Polygenic analysis of ammonia-oxidizing bacteria using 16S rDNA, amoA, and amoB genes

Summary. Finding a unique molecular marker capable of quickly providing rigorous and useful phylogenetic information would facilitate assessing the diversity of ammonia-oxidizing bacteria in environmental samples. Since only one of several available markers can be used at a time in these kinds of studies, the 16S rDNA, amoA and amoB genes were evaluated individually and then compared in order to identify the one that best fits the information provided by the composite dataset. Distance-based neighbor-joining and maximum parsimony trees generated using the sequences of the three mentioned genes were analyzed with respect to the combined polygenic trees. Maximum parsimony trees were found to be more accurate than distance-based ones, and the polygenic topology was shown to best fit the information contained in the sequences. However, the taxonomic and phylogenetic information provided by the three markers separately was also valid. Therefore, either of the functional markers (amoA or amoB) can be used to trace ammonia oxidizers in environmental studies in which only one gene can be targeted. [Int Microbiol 2005; 8(2):103-110]

Key words: ammonia-oxidizing bacteria · 16S rDNA · amoA · amoB · polygenic analysis

Introduction

Environmental and biotechnological interest in ammonia-oxidizing bacteria (AOB) has increased tremendously in recent years. However, their slow growth and the difficulty to be cultured have necessitated the development of a variety of culture-independent techniques for carrying out ecologic and taxonomic studies [18,19,21,28,48,53,54]. These techniques include the use of 16S rDNA and protein-encoding genes to characterize natural AOB populations [4,7,25,42,46] and to analyze their taxonomic and phylogenetic features [1,2,6,37,38]. Nonetheless, although 16S rDNA sequences are suitable for providing a comprehensive long-term evolutionary view of prokaryotic taxonomy, they fail to discriminate among close relatives, such as species within a given group or genus [39]. In addition, considerable variability can be found among organisms with almost identical 16S rDNA genes [3]. Thus, while 16S rDNA has proven useful in the discrimination between nitrosococci and nitrosomonads [5,55], the outcome is confusing when examining a single genus, such as Nitrosospira [41]. For this reason, protein-encoding genes, such as amoA, have been added to the collection of comparative tools used by taxonomists and molecular ecologists for diversity studies [14,46]. Gene amoA codes for the active site of ammonia monooxygenase [30], and it has been extensively used for the detection and study of ammonia oxidizers, particularly in natural environments [1,15,21]. According to Rotthauwe et al. [41], amoA is more useful at a fine-scale than 16S rDNA. By contrast, Ludwig and Schleifer [27] stated that the 16S rDNA gene is the best marker to infer phylogenetic relationships, since the topologies derived from 16S rDNA are in accordance with those obtained using markers with rather diverse functions. This has been recently supported by Purkhold et al. [38], who showed higher resolution using 16S rDNA than amoA within the traditional classification. Recently, amoB has been shown to be a suitable molecular marker for the study of AOB, as it has a high capacity of resolution within genera. In addition, its phylogeny is highly consistent with the current taxonomic outline [6].

The reconstruction of phylogenetic relationships between closely related species requires the use of markers with significant mutation rates; however, the accumulation of recurrent mutations results in the incorporation of large amounts of mutational homoplasy into the molecular data [47]. In addition, when mutations occur repeatedly at the same site, those that occurred later mask the previous ones, rendering the sequences useless for phylogenetic purposes. This phenomenon, known by geneticists as substitution saturation, should be taken into account before proceeding with any type of phylogenetic analysis [12]. It should also be noted that polymorphisms detected in the sequences of a given population reveal not only the mutations experienced by the ancestors but also the consequences of evolutionary forces, such as genetic drift and natural selection. It is therefore essential to check whether the molecular dataset has been affected by evolutionary pressures, especially since the neutral theory has become the standard null hypothesis in the study of molecular evolution [13,22].

In the present work, the 16S rDNA, amoA, and amoB genes were used to determine whether or not a polygenic or single-marker analysis was more suitable for taxonomic studies of AOB. These three markers were evaluated independently in a panel of genetic tests to compare the amount of useful information contained in their respective sequences. The phylogenetic trees constructed from each gene were then weighted against the composite sequence dataset to identify the marker that best reproduced the information resulting from the polygenic tree.

Materials and methods

Sequences. 16S rDNA, amoA, and amoB partial gene sequences from a total of 20 AOB strains of the β- and γ-subclasses of Proteobacteria were obtained from the databases and used in this study (Table 1). The sequences of two methane oxidizers, Methylocystis sp. and Methylosinus trichosporium, from the β- and γ-subclass of Proteobacteria, respectively, were also included and used as outgroups for phylogenetic reconstruction.

]]> Mutational model. Multiple sequence alignments were performed with CLUSTAL W [51] and refined manually. The proportions of variable and conserved positions were calculated with DNAs

amoA and amoB were manually checked by comparing the DNA sequences with the translated amino-acid sequences.

The hypothesis of neutrality in nucleotide substitution was tested using Tajima's D test [50], included in the software MEGA v.2.1 [26]. The test was independently performed for every marker and for the three positions of the codons from the amoA and amoB sequences. Substitution saturation was determined with the index developed by Xia et al. [58] which is included in the DAMBE software [57]. This test is based on the notion of entropy in information theory and yields a critical value permitting the saturation degree of a given set of aligned sequences to be assessed. The same saturation index was calculated for the first, second, and third codon positions in amoA and amoB. In addition, the entire sequences of the three markers were tested individually for saturation.

The nucleotide substitution model best fitting the variations observed in the 16S rDNA, amoA, and amoB partial sequences was determined using the software MODELTEST 3.04 [36]. This program allows the most appropriate among 56 models of nucleotide substitution to be chosen.

Phylogenetic analysis. Neighbor-joining (NJ) trees for 16S rDNA, amoA, and amoB genes were generated from the corresponding matrix of nucleotide divergence between sequences using the program MEGA2 [26]. Maximum parsimony (MP) trees were also constructed for each marker using the software PAup 4.0b [49]. To reduce the computational time required by the parsimony algorithm when carried out with a heuristic search, a TBR branch-swapping value of 100 was used. Confidence in the branching points was obtained with 1000 bootstra

amoA, and amoB sequence datasets provided similar phylogenetic information. The overall NJ and M

Tree topologies were compared using maximum likelihood, minimum evolution, and parsimony criteria. First, the topologies were analyzed according to the modified Kishino and Hasegawa test [44], computing the log-likelihoods per site for each tree and comparing the total log-likelihoods among topologies [11]. Minimum evolution (ME) scores were then compared for each topology. Finally, the number of steps and both the consistency and retention indices of the parsimony analysis for each tree were computed.

Results

Quantitative aspects of gene variation. Sequences of amoA contained the highest proportion of polymorphic sites. Of the 399 sites in this gene, 203 (62.15%) were found to be variable; of these, 11.27% were silent and 50.88% effective. In amoB, 180 (45.80%) out of 393 genes were variable and all of them were effective. Of the 1014 16S rDNA genes analyzed, 278 (27.42%) were variable. There were 231 parsimony-informative sites in 16S rDNA, 232 in amoA, and 172 in amoB. Approximately 50% of the nucleotide substitutions in amoA affected the third position of the codon (Table 2), while in the case of amoB the variable positions were evenly distributed. Furthermore, 44.27% of the polymorphisms detected in the third base-pair of amoB were silent substitutions, i.e., they had no effect on the amino-acid sequence.

Neutrality and substitution saturation. The dataset fit the model of neutral molecular evolution. In fact, the results of Tajima's D test indicated no significant skew in the entire sequences of the three markers in the case of amoB (0.505, 0.421, and 0.623 for the positions 1, 2, and 3, respectively). However, this test revealed a significant excess of polymorphisms in the third position of the codons in amoA (3.131 in contrast with the values 1.091 and 0.315 for the positions 1 and 2, respectively). These results agreed with measurements of substitution saturation, which produced a strong signal in the third position of amoA codons (Fig. 1). The persistent accumulation of changes in these specific sites in amoA may produce a loss of phylogenetic information. No substitution saturation was detected in the other two markers.

Phylogenetic and topology analyses. The evolution of each gene can be described by a distinct substitution model (Table 3). Tamura-Nei 1993 (TrN93) is the nucleotide substitution model including the greatest number of parameters, and the one best fitting the combined dataset. The models Hasegawa-Kishino-Yano 1985 (HKY) and Felsenstein 1981 (F81), obtained for amoA and amoB, respectively, can be considered simplifications of TrN93. TrN93 was applied to all trees based on genetic distances, with a single transition type and a single substitution rate when the selected models were HKY and F81, respectively.

For each marker, an NJ tree was constructed using the appropriate nucleotide substitution model (Table 3, Fig. 2). The trees constructed by MP showed topologies similar to those of their NJ counterparts (data not shown). In all cases, the Nitrosomonas and Nitrosospira radiations grouped together, and the γ-proteobacterial nitrosococci branched separately. This agrees with the classical phylogenetic topology of AOB. Likewise, two different clusters were distinguishable within the β-subgroup of ammonia oxidizers, as Nitrosomonas and Nitrosospira clearly formed two separate clades. Nevertheless, the allocation of Nitrosomonas aestuarii Nm36 was uncertain, since it grouped within the Nitrosospira cluster when using amoB as a molecular marker but fell within the Nitrosomonas group when using 16S rDNA or amoA. The ILD test corroborated (P < 0.001) the incongruence between the phylogenetic information provided by the three markers. However, since under some circumstances combining sequences with different phylogenetic histories can improve the accuracy of phylogenetic analysis [56], polygenic trees were constructed.

The consensus polygenic trees generated by M ]]> Fig. 3. The topologies of the two trees were similar and consistent with both the standard classification of AOB and the results previously obtained with each of the three markers. In this polygenic analysis, Nm. aestuarii Nm36 was considered to be the most divergent Nitrosomonas.

All topological evaluations (likelihood, minimum evolution, and parsimony criteria) indicated that the MP tree obtained from the composite dataset displayed the most probable topology (Table 4). Similar values were obtained for the rest of the trees, which indicated that they were not significantly worse than the best-supported tree.

Discussion

Correct classification of any bacterial group requires the input of different genetic and phenetic characters, which is not possible when using uncultured bacteria from natural environments [39]. Alternatively, a polygenic approach leads to more accurate estimations of the diversity and composition of natural populations [34,35], although care must be taken when combining datasets from different markers [9,56]. Gene amoA encodes the active site of ammonia monooxygenase [20], which makes it difficult for effective mutations to occur in this gene. In amoA, 50% of the mutations were detected in the third base-pair (see Table 2), which showed a large accumulation of nucleotide changes at this position. As a consequence, the third position of the codon in amoA is strongly saturated (see Fig. 1) and deviates from neutrality, suggesting that this position has experienced selective pressures different from those of the other two positions [24]. The persistent accumulation of changes in these specific sites in amoA may produce a loss of phylogenetic information. By contrast, in the case of amoB, ca. 55% of the conserved sites were detected in all codon positions (Table 2). Moreover, ca. 70% of the amino-acid variations observed in the deduced partial amino acid sequences of AmoB proteins are conservative (data not shown).

The general topologies of the constructed trees were almost identical, with one exception. Although taxonomically classified into the genus Nitrosomonas, the strain Nm. aestuarii Nm36 showed significant phylogenetic distances, supported by high bootstra

Nitrosomonas when using 16S rDNA and amoA gene sequences. By contrast, Nm. aestuarii Nm36 grouped together with the Nitrosospira lineage when using amoB, but a considerable phylogenetic distance also distinguished this strain from the rest of the nitrosospiras. In the polygenic tree, Nm aestuarii Nm36 again grouped with the Nitrosomonas cluster. Purkhold et al. also reported the ambiguous phylogenetic arrangement of this species depending on the treeing method employed and the type of sequences used [37]. Therefore, this strain should be further studied in order to clarify its phylogenetic affiliation. Moreover, it would be of interest to determine whether amoB of Nm. aestuarii Nm36 has followed a different pattern of evolution and represents the ancestral state within the Nitrosomonas cluster, or whether it is a case of lateral gene transfer.

amoA: 399 bp; amoB: 393 bp) and the number of parsimonic informative sites, they support 16S rDNA as a good phylogenetic marker, especially concerning the avoidance of ambiguous classifications. Several authors have recently reaffirmed the potential of 16S rDNA sequences for drawing phylogenetic inferences [3,27,38]. Nonetheless, obtaining the 16S rDNA gene from environmental samples is time-consuming and tedious. It requires the cloning of all 16S rDNA genes present in the sample and then distinguishing the 16S rDNA genes belonging to AOB from the rest.

By contrast, environmental population studies based on the analysis of amoA or amoB present some significant advantages: the genes are AOB-specific, are large enough to allow quick fingerprinting of natural communities, and provide a phylogeny consistent with the current taxonomic outlines. Nevertheless, Oved et al. [33] and Nicolaisen and Ramsing [31] reported the amplification of non-AOB sequences when using amoA sequences in a PCR-denaturing gradient gel electrophoresis (DGGE) approach. Our experiments based on amoB amplification combined with DGGE resulted in the establishment of a sensitive and reliable screening method to detect and identify AOB in environmental samples (data not shown). Additionally, the benefit of using amoB in ecophysiology studies is the ability to distinguish methane-oxidizing bacteria from AOB on a simple agarose gel [6].

Based on the results reported here, for taxonomic purposes we strongly recommend sequencing16S rDNA, amoA and amoB genes, and to construct a polygenic tree. Since the third position of the codon in amoA is saturated, and due to the non-AOB sequences retrieved by other authors when using this gene [31,33], the use of amoB is recommended when carrying out environmental ecophysiology studies. amoB allows fingerprinting techniques, such as terminal restriction fragment length polymorphism (tRFLP) and DGGE, to be performed, and results in a reliable phylogenetic profile. Moreover, when using amoB as a marker, the methane-oxidizers present in the sample can be quickly and easily distinguished from AOB, which may be of great hel

Acknowledgements. This work was funded by the Spanish Ministry of Science and Technology (CICYT REN2000). LC is the recipient of a fellowshi