic conditions, and monitored for gas kinetics while incubated at 23 °C. Panel d shows the measured O2 and N2O, together with the cell density as calculated from the initial cell density and growth as calculated from the measured O2 and N2O-reduction, using the yield per mol O2 and N2O determined previously (YO2 = 4 × 1014 cells mol−1 O2, YN2O = 1.7 × 1014 cells mol−1 N2O, Extended Data Fig. 1). Panel e shows the O2 and N2O consumption rates (µmol vial−1 h−1) for each time increment, and panel f shows the rates of electron flow to O2 (aerobic respiration) and N2O (anaerobic respiration), and the total electron flow as a dashed line. In contrast to the results with nutrient broth (Extended Data Fig. 4) there is hardly any depression of the electron flow in response to oxygen depletion, suggesting that FnosZ ~ 1, i.e. all cells switch to respiring N2O (express nosZ). To inspect the validity of this further, the electron flow rates per cell (fmol e− cell−1 h−1) to O2 (VeO2) and N2O (VeN2O) were calculated, and shown in panel g. As expected, VeO2 remained stable around 3 fmol e− cell−1 h−1 until oxygen became limiting, which is close to the maximum rate determined previously (VmaxO2 = 0.72 fmol O2 cell−1 h−1 = 2.88 fmol e− cell−1 h−1), and VeN2O reached ~1.4 fmol e− cell−1 h−1 immediately after O2-depletion, which is close to the maximum rate determined in nutrient broth cultures (VmaxN2O = 0.6 fmol N2O cell−1 h−1 = 1.2 fmol e− cell−1 h−1). All panels show the average of three replicate vials, with standard deviation as vertical lines. The experiment included treatments with 0.8, 3 and 4.6 mL O2 (3 replicates of each), hence with widely different cell densities at the time of O2-depletion, and they all showed FnosZ to be close to 1 (results not shown). In summary, CB-01 express NosZ in all cells (hence no bet-hedging) in response to O2-depletion when grown in digestate, unaffected by the cell density at the time of O2-depletion.
Extended Data Fig. 6 Soil characteristics and changes to soil pH due to application of digestate.
Some key characteristics of the different soil types are listed in the table above. The acid sandy silt soil (S) was taken from an agricultural field in Solør, Norway, dominated by fluvial sandy silt soils. The clay loam soils L, I and N were from different plots within a liming experiment near the Norwegian University of Life Sciences (59°39’48.2”N 10°45’44.8”E), limed in 201441. O was a clay loam soil from the same area (hence with similar mineral components), but with a much higher content of organic C because it had been a wetland prior to cultivation. Soils S, L, N and O were used in the field bucket experiments. Soil I is the soil of the plots used for the field plot experiment. The bar chart shows pH(CaCl2) in the four soils as affected by applying 0.055, 0.11 and 0.165 mL digestate g−1 soil dry weight, which is 50, 100 and 150 % of the amounts added to the soils in the field bucket experiment (0.11 mL digestate g−1 soil dry weight). Water was added (if needed) to reach a water-filled pore space (%) equivalent to that in the buckets after digestate application. *pH(CaCl2) was measured after dispersing 10 g soil in 25 mL of 0.01 M CaCl2. **Tot C = total organic C, Tot N = total organic N, and [NO3] = mg NO3N kg−1 soil dry weight.
Extended Data Fig. 7 The influence of CB-01 inoculation on the soil microbial communities.
Panel a shows an MDS plot based on Jaccard’s dissimilarity measures for microbial communities in the soil of the first field bucket experiment treated with live CB-01 in digestate, killed CB-01 in digestate, and water treatment only (Methods 8). The panels show the results for soils sampled 2, 22, 36 and 92 days after fertilization with digestate containing CB-01. Results for the last sampling (92 days) are encircled, and a single outlier is marked by arrow. Panel b shows the Shannon’s diversity indices for microbial soil communities sampled throughout the field bucket experiment treated with live CB-01 in digestate, killed CB-01 in digestate and water treatment only. Panel c shows the Simpson’s diversity indices for microbial soil communities sampled throughout the field bucket experiment treated with live CB-01 in digestate, killed CB-01 in digestate and water treatment only. Boxes in the box plots shown in panels b and c indicate the interquartile range with a line within the box representing the median. The whiskers extending from the box extend to the furthest datapoint contained within 1.5 times the interquartile range from the boxes and circles past these whiskers denote outlier values (n = 8).
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Hiis, E.G., Vick, S.H.W., Molstad, L. et al. Unlocking bacterial potential to reduce farmland N2O emissions. Nature 630, 421–428 (2024). https://doi.org/10.1038/s41586-024-07464-3
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1038/s41586-024-07464-3
This article is cited by
-
Genomics-based identification of a cold adapted clade in Deinococcus
BMC Biology (2024)
-
Bacteria curb emissions in farmland
Nature Reviews Microbiology (2024)
-
Nitrogen-hungry bacteria added to farm soil curb greenhouse-gas emissions
Nature (2024)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.
