|Home | About | Journals | Submit | Contact Us | Français|
Norway spruce (Picea abies) forests exhibit lower annual atmospheric methane consumption rates than do European beech (Fagus sylvatica) forests. In the current study, pmoA (encoding a subunit of membrane-bound CH4 monooxygenase) genes from three temperate forest ecosystems with both beech and spruce stands were analyzed to assess the potential effect of tree species on methanotrophic communities. A pmoA sequence difference of 7% at the derived protein level correlated with the species-level distance cutoff value of 3% based on the 16S rRNA gene. Applying this distance cutoff, higher numbers of species-level pmoA genotypes were detected in beech than in spruce soil samples, all affiliating with upland soil cluster α (USCα). Additionally, two deep-branching genotypes (named 6 and 7) were present in various soil samples not affiliating with pmoA or amoA. Abundance of USCα pmoA genes was higher in beech soils and reached up to (1.2 ± 0.2) × 108 pmoA genes per g of dry weight. Calculated atmospheric methane oxidation rates per cell yielded the same trend. However, these values were below the theoretical threshold necessary for facilitating cell maintenance, suggesting that USCα species might require alternative carbon or energy sources to thrive in forest soils. These collective results indicate that the methanotrophic diversity and abundance in spruce soils are lower than those of beech soils, suggesting that tree species-related factors might influence the in situ activity of methanotrophs.
Methanotrophic bacteria in aerated soils contribute significantly to the global flux of methane (CH4) and consume about 30 Tg year−1 of atmospheric CH4 (12). Forest soils are considered to be the most effective sinks of CH4 among these soils (8, 48). Atmospheric CH4 oxidation and the methanotrophic communities in forest soils may be affected by various abiotic factors, e.g., pH, water content, and soil temperature (32).
Cultivated methanotrophs belong to the families Methylocystaceae and Beijerinckiaceae (both Alphaproteobacteria), Methylococcaceae (Gammaproteobacteria), and the recently proposed Methylacidiphilaceae (Verrucomicrobia) (43). Most of these cultivated species are restricted to one-carbon compounds as carbon and energy sources (53). Only Methylocella species may grow on multicarbon compounds (9). Isolates of Methylocystis (i.e., strains LR1, DWT, and SC2) utilize atmospheric CH4 (1.8 ppm by volume [ppmv]); however, growth is not apparent at this concentration (3, 16, 28).
Genes of the CH4 monooxygenase (MMO) are well-established functional markers of methanotrophs (38). pmoA encodes the hydroxylase of membrane-bound CH4 monooxygenase (pMMO) while mmoX encodes the corresponding subunit of cytoplasmic CH4 monooxygenase (sMMO). Facultative methanotrophic species of Methylocella exclusively possess sMMO and thus will not be detected in pmoA analyses (10, 38). Atmospheric CH4-oxidizing communities have been analyzed in a variety of forest soils, and Methylocystis species and upland soil cluster α (USCα)-affiliated genotypes are detected in most of these soils (32). Methanotrophs affiliated with USCγ and clusters 1 and 5 are also abundant or even dominant at certain forest sites (30, 31, 33). Numbers of methanotrophs in forest soil range from 0.4 × 106 to 21 × 106 cells per g of dry weight (gDW) as determined by quantitative PCR (qPCR) (29, 33), and these bacteria have resisted cultivation (32).
Coniferous forests can have lower annual atmospheric CH4 uptake rates than deciduous forests (6, 40). European beech (Fagus sylvatica) soils exhibit a lower capacity to oxidize CH4 than Norway spruce (Picea abies) soils, indicating a causative effect of tree species on methanotrophic community (11). In the current study, the methanotrophic community composition and abundance of methanotrophs in soil from forests with adjacent beech and spruce stands were assessed to evaluate the potential effect of tree species on methanotrophic communities.
Soil samples were taken from three temperate forests with European beech (F. sylvatica) and adjacent Norway spruce (P. abies) stands in Germany, i.e., Solling, Steigerwald, and Unterlüß. Soils at any sampling site were classified as dystric cambisols (11). Five soil cores were taken at each of the six sites and stored in open plastic bags at 5°C for up to 4 days until they were separated into depth layers, manually homogenized, and frozen at −80°C. Soil pH values were similar at beech and spruce sites and ranged from 3.9 to 5.4 (11). Samples from the Oa horizon (blackish humic layer) and from mineral soil at a depth of 0 to 5 cm (0- to 5-cm soil) were pooled and used for molecular community analyses since these layers exhibited the highest atmospheric CH4 oxidation rates (11).
DNA was extracted from 0.3 g of frozen soil (0.1 g of Oa horizon plus 0.2 g of 0- to 5-cm mineral soil) according to the protocol of Stralis-Pavese and colleagues (51) by mechanical, chemical, and enzymatic lysis. The following four modifications to the protocol were made. (i) Lysis buffer I used for initial extraction of soil also contained 20 mg ml−1 of polyvinyl pyrrolidone. (ii) Lysis buffer II for reextraction of soil samples was identical with buffer I. However, polyvinyl pyrrolidone and lysozyme were not added. (iii) The DNA pellet was dried for 15 min at 40°C after the washing step. (iv) DNA was eluted with Tris-HCl (pH 8.0). DNA extracts were quantified using a Quant-iT PicoGreen dsDNA kit (Invitrogen, Germany). For each of the 30 soil cores, DNA was extracted in triplicate, i.e., 90 extracts in total.
Gene libraries of pmoA, mmoX, and Verrucomicrobia-related pmoA were set up. DNA extracts in triplicate per soil sample were pooled, and 1 ng of DNA was used as a template in PCRs. PCR premix F (Epicentre) combined with an Invitrogen Taq polymerase recombinant (Invitrogen, Germany) was used. pmoA was amplified with primers A189f (GGN GAC TGG GAC TTC TGG) and A682r (GAA SGC NGA GAA GAA SGC) (500 nM) (20). The initial denaturation step was at 94°C for 5 min, and the final elongation was at 72°C for 10 min. Denaturation and elongation during repeated PCR cycles was at 94°C (1 min) and 72°C (1 min), respectively. An 11-cycle touchdown was performed with a stepwise decrease of the annealing temperature from 65°C to 55°C (1 min), followed by 24 cycles with an annealing temperature of 55°C (1 min). Taq DNA polymerase was added after the initial denaturation step. mmoX was amplified with primers mmoX206f (ATC GCB AAR GAA TAY GCS CG) and mmoX886r (ACC CAN GGC TCG ACY TTG AA) (500 nM) (22). Five precycles with an annealing temperature of 56°C were performed, followed by 35 cycles at 59°C. Genomic DNA of Methylococcus capsulatus was used as positive control for pmoA- and mmoX-specific PCRs. Four novel primers were designed to prove the presence of methanotrophic Verrucomicrobia in the soil samples, i.e., Verruco29f (5′-AAG AYM GRA TGT GGT G-3′), Verruco120f (5′-GCC YAT AGG WGC RAC MT-3′), Verruco391r (5′-GTC CAT AGT ATT CCA C-3′), and Verruco500r (5′-ACD CCH CCN GCA AAR CT-3′). PCR was performed with pooled soil DNA extracts from all soil cores of the six sites (i.e., beech and spruce of Solling, Steigerwald, and Unterlüß). Since no positive control for Verrucomicrobia-related pmoA was available, PCR was performed with different primer combinations and annealing temperatures from 50°C to 70°C, with and without the addition of 0.12% bovine serum albumin (BSA), and as simple or nested assays.
PCR products were purified by agarose gel electrophoresis, and bands of the right size were excised and purified with a Montage Gel Extraction Kit (Millipore). Equal concentrations of all five PCR products of one sampling site were pooled and cloned by AGOWA (Germany). Per sampling site, 48 to 96 clones were sequenced by MacroGen (South Korea).
16S rRNA gene and pmoA gene sequences from 22 pure cultures were retrieved from GenBank (5). The gene fragments corresponded to regions amplified by the commonly used primer pair A189f and A682r (20) and the pair 27F and 1492R (35), respectively, and were further analyzed in ARB (36). Gene fragments were aligned, and pmoA sequences were in silico translated. Pairwise comparisons of all pmoA (DNA), PmoA (amino acids), and 16S rRNA gene sequences were performed to calculate the nucleotide or amino acid difference, D (18). The percentage sequence similarity of pmoA and PmoA sequence pairs was plotted versus the percentage sequence similarity of the 16S rRNA gene sequence pairs of the same strains. The similarity, S, was calculated according to the following equation: S = 1 − D.
pmoA sequences retrieved from the six pmoA gene libraries of the sampled sites were aligned and in silico translated. Amino acid sequences were analyzed with the programs DOTUR (46) and ∫-LIBSHUFF (47) using an uncorrected distance matrix as input data set. The number of operational taxonomic units (OTUs) achieved by DOTUR was calculated for a 7% distance, which is an estimated cutoff value for methanotrophic species-level OTUs (Fig. (Fig.1).1). Coverage, C, of pmoA gene libraries was calculated according to the following equation, where n is the number of OTUs with only one sequence (singletons), and N is the number of clones examined: C = [1 − (n/N)] × 100.
Bacteria were quantified with a modified assay established by Zaprasis and coworkers (55) using primers Eub341f (CCT ACG GGA GGC AGC AG) and Eub534r (ATT ACC GCG GCT GCT GG) (42). New qPCR assays (Q-USCα, Q-C7, and Q-AOB) were developed to assess the abundance of pmoA genes of USCα, pmoA and amoA genes of cluster 7 (C7), and amoA genes of ammonium oxidizing bacteria [AOB] within the Betaproteobacteria) detected in the soils. The following novel primers were designed and used in combination with common published primers (Table (Table1):1): USCα-346f (5′-TGG GYG ATC CTN GCN C-3′), C7-128r (5′-CCA ATG GGG AGC CTA AAT-3′), amoA38f (5′-AAT GGT GGC CGG TKG TNA C-3′), and amoA200r (5′-GAC CAC CAG TAR AAD CCC C-3′). The establishment of a primer system for the specific quantification of cluster 6 did not succeed; i.e., primers yielded unspecific by-products. Specificity of the novel qPCR assays was evaluated by means of nontarget DNA that had at least two mismatches at the binding site of the target group-specific primers. pmoA fragments of the different detected genotypes and of M. capsulatus (NCIMB 11853) were used as nontargets. Dilution series of target DNA quantified with a Quant-iT PicoGreen double-stranded DNA (dsDNA) kit (Invitrogen, Germany) were applied as quantitative calibration standards. A 16S rRNA gene fragment of Escherichia coli amplified with primers 27f and 907mr (35) was used as standard DNA for the Bacteria-targeting assay. For Q-USCα, Q-C7, and Q-AOB, an M13 PCR product of a respective clone insert for pmoA and amoA was used.
All PCRs were performed in triplicate in 20-μl volumes using Thermosprint 96 PCR plates sealed with Thermosprint transparent sealing tapes (Bilatec, Germany). Five microliters of DNA template, i.e., 1 to 6 ng of DNA, were added to 15 μl of Master Mix containing the primers and the iQ SYBR Green Supermix (Bio-Rad, Germany). Final primer concentrations were 1,000 nM for Q-AOB and 500 nM for Q-USCα, Q-C7, and the Bacteria-targeting assay. The initial denaturation step was at 95°C for 8 min. Denaturation and elongation during PCR was at 95°C (30 s) and 72°C (15 s for Q-USCα and Q-C7, 20 s for Q-AOB, and 25 s for the Bacteria-targeting assay), respectively. Fluorescence data were acquired during the elongation step at 72°C since primer dimer formation was not evident by melting curve analysis. More details on the assays are presented in Table Table1.1. qPCR was performed on an iCycler thermocycler with an iQ5 multicolor real-time PCR detection system. Data were analyzed with iQ5 optical system software (version 2.0; Bio-Rad, Germany).
Target group specificity of the assay Q-USCα was further evaluated by cloning (TOPO TA Cloning Kit; Invitrogen, Germany) purified (PCR purification kit; Qiagen, Germany) qPCR products derived from soil extracts and subsequent sequencing of 10 clone insert sequences. All retrieved sequences were pmoA genes and affiliated with USCα (data not shown).
The assay INHIB-CORR using primers T7-Prom (TAA TAC GAC TCA CTA TAG GG, Promega, Germany) and M13r (CAG GAA ACA GCT ATG ACC) (41) was established to calculate a factor correcting for PCR inhibition in each DNA extract caused by coextracted inhibitory compounds, such as humic substances (54). The vector region of the plasmid pCR 2.1-TOPO (Invitrogen, Germany) without a DNA insert amplified with primers M13f and M13r (41) served as a target for the assay INHIB-CORR and was used as inhibition control DNA. Every DNA extract was spiked with 0.2 × 104 inhibition control DNA molecules μl−1. Spiked DNA did not affect the amplification efficiency of the genotype-specific assays Q-USCα, Q-C7, and Q-AOB, and the Bacteria-targeting assay (data not shown). Final primer concentrations of the assay INHIB-CORR were 500 nM, and elongation during PCR was performed for 15 s. Further details are given in Table Table11.
Spiked environmental DNA extracts were measured with genotype-specific assays and with INHIB-CORR. The logarithmic values of the starting quantity measured with the genotype-specific assays [lg(SQ)measured] were corrected for the inhibition factors determined with INHIB-CORR and for the difference in amplification efficiency (E) between the genotype-specific assays and INHIB-CORR (equation 1). Spiked DNA extracts were diluted to keep inhibition factors below a factor of 2. Corrected logarithmic values of the starting quantity [lg(SQ)corrected] were then converted to gene numbers per ng of DNA (ngDNA−1):
For comparison of USCα gene abundances measured using the new Q-USCα assay with those measured using the assay FOREST (34) in a previous work (33), both assays were applied to the same DNA extracts. In silico coverage and measured numbers of USCα genes were compared (see Table S2 in the supplemental material).
pmoA and amoA sequences were deposited at the European Molecular Biology Laboratory (EMBL) database under accession numbers FN564572 to FN564934.
Sequence similarities of pmoAs and PmoAs and 16S rRNA genes of 22 methanotrophic isolates were linearly correlated, as suggested previously (19). This linear correlation was the basis to define distance cutoff values for methanotrophic species-level OTUs. The 97% species cutoff value based on 16S rRNA genes (49) corresponded with similarity cutoff values of 13% and 7% at the DNA and protein levels, respectively, for methanotrophic species-level pmoA-based OTUs (Fig. (Fig.1).1). Ten species-level USCα-affiliated OTUs (USCα 1 to 10) were resolved when these species-level cutoff values were used to analyze all available sequences (i.e., those published and those obtained in the current study) (Fig. (Fig.2).2). Clusters 5 and MHP were composed of nine and five species-level OTUs, respectively. Sequences within cluster AC (aquatic clones) and Methylocapsa acidiphila-related sequences formed separated OTUs. USCα and clusters 5 and MHP might represent three methanotrophic genera.
pmoA sequences were analyzed to resolve the identity of methanotrophic bacteria in pooled Oa horizon and mineral soil samples from the three forests, Solling, Steigerwald, and Unterlüß; 41 to 85 sequences were obtained per sampling site, with a total of 204 and 162 sequences from beech and spruce soils, respectively. Coverage of gene libraries based on the 7% distance cutoff value (Fig. (Fig.1)1) was close to 100% for most gene libraries, indicating sufficient sampling at most sites (Solling beech, 93%; Solling spruce, 82%; Steigerwald beech, 99%; Steigerwald spruce, 100%; Unterlüß beech, 100%; and Unterlüß spruce, 100%). Nine species-level OTUs were retrieved from beech soils (Fig. (Fig.3).3). These sequences affiliated with seven different OTUs within USCα (USCα-1, -2, -6, -7, -8, -9, and -10) and two deep-branching clusters (clusters 6 and 7) (Fig. (Fig.33 and and4)4) that might represent PmoA or AmoA sequences. Cluster 6 includes the environmental sequence EU723753 (retrieved from a forest soil) (26). The novel cluster 7 exclusively contains sequences obtained from Solling beech soil samples. In contrast, only five OTUs were detected in spruce soils (USCα-1, -4, -7, and -8 and cluster 6), indicating that methanotrophic diversity in spruce soils was lower than that in beech soils, as confirmed by statistical analysis (∫-LIBSHUFF, P < 0.05) (47). USCα-1 was the most frequently detected OTU associated with both tree species. USCα-1 includes RA14, the first pmoA genotype found to be potentially associated with atmospheric methane oxidation (21).
amoA sequences of Betaproteobacteria were occasionally found in Solling beech and Unterlüß spruce soils but not in Steigerwald soils. A high proportion of amoA sequences were detected in spruce soil from Solling (Fig. (Fig.3)3) and led to a lower coverage of methanotrophic OTUs than in other sampling sites and probably to an underestimation of methanotrophic diversity at this site. mmoX and Verrucomicrobia-related pmoA genes were not detected in any soil sample.
In silico coverage with the USCα primers used in the FOREST assay (34) approximated 86% of the USCα genotypes detected in the current study, whereas coverage with the new primers used in the new Q-USCα assay was 98% (see Table S2 in the supplemental material). Q-USCα pmoA gene numbers exceeded those detected with FOREST by a factor of up to 8, as revealed from Steigerwald beech soil DNA extracts (see Table S2). Gene numbers of cluster 7 were below the detection limit of the corresponding qPCR assay (8.8 × 10−1 genes ngDNA−1). Abundance of USCα was higher in beech than in spruce soils at all three forest sites (Fig. (Fig.5A)5A) (P > 0.05). USCα pmoA gene numbers ranged from 1.3 × 103 to 3.3 × 103 pmoA genes ngDNA−1, i.e., 0.3 × 108 to 1.2 × 108 pmoA genes gDW−1. amoA genes exceeded pmoA genes by a factor of 10, ranging from 1.3 × 104 to 3.7 × 104 amoA genes ngDNA−1, i.e., 0.3 × 109 to 2.6 × 109 amoA genes gDW−1 (Fig. (Fig.5B).5B). Highest abundance of amoA genes was detected in Solling spruce soil samples. Abundance of bacterial 16S rRNA genes was generally higher in spruce soils than in beech soils (Fig. (Fig.5C).5C). Gene numbers ranged from 0.4 × 106 to 1.1 × 106 16S rRNA genes ngDNA−1, i.e., 6.3 ×109 to 7.2 × 1010 16S rRNA genes gDW−1. Assuming that a methanotrophic cell contains two pmoA copies (50) and that a bacterial cell contains 4.13 16S rRNA gene copies (27), ratios of USCα to total Bacteria cell numbers were up to 2% and 0.5% for beech and spruce soils, respectively.
pmoA numbers and atmospheric CH4 oxidation rates of the six forest soils (11) were used to estimate atmospheric CH4 oxidation rates per cell. This calculated cell-specific activity was higher in beech than in spruce soil samples from Unterlüß and Solling, whereas cell-specific activities were similar in beech and spruce soils from Steigerwald (Table (Table22).
USCα is dominant in a variety of forest soils (24, 30, 32, 33). Consistently, pmoA sequences most frequently related to USCα were detected in pmoA gene libraries of Solling, Steigerwald, and Unterlüß forest soils in the current study. Clone libraries revealed a lower diversity of methanotrophic bacteria in spruce than in beech soils at all three sampling sites (Fig. (Fig.3),3), evidence that tree species affect methanotrophic community structure. USCα-1 was most frequently detected in both beech and spruce soils, indicating that this species-level OTU might constitute a dominant taxon.
Differences mainly occurred in less abundant species-level OTUs. USCα-2, -6, -9, and -10 and cluster 7 were not detected in spruce soils. Coniferous trees release monoterpenes that may reduce CH4 oxidation (1, 37). Atmospheric CH4 consumption rates are often lower in coniferous forest soils than in deciduous forest soils (6, 11, 40). USCα-2, -6, -9, and -10 and cluster 7 might represent genotypes that are more sensitive to monoterpenes than USCα-1, -4, -7, and -8 and cluster 6. Tree species did not affect methanotrophic community composition in a boreal tree succession experiment, as revealed by membrane lipid fingerprinting (39). However, the analysis of pmoA genes in the current study provides a more refined resolution of genotypes and facilitates the identification of methanotrophs on genus and species levels. Differences observed in the methanotrophic community structure between beech and spruce soils might reflect differences in monoterpene concentrations between coniferous and deciduous forests (37).
Methylocystis or alternative atmospheric CH4-oxidizing bacteria, such as those belonging to clusters 1 and 5 (32), were not detected in the forest soils of Solling, Steigerwald, and Unterlüß. mmoX genes were not detected, indicating that facultative methanotrophic species of the genus Methylocella and other sMMO-possessing methanotrophs were either not present or below the detection limit. Methanotrophic Verrucomicrobia bacteria were recently isolated from acidic environments (17, 23, 44). It was hypothesized that these bacteria might have been overlooked in other acidic environments due to their divergent pmoA gene that cannot be amplified with common pmoA primers (17). Nonetheless, Verrucomicrobia-like pmoA genes could not be amplified in the investigated soils with new primers that target pmoA1 and pmoA2 of Verrucomicrobia, suggesting the absence of this taxon. AOB of the genus Nitrosospira were highly abundant in Solling spruce soil (Fig. (Fig.33 and and5).5). This forest soil had a 3- to 13-fold higher ammonium content than the other soils (11), which might explain the high proportion of AOB in the gene library. It cannot be excluded that atmospheric CH4 was oxidized to some extent by AOB since CH4 is a substrate homolog for ammonia monooxygenase or that CH4 oxidation activity was lowered by the competitive inhibition of MMO by ammonium in the spruce soil of Solling forest (4, 13).
Abundance of USCα and its relative contribution to the total bacterial community were slightly higher in beech than in spruce soils (Fig. (Fig.5).5). pmoA gene numbers ranged from 0.3 × 108 to 1.2 × 108 pmoA genes gDW−1, values that exceeded those obtained with the FOREST assay (0.9 × 106 to 6.2 × 106 pmoA genes gDW−1) (29, 33). The primers used with the PCR assay FOREST cover only 81% of the available sequences (see Table S2 in the supplemental material), and the new Q-USCα assay yielded 8-fold higher gene numbers than the FOREST assay (see Table S2). Thus, the lower coverage of primers used in the FOREST assay might contribute to the differences in abundances measured in the previous and present study.
Calculated cell-specific CH4 oxidation rates were slightly higher in beech than in spruce soils (Table (Table2).2). A CH4 oxidation rate of at least 40 × 10−18 mol of CH4 cell−1 h−1 may be required for maintenance metabolism (33). Estimated USCα-specific oxidation rates approximated 800 × 10−18 mol of CH4 cell−1 h−1, suggesting that these organisms conserve enough energy for growth during the oxidation of atmospheric CH4 (33). However, these rates may have been overestimated since pmoA gene numbers were likely underestimated with the FOREST assay. Cell-specific CH4 oxidation rates calculated in the current study ranged from 1 × 10−18 to 14 × 10−18 mol of CH4 cell−1 h−1; these values are lower than required for maintenance of methanotrophic biomass, which raises the question of how atmospheric CH4 oxidizers are capable of growth in aerated soils (14). CH4 production was not measured in soil samples from the six sampling sites under anoxic conditions (11), and, thus, it is unlikely that USCα species grow also on CH4 produced in soil. Another possible survival mechanism would be the utilization of substrates other than or in addition to CH4 (14). It is known that Methylocella silvestris, which was isolated from a forest soil, is able to utilize alternative carbon compounds, such as acetate or methanol (9, 15). This suggests that USCα might rely on additional carbon sources to conserve enough energy for cell maintenance and growth. The capacity to use alternative carbon sources might thus lead to a high relative abundance of USCα (i.e., up to 2% of the total detected bacterial community), as has been observed in this study. Nonetheless, calculations of cell-specific CH4 oxidation rates are based on the detection of DNA and, thus, on the assumption that all detected genes represented active cells. This assumption might be not fully true and might have led to an underestimation of cell-specific CH4 oxidation rates.
Soil pH, temperature, and water content may impact community structure of methanotrophs (6, 8, 32). Nonetheless, different CH4 oxidation rates occur in acidic beech and spruce soils at equal matrix potentials and temperatures, indicating that tree species itself is an important environmental determinant of atmospheric CH4 consumption in forest soils (11). Monoterpenes have been proposed as factors that lower methanotrophic metabolic activities in spruce soils (1, 37). These collective results indicate that the methanotrophic diversity and abundance in spruce soils are lower than those of beech soils, suggesting that tree species-related factors might influence the in situ activity of methanotrophs.
We thank Henri Siljanen and Levente Bodrossy for providing the DNA extraction protocol and Egbert Matzner for support and helpful discussions.
The project was associated with the METHECO project within the EuroDIVERSITY program of the European Science Foundation (ESF). The study was financially supported by the Deutsche Forschungsgemeinschaft (grant number KO2912/2-1), the ESF, and the University of Bayreuth.
Published ahead of print on 26 March 2010.
†Supplemental material for this article may be found at http://aem.asm.org/.