Subcellular localization of an ATPase in anammox bacteria using proteomics and immunogold electron microscopy



Anaerobic ammonium oxidation (anammox) has received significant attention during optimization of waste-water treatment and constitutes an important pathway for the removal of bioavailable nitrogen from natural environments. Studies of key catabolic enzymes indicate that the anammox reaction takes place inside the anammoxosome, an organelle-like membranous compartment of anammox bacteria. The anammoxosome has also been suggested as a site for ATP synthesis. A lipid-based protein immobilization technique, previously used to identify proteins essential for the anammox reaction, was in this study used to select linear epitopes for antibodies specifically targeted against an identified ATPase. The approach of using proteomics and bioinformatics as tools for selecting antibody targets for immunolocalization provides an important alternative to traditional methods for selection of specific antibodies. Immunogold electron microscopy and statistical evaluations indicated that the antibodies against the ATPase were exclusively found associated with the anammoxosome membrane. This provides strong evidence for ATP synthesis by an intracellular proton motive force in anammox bacteria. Within prokaryotes, an ATP synthase associated with an intracellular compartment is a feature unique for anammox bacteria.


Anaerobic ammonium oxidation (anammox) is a globally important pathway for the removal of biologically available nitrogen, and up to 50% of total marine N2 production is suggested to be attributed to anammox (Devol, 2003; Arrigo, 2005). Subsequent to the initial theoretical predictions of anammox (Broda, 1977), experimental evidence of this reaction was first provided by studies in a waste-water reactor (Mulder et al., 1995). Bacteria catalyzing the reaction were identified few years later (Strous et al., 1999). Genomic analysis has implied that anammox bacteria form a monophyletic group of bacteria that belongs to the phylum Planctomycetes (Strous et al., 1999; Schmid et al., 2000). So far, five different genera have been discovered: ‘Candidatus Brocadia’, ‘Candidatus Kuenenia’, ‘Candidatus Anammoxoglobus’, ‘Candidatus Scalindua’, and ‘Candidatus Jettenia’ (Strous et al., 1999; Kuypers et al., 2005; Kartal et al., 2007). Through studies of the metagenomes of ‘Candidatus Kuenenia stuttgartiensis’ (Strous et al., 2006) and ‘Candidatus Scalindua profunda’ (van de Vossenberg et al., 2013), comparative mechanistic studies for metabolic pathways of anammox are now possible for a range of natural environments.

Studies using electron microscopy have identified a membranous compartment, the anammoxosome, within the cells of anammox bacteria (Lindsay et al., 2001; Fuerst, 2005; van Niftrik & Jetten, 2012). This organelle-like compartment has from several studies been suggested as the site for the anammox reaction (Fuerst, 2005; Jetten et al., 2009; Karlsson et al., 2009). The anammox reaction requires a specific set of enzymes, including a nitrite/nitric oxide oxidoreductase that reduces nitrite to nitric oxide, a hydrazine synthase (HZS) that generates hydrazine from ammonium and nitric oxide, and a hydrazine-oxidizing enzyme (HZO) that produces N2 from hydrazine (Strous et al., 2006; Shimamura et al., 2007; Jetten et al., 2009). Observations have shown that a hydroxylamine oxidoreductase (Lindsay et al., 2001) and several cytochrome c-type proteins (van Niftrik et al., 2008a2008b) are localized inside the anammoxosome. These proteins are important for electron transfer during the anammox reaction, which thereby further strengthens the hypothesis that the anammox reaction is closely associated with the anammoxosome. It has also been shown that the anammoxosome divides independently from the rest of the cell, observations that support a differentiated targeting of proteins (van Niftrik et al., 2009). The HZO and a putative HZS (cf hydrazine hydrolase depending on the reaction mechanism and pathways for synthesis or hydrolysis of hydrazine) as well as subunits of an F-ATPase 1 were identified from a membrane preparation of anammox cells using proteomics and mass spectrometry (Karlsson et al., 2009). Additionally, the HZS was through immunogold electron microscopy (EM) determined to be localized to the interior of the anammoxosome (Karlsson et al., 2009).

In an experimental effort to clarify the subcellular localization of one of the ATPases encoded by ‘Candidatus Kuenenia stuttgartiensis’, van Niftrik et al. (2010) raised antibodies against cloned protein subunits. Immunogold EM (Tokuyasu, 1986) revealed positive labeling to both the outer membrane as well as the anammoxosome membrane. Gold particles were also found, although to a lesser degree, associated with the cytoplasmic membrane. Due to the difficulty to visually assign the specific loci of the ATPase, the authors used statistical evaluation to predominantly allocate the protein to the anammoxosome membrane and the outermost membrane. A dual location of the ATPase is, however, contradictory to theoretical predictions (Medema et al., 2010) as well as experimental observations (van Niftrik et al., 2009; van der Star et al., 2010). Thus, the subcellular location of ATPases is not clearly elucidated, nor is the location of energized membranes in anammox bacteria (van Niftrik et al., 2010; Santarella-Mellwig et al., 2013; van Teeseling et al., 2014). A proton gradient with the potential to fuel a membrane-bound ATP synthase can supposedly be generated across the anammoxosome membrane by the anammox reaction, mediated by proteins located inside the anammoxosome, and associated with the anammoxosome membrane (van Niftrik et al., 2004, 2008a2008b; Fuerst, 2005). Experimental support of a significant (c. 1 pH unit) proton gradient across the anammoxosome membrane, similar in magnitude to that associated with mitochondria, was recently provided with 31P-NMR (van der Star et al., 2010). Further, Medema et al. (2010) performed detailed gene sequence analysis which could be used to predict proteins targeted to the anammoxosome. All catabolic functions, for example, the enzymes of the anammox reaction and the major respiratory complexes (including the F-ATPase 1), were found to be targeted to the anammoxosome. To further clarify the location, the main objective of this study was to improve the specific antibody recognition of the target protein. This was achieved using a proteomic-based protocol to select linear epitopes of the F-ATPase 1. Peptides generated from enzymatic digestion of native membrane preparations from anammox bacteria provide crucial information of the linear epitopes that are surface exposed.

Materials and methods

Preparation of proteoliposomes

Proteoliposomes were prepared as previously described (Karlsson et al., 2009). In brief, aggregates of bacterial cells from an anammox waste-water reactor for nitrogen removal (Himmerfjärdsverket, were transferred to a buffer solution (50 mM Tris-HCl, 60 mM EDTA, 0.25 M sucrose, pH 7). The cells were subjected to mild sonication in an ultrasonic bath (c. 5 min; 25% of maximum intensity, 500 W; Ultrasonic) until disruption into single cells, as confirmed by light microscopy. Lysozyme (5 mg mL−1) was added to the single-cell solution prior to rapid cooling to c. 4 °C. After 20 min, the solution was sonicated using a tip-sonicator (500 W; 5 s intervals between 5 s pulses; 20 times, 25% of maximum intensity; Sonics Vibracell, Model 501). The solution was centrifuged (4 °C 10 000 g, 10 min) resulting in a red–orange pellet of membrane material. The membrane material was resuspended in buffer (10 mM Trizma, 300 mM NaCl, pH 8) and heavily sonicated on ice (75% of maximum intensity, 500 W) using the tip-sonicator, whereby proteoliposomes were formed.

Instrumentation and identification of membrane-associated proteins

The proteoliposome solution was injected into the LPI FlowCell (Nanoxis Consulting AB, Göteborg, Sweden) where the proteoliposomes were allowed to attach to the gold-coated surfaces of the LPI FlowCell. The flow cell was subsequently rinsed with buffer (10 mM Trizma, 20 mM NaCl, pH 8) to remove any contaminating proteins. The proteins associated with the immobilized proteoliposomes were digested by incubating the sample in the LPI FlowCell with trypsin (5 μg mL−1 in 10 mM Trizma, 20 mM NaCl, pH 8) for 2 h at 37 °C. The resulting peptides were eluted with 700 μL buffer (10 mM Trizma, 20 mM NaCl, pH 8).

The peptide fraction collected from the LPI FlowCell was subsequently analyzed by liquid chromatography–mass spectrometry (LC-MS/MS) at the Proteomics Core Facility, University of Gothenburg, according to Karlsson et al. (2009).

Production of antibodies targeting the F-ATPase 1

Two peptide sequences (single letter amino acid sequence AIDAMIPIGR from kuste3793, the alpha-subunit of the F-ATPase 1, and QIAELGIYPAVDPLR from kuste3795, the beta-subunit F-ATPase 1) were chosen as linear epitopes for immunogold labeling targeting the F-ATPase 1. The peptides were selected due to their high abundance with multiple hits in the MS-database analysis (Supporting Information, Data S3). Synthetic duplicate peptides and antibodies against these peptides were produced by Innovagen AB (Lund, Sweden). The peptides were purified by reversed-phase HPLC, using a C18-column, and analyzed for purity and quality by mass spectrometry. The peptides were linked by a cysteine residue to keyhole limpet hemocyanin (KLH). The production of the specific antibodies named anti-kuste3793 and anti-kuste3795 was performed by immunization of specific pathogen-free (SPF) rabbits following injection of KLH with linked specific peptides. The antibodies were purified and subjected to an ELISA test according to standard protocols (Supporting Information, Data S1 and S2).

EM immunocytochemistry

Aggregates of bacterial cells (c.80% ‘Candidatus Brocadia anammoxidans’, < 20% ‘Candidatus Kuenenia stuttgartiensis’; Karlsson et al., 2009) from the anammox waste-water reactor were fixed for 2 h in a mixture of 2% formaldehyde and 0.2% glutaraldehyde in 0.1 M phosphate buffer (PB, pH 7.4). Cells were processed for cryosectioning according to Tokuyasu (1986). After washing in phosphate-buffered saline (PBS; 137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 2 mM KH2PO4, pH 7.4) with the addition of 20 mM glycine for aldehyde group quenching, colonies were embedded in 12% gelatine. Small cubes of gelatine-enclosed colonies were infused over night with 2.3 M sucrose and frozen in liquid N2. Dry sectioning was performed at −110 °C in a Leica EM UC7 cryoultramicrotome with a Diatome cryo knife. Section ribbons were transferred to copper grids covered with Formvar support film and were subjected to an immunolocalization incubation sequence. Two independent experimental series were performed, using either pre-immune serum or an affinity-purified rabbit antibody targeting an antimicrobial peptide (anti-CRAMP; polyclonal antibody raised against the synthetic peptide corresponding to aa 135–173 of the mouse cathelin-related antimicrobial peptide CRAMP) as controls, respectively.

Sections were pretreated 2 × 2 min with 20 mM glycine in PBS followed by 1% bovine serum albumin in PBS with 0.02% Na azide (incubation medium) for 2 min. Subsequently, the sections were incubated with primary antibodies in the incubation medium (anti-kuste3793 and anti-kuste3795 diluted 1 : 100), rabbit polyclonal anti-CRAMP at 1 : 100 (control), or pre-immune serum (control) for 20 min. All primary antibodies (affinity-purified polyclonal rabbit antibodies) were produced and prepared in similar ways by the same supplier (Innovagen AB). Secondary detection was achieved with protein A conjugated with 10-nm gold particles (Aurion, NL) at 1 : 100 dilution with incubation medium for 20 min. Sections were rinsed with PBS between incubation steps and were postfixed with 1% glutaraldehyde. Finally, cryosections were contrasted with 2% uranyl acetate/oxalate and stabilized with methyl cellulose and uranyl acetate followed by air-drying. They were examined in a LEO 912AB transmission electron microscope equipped with an Olympus-SiS Veleta CCD camera for digital image capture.

Distribution of nanogold particles

Measurements of nanogold particle distributions were performed with tools in the item software: (1) random points over anammox colonies were memorized in the store position library of the microscope control system at low magnification (4000×) where gold particles are invisible. Digital images of the predetermined areas were collected at 25 000×, and the number of gold particles as well as the number of cells with an identifiable anammoxosome were quantified. Ten image fields were observed at magnification 25 000×, each with an area of 14.3 μm2 (i.e. the total observed area was 143 μm2); (2) detailed examination of anammox colonies were made at 50 000×, and all cell profiles with a gold particle and a transected anammoxosome were saved consecutively. The distance from each gold particle to the nearest anammoxosome membrane was measured and recorded. The length of the rabbit-antibody-protein-A gold complex was set to be 25 nm (van Niftrik et al., 2010). A zone of acceptance (Mayhew & Lucocq, 2008) of 25 nm of either side of the anammoxosome membrane was therefore used to assign the gold particles as bound to the membrane. For each antibody (anti-kuste3793 and anti-kuste3795), at least 70 labeled cells were analyzed. Graphical techniques were used for statistical analysis and to explore data (box-and-whisker plots, IBM spss Statistics).

Results and discussion

Protein identification and determination of linear epitopes

In the genome of ‘Candidatus Kuenenia stuttgartiensis’, four different gene clusters have been identified to encode membrane-bound complexes of ATPase (van Niftrik et al., 2010). All of these can function either as an ATP-dependent proton pump or as an ATP synthase. The gene cluster of the F-ATPase 1 (open-reading frames, ORF; kuste37873796) shows similar protein sequences as compared to an F-ATPase 1 in E. coli with an experimentally verified function of ATP synthase (van Niftrik et al., 2010). To elucidate the expression of the putative ATPases in the bacterial cells, efforts have been made to study the transcriptome (mRNA) and proteome (peptides) (van Niftrik et al., 2010). Only the F-ATPase 1 (ORF; kuste37873796) was found to be significantly expressed. Immunolocalization suggested a diffuse subcellular localization of the F-ATPase 1, with several possible loci within the anammox cell. Strongest labeling was found in the outermost membrane and in the anammoxosome membrane (van Niftrik et al., 2010). This observation is quite in contrast to the genomic prediction that rather suggests that the anammoxosome has a targeted subproteome (Medema et al., 2010). In accordance, the F-ATPase 1 would only be associated with the anammoxosome membrane.

This study utilized a recently developed technique to select and experimentally verify linear epitopes of the F-ATPase 1 (Karlsson et al., 2009). Peptides found in the MS analysis were therefore regarded as suitable substrates for antibodies (Fig. 1). One of the main advantages of using the lipid-based protein immobilization (LPI) technology in a flow cell format is the ability to use short digestion times which enables a surface shaving of the exposed proteins. This was recently illustrated by performing a sequential digestion approach to maximize the number of protein hits in a protein profiling study (Jansson et al., 2012).

Figure 1.

Schematic illustration of the differences in work-flow using in silico prediction tools vs. proteomics-based generation of experimentally verified linear epitopes for antibody generation. Using software tools, it is possible to predict sequence regions which are most likely buried in a cellular membrane and therefore not suitable for the selection of linear epitopes. When selecting epitopes predicted to be in exposed extracellular loops, there is always the risk of generating antibodies against target sequences hidden in the three-dimensional protein structure. Using a proteomic-based approach followed by bioinformatics enables the selection of experimentally verified linear epitopes (peptides) that are exposed on the surface of the protein structure. Such exposed linear epitopes are subsequently used for antibody generation, thus increasing the probability of a specific binding event.

As the ATP-generating ATPase is one of the most conserved proteins in bacteria (Deckers-Hebestreit & Altendorf, 1996), peptides found in the MS/MS analysis may not be unique to anammox bacteria. Following analysis, the blast network service (, Altschul et al., 1997) was therefore used to screen for homologue peptides originating from bacteria other than ‘Candidatus K. stuttgartiensis’. In the database search for peptides from the beta-subunit of the F-ATPase 1, there were hits from the anammox bacterial strain KSU-1 (Shimamura et al., 2007) and from uncultured Planctomycetes. Altogether, the sequence overlap between peptides from the beta-subunit of anammox bacteria and other bacteria was quite low (Karlsson et al., 2009). In the search for peptides originating from the alpha-subunit, however, there were complete sequence overlaps with many species of bacteria. To retain the structure and function of the ATPase during evolution, key peptide sequences are likely preserved. As many hits were found in the blast search, peptides arising from the alpha-subunit are probably peptides that originate from such highly conserved regions of the ATPase. The unique morphology of anammox bacteria still facilitated a positive localization of the alpha-subunit of the F-ATPase 1.

Two peptides were used as linear epitopes for raising antibodies in rabbits (Innovagen AB). AIDAMIPIGR (single letter amino acid sequence) was the target peptide in the alpha-subunit of the F-ATPase 1 (kuste3793), and QIAELGIYPAVDPLR was the target peptide in the beta-subunit (kuste3795). The generated antibodies were thus named anti-kuste3793 and anti-kuste3795, respectively.

EM immunocytochemistry and subcellular localization of the F-ATPase 1

The subcellular organization of anammox bacteria was well preserved after cryopreparation, whereby the cell structure, including the anammoxosome membrane, was clear and distinct (Fig. 2). In general, there was a low but favored labeling of gold particles associated with the anammoxosome membrane after use of the two specific antibodies. In contrast, the control antibody (anti-CRAMP) resulted in extremely few nanogold particles over cell profiles. The background using pre-immune serum was for unknown reasons rather high. Results from the immunocytochemistry are exemplified in Fig. 2.

Figure 2.

EM immunocytochemistry and localization of the F-ATPase 1. (a) Schematic illustration of the protocol for immunogold labeling. Target peptides from either the alpha (kuste3793)- or beta (kuste3795)-subunits of the identified F-ATPase 1 were chosen from the MS/MS analysis as linear epitopes. Antibodies against these peptides (anti-kuste3793 and anti-kuste3795) were raised in rabbits. During the immunogold localization, the antibodies bound to their target proteins in cryo-sectioned anammox colonies. Protein A labeled with gold particles (10 nm) was used to localize the antibodies, thereby revealing the location of the F-ATPase 1. (b–c) Anammox cells incubated with the antibodies (anti-kuste3793 and anti-kuste3795). In each image, one or two 10 nm gold particles were associated with the anammoxosome membrane. Width of images is 500 nm. (d) Colony of anammox bacteria after immunogold incubation with anti-CRAMP antibody (control). The cytoplasmic membrane (white arrow, CM) and the anammoxosome membrane (black arrow, AM) are indicated. (e) Schematic illustration of the cellular organization of anammox bacteria (adapted from

van Niftrik et al., 2008b).

The frequency of labeling was estimated by measurements of nanogold particle distributions over anammox cells using predetermined areas (ten images, each with an area of 14.3 μm2). The labeling frequency (described by the number of labeled cells vs. total number of analyzed cells) was highest for anti-kuste3793 (23%), somewhat lower for anti-kuste3795 (13%), and almost negligible for the control antibody (anti-CRAMP; 1%). Due to the low labeling frequency of the anti-CRAMP antibody, the total number of data points for this control was low.

In another set of experiments, a higher magnification of 50 000× was used to observe gold labeling of the anammoxosome membrane. A zone of acceptance of 25 nm to either side of the anammoxosome membrane was used to indicate a positive labeling of the membrane (Mayhew & Lucocq, 2008; van Niftrik et al., 2010). Incubations of the antibodies targeted against the alpha- and beta-subunit of the F-ATPase 1 (anti-kuste3793 and anti-kuste3795) gave rise to a repeated gold particle distribution at or in the immediate proximity of visible anammoxosomes (Fig. 2 and 3). Immunogold labeling against the alpha-subunit of the F-ATPase 1 (anti-kuste3793) showed a high specificity. The relative frequency of immunogold labeling within the zone of acceptance was 93% and 87%, respectively, for the two experimental series. For the beta-subunit of the F-ATPase 1 (anti-kuste3795), the relative frequency was lower (64% and 69%, respectively). Frequency distributions of measured distance of gold particles from the anammoxosome membrane are shown in Fig. 3. Statistical evaluation by box-and-whiskers plots clearly visualized the high specific labeling of the anammoxosome membrane-associated F-ATPase 1 using the anti-kuste3793 and anti-kuste3795 antibodies, as compared to the two negative controls (Fig. 4).

Figure 3.

Frequency distributions of measured distance of gold particles from the anammoxosome membrane. The zone of acceptance (± 25 nm) is indicated by filled bars. (a) Anti-kuste3793. (b) Anti-kuste3795. (c) Control. Left panels illustrate experimental series with pre-immune serum as control. Right panels illustrate experimental series with anti-CRAMP antibody as control.

Figure 4.

Measured distance of the immunogold labeling from the anammoxosome membrane (open circles) for experimental series using pre-immune serum (a) and anti-CRAMP antibody (b) as control. Statistical evaluation of the immunogold labeling is illustrated by box-and-whisker plots. The line within the box denotes the median distance of the positive immunogold labeling from the anammoxosome membrane, the box represents the standard error, and the whiskers indicate the standard deviation for the mean distance measured. Values more than three times the height of the box are statistically considered as distant outliers (stars).

The specific association to the anammoxosome membrane was somewhat less clear for the anti-kuste3795 antibody as compared to the labeling with anti-kuste3793. Affinity data indicate similar binding properties and specificities of both antibodies (Figs S1 and S2). It could be speculated that the anti-kuste3795 antibody detects a portion of the protein that at instances might be relocated (smeared) during cryosectioning. However, further investigations are needed to clarify this eventual difference between the two antibodies.

The specific, however low, degree of labeling suggests a low expression level (i.e. low number of ATPase copies). From studies using cultures of pure anammox bacteria, it has been inferred that the anammox bacteria grow very slowly, dividing only once in every 2 weeks (Strous et al., 2006). Recently, one of the central enzymes of the anammox reaction, the enzyme complex of the HZS, was isolated (Kartal et al., 2011). Hydrazine synthesis by the catalytic part of the HZS (open-reading frames kuste2859–2861) was investigated using a coupled assay with another enzyme, kustc1061, which converted the generated hydrazine to nitrogen gas. Using inline image and 14NO as substrates for the HZS, the generation of 29N2 was measured using gas chromatography–mass spectrometry (GC-MS). Supporting the idea of slow growth rates, the activity of the HZS was observed to be low under the experimental conditions applied (Kartal et al., 2011, 2013). The slow growth infers a low catabolism that would restrict the number of protons generated inside the anammoxosome per unit time. The expression of ATP synthase is tightly regulated, as synthesis of such a major protein complex is energetically costly for the cell (Kramarova et al., 2008). For example, in brown adipose tissue mitochondria, where the energy is released as heat through an uncoupling protein (UCP1), the proton gradient is uncoupled from the ATP synthase complex. The copy number of ATP synthase in brown adipose tissue mitochondria is therefore kept low (Kramarova et al., 2008). Further, the bacteria clusters used in this study were sampled from a waste-water treatment facility, exposing the bacteria to conditions different from those in laboratory monocultures. Also, when performing immunogold EM, it is likely that the number of exposed ATP synthases available for immunogold staining is limited in a thin two-dimensional section arising from a three-dimensional anammoxosome membrane. The curved membrane of the anammoxosome, arising from a maximization of membrane area available for the catabolic processes (van Niftrik et al., 2008a), might also affect the steric availability of the target for the antibody. These aspects could, at least partly, explain the relatively low labeling observed in the EM images.

In conclusion, the results from immunogold labeling allowed for a highly specific localization of the alpha- and beta-subunits of the identified F-ATPase 1 to the anammoxosome membrane. The use of the experimentally verified linear epitopes (peptides) found in the proteomics analysis, proved successful in generating highly specific antibodies for the F-ATPase 1 under investigation. The high sequence similarity of the F-ATPase 1 to an F-ATPase 1 with a verified function as an ATP synthase, together with the presence of catabolic enzymes inside the anammoxosome (van Niftrik et al., 2008b; Karlsson et al., 2009) and experimental evidence of a proton gradient across the anammoxosome membrane (van der Star et al., 2010), confirm that the anammoxosome membrane is energized and a site for ATP synthesis. An intracellular membranous compartment containing a membrane-bound ATP synthase has previously not been encountered in any other bacteria.


Financial support was obtained from The Swedish Research Council (VR), The Foundation for Strategic Environmental Research (MISTRA), and The Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (FORMAS). S. Stridh and M. Tuvesson (SYVAB) provided access to the anammox pilot waste-water reactor. The EM equipment has been defrayed by grants from the Lundberg Research Foundation. The technical support by Mrs Yvonne Josefsson is acknowledged. The expertise at the Proteomics Core Facility, University of Gothenburg, as well as Innovagen AB, Lund, is acknowledged. The authors A.K. and R.K. are employed by Nanoxis Consulting AB, a Company which holds a patent for the LPI technology (WO2006068619). We acknowledge Stina Lindqvist for help with statistical evaluations.