16S rRNA Gene-Based Oligonucleotide Microarray for Environmental Monitoring of the Betaproteobacterial Order “Rhodocyclales”

Tables 1 and 2 list the 12 pure cultures and the 17 16S rRNA gene-containing clones that were used to evaluate the RHC-PhyloChip.

Activated sludge samples were collected in June 2002 from the intermittently aerated nitrification-denitrification basin of an industrial WWTP. This treatment plant received its sewage from a rendering plant (Tierkörperbeseitigungsanstalt Kraftisried, Kraftisried, Germany). For DNA isolation, aliquots (4 ml) of the samples were pelleted by centrifugation (5,000 rpm for 2 min) at the treatment plant, immediately put on ice, and stored at −20°C upon arrival at the laboratory. For fluorescence in situ hybridization (FISH), an activated sludge sample was fixed at the WWTP with paraformaldehyde as outlined previously (15).

Genomic DNA was isolated from reference organisms by using the FastDNA kit (Bio101, Vista, Calif.). DNA from Kraftisried activated sludge was extracted by using a modification (35) of a previously described protocol (21).

For DNA microarray hybridization, 16S rRNA gene fragments from DNA of “Rhodocyclales” reference pure cultures were amplified by using the bacterial primer pair 616V and 630R (Table 3), whereas 16S rRNA gene inserts of reference clones were amplified with cloning vector-specific primers M13F(−20) (5′-GTAAAACGACGGCCAG-3′) and M13R (5′-CAGGAAACAGCTATGAC-3′) (Invitrogen Corp., San Diego, Calif.). Amplification of 16S rRNA gene fragments from DNA of the activated sludge sample was performed by using the bacterial primer pairs 616V and 630R and 616V and 1492R or the newly designed primer pairs A, R, or Z, each targeting different “Rhodocyclales” subgroups (Table 3).

Positive controls containing purified DNA from suitable reference organisms were included in all of the PCR amplification experiments along with negative controls (no template DNA added). For 16S rRNA gene amplification, reaction mixtures (total volume, 50 μl) containing each primer at a concentration of 25 pM were prepared by using 10× Ex Taq reaction buffer and 2.5 U of Ex Taq polymerase (Takara Biomedicals, Otsu, Shiga, Japan). Additionally, 20 mM tetramethylammonium chloride (Sigma, Deisenhofen, Germany) was added to each amplification mixture to enhance the specificity of the PCR (31). Thermal cycling was carried out by using an initial denaturation step of 94°C for 1 min, followed by 30 cycles of denaturation at 94°C for 40 s, annealing at temperatures ranging from 52°C to 60°C (depending on the primer pair [Table 3]) for 40 s, and elongation at 72°C for 1.5 min. Cycling was completed by a final elongation step of 72°C for 10 min.

The presence and sizes of the amplification products were determined by agarose (1%) gel electrophoresis of the reaction product. Ethidium bromide-stained bands were digitally recorded with a video documentation system (Cybertech, Hamburg, Germany).

16S rRNA-targeted oligonucleotides were designed in silico by using the ARB probe design and probe match tools (37) and obtained from MWG Biotech (Ebersberg, Germany). Table 4 lists the sequence, specificity, and microarray position of all oligonucleotide probes. The theoretical melting temperatures (T_m) of the probes were calculated according to the nearest neighbor method by using the OligoAnalyzer 3.0 software with default settings (http://biotools.idtdna.com/analyzer/oligocalc.asp ). For each probe and each reference organism, the free energies, ΔG, of the perfectly matched and the mismatched (up to 4.5 weighted mismatches; as determined by using the ARB probe match tool) probe-target hybrids were calculated. ΔG calculation was performed with the two-state hybridization server (concentration of Na⁺ and temperature were set to 0.829 M and 42°C, respectively) at the mfold website (http://www.bioinfo.rpi.edu/applications/mfold/ ) (63). Additional information on RHC-PhyloChip probes can be viewed at the probeBase website (http://www.microbial-ecology.net/probebase ) (34).

Each oligonucleotide probe contained a spacer element consisting of 15 dTTPs at the 5′ end and was aminated at the 5′-terminal nucleotide to allow covalent coupling to aldehyde group-coated CSS-100 glass slides (CEL Associates, Houston, Tex.). Fluorescence labeling of PCR amplicons, manufacturing and processing of microarrays, and reverse hybridization on microarrays were performed as outlined previously (35). The concentration of oligonucleotide probes before printing was adjusted to 50 pmol μl⁻¹ in 50% dimethyl sulfoxide to prevent evaporation during the printing procedure. RHC-PhyloChips with triplicate spots for each probe were printed by using a GMS 417 contact arrayer (Affymetrix, Santa Clara, Calif.). Spotted DNA microarrays were dried overnight at room temperature in the dark to allow efficient cross-linking. Free aldehyde groups at the slide surface were reduced with sodium borohydride solution (35). For each reference organism, a separate microarray was hybridized, washed, and scanned under identical conditions and settings.

Fluorescence images of the RHC-PhyloChips were recorded by scanning the slides with a GMS 418 array scanner (Affymetrix). The fluorescence signals were quantified by using the ImaGene 4.0 software (BioDiscovery, Inc., Los Angeles, Calif.). A grid of individual circles defining the location of each spot on the array was superimposed on the image to designate each fluorescent spot to be quantified. The mean signal intensity of each spot and the local background area surrounding each spot was determined. Subsequently, for each probe the signal-to-noise ratio (SNR) was calculated according to the following formula:

SNR = [I_P − (I_N − I_NLB)] × I_PLB⁻¹

where I_P is the mean pixel intensity of all replicate probe spots, I_N is the mean pixel intensity of all nonsense probe spots, I_NLB is the mean pixel intensity of the local background area around all nonsense probe spots (note that I_N − I_NLB must always have a lower value than I_P), and I_PLB is the mean pixel intensity of the local background area around all replicate probe spots. Probes for which the SNR was equal to or greater than 2.0 were considered positive (35). Furthermore, in the reference strain evaluation experiments the SNR of each probe was normalized against the SNR of the bacterial EUB338 probe, recorded on the same microarray, according to the following formula:

nSNR = SNR × {[I_EUB − (I_N − I_NLB)] × I_EUBLB⁻¹}⁻¹

where nSNR is the normalized SNR of the specific probe, I_EUB is the mean pixel intensity of all EUB338 probe spots, and I_EUBLB is the mean pixel intensity of the local background area around all EUB338 probe spots.

Prior to cloning, the PCR amplification products were purified by low-melting-point agarose (1.5%) gel electrophoresis (NuSieve 3:1; FMC Bioproducts, Biozym Diagnostics GmbH, Oldendorf, Germany) and stained in SYBR Green I solution (10 μl of 10,000× SYBR Green I stain in 100 ml of TAE buffer [40 mM Tris, 10 mM sodium acetate, 1 mM EDTA, pH 8.0]; Biozym Diagnostics GmbH) for 45 min. Bands of the expected size were excised from the agarose gel with a glass capillary and melted with 80 μl of double-distilled water for 10 min at 80°C (this procedure was also done for the amplicon obtained with primer pair A, although no band was visible). Four microliters of each solution was ligated as recommended by the manufacturer (Invitrogen Corp.) into the cloning vector pCR2.1 of the TOPO TA cloning kit. Nucleotide sequences were determined by the dideoxynucleotide method (46) following a previously published protocol (42).

All phylogenetic analyses were performed by using the alignment and treeing tools implemented in the ARB program package (37). All almost-full-length 16S rRNA sequences (>1,300 bases) which have been assigned to the order “Rhodocyclales” in the preview release of the RDP II database (version September 2003) (12) and all 16S rRNA sequences obtained from the activated sludge samples in this study were added to an ARB alignment of about 20,000 small-subunit rRNA sequences by using the alignment tool ARB_EDIT. Alignments were refined by visual inspection. Chimeric “Rhodocyclales” sequences were identified by independently subjecting base positions 1 to 513, 514 to 1026, and 1027 to 1539 (Escherichia coli numbering) of the 16S rRNA sequence to phylogenetic analysis. Inconsistent affiliation of the gene fragments in the phylogenetic trees was interpreted as being caused by a chimeric sequence. In addition, the CHECK_CHIMERA program of the RDP II was used for confirmation. Ambiguous base positions were excluded during calculation of 16S rRNA sequence similarities.

16S rRNA phylogenetic analyses were performed by applying distance-matrix, maximum-parsimony, and maximum-likelihood methods (36). A representative assortment of type strain sequences of different orders of the Beta- and Gammaproteobacteria was used as the outgroup for treeing. Variability of the individual alignment positions was determined by using the ARB_SAI tools and used as a criterion to remove or include variable positions for phylogenetic analyses. The neighbor-joining method combined with a Jukes-Cantor correction was used to infer distance-matrix trees. Maximum-likelihood trees were calculated by Tree-puzzle (54) and by applying the new A(x)ccelerated Maximum-Likelihood (AxML) algorithm (52). Maximum-parsimony analyses (treeing and bootstrapping) were performed with the Phylogeny Inference Package (PHYLIP, version 3.57c, J. Felsenstein, Department of Genetics, University of Washington, Seattle). For parsimony bootstrap analysis, 100 resamplings were used. All phylogenetic consensus trees were drawn according to recommendations outlined previously (36).

The abundance of selected “Rhodocyclales” groups in the activated sludge sample was determined by FISH combined with subsequent image analysis (14, 47). Fluorescently labeled oligonucleotide probes (Table 5) (34) were purchased from Thermo Hybaid (Ulm, Germany). Hybridization under optimal conditions was performed as described previously (27, 38).

The names of the bacterial taxa were used in accordance with the prokaryotic nomenclature proposed in the latest taxonomic outline of the second edition of Bergey′s Manual of Systematic Bacteriology (http://dx.doi.org/10.1007/bergeysoutline200310 ) (20).

The sequences determined in this study were deposited at GenBank under accession numbers AY689085 to AY689093 .

The latest taxonomic outline of Bergey′s Manual of Systematic Bacteriology lists 30 validly published species assigned into the 12 recognized genera of the “Rhodocyclaceae” (Azoarcus, Thauera, Zoogloea, Azovibrio, Azospira, Rhodocyclus, Propionivibrio, Dechloromonas, Ferribacterium, Quadricoccus, Azonexus, and Sterolibacterium), the only family within the betaproteobacterial order “Rhodocyclales.” The order additionally encompasses the species “Denitromonas aromaticus” and the as yet uncultured Candidatus species “Accumulibacter phosphatis” (25), both of which still await valid description. In addition, 92 isolates and 104 environmentally retrieved 16S rRNA sequences affiliated to this order were included in the analysis.

To establish a robust and detailed phylogenetic backbone for subsequent design of microarray probes and for future taxonomic and environmental studies, an evaluation of the phylogeny of cultivated and yet uncultivated “Rhodocyclales” was performed. Initially, 10 sequences from environmental 16S rRNA clones (GenBank accession numbers AJ009452, AF245350, AF280861, AF281119, AY118150, AF529340, AY082472, AB089100, AB089101, and AF204249 ) were identified as chimeras and omitted from all subsequent analyses. The remaining 218 almost-full-length 16S rRNA sequences of “Rhodocyclales” were phylogenetically analyzed by calculating similarities and applying distance-matrix, maximum-parsimony, and maximum-likelihood methods for treeing.

The minimum 16S rRNA sequence similarity of two members of the “Rhodocyclales” was 88.1%. This is in the range of minimal similarities previously reported for other bacterial families e.g., 83% each for “Desulfobacteraceae” and “Desulfovibrionaceae” (33), 89% for “Nitrosomonadaceae” (30), and 90% for Chlamydiaceae (18), and therefore legitimates subclassification of all “Rhodocyclales” into the single family “Rhodocyclaceae” (http://dx.doi.org/10.1007/bergeysoutline200310 ) (20) from an rRNA-based point of view.

Independently of the phylogeny inference method applied, members of the “Rhodocyclaceae” could be subdivided into nine different monophyletic lineages (Fig. 1). The phylogenetic positions of these lineages to other betaproteobacterial orders could not be unambiguously determined, as shown by a polytomic tree topology (Fig. 1). With the exception of the anaerobic consortium clone SJA-109 lineage, designated according to an environmental clone sequence from an anaerobic, trichlorobenzene-transforming microbial consortium (58), each lineage was represented by at least one validly described or cultured species, indicating that lineage-level biodiversity of “Rhodocyclaceae” is well reflected by available cultured strains. Detailed phylogenetic trees showing the affiliation of all members of each “Rhodocyclaceae” lineage can be downloaded at http://www.microbial-ecology.net/supplements.asp (supplemental Fig. S1).

In general, the same strategies for in silico development and the same technical set-up for fabrication and hybridization of the RHC-PhyloChip were used as for the development of a 16S rRNA-targeted oligonucleotide microarray for detection of all lineages of recognized sulfate-reducing prokaryotes (SRP-PhyloChip) (35).

Initially, the specificities of all previously published 16S rRNA-targeted probes and primers for “Rhodocyclales” (13, 24-27, 43, 45) were reevaluated with the updated 16S rRNA database. Eighteen probes were found to target “Rhodocyclales” only and were therefore included on the RHC-PhyloChip, although not all of them target monophyletic “Rhodocyclales” groups (Table 4 and supplemental Fig. S2). In addition, 60 oligonucleotide probes were designed according to the “multiple probe concept” (3, 8) to specifically target “Rhodocyclales” at hierarchical and identical phylogenetic levels (Table 4 and Fig. S2). Because multiple nested probes can at least partly compensate for unspecificities of individual probes (32, 35), this probe design strategy is particularly valuable if microarray formats are used which only allow hybridization or washing at a single stringency. In summary, the RHC-PhyloChip probe set consists of 78 specific probes covering the complete diversity of “Rhodocyclales” known so far (see above), two probes targeting betaproteobacterial groups at a broader specificity (BONE663 and BTWO663) (4), six bacterial or universal probes, and two probes as positive and negative hybridization controls (CONT and NONSENSE) (Table 4). All probes were designed to have the same length (18 bases, excluding the T-spacer), but the G+C contents of the probes varied between 38.9 and 77.8%. To attenuate the influence of differing G+C contents of probe-target duplexes on duplex stabilities, 3M tetramethylammonium chloride was added to the washing buffer (8, 35, 39).

The specificity of the individual probes was tested under monostringent conditions (i.e., the same hybridization and washing conditions for all probes and microarrays). Cy5-labeled 16S rRNA gene amplicons of each pure culture and each 16S rRNA gene-containing clone (n = 29) were hybridized with a separate RHC-PhyloChip. For 60 “Rhodocyclales”-specific probes, this set of reference 16S rRNA genes contained at least one sequence with a perfectly matched target site. None of these probes showed false-negative signals. Out of the 18 probes for which no perfectly matched reference sequence was available, 10 yielded positive signals with mismatching reference 16S rRNA gene amplicons, indicating that the respective probe-target sites were accessible for hybridization. A detailed list of the individual hybridization results can be downloaded at http://www.microbial-ecology.net/supplements.asp (supplemental Table S1). Seven of the probes hybridized nonspecifically with many reference organisms not having fully complementary probe target sites and were therefore excluded from the final version of the RHC-PhyloChip (listed separately in Table 4 and supplemental Table S1).

In order to compare the hybridization efficiency of the different RHC-PhyloChip probes, the signal-to-noise ratios (SNRs) of the probes were normalized against the SNR of the bacterial probe EUB338 recorded on the same microarray. The resulting nSNRs for perfectly matched probe-target duplexes ranged from 0.2 (for probe A33-587 with Kraftisried WWTP clone A33) to 48.0 (for probe BTWO663 with Azonexus fungiphilus), demonstrating that the signal intensities of individual probes vary strongly if excess target DNA is added. It has been observed previously that the duplex yield of different rRNA gene-targeted microarray probes can differ considerably (32, 35, 41), and on the RHC-PhyloChip the duplex yield was significantly positively correlated with the theoretical T_m of the probe (Spearman nonparametric correlation test: R = 0.342, P = 0.013). The latter point demonstrated that the addition of tetramethylammonium chloride did not completely abolish the influence of the G+C content on the duplex yield of different probes.

Of the 2,291 different hybridizations (each reference DNA with each probe) which were performed in total, 208 (9.1%) were false-positive (positive probe signal with a nontarget organism having one or more mismatches in the probe target site), while no hybridizations were false-negatives (supplemental Table S1). The occurrence of some false-positive results is almost impossible to avoid with a monostringent microarray hybridization approach, because the stability of mismatched probe-target hybrids is difficult to predict in silico and influenced by many factors, such as (i) the number of mismatches, (ii) the nature of the mismatching nucleotides, (iii) the position of the mismatches in the probe target site, and (iv) possible stacking interactions of nucleotides adjacent to the mismatches (19, 53, 56, 57). However, specific identification of target organisms is still possible with the RHC-PhyloChip because of the multiple probe concept (the theoretical specificities of the nested probes are depicted in supplemental Fig. S2). Nevertheless, future microarray probe design would be further improved if oligonucleotide probe parameters were available which allow estimation of the hybridization behavior of each probe in silico (53).

One hybridization parameter that might be a suitable candidate is the free energy, ΔG, of a given perfectly matching or mismatching probe-target hybrid (23, 55). On the RHC-PhyloChip, the ΔG values of most (88%) of the false-positive probe-target hybrids with one or two mismatches were in a similar range (−22 to −16 kcal mol⁻¹) to ΔG values of all perfectly matched probe-target pairs (−25 to −17 kcal mol⁻¹), providing an explanation of why discrimination was not successful under monostringent hybridization conditions. As can be inferred from Fig. 2, most of the positive probe-target hybrids (84%) (including false-positive signals) had a ΔG below −16 kcal mol⁻¹. Additionally, only 3% of all probe-target combinations that yielded no hybridization signal had also a ΔG below −16 kcal mol⁻¹. Therefore, a ΔG threshold of −16 kcal mol⁻¹ could provide useful guidance for future preselection of appropriate probes in silico (Fig. 2) but does not abolish the need for extensive empirical testing of the hybridization behavior of microarray probes. It should be further stressed that the ΔG threshold of −16 kcal mol⁻¹ might only apply to the microarray set-up and the hybridization conditions used in this study.

To demonstrate the applicability of the RHC-PhyloChip for rapid screening of “Rhodocyclales” diversity in environmental samples, activated sludge from the nitrifying-denitrifying WWTP Kraftisried was analyzed. Kraftisried was chosen as a model system because in 1996 members of diverse lineages of “Rhodocyclales” comprised more than one third of the entire bacterial biovolume in this WWTP (27).

Initially, 16S rRNA genes were amplified from Kraftisried DNA by using standard bacterial primers, fluorescently labeled, and hybridized with the RHC-PhyloChip. Surprisingly, the hybridization pattern obtained did not indicate the presence of members of the “Rhodocyclales” below the lineage level [some probes (BTWO663, ATD1459, RHC630, RHC175a, RHC143, RHC222, and RHAC855) targeting “Rhodocyclales” at broader phylogenetic levels showed positive signals but almost all probes of higher specificity were negative] (Fig. 3A). To find an explanation for this unexpected result, the relative abundance of “Rhodocyclales” in this WWTP was analyzed quantitatively by FISH. Compared to 1996, the relative abundance of “Betaproteobacteria” in the activated sludge from 2002 decreased from 47 to 18% of all bacteria detectable by FISH (Table 5). Similarly, the abundance of members of the “Rhodocyclales” decreased dramatically between the two samples. While the activated sludge from 1996 contained significant amounts of “Rhodocyclales” detectable by probes AT1458, S-*-OTU1-1415-a-A-20, and S-*-OTU3-0445-a-A-20, [each targeting a “Rhodocyclales” group found previously in this WWTP (27)] (Table 5), less than 1% of the cells hybridized with these probes in the WWTP sample from 2002.

To increase the sensitivity of the RHC-PhyloChip, three “Rhodocyclales”-subgroup-selective primer pairs called A, R, and Z (together targeting almost all “Rhodocyclales”) (Table 3) were designed and applied for amplification of 16S rRNA genes prior to microarray hybridization. Although these new primers were selected to amplify 16S rRNA gene fragments of the maximum possible length, the target sites of some RHC-PhyloChip probes are outside the amplified 16S rRNA gene region (Table 4), and these probes must thus be ignored during interpretation of hybridization patterns.

Each “Rhodocyclales”-subgroup-selective primer pair was used separately for amplification of Kraftisried activated sludge DNA at low stringency (Table 3) to allow potential primer binding to 16S rRNA genes of “Rhodocyclales” having mismatches in the primer target sites. PCR products of the expected length were obtained for primer pairs R and Z, but no primer pair A PCR product was observed after gel electrophoresis. The “Rhodocyclales” subgroup-selective PCR amplicons obtained were fluorescently labeled and hybridized with two separate RHC-PhyloChips. The RHC-PhyloChip hybridization patterns of the R and Z amplicons differed from each other and from the pattern obtained by using general bacterial primer pairs (Fig. 3A). In more detail, the hybridization pattern obtained with primer pair R indicated the presence of bacteria related to the genera Ferribacterium and Dechloromonas, whereas the hybridization pattern obtained with primer pair Z pointed to the presence of Zoogloea species.

A composite microarray fingerprint of the “Rhodocyclales” community present in activated sludge from Kraftisried was created by merging the separate microarray hybridization patterns obtained with the “Rhodocyclales”-subgroup-selective and the common bacterial 16S rRNA gene amplicons (Fig. 3A). Besides Ferribacterium/Dechloromonas-related bacteria and Zoogloea species, this composite microarray fingerprint additionally indicated the presence of members of the Sterolibacterium lineage (Fig. 3B).

The microarray results were confirmed independently by cloning and sequencing of the 16S rRNA gene PCR products obtained with the three “Rhodocyclales” subgroup-selective primers. It should be noted that cloning of PCR products amplified from Kraftisried DNA with primer pair A was successful, although only small amounts of PCR product could be retrieved (see above). All 16S rRNA gene clones obtained with primer pairs Z and A were closely related to clones already found in the Kraftisried WWTP in 1996 (27) and belonged to the genus Zoogloea and the Sterolibacterium lineage, respectively (Fig. 4). In contrast, all 16S rRNA gene sequences obtained by using primer pair R clustered with members of the genera Dechloromonas and Ferribacterium (Fig. 4), which were not detected by the 16S rRNA full-cycle approach in Kraftisried WWTP samples from 1996 (27). The phylogeny of all retrieved 16S rRNA gene sequences was in perfect agreement with the microarray results. Furthermore, the sequenced 16S rRNA genes have perfectly matched target sites for the probes that showed a positive signal in the RHC-PhyloChip analyses (Figs. 3 and 4).

The microarray hybridizations, the retrieved 16S rRNA sequences, and the quantitative FISH data collected in this study provide corroborating evidence that substantial changes have occurred within the “Rhodocyclales” community in Kraftisried since the first bacterial community analysis of this WWTP (27). The dramatic decline in “Rhodocyclales,” assumed to be the major denitrifiers in this system (27), from 35% of the total bacterial biovolume in 1996 to less than 1% in 2002 may have been caused by the seasonal implementation of a partial ammonium stripping step prior to biological nitrogen removal in 1999. This physical sewage treatment step reduces the ammonia concentration and increases the salt concentration in the sewage and probably had dramatic consequences for the population structure of nitrifiers (data not shown) and potentially denitrifying heterotrophs in the activated sludge.