Meta-Analyses of Dehalococcoides mccartyi Strain 195 Transcriptomic Profiles Identify a Respiration Rate-Related Gene Expression Transition Point and Interoperon Recruitment of a Key Oxidoreductase Subunit

Abstract

A cDNA-microarray was designed and used to monitor the transcriptomic profile of Dehalococcoides mccartyi strain 195 (in a mixed community) respiring various chlorinated organics, including chloroethenes and 2,3-dichlorophenol. The cultures were continuously fed in order to establish steady-state respiration rates and substrate levels. The organization of array data into a clustered heat map revealed two major experimental partitions. This partitioning in the data set was further explored through principal component analysis. The first two principal components separated the experiments into those with slow (1.6 ± 0.6 μM Cl⁻/h)- and fast (22.9 ± 9.6 μM Cl⁻/h)-respiring cultures. Additionally, the transcripts with the highest loadings in these principal components were identified, suggesting that those transcripts were responsible for the partitioning of the experiments. By analyzing the transcriptomes (n = 53) across experiments, relationships among transcripts were identified, and hypotheses about the relationships between electron transport chain members were proposed. One hypothesis, that the hydrogenases Hup and Hym and the formate dehydrogenase-like oxidoreductase (DET0186-DET0187) form a complex (as displayed by their tight clustering in the heat map analysis), was explored using a nondenaturing protein separation technique combined with proteomic sequencing. Although these proteins did not migrate as a single complex, DET0112 (an FdhB-like protein encoded in the Hup operon) was found to comigrate with DET0187 rather than with the catalytic Hup subunit DET0110. On closer inspection of the genome annotations of all Dehalococcoides strains, the DET0185-to-DET0187 operon was found to lack a key subunit, an FdhB-like protein. Therefore, on the basis of the transcriptomic, genomic, and proteomic evidence, the place of the missing subunit in the DET0185-to-DET0187 operon is likely filled by recruiting a subunit expressed from the Hup operon (DET0112).

INTRODUCTION

The pervasiveness of recalcitrant halogenated organics in hazardous waste cleanup sites (1) makes bioremediation utilizing Dehalococcoides mccartyi populations an attractive tool (2). All known D. mccartyi strains are obligate hydrogenotrophs and organohalide respirers that reductively dehalogenate a wide spectrum of halogenated organics (3), suggesting broad applicability for the bioremediation of various contaminants. However, despite the isolation and genome sequencing of several strains (D. mccartyi strains 195 [4, 5], CBDB1 [6, 7], VS [8–10], GT [11], BAV1 [10, 12, 13], BTF08 [14], and DCMB5 [14]), the complete set of enzymes participating in the central electron transport chain (ETC) and the organization of the ETC remain unknown. Previous studies have identified two main classes of enzymes involved in the ETC: reductive dehalogenases (RDases) (5), which pass an electron to halogenated organics, and hydrogenases (H₂ases), which oxidize the sole known electron donor of D. mccartyi, hydrogen (15, 16). Although expression has been confirmed at both the mRNA and protein levels for RDases (differing with the strain), H₂ases (Hup, Hym-1, Hym-2, Vhu, Ech, and Hyc [5, 17, 18]), and other putative membrane-bound oxidoreductase proteins, the interactions among these proteins are not well described (19). Three of the other notable putative oxidoreductases are DET0187, a predicted (but not biochemically confirmed) molybdopterin-containing protein annotated as a formate dehydrogenase that maintains an atypical serine in its active site rather than the typical cysteine or selenocysteine (15); Nuo, a NADH-ubiquinone oxidoreductase that lacks the NADH-receiving subunits (NuoEFG) in its predicted operon; and Mod, an additional predicted (but not biochemically confirmed) molybdopterin-containing oxidoreductase.

The main goal of this study was to observe broad transcriptional expression patterns during the steady-state respiration of chloroorganics in order to generate and test hypotheses about which proteins are working in concert to achieve energy-conserving organochlorine respiration in D. mccartyi. Previous microarray studies performed on D. mccartyi include work characterizing genomic variability across strains or cultures (20–23) and monitoring the transcriptomic responses of both pure and mixed cultures to a range of conditions (24–26, 55, 56). However, these studies were performed in batch reactors and on limited numbers of samples. Because our primary interest was in the enzymes involved in the putative ETC shuttling electrons from H₂ to chlorinated organics, studies that controlled the cellular respiration rate were required.

To avoid the dynamic limitations inherent to batch experiments, this study investigated cultures that were continuously fed. Previous experiments investigating continuously fed cultures tied the abundances of preselected transcripts, determined via reverse transcription-quantitative PCR (RT-qPCR) assays, to the respiration rate of D. mccartyi strain 195 (27–30). Those studies found the strongest correlations with the respiration rate for the tceA and hupL transcripts. In the current study, we used two-color microarrays to monitor the transcriptome of the sole organohalorespiring D. mccartyi strain, strain 195, in the well-characterized D2 mixed-community culture (27–32) (the metagenome is freely available at JGI under taxon identification [ID] number 2088090019) in a total of 53 experimental cultures and 2 control cultures (the control-control hybridizations). Experimental cultures were grown under conditions with varying electron acceptor feed rates (from 0 to 85 μM Cl⁻/h), electron acceptor types (tetrachloroethene [also known as perchloroethylene {PCE}], trichloroethene [TCE], cis-dichloroethene [cDCE], 2,3-dichlorophenol [DCP], or no electron acceptor), electron donor-to-acceptor ratios (from 0 to 17 on an H₂ electron equivalence basis), and electron donor types (butyrate, lactate, yeast extract, fermented yeast extract, pure H₂, or endogenous biomass decay) as described previously (29). By performing meta-analyses of the full transcriptome rather than a limited number of preselected RT-qPCR targets, genome-wide expression responses were noted, suggesting novel connections between annotated proteins. From these displayed transcriptional relationships, one hypothesis on the complexation of key oxidoreductases was further explored in comparative genomic analyses and proteomics-based assays.

MATERIALS AND METHODS

Culture growth conditions.

All 100-ml subcultures studied in this experiment were derived from a previously described 6-liter mixed-community culture containing D. mccartyi strain 195 (the D2 culture [27–32]) (the sequenced D2 metagenome is freely available at JGI's Integrated Microbial Genomes database under taxon ID 2088090019). The culture contains a D. mccartyi strain 195 population (∼5 × 10⁸ cells/ml; 60% of the total community population) mixed with methanogens and organisms that ferment organic substrates. Three-day-starved biomass from the main D2 reactor served as the inoculum for all experiments (and the reference control in the microarray analysis).

A total of 53 experimental cultures were run; 50 were continuously fed. Three cultures were not fed (the no-electron-acceptor, no-electron-donor series) and represent starving cultures. The shorthand abbreviation for experiments is formatted as follows: Electron acceptor type_Electron donor type_replicate number_expected electron acceptor feed rate (in μM Cl⁻/h). For example, the label for the second replicate culture fed PCE and butyrate at 40 μM Cl⁻/h would be PCE_Butyrate_2_40. The cultures were analyzed as either biological duplicates or triplicates.

The setup for the 50 continuously fed cultures has been described previously (27, 30). In brief, Pressure-Lok syringes (VICI Precision Sampling, Baton Rouge, LA) were loaded onto a Cole Parmer (Vernon Hills, IL) syringe pump, model 74900, to deliver the electron acceptor and donor stocks at the desired rates. The continuous-feed experiments for these 50 cultures were established with varying electron acceptor feed rates (0 to 85 μM Cl⁻/h or 0 × 10⁻¹⁰ to 3.4 × 10⁻¹⁰ μM Cl⁻/cell/h), electron acceptor types (tetrachloroethene [PCE], trichloroethene [TCE], cis-dichloroethene [cDCE], 2,3-dichlorophenol [DCP], or no electron acceptor), electron donor-to-acceptor ratios (from 0 to 17 on an H₂ electron equivalence basis), and electron donor types (butyrate, lactate, yeast extract, fermented yeast extract, pure H₂, or endogenous biomass decay). The complete setup for each of these experiments is detailed in the literature (29, 30) and in Table S1 in the supplemental material.

The three starved cultures were established with 100 ml of D2 in a 160-ml sealed crimp vial. These cultures were incubated as described above, but no amendments were added; the cultures were sampled 3 days (one culture) and 7 days (two cultures) after setup.

GC.

Chlorinated ethenes were assayed from headspace samples by utilizing a gas chromatography (GC) with flame ionization detection (FID) method described previously (27, 28, 31). To assay the chlorinated phenols, 1 ml of liquid sample was withdrawn from the subcultures and was frozen. The thawed samples were transferred to a 2-ml glass vial. The samples were acidified; 100 μl of hexane (99+%; Acros Organics, Geel, Belgium) was added to each sample; and the vials were shaken at 300 rpm for 6 h. Five microliters of the hexane sample was injected onto a GC-FID with an HP-1 column (5 m by 0.53 mm with 2.65-μm film; Hewlett Packard, Palo Alto, CA) at a constant FID oven temperature of 160°C (33). The injector and detector temperatures were 220°C and 240°C, respectively.

Calculation of respiration rates.

In D. mccartyi strain 195, vinyl chloride (VC) is considered the respiration endpoint for chlorinated ethenes (VC is dehalogenated to ethene cometabolically) (27). Therefore, 3, 2, and 1 chloride ion (Cl⁻) are respired in the reduction of PCE, TCE, and cDCE, respectively, as described previously (28). In the DCP experiments, monochlorophenol (MCP) is the terminal product (34, 35). Therefore, D. mccartyi strain 195 respires 1 Cl⁻ ion per mol of DCP. The respiration rate for DCP (r__DCP), expressed in μM Cl⁻/h, is calculated as d[MCP]/dt, where [MCP] is the nominal molar concentration of MCP in μmol/liter and t is the time in hours.

For each experiment, two calculations of respiration were made: the whole-experiment respiration rate and the instantaneous respiration rate. The whole-experiment respiration rate is the respiration rate averaged for all sampling time points. The instantaneous respiration rate is calculated as the respiration rate at the final time point (when RNA was sampled). Overall steady-state respiration was established within 6 h; RNA sampling occurred at 24 h to 168 h.

Microarray design.

The microarray was an Agilent Technologies two-color, 15K, 60-mer, 8-plex array that was manufactured using phosphoramidite chemistry printing technology (Agilent Technologies, Santa Clara, CA). The array probe set targeted D2 sequences for all D. mccartyi strain 195 annotated protein-encoding transcripts, community member 16S rRNA sequences, D. mccartyi strain 195 non-protein-encoding RNA transcripts (rRNAs and tRNAs), and luciferase (as a control). The probes were designed according to the temperature-matching (Tm) method provided in the eArray software suite (Agilent). However, if a base composition (BC) score (a quality metric ranging from 1 [good] to 4 [poor]) of >3 was reported for a transcript, multiple probes around the melting temperature of 80°C were designed. The probe with the best BC score nearest the 80°C temperature was selected. To confirm the specificity of the probes, the probes designed were searched (using the Basic Local Alignment Search Tool [BLAST]) against both the National Center for Biotechnology Information (NCBI) nonredundant nucleotide collection and the assembled D2 mixed-community metagenome (JGI ID 2088090019). The microarray platform design is freely available at the NCBI Gene Expression Omnibus (GEO) database (GPL10023) (http://www.ncbi.nlm.nih.gov/geo/). Additionally, this probe set was included on a previously validated D. mccartyi pangenome array (36).

Microarray probe set validation.

The probe set from this study was included on the pangenome array published by Hug et al. (36). In that study, the probe set was challenged with labeled DNA from multiple pure strains of D. mccartyi. All probes in the current probe set were shown to capture DNA signals from a pure culture of D. mccartyi strain 195. However, when D. mccartyi strain 195 DNA from D2 was run on the pangenome microarray in that study, a subsection of probes did not fluoresce with significant signal intensity—suggesting that the D2 populations of D. mccartyi strain 195 lack some of the genes of the isolated strain 195. The D. mccartyi strain 195 genes that were annotated as protein encoding and passed the threshold analysis reported in that study were used in this analysis (n = 1,451).

RNA-cDNA handling for microarray assays.

After the experiment was concluded (24 h to 168 h after the start), 50 ml of liquid culture was centrifuged at 14,190 × g for 10 min and was then frozen at −80°C for <1 week. Each centrifuged sample was thawed and split into eight individual RNA extracts using the RNeasy minikit (Qiagen, Hilden, Germany) as outlined previously (28). The eight distinct RNA extracts were recombined on one spin filter before the first wash in this kit. The luciferase RNA addition, DNase cleanup (DNase I and Turbo DNase; Life Technologies, Grand Island, NY), SuperScript II (Life Technologies) aminoallyl cDNA synthesis (GE Healthcare, Buckinghamshire, United Kingdom), cDNA cleanup, and cDNA labeling with cyanine 3 or cyanine 5 (Cy3 or Cy5, respectively; GE Healthcare) followed a method outlined previously (23). The quality and quantity of the RNA pool was determined using an RNA 6000 Nano assay on an Agilent 2100 bioanalyzer (Agilent Technologies). The quantity of resulting cDNA was determined by using the Quant-iT OliGreen ssDNA assay kit (Life Technologies) and a NanoDrop 2000c spectrophotometer (Thermo Scientific, Waltham, MA). A common control RNA pool sampled from the main D2 reactor after 3 days of starvation (representing the preexperiment condition for all experimental cultures) was labeled with Cy3.

Microarray hybridization and scanning.

For each experiment, 400 ng of Cy5-labeled experimental cDNA was hybridized against 400 ng of Cy3-labeled control cDNA. Additionally, two associated control-control hybridizations (labeling a common control pool in both the Cy3 and Cy5 channels) were run to check for probe label incorporation and dye biases. The hybridization, washing, and scanning of the microarray samples were performed by the Cornell University Microarray Core Facility and followed the methods outlined by the manufacturer (37). The arrays were scanned with an Agilent Technologies scanner, model G2505C, at a 5-μm resolution.

Microarray data processing.

Variability among arrays was addressed by comparing the Cy5 experimental channel with a common Cy3 reference control. Microarray images were analyzed using Feature Extraction (version 10.5) image analysis software (Agilent). The Feature Extraction software linearly subtracted the background intensity, performed a within-array modified LOESS (locally estimated scatterplot smoothing) normalization between the Cy5 and Cy3 signals, and calculated a base-10 log ratio between normalized signals from the Cy5 and Cy3 channels (38, 39). A complete guide to the detailed methods employed by the Agilent Feature Extraction software can be found in the user manual (38). The method described above was employed for each probe spot; every probe designed was replicated in 6 to 20 spots across the microarray. Therefore, replicate spots for identical probes were geometrically averaged, and the variability within replicate spots was additionally recorded.

The average positive signal had to exceed the mean of the array background (i.e., negative-control spots) plus 3 standard deviations in order to be considered detected. Values below the background intensity were replaced with the background value. Additionally, the signal had to be 10% below saturation (average saturation value, 7 × 10⁶ processed fluorescence units [PFU]) to be considered quantifiable (40, 41).

Statistical analysis of the data set.

The heat map with associated hierarchical clustering was constructed in R, version 3.0.2 (http://www.r-project.org/), with the heatmap.2 function. Hierarchical clustering was run with a Euclidean distance matrix, and the data were subsequently Ward clustered (42). Principal component analysis (PCA) was conducted using the princomp function in R, version 3.0.2. The data set considered in this PCA was the log ratios of the microarray gene expression data across the 53 experiments relative to the control and the two control-control arrays. The resulting PCA develops two matrices: the loading matrix (representing the experimental conditions) and the score matrix (representing the transcript log ratios). The PCA was run using the default Pearson correlation matrix generated in the princomp function.

Native gel separation and proteomic analysis.

Blue native gel separations were performed by previously published methods (43). In brief, a stable pool of proteins was established by growing a 5-liter culture derived from the D2 mixed community at a continuous respiration rate of 2.5 μM Cl⁻/h for 800 h. The culture was centrifuged and was frozen at −80°C. Two hundred microliters of 1% (wt/vol) digitonin provided in the NativePAGE Sample Prep kit (Life Technologies) was added to the thawed samples (each representing the collected mass of 125 ml of culture). Each sample received 100 μl of 75-μm-diameter glass beads and was sealed in a 1.5-ml screw-cap tube. The samples were shaken in a bead beater for 10 s (intensity, 4; FastPrep DNA extractor; Savant Instruments, Holbrook, NY) and were centrifuged at 13,000 × g for 10 min to remove cell debris. One-fifth of the resulting crude extract (20 to 25 μl, representing 25 ml of the original culture) was reacted with the Coomassie blue G-250 sample additive (Life Technologies) at a ratio of 5% (wt/vol).

The stained sample was loaded into the lanes of the 4-to-16% NativePAGE Bis-Tris gel (Life Technologies) in a room at a constant temperature of 11°C. Anode and cathode buffers were prepared according to the manufacturer's specifications. The gel was loaded into an XCell SureLock minicell (Life Technologies). A NativeMark (volume, 5 μl; Life Technologies) unstained protein standard was run as a control. After the run, the gel lanes were separated, and two experimental lanes and one ladder lane were stained with silver nitrate (44) to visualize the gel bands. Slices corresponding to dominant bands were excised from the two non-silver-nitrate-developed lanes, stored in glacial acetic acid, and later destained of Coomassie blue. They were subsequently trypsin digested and were analyzed via liquid chromatography-tandem mass spectrometry at the Proteomics Facility at the Helmholtz Centre for Environmental Research, Leipzig, Germany, by following the methods and Mascot protein-scoring algorithms described by Tang et al. (43).

Microarray data accession number.

The microarray data determined in this study have been uploaded and are freely available at the NCBI BioProject database (accession number GSE26287).

RESULTS

Overall microarray characteristics.

The D. mccartyi strain 195-specific microarray contained 1,510 probes that were designed for the putative protein-encoding genes. Of these 1,510 designed probes, 1,487 had been shown previously by DNA hybridization analysis to be specific and sensitive to the D. mccartyi strain 195 population in the D2 community (36). The 23 genes that failed the analysis (not detected above a threshold) corresponded to DET0251 to -0272 and DET0275, a region that is triplicated in the isolated D. mccartyi strain 195 genome. This region was also absent from the metagenome sequences assembled from the D2 community (see Fig. S1 in the supplemental material). Each additional probe matched a gene from the D. mccartyi population in D2. Additionally, the assembled genome of the D. mccartyi population in the D2 metagenome did not contain any unique genes outside the sequenced, purified D. mccartyi strain 195 genome, suggesting that all relevant protein-coding genes were represented on the array.

Of the 1,487 probes specific for putative protein-encoding genes, the percentage significantly detected in this microarray study ranged in each experiment from 14.6% ± 5.9% in the decay studies to 87.5% ± 0.1% in the fast-respiration experiments. In experimental samples, the minimum, first-quartile boundary, second-quartile boundary, third-quartile boundary, and maximum of the processed intensity (for the Cy5 channel) across all arrays and all probes were 3.0, 34, 140, 580, and 390,000 processed fluorescence units (PFU), respectively. The average intensity of the Cy5 channel (1,180 PFU) exceeded the boundary of the third quartile, reflecting the log-normal nature of microarray data. Across all arrays, 1,451 of the 1,487 probes considered (97.5%) displayed confident transcript expression in at least one experiment. The 36 transcripts that were never confidently detected were present on the D. mccartyi genomic DNA, suggesting that these transcripts either were transcriptionally inactive or were misidentified as gene transcripts. These 36 transcripts were filtered out of the data set prior to further analysis (see Table S2a in the supplemental material). In the duplicate control-control hybridization analyses (see Fig. S2 in the supplemental material), 85.4% ± 7.4% (standard error) of the remaining genes displayed a log ratio of 0 ± 0.15. Additionally, the histogram observed is representative of a LOESS-normalized data set. Only 20 transcripts displayed an average log ratio higher than 0.3 or lower than −0.3; these transcripts were therefore not considered in the discussion.

To assess any variability across arrays or across analysis days, the Cy3 channel was assigned as a reference control consisting of pooled D2 samples from the seed culture (prefeed). For the experimental microarrays, each Cy3 channel was compared with every other Cy3 channel for the 53 experimental cultures (total number of comparisons, 2,809). The resulting average power law R² for all 2,809 comparisons was 0.88 (standard error, ±0.08; median, 0.92), suggesting reproducible between-array performance.

The 16S rRNA-normalized microarray intensity of tceA was plotted on a log-log scale against the 16S rRNA-normalized RT-qPCR tceA results as published previously by Rowe et al. (29) (see Fig. S3 in the supplemental material). The resulting power relationship fit provided an R² value of 0.73 for the 42 experiments sharing microarray and RT-qPCR values, further validating the trends from the microarray data.

Expression profile comparison across experiments.

The transcriptomes of the 53 experimental cultures (50 continuously fed, 3 starved) and 2 control-control hybridization arrays (n, 55 total) are presented in a heat map that was clustered in both the experimental (y axis) and gene (x axis) dimensions based on a Euclidean distance matrix of the log ratios (experimental/control) (Fig. 1). The majority of experiments and biological replicates clustered as expected, supporting the reproducibility of the genome-wide transcriptional response as captured by the microarray data (Fig. 1). However, two fast PCE-fed cultures that were inhibited because of solvent toxicity were binned into different clusters. Both cultures experienced an initially high respiration rate (∼50 μM Cl⁻/h); however, PCE accumulated above its aqueous solubility limit, leading to solvent toxicity that inhibited the respiration of PCE to a low rate by the final time point (2.4 and 3.9 μM Cl⁻/h for PCE_Butyrate_1_195 and PCE_Butyrate_2_195, respectively). These two experiments fall into separate clusters, and this separation suggests that the cultures experienced different stress conditions or were at different stages of their dynamic response to the stress.

FIG 1.

Heat map of the log ratios of gene expression with experiments (y axis) and gene transcripts (x axis) clustered. The central heat map ranges from blue (negative) to red (positive) log ratios (see the color key at the top). A dendrogram of the resulting hierarchal cluster for experimental conditions is presented on the primary y axis and is divided into two major partitions. A dendrogram of the transcripts is presented on the primary x axis and is divided into three major partitions; the A and B regions in partition 2 highlight transcripts responding to slow respiration. A legible version of this dendrogram is presented in Fig. S4 in the supplemental material. Genes identified either as a catalytic subunit of an RDase (orange) or as respiration (Resp.) linked (purple) are highlighted as bars below the gene dendrogram on the x axis. A bar chart of the overall (gray) and instantaneous (black) respiration rates is presented on the secondary, left-hand y axis. Experiments are notated according to the following format: Electron acceptor_Electron donor_replicate number_acceptor feed rate. Zeroes in the notation indicate that there is no donor or no acceptor.

The heat map shows two main partitions of experiments. Notably, the two clusters were separated primarily by the instantaneous respiration rate into fast- and slow-respiring cultures. The instantaneous respiration rates, averaged across the fast- and slow-respiration clusters of experiments, were 22.9 ± 9.6 and 1.6 ± 0.6 μM Cl⁻/h, respectively, with ranges of 2.7 to 73.6 μM Cl⁻/h and 0 to 5.3 μM Cl⁻/h, respectively. An instantaneous respiration rate breakpoint that triggers a substantial shift in the expression pattern of D. mccartyi between the two clusters therefore appears to exist between 2.7 and 5.3 μM Cl⁻/h, because this is the overlapping range of the two major clusters. Two biological replicates that had respiration rates in the range of the transition point (DCE_Butyrate_1_4.8 and -2_4.8) appeared in different clusters. The instantaneous respiration rates for the DCE_Butyrate_1_4.8 and DCE_Butyrate_2_4.8 cultures were 4.9 and 3.8 μM Cl⁻/h, respectively. These cultures were clustered in the fast- and slow-respiration-linked clusters, respectively (Fig. 1), emphasizing the existence of a respiration breakpoint in this range. The gene transcripts that displayed the highest fold difference (>100) between these two replicates were DET1569 (a hypothetical protein) and DET1173 (encoded by a member of a Fec transporter operon).

Several notable relationships were evident in the clustering of transcripts into three main clusters (Fig. 1, x axis; see also Fig. S4 in the supplemental material). First, the RDase-encoding gene with the highest expression level, tceA (DET0079), clustered closely with the gene encoding RDase DET1559 (cluster 3 in Fig. 1 and in Fig. S4 in the supplemental material). RDase catalytic subunits linked to slow respiration in previous studies (DET0173, -1535, -1538, and -1545) (24) were distant from the tceA cluster, displaying an approximately inverse expression profile on the heat map (cluster 1). Second, one subcluster of transcripts was composed almost entirely of respiration-linked oxidoreductases (a subcluster of cluster 3 [Fig. 1 inset]). These transcripts included members of the operons for Hup (DET0110 to -0112), Hym (DET0145 to -0148), DET0186-DET0187, Nuo, and ATPase. Third, and finally, the gene transcripts that were most highly upregulated in the slow-respiration cluster relative to the fast-respiration cluster fell into two groups (groups A and B of cluster 2). Groups A and B contained 70 and 63 gene transcripts, respectively, that passed all filters. Notable gene transcripts in group A included nine reductive-dehalogenase-associated gene transcripts (DET0172, -0176, -0237, -0306, -1170, -1523, -1530, -1537, and -1542) and members of the putative nif operon (DET1148, -1151, -1153, and -1156); notable members of group B included gene transcripts encoding DNA replication/repair proteins (DET0442, -0535, -0572, -1109, -1112, and -1464) and virus-related gene transcripts (DET0063, -0068, -0072, -0323, -1085, -1086, -1091, -1092, -1475, and -1476). A detailed list of these two clusters with gene transcript descriptions is provided in Table S3 in the supplemental material.

Compared to the strong effect of the respiration rate, experimental variables such as the electron donor type, electron acceptor type, and electron acceptor-to-donor ratio did not display significant overall effects on transcriptomic profiles. However, the electron acceptor type affected RDase expression profiles. For example, DET0318 (encoded by pceA) and DET1559 were more abundantly detected on chlorophenols than on chloroethenes (when fed at similar electron acceptor feed rates). These findings support previous observations and further support the hypothesis that DET0318 is a bifunctional RDase that can dechlorinate both PCE and chlorinated aromatics (45).

Principal component analysis.

Figure 2 displays a scree plot of the proportions of gene expression variance explained by the first 10 principal components and a plot of the loading matrix (experimental conditions) expression data projected onto the first two principal components. The scree plot shows that the first two principal components (PCs) capture 42.5% and 18.7% of total variability, respectively (Fig. 2A). The first two PCs separate the experiments into two distinct clusters: slow-respiring (with chloroethenes and chlorophenols) and fast-respiring (with chloroethenes) cultures (Fig. 2B). The PCA therefore supports the noted transition value in the clustering analysis.

FIG 2.

(A) Scree plot displaying the percentage of variance captured by each of the first 10 principal components. (B) The data for each experiment (the loading matrix) cluster into two groups when projected onto the first and second principal components. Data points representing the experiments are color-coded blue (slow-respiration cluster in Fig. 1), red (fast-respiration cluster in Fig. 1), orange (the three starved cultures), or black (the two control-control hybridizations).

In addition to the consideration of the experiments in the principal component analysis, we also projected the score matrix (the transcript log ratio profiles) onto the first two principal components (Fig. 3a). Figure 3b to d present selected subsets of transcripts from Fig. 3a: transcripts for the genes identified by Seshadri et al. (5) encoding either RDase catalytic subunits (Fig. 3b), Hym, Hup, and DET0185 to -0187 (Fig. 3c), or ATPase subunits (Fig. 3d). The loading matrix for all experiments and the score matrix for all transcripts analyzed across the first four PCs are presented in Table S4 in the supplemental material. In combination, Fig. 2B and 3a can be represented as a single biplot of the data set analyzing the coefficient of the variables (experimental conditions) and the observational loadings (the sum of the transformed log ratio microarray gene expression values weighted by the coefficient of variables); however, in order to maintain visual clarity, these plots were separated. Notably, the scales of the two axes are different because Fig. 2B represents the loading matrix whereas Fig. 3 represents the score matrix. However, the directionality is preserved because of the relationship between the score matrix and the loading matrix.

FIG 3.

Plots of the PCA scores for the gene transcripts in the following categories: (a) all, (b) RDase catalytic subunits, (c) Hup (DET0110 to -0112), Hym (DET0145 to -0148), and DET0185 to -0187, and (d) the ATPase operon (DET0558 to -0565). In all four plots, the numbers represent the DET gene number on the genome. The bottom left quadrant in all four of these plots is related to high respiration because the principal component scores displayed were derived from the loadings shown in Fig. 2B.

The RDases displayed an expected trend, with tceA (DET0079) plotting close to the location of the fast-respiration-linked cluster (compare Fig. 3b and 2B). DET1559 (an RDase with an unknown function) plots near tceA, and tceA and DET1559 display the highest and second-highest loadings in PC1 in the direction of the fast-respiration experiments. Other RDases, such as DET1545, DET1535, and DET1538, plot in the inverse quadrants on the biplot. This distant location indicates that these transcripts are anticorrelated both with the expression of tceA and with fast respiration.

Like tceA and DET1559, transcripts for the other respiration-linked enzymes (Hym, Hup, and DET0185 to -0187) also clustered into the bottom left quadrant, which contained the fast-respiration experiments. Notably, DET0186-DET0187 and members of the ATPase (DET0559-DET0560) displayed some of the strongest loadings for PC2 and PC1, respectively. Overall, the loadings of numerous respiration- and growth-linked genes, such as those encoding a chromosomal replication initiator protein (DET0001) and subunits of a cobalamin transporter (DET0650 to -0653) (cobalamin derivatives are critical cofactors for RDase enzymes), placed them into the quadrant with the fast-respiration cluster.

Additionally, transcripts of tceA, discussed above (encoding the major RDase DET0079), and of the ATPase operon (e.g., atpF, encoding DET0560) were anticipated to display a trend with the respiration rate. When the expression profiles of these two transcripts were compared directly to the two calculated respiration rates, the significance of the power relationship between the respiration rate and the log ratios for the atpF or tceA transcript was higher for the instantaneous respiration rate than for the whole-experiment-averaged respiration rate (the R² values were 0.84 and 0.53, respectively, for atpF and 0.60 and 0.30, respectively, for tceA) (see Fig. S5 in the supplemental material). The correlation of the log ratio of tceA transcript expression in the microarray to the log-instantaneous respiration rate (R² = 0.60) is similar to the correlation reported previously for the power relationship between the overall respiration rate and the absolute transcript abundance of tceA as determined by RT-qPCR (R² = 0.74) (28).

Gel separation of putative respiration-linked enzyme complexes.

The tight clustering of subunits of Hup, Hym, and DET0186-DET0187 in Fig. 1 raises the possibility that these three enzymes form a protein complex, and the plotting of these transcripts in the bottom left quadrant in Fig. 3c links them with fast respiration. Proteins in crude extracts from the culture were separated on a native PAGE gel (Fig. 4A). Dominant bands were visualized, excised, digested with trypsin, and sequenced. All the D. mccartyi-specific proteins detected are listed in Table S5 in the supplemental material. The Mascot-determined protein scores for the detected subunits of the Hup, Hym, and DET0186-DET0187 operons are displayed in Fig. 4B. Although subunits of Hup, Hym, and DET0186-DET0187 were all detected in the top band, this is the location in which cell debris (including membrane fragments) and contaminants that could not migrate into the gel remain. Also, each of these subunits achieved a higher protein score in other gel bands. Therefore, the presence of these proteins in this higher gel band likely indicates the location of denatured proteins.

FIG 4.

Proteomic exploration of the respiration-linked oxidoreductases. (A) Gel image of the separated crude-extract proteins with a ruler and a molecular mass guide. (B) Protein scores for members of operons for Hup (DET0110, DET0112), Hym (DET0147), and DET0187 detected in gel slices from different depths. (C) Genome annotations of both Hup (top) and DET0185 to -0187 (bottom) operon homologs in Dehalococcoides, Dehalogenimonas (51), and Dehalobacter (54). The FdhB/FdhB-like subunits in the annotated operons are shown in red. U, S, and C denote selenocysteine, serine, or cysteine residues at the predicted active sites of the FdhA/FdhA-like proteins. (D) Enzyme cartoon displaying the potential interactions of DET0112 with the DET0186-DET0187 complex. TAT denotes subunits with twin-arginine transport signal peptides. Fd (ferredoxin), Qu (quinone), and NAD(P)H denote some possible electron carriers.

Further down in the gel, Hym and DET0187 overlapped in two bands near the 1,000-kDa range. Additionally, a subunit of Hup (DET0112; 31 kDa) comigrated with DET0187 (106 kDa) in one gel band (slightly larger than the 146-kDa standard). The summed molecular masses of DET0112 and DET0187 are 137 kDa (if the predicted membrane subunit DET0186 is present but below the detection limit, the total mass, excluding cofactors in the complex, would be 178 kDa). The other member of the Hup operon detected (DET0110) resides in lower gel bands (with lower molecular masses).

DISCUSSION

Genomic differences between a pure culture of D. mccartyi strain 195 and the population of D. mccartyi strain 195 in the mixed culture D2.

Twenty-three probes for protein-coding genes failed to show detectable signals when hybridized with DNA from the D2 mixed culture. The missing region on the microarray (DET0251 to -272, DET0275) is part of a transposable region on the genome of the D. mccartyi strain 195 pure culture. This entire region was also absent from the metagenome (members of this region are triplicated on the D. mccartyi strain 195 genome: DET0253 to -272, DET0276 to -0295, and DET0886 to -0905) except for DET0270 (or the paralogs DET0293 and DET0903), suggesting that these sequences are absent or transient in the genome of the D. mccartyi strain 195 population residing in D2. DET0270 (or DET0298 and DET0903) was detected in the metagenome; however, the average nucleotide identity of the metagenome sequences to the pure-strain isolate sequences across the gene was 0.84, compared to an average nucleotide identity of 0.99 for all D. mccartyi strain 195 genes that were detected on the array (n = 1,487). Considering the transient nature of the transposable elements and the genome-encoded viruses found in multiple D. mccartyi strains, this is not a surprising finding (10).

The partitioning of the experiments depended on the respiration rate.

In the heat map clustering, two groupings of experiments emerged. Faster-respiring cultures (range, 2.7 to 73.6 μM Cl⁻/h; average, 22.9 ± 9.6 μM Cl⁻/h) clustered together, as did slower-respiring cultures (range, 0 to 5.3 μM Cl⁻/h; average, 1.6 ± 0.6 μM Cl⁻/h) (Fig. 1). From the PCA for experiments (Fig. 2), the respiration rate appeared to describe the variance captured in the first two PCs. Therefore, the instantaneous respiration rate influences the majority of the variance in the data set. A more descriptive representation of the respiration rate would be expressed in terms of per-cell respiration. As reported previously, the D. mccartyi strain 195 populations were stable over time in these experiments, at about 5 × 10⁸ cells/ml (29); per-cell respiration rates differed proportionately. Therefore, a major shift in the transcriptome is triggered at respiration rates in the range of 5.4 × 10⁻¹² to 10.6 × 10⁻¹² μmol Cl⁻/cell/h.

From the PCA (Fig. 2 and 3), several transcripts for genes previously suggested to be respiration linked (5) were mapped primarily into the region of fast respiration; these were members of the ATPase operon encoding DET0558 to -0560, with PC1 and PC2 scores of −9.57 and −3.74, −11.64 and −3.71, and −10.37 and −5.70, respectively. Additionally, two RDases, DET0079 (encoded by tceA) (PC1 and PC2 scores, −14.80 and −4.32, respectively) and DET1559 (−10.37 and −5.70), clustered in the direction of the fast-respiring cluster. Taken together, these findings suggest that the binning of the experiments depends substantially on key RDases and respiration-linked enzymes in the data set. Other genes, such as the chromosomal replicator DET0001 (PC1 and PC2 scores, −7.80 and −7.32, respectively), a member of the TAT secretion system, DET1600 (−11.29 and −1.94), and a cobalamin transporter, DET0650 (−0.33 and −11.49]), displayed high loadings in the direction of the fast-respiring cultures and were highlighted by the PCA. Additionally, the transcripts encoding Hup, Hym, and DET0187 clustered closely on the plot, promoting an investigation into possible relationships among these enzymes.

Membrane-associated oxidoreductases.

Through screening of the transcripts in the gene clusters of the heat map in Fig. 1, a tight cluster containing transcripts encoding subunits of Hup, Hym, and DET0186-DET0187 was found. The [NiFe] hydrogenase Hup is the predominant hydrogenase predicted to be involved in the oxidation of hydrogen (5, 15). At the protein and transcript levels, Hup subunits have often been detected at levels similar to those of DET0187 (a formate dehydrogenase-like protein) in multiple D. mccartyi strains (15, 16, 46). Both of these enzymes are potentially involved in electron transport processes, passing a low-potential electron to a Fe-S small protein (47). Additionally, both of these enzymes are predicted to face the periplasm and are therefore able to interact directly with extracellular hydrogen or, in the case of DET0187, with an as yet unknown electron donor. The [Fe] hydrogenase Hym is also membrane bound but faces the cytoplasm. Transcripts for all three of these enzyme complexes plotted into the same quadrant as the fast-respiring cultures in the PCA, and DET0186-DET0187 displayed some of the highest loadings in PC2, suggesting that these enzymes are critical in the respiratory process of D. mccartyi, despite their unknown catalytic role. DET0187 maintains a serine in its active site (rather than selenocysteine or cysteine, which are typical of biochemically validated formate dehydrogenases), which may explain its currently unknown function (48). An orthologous selenocysteine-containing Fdh in Escherichia coli was recently shown to have hydrogenase functionality (49); the authors of that study speculated that the seryl-mutant DET0187 possibly functions as an additional hydrogen uptake enzyme in D. mccartyi. Alternately, DET0187 may use a different electron donor altogether.

Closer inspection of the annotation information available for the hup (DET0109-to-DET0112) and DET0185-to-DET0187 operons provides suggestions for the functions of the protein complexes and protein-protein interactions. Surprisingly, our reexamination of all available D. mccartyi genome sequences found that the Fdh-like operon (DET0185 to DET0187) is consistently missing a potential key subunit, an FdhB-like protein (Fig. 4C). In conventional Fdh proteins, the beta subunit serves as the bridge, providing the [4Fe-4S]-based electron channel between the molybdopterin-containing major subunit (potentially DET0187 in D. mccartyi) and the membrane-bound electron exchanger subunit (DET0186; FdhM or FdhG; additionally annotated in public databases as a polysulfide reductase membrane subunit) (50). In all strains of D. mccartyi, the only transcript encoding a potential homolog of the beta or transmembrane subunit is found on the hup operon (DET0112 in D. mccartyi strain 195; others noted in Fig. 4C). Notably, the other member of the Dehalococcoidaceae family with a sequenced genome, Dehalogenimonas lykanthroporepellens BL-DC-9, maintains multiple Fdh and Hup operons on its genome; one of these Hup operons encodes an “FdhB”-like protein similar to DET0112, encoded by the Hup operon of D. mccartyi (Fig. 4C). Additionally, the presence of a selenocysteine-containing database-annotated Fdh encoded in the genome of a Dehalogenimonas species is surprising, because this organism, like D. mccartyi, cannot use formate as an electron donor (51).

Data from the nondenaturing-gel separation of protein complexes support the hypothesis that DET0112 complexes with the FdhA-like protein DET0187. Surprisingly, from the proteomic data displayed in Fig. 4B, the only gel location in which DET0112 was found was the band showing the highest protein score for DET0187 (near the 146-kDa protein standard). Additionally, HupL (DET0110) appeared in lower bands, suggesting that DET0110 and DET0112 do not comigrate. This finding is supported by a previous investigation that used blue native PAGE (BN-PAGE) gels to separate RDases from the D. mccartyi-containing KB-1 consortia in order to monitor reductive dechlorination activity (43). For the TCE-induced consortia in that study, proteomic sequencing of the active band identified homologs of DET0187 and DET0186 with the KB1 homolog of DET0112. No additional members of Hup were identified in those gel slices. Conventional architecture of Fdh enzymes in E. coli studies suggests that an FdhB is required and comigrates with the FdhA subunit (52). A Dehalobacter species, an organism that is more physiologically similar to Dehalococcoides species than E. coli, maintains a conventional Fdh operon, as displayed in Fig. 4C. A study investigating the blue native PAGE separation of proteins for this organism also found that FdhB comigrated with the FdhA subunit (53). The combined evidence from the genomic, transcriptomic, proteomic, and homology investigations suggests that DET0112 is recruited by the DET0186-DET0187 enzyme. This potential complex of DET0186-DET0187 with DET0112 is shown in Fig. 4D, illustrating the potential role of DET0112 as the neck between the catalytic and membrane subunits of DET0186-DET0187. Additional cross-linking studies could further resolve such protein-protein interaction questions.

In summary, we generated and analyzed a transcriptomic data set from a range of continuously fed experimental cultures to develop specific, testable hypotheses for D. mccartyi strain 195. Experimental cultures were grown under conditions with varying electron acceptor feed rates, electron acceptor types, electron donor-to-acceptor ratios, and electron donor types. In the meta-analyses of the full transcriptome, genome-wide expression responses were noted and suggested novel connections between annotated proteins. In analysis of the data set, the experiments were found to be divided into fast- and slow-respiration-linked clusters. Several oxidoreductases were noted to be associated with the fast-respiring cultures, and a specific investigation into their behavior allowed us to propose the hypothesis that a subunit of the Hup operon is recruited by the DET0186-DET0187 enzyme. This hypothesis was supported by subsequent proteomic assays.