Identification of Unknown Carboxydovore Bacteria Dominant in Deciduous Forest Soil via Succession of Bacterial Communities, coxL Genotypes, and Carbon Monoxide Oxidation Activity in Soil Microcosms

Abstract

Surveys of the coxL gene, encoding the large subunit of the CO dehydrogenase, are used as a standard approach in ecological studies of carboxydovore bacteria scavenging atmospheric CO. Recent soil surveys unveiled that the distribution of coxL sequences encompassing the atypical genotype coxL type I group x was correlated to the CO oxidation activity. Based on phylogenetic analysis including the available coxL reference genome sequences, this unusual genotype was assigned to an unknown member of the Deltaproteobacteria, with the coxL sequence from Haliangium ochraceum being the sole and closest reference sequence. Here we seek to challenge the proposed taxonomic assignation of the coxL group x genotype through the monitoring of CO consumption activity and microbial community successions during the colonization of sterile soil microcosms inoculated with indigenous microorganisms. In our study, we established that the estimated population density of Deltaproteobacteria was too small to account for the abundance of the coxL group x genotype detected in soil. Furthermore, we computed a correlation network to relate 16S rRNA gene profiles with the succession of coxL genotypes and CO uptake activity in soil. We found that most of the coxL genotypes for which the colonization profile displayed covariance with CO uptake activity were related to potential carboxydovore bacteria belonging to Actinobacteria and Alphaproteobacteria. Our analysis did not provide any evidence that coxL group x genotypes belonged to Deltaproteobacteria. Considering the colonization profile of CO-oxidizing bacteria and the theoretical energy yield of measured CO oxidation rates in soil microcosms, we propose that unknown carboxydovore bacteria harboring the atypical coxL group x genotype are mixotrophic K-strategists.

INTRODUCTION

Carbon monoxide (CO) is a trace gas in the atmosphere, with a typical mixing ratio of 60 to 150 ppb by volume (ppbv) (1). Combustion of fossil fuels and biomass burning are the main sources of this gas (2). The hydroxyl radicals (OH˙) are responsible for 85% of the global sink of CO, while specialized aerobic microorganisms in the soil consume the remaining 15% (3, 4). These CO-oxidizing bacteria represent the most uncertain term for the budget of atmospheric CO. The diversity and ecophysiology of this ubiquitous functional group present in the environment are poorly documented and impair the fate prediction of the biological sink and the atmospheric burden CO in response to global change.

CO-oxidizing bacteria possess a CO dehydrogenase enzyme (CODH) catalyzing the following reaction: CO + H₂O → CO₂ + 2H⁻ + 2e⁺. The enzyme is composed of a dimer of heterotrimers with CoxS, CoxM, and CoxL as small, medium, and large subunits, respectively (5). The CoxL subunit contains the active site of the enzyme and is used as a molecular marker to assess CO-oxidizer diversity in the environment (6). Furthermore, the coxL gene sequences encompass two different phylogenetic groups: the type I group that has been genetically and biochemically characterized and the putative type II group. There is compelling evidence suggesting that type II CODH is not functional or uses another substrate than CO (7, 8), and thus coxL sequences encompassing this group are not considered in an environmental survey of CO-oxidizing bacteria. CO-oxidizing bacteria harboring coxL type I comprise two different physiological groups. They are composed of the carboxydotrophic bacteria, displaying low affinity for CO and the ability to grow using CO as the sole carbon and energy source, and the carboxydovore bacteria, displaying high affinity for CO, enabling the capacity to scavenge the trace amounts present in the atmosphere (9). Only a few carboxydovore bacteria have been extensively characterized, encompassing Alphaproteobacteria (Bradyrhizobium spp., Stappia spp., Aminobacter sp., and Ruegeria pomeroy DSS-3) (10–12), Betaproteobacteria (Burkholderia spp.) (13), Gammaproteobacteria (Alkalilimnicola ehrlichii and Streptomonas sp. strain LUP) (10, 14), Actinobacteria (Mycobacterium spp.) (10), and Chloroflexi (Ktedonobacteria) (15). However, the environmental factors determining their activity and distribution remain elusive (9).

Previously published environmental surveys of coxL sequences have been done in volcanic deposits (6, 16, 17), hot springs (18), and aquatic environments (19). Even though forest ecosystems exert a dominant role in the global sink of atmospheric CO (20), the carboxydovore bacteria thriving in forest soil have received little attention. Recently, a soil survey of coxL type I sequences in deciduous forest have led to the identification of a new coxL group, referred to as coxL group x in this paper, for which the distribution was correlated with CO soil uptake activity. A phylogenetic analysis identified the reference coxL sequence from Haliangium ochraceum as the closest relative of the coxL group x genotype suggesting the presence of unknown CO-oxidizing Deltaproteobacteria in soil (21). However, this assumption is highly uncertain because of the potential lateral transfer of the coxL gene and the relatively low identity score (less than 75%) between coxL group x sequences retrieved from soil and the reference coxL sequence in H. ochraceum. To address the problematic nature of this assumption, our study aimed to challenge the hypothetical affiliation of the coxL group x genotype with Deltaproteobacteria. The succession of bacterial communities, coxL genotypes, and the CO oxidation activity were monitored in sterile deciduous forest soil microcosms inoculated with the indigenous microorganisms. Also, we verified the hypothesis that the maturation of the CO oxidation activity would be related to the concomitant emergence of the coxL group x genotype and 16S rRNA gene sequences from Deltaproteobacteria.

MATERIALS AND METHODS

Soil sample and incubation.

The soil sample was collected from a deciduous forest located about 40 km from Montreal on the south shore of the St. Lawrence River (45°67′N, 73°32′W). Soil texture was determined by the hydrometer method, and particle size distribution assigned soil samples to the silt loam textural class (22). The same site was visited to investigate the impact of land use change, soil nutrients, and physicochemical properties on CO-oxidizing bacteria (21). The soil sample was homogenized (2-mm-pore sieve) and used for the preparation of the colonization substrates and inocula. For the colonization substrate, soil was sterilized two times, separated by a 24-h interval. Sterilization was done in 500-ml Wheaton glass bottles closed with a foam cap. Three bottles were prepared containing 72 g of soil. After the sterilization procedure, the bottles were kept on the laboratory bench with a foam cap for a 5-day period in order to evacuate CO produced during autoclaving. Sterile soil microcosms were inoculated with 8 g of homogenized nonsterile soil (i.e., indigenous soil microorganisms), corresponding to a 10% (vol/vol) inoculum. Three bottles were treated in the same manner with 80 g of sterile soil without inoculation for the controls. The sterile foam cap was kept during the incubation at 20°C to allow gaseous exchanges in the microcosms. After 0, 4, 9, 12, 15, 22, 36, and 55 incubation days, subsamples of 5.5 g were collected in the six microcosms for DNA extraction (0.5 g), and moisture content (5 g) was determined using the standard gravimetric method. Soil water content was maintained at approximately 35% during the incubation period. This water level was selected to avoid diffusion limitation of the CO uptake activity, while supporting bacterial metabolism and growth (23, 24). Soil samples dedicated to DNA extraction were frozen in a 2-ml microtube containing 700 μl TPM buffer (50 mM Tris-HCl [pH 7], 1.7% polyvinylpolypyrrolidone, 20 mM MgCl₂), 35 μl 20% SDS, and 0.5 g glass beads (0.1-mm diameter).

CO oxidation rate in soil.

Before measurement of CO oxidation activity, the foam cap was replaced with a butyl septum cap. A CO gas mixture (508 ± 10 ppm by volume [ppmv] CO [GTS-Welco, PA, USA]) was injected to get an approximately 4-ppmv initial concentration in the static headspace. Headspace samples (10 ml) were collected with a Pressure Lok gastight glass syringe (VICI Precision Sampling, Inc., Baton Rouge, LA, USA) and injected in a gas chromatograph equipped with a reduction gas detector (ta3000R; Ametek Process Instruments, DE, USA). The first-order oxidation rates were calculated by integrating the CO mole fraction time series measured over a 2-h period, using at least four concentration points for data integration. After each oxidation rate measurement, a foam cap was put back onto the bottles for incubation. Reproducibility of the CO analyses was assessed before each set of measurements by repeated analyses of certified CO standard gas (2.05 ± 0.10 ppmv [GTS-Welco, PA, USA]), and standard deviations were lower than 5%.

DNA extraction and qPCR assays.

Genomic DNA was extracted using a combination of chemical and mechanical cell lysis procedures (25). DNA was precipitated with 2 volumes 96% ethanol, and a polyvinylpolypyrrolidone spin column was used for final purification (26). Purified DNA extracts were eluted in 200 μl nuclease-free water. Purified DNA was used for quantitative PCR (qPCR) of bacterial 16S rRNA gene, coxL type I, and coxL type I Deltaproteobacteria (represented as coxL group x in this article) using previously described procedures (21). The reactions were performed using 5 μl of diluted genomic DNA (1:50).

coxL genotyping.

Genomic DNA extracted after 4, 22, 36, and 55 incubation days was used for coxL pyrosequencing. PCR amplification of coxL genes using the universal assay (21), library preparation, and 454 pyrosequencing were performed at McGill University and Génome Québec Innovation Centre. A total of 45,000 raw sequences were obtained from the 12 libraries. The sequences were analyzed using the software mothur (27). Sequences with a quality score (“qaverage”) lower than 25, with more than 8 consecutive homopolymers, and with ambiguous bases were all removed. The average length of sequences was 336 bases. Sequences were aligned against a coxL sequences database previously built (21). An arbitrary identity cutoff of 0.04 was used for the clusterization of nucleotide sequences into operational taxonomic units (OTU) representing coxL sequences belonging to the same bacterial species (21). Libraries were standardized to the sequencing effort of the smallest coxL gene library (i.e., 641 sequences) to avoid biases in comparative analyses introduced by the sampling depth. At this stage, hypothetical coxL type II OTU identified by the canonical amino acid signature of their active site (AYRGAGR) were removed from the data set. The resulting OTU table was used for rarefaction and diversity analyses. OTU with less than 7 sequences (0.1% of total sequences) were eliminated, and 136 OTU were left for phylogenetic and correlation network analysis. Nucleic acid sequences were imported in the software MEGA (28) for phylogenetic analysis using the reference database and procedure previously described (21) to which sequences were added.

Bacterial 16S rRNA gene profiling.

Genomic DNAs extracted after 4, 22, 36, and 55 incubation days were used for 16S rRNA gene profiling. PCR amplification of the V6 to V8 regions of 16S rRNA, library preparation, and Illumina MiSeq 250-bp paired-end sequencing reactions were performed by the technical staff of McGill University and Génome Québec Innovation Centre. The two paired-end reads (forward and reverse) were merged using Flash (29) with a minimum overlap length between the two reads of 20 bases and a maximum overlap of 250 bases. The maximum allowed ratio between the number of mismatched base pairs and the overlap length was set to 0.3. Reads were truncated to a uniform length of 420, corresponding to the length of more than 97.5% of the total reads. Reads with a low-quality score were removed using a maximum expected error value of 2.0 (30). The remaining reads (4,058,447) were processed using the software QIIME version 1.8.0 (31). Reads were dereplicated and then sorted by abundance. Unique reads (1,214,627) were then clustered at 97% identity using USEARCH (30) as the OTU picking method, with a minimum of two sequences for the OTU to be considered. The resulting OTU were given an extra check for chimeras using the Gold reference database (http://drive5.com/uchime/uchime_download.html), and those that were flagged as chimeric (10,339) were removed. The remaining 2,060 nonchimeric OTU were assigned using the “assign taxonomy” command (QIIME) run against the Greengenes database (version gg_13_8) (32). Reads were rarefied to the sequencing effort of the smallest 16S rRNA gene library (206,201 reads) to avoid biases in comparative analysis introduced by the sampling depth, giving a final number of 613 OTU.

Statistical analysis.

Statistical analyses were all performed using the software R (33). The package “Hmisc” was used to measure Spearman correlation between coxL/16S rRNA gene ratio estimated by qPCR and soil CO uptake (34). The package “Vegan” was used to compute α diversity indexes and β diversity with homogeneity multivariate dispersion (35, 36). The package Weighted Gene Co-Expression Network Analysis (WGCNA) was used to identify modules of highly correlated 16S rRNA OTU (37). The data matrix used for this analysis was expressed in 16S rRNA gene copies per gram of soil dry weight. For this purpose, the relative frequency of individual 16S rRNA gene OTU as determined through sequencing was multiplied by the absolute copy number of the 16S rRNA gene as determined by qPCR. A soft power of 12 was used to raise the Spearman correlation matrix into an adjacency matrix. This matrix was then converted into a topological overlap (TO) matrix by the TOM similarity algorithm, and genes were hierarchically clustered based on TO similarity. Modules comprising less than 10 OTU were not considered, and modules having an eigengene dissimilarity of <0.45 were merged. Spearman correlations were computed between the modules eigengene and CO uptake rate and the absolute abundance of coxL OTU. The absolute abundance of coxL OTU (expressed in copies per gram of soil dry weight) was computed by multiplying the relative abundance of individual OTU obtained from pyrosequencing profiles by the absolute abundance of coxL genes determined by qPCR.

Nucleotide sequence accession number.

Raw sequences have been deposited in the Sequence Read Archive of the National Center for Biotechnology Information under BioProject no. PRJNA299500.

RESULTS

Maturation of soil CO uptake activity.

The maturation of CO uptake activity was detected after a lag phase of 10 days (Fig. 1). We observed that the activity then increased continuously, reaching 3,466 ± 800 pmol g dry weight⁻¹ h⁻¹ after 55 days. This activity is comparable to the oxidation rate of 3,243 pmol g dry weight⁻¹ h⁻¹ measured in the native soil (21), indicating that activity was restored at the end of the incubation period. Furthermore, we found that no significant CO uptake activity was observed in the control experiment comprising noninoculated sterile soil (Fig. 1).

FIG 1.

Time series of the CO oxidation activity measured in inoculated (black dots) and sterile (gray dots) soil microcosms. Averages are presented with standard deviations derived from three independent replicates. The gray line represents the CO oxidation rate measured in native soil. The black arrows indicate selected soil samples for high-throughput sequencing of PCR-amplified coxL and 16S rRNA genes. dw, dry weight.

Abundance of bacterial 16S rRNA and the coxL gene.

Using qPCR, we determined that the abundance of 16S rRNA gene increased from (3.3 ± 0.5) × 10⁹ copies g dry weight ⁻¹ at the beginning of the experiment to (4.4 ± 1.4) × 10¹⁰ copies g dry weight ⁻¹ after 4 days of incubation (see Fig. S1 in the supplemental material). No significant increase was observed between 22 and 55 incubation days, with (2.6 ± 0.5) × 10¹¹ copies g dry weight⁻¹ at the end of the incubation period. Quantification of the coxL gene followed the same trend as the 16S rRNA gene, while coxL group x increased exponentially after a lag phase of 15 days (see Fig. S1). The relative abundance of coxL genotypes in soil was estimated by computing the coxL/16S rRNA gene ratio (Fig. 2). We established that the relative abundance of coxL increased by 5.0% ± 0.94% after 4 days and then decreased exponentially, reaching 2.2% ± 0.37% at the end of the incubation. On the other hand, the relative abundance of coxL group x increased exponentially from 0.05% ± 0.01% at day 15 to 0.7% ± 0.09% after 55 days of incubation. The relative abundance of coxL group x was positively correlated with CO uptake rate (Spearman, P < 0.001), while the abundance of neither the coxL gene nor 16S rRNA gene showed a significant relationship to CO oxidation activity.

FIG 2.

Time series of the relative abundance of (A) coxL and (B) coxL group x genotypes in soil microcosms as estimated by qPCR.

Bacterial 16S rRNA genotypes.

The sequencing effort was sufficient to cover most of the diversity of 16S rRNA genes, as assessed by rarefaction curves that reached a saturation plateau (data not shown). In total, 613 OTU were identified. Interestingly, the incubation of soil microcosms resulted in an overall decrease in species richness and species evenness (Table 1), indicating that a fraction of the bacteria present in the native soil inoculum demonstrated the ability to colonize sterilized soil with some disproportionate single-species contribution to diversity. The β diversity analysis showed that the structure of microbial communities was significantly more heterogeneous at the beginning of the incubation than observed in the last three subsamples, indicating reproducible enrichment of the subset of the initial bacterial community in replicated microcosms (see Fig. S2A in the supplemental material). At the end of the incubation period, we found that soil microbial communities were dominated by the phyla Actinobacteria (65% ± 4%), Acidobacteria (17% ± 1%), and Proteobacteria (16% ± 3%). Proteobacteria were dominated by Alphaproteobacteria (16% ± 2.6%), followed by Betaproteobacteria (0.44% ± 0.43%), Gammaproteobacteria (0.12% ± 0.12%), and Deltaproteobacteria (0.008% ± 0.002%). Considering the abundance of the 16S rRNA gene determined by qPCR and assuming three rRNA gene copies per genome (38), we evaluated that the population density of Deltaproteobacteria reached 7.0 × 10⁴ cell g dry weight⁻¹ in the soil microcosm. This is well below the absolute abundance of coxL group x sequences determined by qPCR (1.8 × 10⁹ copies g dry weight⁻¹).

TABLE 1.

Specific richness and evenness of coxL and 16S rRNA gene OTU in soil microcosms

Incubation time (day)	Diversity by index:
	16S rRNA gene		coxL
	Shannon index	Simpson index	Shannon index	Simpson index
4	4.57 ± 0.10	0.980 ± 0.001	4.43 ± 0.14	0.978 ± 0.004
22	3.45 ± 0.08	0.920 ± 0.007	4.07 ± 0.21	0.953 ± 0.004
36	3.66 ± 0.04	0.930 ± 0.006	3.87 ± 0.23	0.924 ± 0.018
55	3.84 ± 0.12	0.937 ± 0.006	3.91 ± 0.23	0.930 ± 0.018

coxL genotypes.

Rarefaction analysis showed that the sequencing effort was sufficient to cover coxL diversity (data not shown), with 136 OTU identified. As observed in the 16S rRNA gene profile, coxL richness and evenness decreased during soil colonization (Table 1). Homogenization of coxL genotype profiles in replicated microcosms was slower than that in 16S rRNA genes, and they exhibited more heterogeneity in soil subsamples collected after 4 and 22 days than was observed in the replicated microcosms sampled at days 36 and 55 (see Fig. S2B in the supplemental material). The coxL OTU were assigned to one of the three reference genotypes in a phylogenetic analysis (Fig. 3). At the end of the incubation, 53% ± 3% of coxL sequences were assigned to coxL group x, while Actinobacteria and Alphaproteobacteria represented 14% ± 3% and 31% ± 3% of retrieved coxL sequences. We noticed that few OTU dominated the three genotypes. One OTU closely related to Bradyrhizobium japonicum USDA (Alphaproteobacteria [OTU 1]) represented 70% of total sequences encompassing Alphaproteobacteria genotype and 18% of total coxL sequences (see Fig. S3A in the supplemental material). The most abundant OTU belonging to Actinobacteria was closely related to Arthrobacter sp. strain FB24 (OTU 4) and represented 24% of the sequences comprising this genotype (see Fig. S3B). OTU 2 and OTU 3 represented 21 and 11% of total sequences encompassing the coxL group x genotype, respectively (see Fig. S3C). The only reference bacterium present in the cluster coxL group x defined by the phylogenetic analysis is H. ochraceum. The percentage of identity between this reference sequence and the closest OTU was 80%. All of the other OTU assigned to coxL group x were more distantly related to H. ochraceum, with an identity score of between 46 and 59%.

FIG 3.

Maximum likelihood phylogenetic analysis of coxL-inferred amino acid sequences of 136 OTU and reference sequences from NCBI (147 residues). (A) The three main subclasses of sequences belonging to coxL are represented in the three clusters. The number of sequences of each OTU or a cluster is indicated in square brackets as “[OTU number; number of sequences].” The percentages of replicated trees in which the associated CoxL sequences clustered together in the bootstrap test (1,000 replicates) are shown for nodes supported by ≥50% of the replicates. (B to D) Time series of coxL sequence relative abundance in (B) coxL group x, (C) coxL Actinobacteria, and (D) coxL Alphaproteobacteria clusters.

Correlation network analysis.

The CO oxidation activity, coxL genotypes, and 16S rRNA gene profile databases were combined to correlate the soil genetic profile with the CO oxidation activity and to test the hypothesis that coxL group x sequences positively correlated to activity belonged to Deltaproteobacteria. For this purpose, a correlation network was first constructed using the time series of 16S rRNA profiles. This analysis led to a classification of the OTU into five modules according to the covariance of their soil colonization profile (Fig. 4). The first module (module 1) was dominated by OTU encompassing Actinobacteria (62%), Acidobacteria (22%), and Proteobacteria (15%). The second module (module 2) was mostly composed of Firmicutes OTU (90%). Module 3 was the most taxonomically diverse, comprising 12 different phyla. Module 4 was mainly represented by OTU affiliated with Actinobacteria (60%), Proteobacteria (26%), and Actinobacteria (9%). Finally, module 5 was composed mostly of Proteobacteria (35%), Actinobacteria (26%), and Chlamydia (16%). Correlation of modules' eigengenes with the CO oxidation rates measured during the incubation of soil microcosms unveiled that the relative abundance of OTU encompassing module 1 was significantly correlated to the maturation of CO uptake activity in soil microcosms (Spearman, P < 0.001). Thus, the OTU constituting module 1 were from presumptive carboxydovore bacteria and other bacteria displaying a similar colonization profile but incapable of CO oxidation activity.

FIG 4.

Taxonomic composition of the five modules defined by the covariance of the 16S rRNA OTU time series in the soil. (A) The heat map shows the number of OTU representing each family detected in the modules. (B) The OTU assigned to the same bacterial phyla were clustered together, and the number of sequences comprising each group was computed to report their relative abundance in the five modules. The color key used to identify each bacterial phylum in the pie charts is defined along the side of the heat map. The total number of sequences comprised in the modules is shown at the bottom of each pie chart.

The 16S rRNA gene OTU assigned as Acidobacteriaceae, Actinospicaceae, Microbacteriaceae, Bradyrhizobiaceae, and Acetobacteraceae dominated the composition of module 1 with 22, 36, 21, 7, and 3% of total sequences in the module, respectively (Table 2). Most of the taxonomic groups (family level) comprising module 1 are represented by at least a reference genome or bacterium possessing coxL sequence (Table 2). Even though no coxL genes were found in the reference genomes from Actinospicaceae, Microbacteriaceae, and Acetobacteraceae, the colonization profile of OTU assigned to these taxonomic groups was significantly correlated with CO oxidation activity (Table 2).

TABLE 2.

Composition of module 1 comprising 16S rRNA OTU displaying colonization profiles related to maturation of CO oxidation activity in soil^a

16S rRNA family	Abundance in module 1 (%)	Correlationb	Presence of coxL	Example accession no.c
Acidobacteriaceae	22.23	0.21*	+	WP_035348617
Unknown Acidimicrobiales	0.02	0.37	+	WP_035348617
Actinospicaceae	35.44	0.71**	−	NAd
Cellulomonadaceae	0.08	0.88***	−	NA
Conexibacteraceae	0.03	0.83***	+	ADB51446
Frankiaceae	0.42	0.29	−	NA
Intrasporangiaceae	0.01	0.09	−	NA
Microbacteriaceae	21.12	0.82***	−	NA
Micromonosporaceae	0.13	0.68*	+	WP_033364885
Mycobacteriaceae	0.97	−0.22	+	AFP37210
Nocardiaceae	0.31	0.22	+	WP_031936578
Pseudonocardiacea	0.09	0.78**	+	WP_028847129
Streptomycetaceae	0.19	0.05	+	WP_033825254
Streptosporangiaceae	0.46	0.59*	+	WP_020541099
Unknown Actinomycetales	2.70	0.61	+	WP_033364885
Unknown Solirubrobacterales	0.33	−0.19	+	ADB51446
Ktedonobacteraceae	0.29	−0.33	+	WP_007909428
Thermogemmatisporaceae	0.10	0.54	+	WP_052889641
Alicyclobacillaceae	0.03	0.28	+	WP_026962960
Acetobacteraceae	2.76	0.76**	−	NA
Beijerinckiaceae	2.39	0.45	+	WP_020174914
Bradyrhizobiaceae	7.43	0.79**	+	WP_014491996
Caulobacteraceae	1.27	0.36	−	NA
Hyphomicrobiaceae	0.13	0.85***	−	NA
Methylocystaceae	0.03	−0.63*	−	NA
Rhodospirillaceae	0.06	0.42	+	WP_028465373
Unknown Rhizobiales	0.07	0.93***	+	WP_014491996
Burkholderiaceae	0.50	0.60*	+	ABE35958
Coxiellaceae	0.05	0.22	−	NA
Sinobacteraceae	0.04	0.59*	−	NA
auto67_4W	0.03	0.31	NA	NA
Unknown WPS-2	0.20	0.84***	NA	NA
Unassigned	0.06	0.80**	NA	NA

^aThe relative abundance of sequences encompassing each family and the presence or the absence of the coxL gene in the reference genome of each taxonomic group are presented. Spearman correlations between the absolute abundance of OTU grouped at the family level and CO oxidation rate are presented.

^bCorrelation significance levels are indicated by asterisks: *, P < 0.05; **, P < 0.01; and ***, P < 0.001.

^cFound by sequence identity with blastp (49) using the coxL sequence in Burkholderia xenovorans LB400 (accession no. ABE35958).

^dNA, not applicable.

Correlation of modules' eigengenes with time series of coxL genotypes led to the identification of 42 coxL OTU whose colonization profiles were significantly correlated with those of 16S rRNA gene OTU comprising module 1 (Fig. 5). Among these OTU, 27 encompassed coxL group x, 5 belonged to Actinobacteria, 8 comprised Alphaproteobacteria, and 2 were outside these groups and affiliated with a coxL sequence from Deinococcus-Thermus. Convergence of the taxonomic affiliation and the relative abundance of some rRNA gene OTU in module 1 and coxL OTU (i.e., Bradyrhizobium spp.) supported the phylogenetic analysis of coxL genes (see Table S1 in the supplemental material). No members of the Deltaproteobacteria were present in module 1, impairing the assignation of coxL group x genotype to this taxonomic group. At the end of the incubation, the absolute abundance of the 27 coxL group x OTU correlated with the CO uptake rate was (2.4 ± 0.34) × 10⁹ bacteria g dry weight⁻¹, as estimated by multiplying the relative abundance of the coxL OTU obtained from pyrosequencing profiles by the absolute abundance of coxL group x genes determined by qPCR. Considering the specific richness and the abundance of OTU encompassing the coxL group x genotype, module 1 was searched for potential 16S rRNA gene OTU candidates to infer a taxonomic affiliation with unknown presumptive carboxydovore bacteria. 16S rRNA OTU belonging to Actinospicaceae and Microbacteriaceae were the best coxL group x candidates, as predicted using the correlation network analysis. These two bacterial families were the sole candidates for which the absolute abundance and species richness of 16S rRNA OTU were congruent with those of coxL group x OTU in module 1 (see Table S1).

FIG 5.

Spearman correlations between the coxL OTU with the eigengenes of the five modules comprising 16S rRNA OTU displaying a similar soil colonization profile. Only the eigengene of module 1 was correlated with the CO uptake rate measured in soil microcosms. Significant correlations between the time series of individual coxL OTU and the eigengene of module 1 are marked with asterisks along the side of the heat map (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

DISCUSSION

Enrichment of microorganisms utilizing atmospheric trace gas is challenging because the energy potential of these substrates usually supports mixotrophic survival metabolism, where reduced inorganic and organic compounds are simultaneously used as energy sources for maintenance or survival. Furthermore, increase of energy potential through increased gas concentration is not recommended because it will result in the enrichment of microbes that are incapable of scavenging atmospheric trace gas due to their low substrate affinity. Most high-affinity CO-oxidizing bacteria were identified through screening of isolates obtained from environmental samples or public collections after detection on coxL in sequenced genome. One notable exception was the isolation of the first carboxydovore bacterium, Aminobacter sp. strain cox1, after long-term exposure of soil to 40 to 400 ppmv CO (39). Recent investigations into the diversity of soil carboxydovore bacteria led to the identification of an atypical coxL group x genotype for which the distribution was significantly related to the CO uptake rate in soil (21). Before investment of cultivation efforts to isolate these unknown carboxydovore bacteria, we sought to challenge the hypothetical taxonomic affiliation of atypical coxL genotype to Deltaproteobacteria through the monitoring of microbial succession and maturation of CO uptake activity in soil microcosms.

Successions of 16S rRNA and coxL genes analyzed by qPCR (Fig. 2) unveiled that coxL carboxydovore bacteria generally are opportunistic r-strategist bacteria using labile nutrients in soil. They also increase their relative abundance early in the incubation period. Our data showed that their enrichment was transient, suggesting that their populations were not maintained upon soil nutrient depletion. Interestingly, the CO-oxidizers represented by coxL group x displayed the K-strategist colonization profile, with slow growth of their population sustained by the efficient use of limited resources in soil. At the end of the incubation, total coxL and coxL group x genotypes represented 2.2% and 0.7% of the total bacterial population, respectively. It is believed that atmospheric CO can support mixotrophic metabolism of these carboxydovore bacteria (9). To address that question, we combined our qPCR time series with thermodynamic models. The theoretical population size of high-affinity CO-oxidizing bacteria in soil for which CO oxidation fully supplies maintenance energy requirements was estimated based on the theoretical maintenance energy consumption rate and the free energy of atmospheric CO oxidation, using the following model formulated by Conrad (40): N = [1.4 × 10¹⁴ (−ΔG) d]/mE, where N is the theoretical population size of CO-oxidizing bacteria in soil (expressed as the number of cells per gram of soil dry weight), 1.4 × 10¹⁴ is a constant expressing the density of bacterial cells containing 1 mol carbon (number of cells per mole of C biomass), ΔG is the Gibbs free energy of atmospheric CO oxidation (−235 kJ per mol of CO), d is the measured CO oxidation rate (moles of CO per gram of soil dry weight), and mE (4.5 kJ per mol of C biomass) is the energy maintenance requirement of the population (41). This model estimated that measured CO uptake activity would provide sufficient energy to support the survival of a maximal population 2.5 × 10⁷ carboxydovore bacteria g dry weight ⁻¹. According to our qPCR results, the abundance of total coxL and coxL group x sequences reached (5.5 ± 0.97) × 10⁹ and (1.8 ± 0.20) × 10⁹ copies g dry weight in soil at the end of the incubation, respectively. Taken together, these observations imply that oligotrophic cultivation media comprising both organic carbon and CO would be necessary to isolate unknown carboxydovore bacteria harboring the atypical coxL group x genotype that are expected mixotrophic K-strategists.

Although microbial succession was shaped by stochasticity in the colonial growth dynamic of individual cells (42) and alteration of the chemical structure and bioavailability of soil organic carbon due to autoclaving (43), the taxonomic affiliation of coxL genotypes inferred from phylogenetic analysis resulted in similar distributions compared with the observations made using native deciduous soil (21). Actinobacteria, Alphaproteobacteria, and coxL group x were the dominating coxL genotypes, representing 31% ± 3%, 14% ± 3%, and 53% ± 3% of the total sequences, respectively. Very few studies have reported the taxonomic composition of carboxydovore bacteria in soil. A soil survey along a succession of volcanic deposits unveiled an enrichment of coxL sequences closely related to Burkholderia isolates known to consume atmospheric CO (10) as a function of vegetation regeneration and CO uptake activity (17, 44). In this study, two coxL OTU (OTU 79 and 100) related to Burkholderia xenovorans LB400 coxL sequence were detected, but their temporal distribution profile was not related to the maturation of CO uptake activity in soil. It has been reported that some species of mycobacteria from the Mycobacteriaceae family in the Actinobacteria phylum are capable of consuming atmospheric CO (45, 46). In this study, 14 coxL OTU closely related to Mycobacterium sequences were detected, but none displayed significant correlation with the CO oxidation rate in soil. The high abundance of coxL sequences closely related to Bradyrhizobium spp. and the correlation of their distribution profile with CO uptake activity suggest implication of these Alphaproteobacteria in the activity measured in the microcosms. Although they were not dominant in previous soil surveys, their capacity to consume atmospheric CO has already been reported using environmental isolates (10). Sequences encompassing coxL group x were abundant, with H. ochraceum as the only representative. CO uptake activity has been confirmed in this bacterium (21), but no coxL sequences have been noticed in sequenced genomes of the other aerobic Deltaproteobacteria encompassing the orders Myxococcales and Bdellovibrionales. The combination of qPCR and high-throughput sequencing of 16S rRNA and coxL genes led to the conclusion that it is very unlikely that Deltaproteobacteria harbor the coxL group x genotype. If unknown Deltaproteobacteria harbored the atypical coxL genotype, then a strong primer bias has impaired their detection or considerably underestimated their abundance.

Network correlation analysis led to the identification of taxonomic groups of bacteria for which the distribution was significantly correlated with the distribution of OTU encompassing coxL group x. Although the analysis does not imply any causal effect, it supported the suggestion that unknown carboxydovore bacteria are mixotrophic K-strategists, which is critical for the establishment of the best isolation strategy. Among the taxonomic groups detected in module 1 for which the eigengene was correlated to CO uptake rate, bacteria encompassing the Actinospicaceae and Microbacteriaceae families were the most probable taxonomic assignment of unknown carboxydovore bacteria harboring the coxL group x genotype. This taxonomic inference will need validation through isolation efforts since no representative of these two families was reported as CO-oxidizing bacteria. Interestingly, microorganisms that belong to these taxonomic groups are generally K-strategists, with some of them having the ability to use recalcitrant carbon such as cellulose (47). Furthermore, soil amendment experiments unveiled that cellulose stimulates CO uptake activity in soil, while glucose exerts the inverse effect (I. Lalonde, unpublished data). Bioprospection efforts are currently being used by the authors to identify and characterize unknown carboxydovore bacteria harboring the coxL group x genotype. Alternatively, single-cell sequencing (48) could be envisaged if cultivation efforts are unsuccessful.

Funding Statement

The NSERC funding was via a Discovery grant.