Haloferax volcanii archaeosortase is required for motility, mating, and C-terminal processing of the S-layer glycoprotein

Authors


For correspondence. E-mail pohlschr@sas.upenn.edu; Tel. (215) 573 2283; Fax (215) 898 8780.

Summary

Cell surfaces are decorated by a variety of proteins that facilitate interactions with their environments and support cell stability. These secreted proteins are anchored to the cell by mechanisms that are diverse, and, in archaea, poorly understood. Recently published in silico data suggest that in some species a subset of secreted euryarchaeal proteins, which includes the S-layer glycoprotein, is processed and covalently linked to the cell membrane by enzymes referred to as archaeosortases. In silico work led to the proposal that an independent, sortase-like system for proteolysis-coupled, carboxy-terminal lipid modification exists in bacteria (exosortase) and archaea (archaeosortase). Here, we provide the first in vivo characterization of an archaeosortase in the haloarchaeal model organism Haloferax volcanii. Deletion of the artA gene (HVO_0915) resulted in multiple biological phenotypes: (a) poor growth, especially under low-salt conditions, (b) alterations in cell shape and the S-layer, (c) impaired motility, suppressors of which still exhibit poor growth, and (d) impaired conjugation. We studied one of the ArtA substrates, the S-layer glycoprotein, using detailed proteomic analysis. While the carboxy-terminal region of S-layer glycoproteins, consisting of a putative threonine-rich O-glycosylated region followed by a hydrophobic transmembrane helix, has been notoriously resistant to any proteomic peptide identification, we were able to identify two overlapping peptides from the transmembrane domain present in the ΔartA strain but not in the wild-type strain. This clearly shows that ArtA is involved in carboxy-terminal post-translational processing of the S-layer glycoprotein. As it is known from previous studies that a lipid is covalently attached to the carboxy-terminal region of the S-layer glycoprotein, our data strongly support the conclusion that archaeosortase functions analogously to sortase, mediating proteolysis-coupled, covalent cell surface attachment.

Introduction

Prokaryotes secrete a wide variety of proteins across the cytoplasmic membrane. Some secreted proteins, such as many toxins and polymer-degrading enzymes, are released into the extracellular environment; however, most remain attached to the cells, by intercalating a terminal transmembrane segment into the cell membrane; through a covalent link between an amino-terminal cysteine and the lipid bilayer; or by associating with other cell-bound proteins (Szabo and Pohlschroder, 2012). Recent in silico data have suggested that archaea also anchor a portion of membrane-bound proteins through a covalent association of their carboxy-termini with the lipid bilayer (Haft et al., 2012).

The genomes of some euryarchaeal species contain a gene that encodes a predicted transpeptidase with multiple transmembrane segments, archaeosortase A (Haft et al., 2012). This archaeosortase is distantly related to a predicted bacterial exosortase, which, despite a lack of sequence similarity, is believed to show functional similarities to sortase (Haft et al., 2006). All three proteins – sortase, exosortase, and archaeosortase – are predicted to process target proteins, hereafter called their substrates, with removal of a short, hydrophobic segment. As was first shown for sortase A from Staphylococcus aureus (Mazmanian et al., 1999), sortases recognize a tripartite structure at the C-terminus consisting of a signature motif, a transmembrane alpha helix domain, and a cluster of basic residues.

Sortase A recognizes the specific signature motif (LPXTG), which is located immediately upstream of a transmembrane alpha helix domain in the substrate (Mazmanian et al., 1999). To date, the majority of Gram-positive bacterial species examined encode homologues of this sortase, which are now known to play a critical role in several basic biological processes (Spirig et al., 2011).

Sortase A substrates have amino-terminal signal peptides that are recognized by the machinery of the Sec protein transport pathway, which carries proteins to the cytoplasmic membrane where they exit the cytoplasm through the Sec pore (Kline et al., 2009). The hydrophobic stretch of a sortase substrate inserts laterally into the lipid bilayer, temporarily anchoring the protein to the exterior of the cytoplasmic membrane, where the sortase recognizes the conserved signature motif and transfers its substrate to the peptide portion of the peptidoglycan precursor lipid II (Perry et al., 2002). This transfer involves proteolytic cleavage of the substrate protein near the carboxy-terminus and the formation of a covalent intermediate with the sortase via a thioester bond (Ton-That et al., 1999). This intermediate is subject to nucleophilic attack by the free amino group of lipid II, releasing the substrate protein from the sortase and covalently linking it to lipid II (Ton-That and Schneewind, 1999). The lipid II-linked sortase substrate is incorporated into the cell wall, anchoring it to the cell surface (Spirig et al., 2011).

Additional types of sortase, which recognize substrates carrying distinct signature motifs as part of the tripartite structure, have since been identified in many Gram-positive bacterial species (Schneewind and Missiakas, 2012). These minor sortases may not act exclusively as transpeptidases to peptidoglycan, but also cross-link pilin precursors. Sortases appear largely limited to the Gram-positive bacteria, although a dedicated system (one sortase for one substrate) occurs in a number of marine proteobacterial species (Comfort and Clubb, 2004; Haft et al., 2013). However, Haft et al. identified a number of Gram-negative bacteria that have families of Sec substrates sharing a homology domain structurally similar to the tripartite structure of known sortase substrates (Haft et al., 2006). Each genome encoding these substrates always encode at least one member of a well-defined protein family which henceforth is referred to as exosortase. Exosortases, which are multiple membrane-spanning proteins, lack any homology to known sortases, but contain conserved cysteine, arginine, and histidine residues, which are also found in the catalytic triad of the sortase family. Based on this observation, it was proposed that the predicted exosortases may perform functions similar to those of sortases, while recognizing a different motif, Pro-Glu-Pro (PEP-CTERM). Members of the exosortase family, such as EpsH from the methanolan biosynthesis cassette of Methylobacillus sp strain 12S, frequently are found within a locus of exopolysaccharide production genes, (hence the designation ‘exosortase’), despite not being involved in exopolysaccharide biosynthesis per se (Yoshida et al., 2003).

Although in silico analyses have confirmed that the patterns of the distribution of exosortases and its putative substrates in Gram-negative species are identical, exosortase processing of substrates containing the PEP-CTERM motif has not been confirmed in vivo.

A recent study by Haft et al. has revealed that many euryarchaeal genomes encode distant homologues of the bacterial exosortase (Haft et al., 2012). Consistent with this finding, these archaeal genomes also encode potential substrates having the typical (exo)sortase tripartite structure, which includes the signature motif, the transmembrane domain, and positively charged residues at the carboxy-terminus (Haft et al., 2012). These homologues of exosortases, which were termed archaeosortases, were classified into distinct subfamilies, and for each subfamily, substrates with a specific signature motif were proposed by genome analysis. Archaeosortase A (ArtA), which is found in many euryarchaea, probably recognizes a Pro-Gly-Phe motif (PGF-CTERM). Surprisingly, one of these predicted ArtA substrates is the S-layer glycoprotein. The S-layer glycoprotein forms a paracrystalline lattice that functions as the euryarchaeal cell wall. This protein was previously believed to be anchored to the cell membrane by the intercalation of a transmembrane segment (Sumper, 1987; Sumper et al., 1990). However, previous mass spectrometric analyses of S-layer proteins have failed to identify peptides at the carboxy-terminus of the protein, which was taken as evidence that the transmembrane segment may not be present in the mature protein (Haft et al., 2012). Moreover, at least a portion of the Haloferax volcanii S-layer glycoprotein has been found to be lipid-modified (Kikuchi et al., 1999; Konrad and Eichler, 2002; Kandiba et al., 2012). This may indicate that archaeosortases, and thus possibly also exosortases, process the carboxy-terminus of its substrates by proteolytic cleavage and transpeptidation to lipids, mirroring the proteolytic cleavage and transpeptidation to peptidoglycan catalysed by sortase A. To confirm that the archaeosortase plays a role in anchoring proteins, such as the S-layer glycoprotein, to the cell surface, we characterized the archaeosortase in the model euryarchaeon H. volcanii. The created ΔartA mutant, although viable, is impaired in growth, motility, and conjugation, and possesses an altered cell morphology. This points to the biological importance of the archaeosortase ArtA. We provide compelling initial evidence that the altered cell morphology can be attributed to incomplete maturation of the S-layer glycoprotein.

Results

The H. volcanii ΔartA strain has a growth defect

Haft et al. reported the identification of eight predicted ArtA substrates, including the S-layer glycoprotein (Haft et al., 2012 and Table 1). In order to examine the role of H. volcanii ArtA in vivo, we used homologous recombination to construct deletion strains that lack artA. We verified the deletion of artA by PCR, using primers designed to amplify the coding region as well as the DNA regions upstream and downstream of artA. A PCR product for the coding region was obtained for the parent strains (H53 and H98) but not for the ΔartA strain. The PCR product obtained with DNA from the deletion strain was significantly smaller than the product with DNA from the parent strain (Fig. S1). Most of the experiments were performed with strain H53 and derivatives thereof. H98 and its ΔartA derivative was only used for conjugation assays.

Table 1. H. volcanii predicted ArtA substrates
ProteinThr-rich regionPGF-CTERM domainaAnnotation, conserved domains
  1. aPGF-motif in bold; hydrophobic stretches are underlined, positive charges are in italics.
  2. bHypothetical protein.
  3. cCell adhesion related domain found in bacteria.
Hvo_2533AANLNPTRTISVNQTREDVELRVEDFENGDGGDAPRETAYEPTETTSSAPGFGPLVAVVGLLAVLVARRWSKbhypothetical; COG4885
Hvo_2160SRQAVTTTTPDDDGTTTTTTPDDGTTTTGTAGFDDETSTAASGTTDSPVPGFGAALAVVALLAAALLAVRRNDMuc19-precursor, cCARDB-domain
Hvo_2006TVEGTMSETTTDEPTDSMTDEPTETMTDEPATEMTDESMETTTESSTDAPGFGLVVALVALVAAALVAARRRhypothetical, CARDB-domain
Hvo_1110LETESVTASTESATETTTAADATATDATETESTEATTTESSDSTTESDAPGFGVVASLAAIGAAFLLARRRCobalamin-binding protein
Hvo_1095PIYEENATLVTYYMPENSTLTPDIRDAATAETTSASTATTADSSSGIDVPGFGAGVGVAALLVASFVAARGLhypothetical
Hvo_A0263LSTFVIGSTPVANTTDTTITTDESAATTESDITDTIDTTDEPATTASDSPGFGVVLTVIAMLVGTFVVRRARCRGluG-domain, fibronectin-like domain, cadherin-repeat
Hvo_B0206DDTGDDSDDSNDGDDAGDGSDGSDGDSGGTPETTTDDNTPATTTTDTPVPGFGVSVALAALVIGSVLLARRRAThypothetical, collagen triple helix domain
Hvo_2072SEREDTTTSSDNATDTTTTTDGPTETTTTAEPTETTEEPTEETTTSSNTPGFGIAVALVALVGAALLALRRENS-layer glycoprotein

The H53 ΔartA strain was viable, but displayed a growth defect in media containing a low (14%), standard (18%), or high (23%) salt concentration (Fig. 1). The most notable growth defect was observed in 14% salt water concentration, while the artA deletion strain exhibited the least severe defect at 23%. These results are consistent with the cell wall of ΔartA being defective, because H. volcanii cells, which maintain a high internal concentration of KCl to counter the high external NaCl concentration, are exposed to the highest turgor pressure in low salt environments. The artA deletion was complemented by cloning the artA gene on a low-copy plasmid under the control of a tryptophan-inducible promoter. When complementation strains were grown in medium containing tryptophan, growth was very similar to that of H53 (Fig. 1).

Figure 1.

Depletion of ArtA results in a growth defect. H. volcanii H53 (wt), ΔartA, and the motility suppressor (ΔartA*), all transformed with an empty pTA963 plasmid, as well as H53 and ΔartA expressing artA expressed from pTA963, (wt + artA and ΔartA + artA respectively) were inoculated in liquid semi-defined Casamino Acids (CA) media supplemented with tryptophan (final concentration 50 μg ml−1) at low (14%), normal (18%) and high (23%) salt concentration.

The S-layer glycoprotein is carboxy-terminally processed and ArtA is required for this post-translational modification

To confirm that it is indeed processed at its carboxy-terminus, the S-layer glycoprotein was isolated and purified from the supernatant fractions of cultures from ΔartA and its parent H53. We observed that significantly more protein was released into the supernatant in the ΔartA strain than in the H53 strain (data not shown), which may indicate reduced stability of the S-layer. The glycoprotein was in-gel digested with trypsin and peptides were characterized using LC-MS/MS. Peptide identification of the samples revealed 119 distinct peptides at a false discovery rate below 1% (Fig. 2A). The abundance of peptides was determined using the AMT (accurate mass and time tag) approach (Smith et al., 2002; Pasa-Tolic et al., 2004). For nearly all of the peptides (117 of 119), the signal intensity was very similar for samples from the ΔartA strain and its parent. However, the identification of two overlapping peptides near the extreme C-terminus (VALVGAALLALR and VGAALLALR) was restricted to ΔartA samples. These peptides were found in all ΔartA samples but completely absent from the H53 samples (Fig. 2B). To confirm this finding, the peptide/spectrum matches for both these peptides were further verified manually. The precursor ion identified in survey scans had a perfect isotopic correlation with the theoretical and < 5ppm mass error (Fig. 2C). The fragmentation spectra for these precursors showed strong b/y ladder representation with low ppm error (Fig. 2D). Therefore, the identification of the two carboxy-terminal peptides in ΔartA samples was highly confident. No tandem mass spectra from these two peptides were identified in H53 samples. Moreover, the AMT quantification strategy did not find any parent ion features that matched these peptides as parent ions. As extra validation, we generated an extracted ion chromatogram (XIC) for the m/z values of the peptides discovered in the ΔartA samples. These XICs showed zero intensity for the expected LC retention window, confirming that the carboxy-terminal peptides VALVGAALLALR and VGAALLALR were not present in the H53 samples. Thus, the wild-type S-layer glycoprotein was indeed proteolytically processed and the extreme carboxy-terminal region was not present in the mature protein. Moreover, the overlapping carboxy-terminal peptides identified in the ΔartA samples were the first to be identified in the carboxy-terminal S-layer glycoprotein region by any mass spectrometric analysis, strongly suggesting that carboxy-terminal processing of this target occurs in the wild-type and is prevented in the ΔartA strain.

Figure 2.

High-resolution mass spectrometry analysis of S-layer glycoprotein.

A. Protein sequence coverage from the identified peptides identified from in-gel tryptic digestion. Red typeset was used to indicate regions covered by identified peptides from H53 strain samples; underline regions covered from ΔartA samples. Green underline was used for the overlapping TM peptides that were found exclusively in ΔartA samples. The PGF motif was highlighted yellow. The signal peptide was marked grey. Incomplete sequence coverage was likely due to an uncharacteristically sparse distribution of cleavage residues Lys and Arg in the protein sequence and to post-translational modifications.

B. Ratio of signal intensity from H53 and ΔartA samples for 119 peptides covering the S-layer glycoprotein; two carboxy-terminal peptides (green spots) were observed and identified exclusively in the ΔartA sample. The black spot indicated a low-intensity peptide of 34 amino acids, which was found in both samples but was subjected to MS/MS analysis only in 1 of 8 analysed samples, the one being from ΔartA. For the carboxy-terminal 19 of this 34 amino acid peptide, no other overlapping identification was obtained and thus this region is underlined and in black typeset in panel A.

C. The overlay between the theoretical (red) and observed (black) isotopic distributions for peptide VALVGAALLALR.

D. The fragments from tandem mass spectra matching the proposed peptide. Dots indicate that a fragment peak was found (e.g. b3) and the colour of the dot indicates the mass measurement accuracy.

We noted that the abundance of the two carboxy-terminal peptides VALVGAALLALR and VGAALLALR (5.80 and 5.61 respectively) agreed with the bulk of the peptide intensities for the ΔartA samples (mean 5.82, standard deviation 0.70). This suggested an even stoichiometry for all the identified peptides, meaning that there was likely only one form of the S-layer glycoprotein identified in the ΔartA samples: an isoform that was incompletely processed.

The only other peptide exclusively identified in ΔartA samples was DSAIGDGSLPSGPSNGATLNDLTGYLDTLDQNNN. While this low-abundance peptide was never selected for fragmentation in H53 samples, it was detected in these samples with a maximum signal intensity ratio of 0.85, consistent with the observed mean 0.91 and standard deviation 0.07 for the other 116 peptides (Table S2).

H. volcanii ΔartA cells are significantly shorter than those of the parent strain

Considering that the S-layer is thought to be critical for the determination and maintenance of the cell shape, we examined samples taken from H53 and ΔartA strain cultures for changes in cell shape and noticeable defects in ΔartA cell morphology using phase contrast microscopy (Fig. 3). Although H. volcanii H53 cells do exhibit a degree of morphological variation, the majority of early log phase parent and ΔartA cells were rod-shaped. However, while the cells of the parent strain had an elongated appearance (mean length: 4.7 μm; SD: 0.9 μm), cells from the ΔartA cultures were significantly shorter (mean length: 2.6 μm; SD: 0.6 μm). wt + artA cells were slightly more elongated (mean length: 5.1 μm; SD: 1.1 μm) than the parent strain, and ΔartA +artA cells possessed the longest rod-like shape (mean length: 5.8 μm; SD: 0.5 μm). The cell width of all the strains measured was comparable. Viability counts revealed no significant differences among the strains.

Figure 3.

Phase contrast microscopy of H. volcanii H53 and ΔartA in liquid culture. The majority of H. volcanii parent and mutant strains were rod-shaped. However, cells of the parent strain (wt) were nearly twice as long as cells of ΔartA cultures. Morphology of a ΔartA motility suppressor (ΔartA*) was comparable to that of the non-motile ΔartA. Both the wt + artA and ΔartA + artA exhibited more elongated rod-like shapes than did the wild-type, with the ΔartA + artA displaying the largest average length of all strains observed. Size bar: 10 μm.

To directly analyse the structure of the S-layer, the cell walls of wild-type and ΔartA cells were examined by transmission electron microscopy (Fig. 4). Cells were prepared by high pressure freezing followed by low temperature acetone, 2% OsO4, and 0.1% uranyl acetate substitution, before being embedded into EPON resin. The most obvious difference between wild-type and ΔartA cells was their cell morphologies. The cytoplasm of the ΔartA strain also stained darker. Although the assembly of an S-layer surface structure was found in H53 and the ΔartA strains, distinct differences were evident at higher magnification. In general, the ΔartA strain S-layer had a less compacted appearance and, with an S-layer width average that ranged from 11 nm to 15 nm, was wider than the H53 S-layer, which had a width that averaged from 6 nm to 10 nm. The measurements of the S-layer thickness were performed for five wild-type and mutant cells, and they revealed very consistent results. These observations are consistent with the S-layer glycoprotein processing defect identified in the ΔartA strain by MS analysis.

Figure 4.

Transmission electron microscopy. H. volcanii H53 (wt) or ΔartA cells were preserved using a high pressure freezing technique and sliced into thin sections.

A. H53 preparations reveal mostly rod-shaped wild-type cells, while ΔartA cells are significantly thinner and have irregular shapes. Size bar: 50 nm.

B. While H53 and ΔartA strains assemble an S-layer (arrows), the ΔartA strain S-layer has a less compacted appearance and is wider (11–15 nm) than the H53 S-layer (6–10 nm). Size bar: 50 nm.

Deletion of artA impairs motility and conjugation

Next to the S-layer glycoprotein, seven additional predicted H. volcanii Sec substrates contain a similar tripartite C-terminal region (Table 1) suggesting the involvement of ArtA in cellular processes other than cell wall biosynthesis. Interestingly, when stab-inoculating H53, ΔartA, as well as the ΔartA complementation strains into (0.3%) motility agar, H53 cells produced a growth halo between 3 and 5 days after stab-inoculation, which indicated swimming motility. No growth halo was observed in the ΔartA strain during the same incubation period. However, the ΔartA complementation strains expressing the tryptophan-inducible artA in trans not only restored the ΔartA swimming motility, it exhibited a greater motility than H53. Hypermotility of the complemented ΔartA strain was evident from the larger growth halo 5 days after inoculation (Fig. 5). The empty expression plasmid pTA963 itself had no effect.

Figure 5.

The ΔartA strain exhibits a motility defect. H. volcanii H53 (wt), ΔartA, and the motility suppressor (ΔartA*), all transformed with an empty pTA963 plasmid, as well as H53 and ΔartA expressing artA expressed from pTA963 (wt + artA and ΔartA + artA respectively), were stab inoculated on motility plate (0.3% agar). While H53 produced a growth halo after 3–5 days, no growth halo was observed for the ΔartA strain during the same incubation period. Complementing the deletion strain with artA expressed from a plasmid not only restored motility but resulted in hypermotility compared with the H53 strain.

Notably, when colonies of the ΔartA strains were stab-inoculated on the motility plate and incubated for an extended period (> 10 days), all of the colonies ultimately suppressed the motility defect and exhibited swimming motility. Upon restabbing, the motile ΔartA cells started moving after 4–6 days (Fig. 5). Using PCR analysis, we confirmed that the motility was not due to reversion of the ΔartA genotype (Fig. S1). Also, while the motility defect was partially suppressed in these strains, the growth defect remained (Fig. 1). This, as well as the fact that the cells had a size similar to that of the non-motile ΔartA strain (Mean length: 2.7 μm; SD: 0.7 μm) (Fig. 3), suggested that the suppression was independent of the S-layer.

Haloferax volcanii conjugation occurs independently of the presence of either type IV pili or flagella (Tripepi et al., 2010). Several predicted ArtA substrates contain cadherin and fibronectin-like domains (Table 1), which are known to facilitate interactions with other cells (Harrison et al., 2010; Brasch et al., 2012; Chagnot et al., 2012). H. volcanii conjugates by forming cytoplasmic bridges (Rosenshine et al., 1989). Therefore, we tested whether the artA deletion had an influence on conjugation by modifying a published conjugation assay (Mevarech and Werczberger, 1985). In brief, we co-cultured H. volcanii tryptophan (H53, ΔartA) and thymidine (H98, ΔartA) auxotrophs, each harbouring an artA deletion, in a selective medium that lacked both tryptophan and thymidine. In the absence of these compounds, independent growth of either auxotroph was not possible. We also plated diluted cultures on complex Modified Growth Medium (MGM) supplemented with thymidine to obtain viable counts. It should be noted that both H98 wild-type and mutant strains exhibited comparable growth to the wild-type and mutant H53 background strains (data not shown).

The conjugation frequency for the parent strains was approximately 4 × 10−3, while the average conjugation frequency for the ΔartA derivatives was only 3.7 × 10−4. Thus, we observed that conjugation rates between ΔartA strains were an order of magnitude lower than those of the parent strains when the strains were co-cultured on a filter placed on a MGM plate and subsequently co-cultured in CA media lacking tryptophan and thymidine. In this experiment, artA was deleted in both strains, because surface adhesin-mediated contact might have been possible even if only one strain expressed the adhesin.

In summary, we showed that various biological phenomena like growth, cell shape, motility, and conjugation were critically dependent on an intact artA gene, even though this gene proved not to be essential.

Discussion

Following up on in silico data, we obtained the first in vivo evidence of the function of archaeosortase. Our data strongly support the functional similarity of archaeosortases with bacterial sortases, namely, proteolytic removal of a carboxy-terminal transmembrane segment as a late step in protein processing, and not by hydrolysis, but by a transpeptidation that anchors proteins covalently to the cell surface. The signature motif for sortase recognition, LPXTG, sits four or five residues away from the transmembrane segment, while the analogous PGF motif in archaeosortase targets lies flush to the transmembrane segment. This contrast follows the model that archaeosortase attaches target proteins covalently with a lipid found in the membrane, while sortase attaches proteins to the growing cell wall.

Our in vivo data also support the in silico prediction that archaeosortases share similar functionality with the distantly related bacterial homologue, exosortase, for which in vivo data have – to our knowledge – not yet been obtained. Thus, prokaryotes independently evolved at least twice a similar mechanism to recognize, process, and anchor to the cell surface proteins that are transported across the cytoplasmic membrane and contain a conserved carboxy-terminal tripartite structure.

We showed that while it was possible to make an artA deletion strain, H. volcanii lacking this gene exhibited several phenotypes: a growth defect, significantly smaller cells that are more fragile, and impaired motility and conjugation. The effects on cell shape and fragility are consistent with a destabilization of the S-layer for H. volcanii lacking ArtA. Our data strongly support the hypothesis that this is due to the involvement of ArtA in carboxy-terminal processing of the S-layer glycoprotein. Concerning the motility and conjugation phenotypes, these may be indirect effects of an altered S-layer or may be attributed to other predicted ArtA substrates.

Direct evidence that ArtA is involved in the processing of the S-layer carboxy-terminus was obtained by LC-MS/MS analysis of this glycoprotein purified from wild-type and ΔartA cells. We readily identified two overlapping peptides from the carboxy-terminal TM domain of S-layer glycoprotein purified from the ΔartA strain. Conversely, we failed to identify this same peptide from the wild-type, strongly suggesting that the wild-type carboxy-terminus is indeed processed. The failure to identify the peptide in the wild-type strain cannot be attributed to technical problems, as our peptide quantification data show that the other 117 identified peptides from the S-layer glycoprotein are found in similar quantities whether purified from the wild-type or from the ΔartA strain. In addition, the two overlapping peptides from the carboxy-terminal TM domain were found in a quantity similar to that of the other 117 peptides from that same strain. These data clearly show that the H. volcanii S-layer glycoprotein is processed and that ArtA is involved in this processing mechanism.

Similar to the sortase substrates, it can be assumed that the S-layer glycoprotein is subject to two types of post-translational modification at its carboxy-terminus: lipid attachment and proteolytic cleavage. While lipid attachment in the carboxy-terminal region has been previously demonstrated (Kikuchi et al., 1999; Konrad and Eichler, 2002; Kandiba et al., 2012), carboxy-terminal proteolytic cleavage had previously not been confirmed. In contrast, it had been proposed that the S-layer glycoprotein is anchored by its carboxy-terminal transmembrane domain (Sumper, 1987; Sumper et al., 1990). Interestingly, a recent paper suggested that two S-layer glycoprotein forms coexist in wild-type H. volcanii (Kandiba et al., 2012). In addition to a membrane-associated fraction that requires detergent for removal from the membrane, Kandiba et al. identified a more weakly membrane-associated fraction that can be released by EDTA treatment. Moreover, mass spectrometry analysis of the base-treated, weakly interacting fraction identified a lipid, which was covalently linked in an alkaline-sensitive manner. As lipid attachment should strengthen and not weaken the membrane interaction, these data support our experimental results, which show that the S-layer glycoprotein undergoes carboxy-terminal proteolytic processing, a coupling reminiscent of the coupling performed by sortases.

At this stage, we do not have a positive identification of the processed carboxy-terminal peptide from the S-layer glycoprotein in the wild-type strain, despite more extensive attempts for its identification. We searched for peptides that carry a modification of uncertain mass using the USTags approach (Shen et al., 2008) (data not shown) but did not find any promising leads in yet unidentified mass peaks, and failed to identify the corresponding peptide. Any proteolytic peptide from this region is supposedly a glyco-lipo-peptide, and it is known that identifying such peptides are associated with extreme experimental challenges. Consistent with this assumption, not a single peptide from this Thr-rich region could be identified, neither in this current study nor in any other mass spec analysis of an archaeal S-layer glycoprotein tested (Haft et al., 2012). Although the underlying technical problems have yet prohibited the positive identification of the carboxy-terminally processed peptide, we are certain that the data already presented unambigously show that ArtA is involved in carboxy-terminal post-translational modification of the S-layer glycoprotein and strongly support that proteolytic cleavage of the carboxy-terminal transmembrane domain occurs.

While we primarily concentrated on the effect of the ΔartA mutant on the S-layer glycoprotein, the additional phenotypic effects described in the current manuscript point to the biological importance of the archaeosortase. The reduced growth, impaired motility, and reduced conjugation efficiency of the ΔartA mutant may be an indirect effect due to a weakened S-layer or a direct effect due to the absence of a distinct ArtA substrate.

The observed growth defect may be due to poor cell wall stability. The most severe growth defect is observed when cells are grown in media containing low (14%) rather than moderate (18%) or high (23%) salt concentration (Fig. 1), probably due to an increased cell turgor under these growth conditions. We observed some variation in growth in these experiments. Occasionally the ΔartA+artA strain grew to higher densities than the parent strain, and sometimes the parent strain wt + artA strain grew to higher densities. There seems to be a narrow range of artA expression that is optimal for growth under the conditions tested.

The effect on cell shape can be directly attributed to the S-layer glycoprotein. The growth defect may also be due to reduced cell stability resulting from an impaired S-layer. However, it should also be noted that under certain growth conditions, an S-layer independent ArtA substrate may add to the growth defect. For example, one of the substrates is a cobalamin uptake protein that may be critical when cells are grown on defined media.

It is currently enigmatic if the effect on motility is a secondary effect also mediated by an impaired S-layer or if another ArtA substrate is responsible for this effect. A motor must be anchored to the cell membrane and interaction with the S-layer may participate in anchoring (Jarrell and McBride, 2008; Ghosh and Albers, 2011). However, as the molecular details of the archaeal motor are still largely unresolved, a distinction between direct and indirect effects is currently not possible. However, it is unlikely that the motility defect is solely due to S-layer instability, as suppressors have been obtained that are motile, albeit not at wild-type levels, yet have a severe growth defect. In this respect, it will also be interesting to analyse why a probable artA overexpressor (the complementation strain with artA transcription from Trp-inducible pAF9 plasmid) is hypermotile.

Finally, it is currently not clear why the ΔartA mutant demonstrates impaired conjugation. In fact, a destabilized S-layer would be expected to facilitate rather than prevent DNA uptake. Thus, other predicted ArtA substrates, some of which contain domains reminiscent of eukaryotic cell attachment molecules, may be responsible for the observed effect.

In summary, we have shown experimentally that ArtA is important for several biological phenomena. The comparison of the S-layer glycoprotein from wild-type and the ΔartA mutant strains strongly supports the hypothesis that the S-layer glycoprotein is not attached via a carboxy-terminal transmembrane region, but rather that the carboxy-terminal region is cleaved off. The S-layer glycoprotein is then anchored to the lipid bilayer. It is highly likely that ArtA is directly involved in this post-translational processing step. The availability of the mutant will allow a deeper experimental analysis of the archaeosortase activity. It needs to be analysed if all archaeosortase targets carry a lipid covalently attached to the carboxy-terminal region. Moreover, the multiple effects of the deletion of artA indicates that analysis of this enzyme and its substrates will yield a better understanding of processes ranging from DNA transfer to motility and nutrient uptake.

Experimental procedures

Reagents

All enzymes used for standard molecular biology procedures were purchased from New England BioLabs, except for iProof High-Fidelity DNA polymerase, which was purchased from Bio-Rad. The MF membrane filters (0.025 μm) and Ultracel-3K membrane were purchased from Millipore. DNA and plasmid purification kits were purchased from Qiagen. NuPAGE gels, buffers, and reagents were purchased from Life Technologies (Invitrogen). Difco agar and Bacto yeast extract were purchased from Becton, Dickinson, and Company. Peptone was purchased from Oxoid. 5-Fluoroorotic acid (5-FOA) was purchased from Toronto Research Biochemicals. All other chemicals and reagents were purchased from either Fisher or Sigma.

Strains and growth conditions

The plasmids and strains used in this study are listed in Table 2. H. volcanii strains were grown at 45°C in liquid or on solid complex (MGM) (Dyall-Smith, 2004) or semi-defined (CA) (Dyall-Smith, 2004) medium. Plates contained 1.5% agar unless mentioned otherwise. To ensure equal concentrations of agar in all plates, the agar was completely dissolved prior to autoclaving, and the autoclaved medium was stirred before plates were poured. H. volcanii strain H53 (Allers et al., 2004) was grown in MGM medium without any supplements. H. volcanii strain H98 (Allers et al., 2004) was grown in MGM medium supplemented with thymidine (40 μg ml−1 final concentration). For the selection of the ΔartA (HVO_0915) mutant (see below), 5-FOA was added at a final concentration of 150 μg ml−1 in CA medium, and uracil was added at one-fifth its normal concentration (i.e. a 10 μg ml−1 final concentration) during 5-FOA selection. H. volcanii strain H53 and the ΔartA mutants transformed with pTA963 or the recombinant derived from it were grown in CA medium supplemented with tryptophan (50 μg ml−1), while the transformed H. volcanii strain H98 and ΔartA were grown in CA medium supplemented with thymidine (40 μg ml−1), hypoxanthine (40 μg ml−1), and tryptophan (50 μg ml−1). Escherichia coli DH5α and DL739 strains were grown at 37°C in NZCYM medium supplemented with ampicillin (200 μg ml−1), when necessary (Blattner et al., 1977).

Table 2. Plasmids and strains
Strain or plasmidRelevant characteristic(s)Reference or source
Plasmids  
pTA131Ampr; pBluescript II with BamHI-XbaI fragments from pGB70 harbouring pfdx-pyrE2Allers et al. (2004)
pTA963Ampr, pyrE2 and hdrB markers, Trp-inducible (ptna) promoterAllers et al. (2010)
pAF3pTA131 carrying chromosomal flanking regions of artAThis study
pAF9pTA963 carrying artAThis study
Strains  
DH5αE. coli, F − ϕ80dlacZΔM15 (lacZYA-argF)U169 recA1Invitrogen
endA1 hsdR17(rK−mK−)phoA supE44 thi-1 gyrA96 relA1
DL739E. coli, MC4100 recAdam-13::Tn9Blyn et al., 1990
H53H. volcanii, ΔpyrE2 ΔtrpAAllers et al. (2004)
H98H. volcanii, ΔpyrE2 ΔhdrBAllers et al. (2004)
AF103H53, ΔartAThis study
AF109AF103 containing pAF9This study
JZ101H53, containing pAF9This study
AF110H98, ΔartAThis study
JZ102AF110, containing pAF9This study
JZ103H98, containing pAF9This study
ΔartA*H53, ΔartA motility suppressor mutantThis Study

Generation of a chromosomal artA deletion in H53 and H98

Chromosomal deletions were generated by using a homologous recombination (pop-in pop-out) method previously described (Allers et al., 2004). Plasmid constructs for use in the pop-in pop-out knockout process were generated by using overlap PCR as described previously (Tripepi et al., 2010) with the following modification: approximately 685 nucleotides flanking the artA gene were PCR amplified and cloned into the haloarchaeal suicide vector pTA131. The final construct, pAF3, contained upstream and downstream artA flanking region inserts (primers used are listed in Table S1 in the Supporting information). The insertion of the correct DNA fragment into the cloning site of the recombinant plasmid was verified by sequencing using the same primers. To confirm the chromosomal replacement event at the proper location on the chromosome, colonies derived from these techniques were screened by PCR using primers listed in Table S1. The artA deletion mutant generated in strains H53 and H98 were designated AF103 and AF110 respectively.

Construction of expression vectors encoding the ArtA protein

The artA gene coding region was amplified by PCR and cloned into the low-copy expression vector pTA963 under the control of an inducible tryptophanase promoter (ptna) to induce the expression of this gene. The recombinant pTA963 carrying the artA gene was designated pAF9. pAF9 was isolated from E. coli DH5α and transformed into E. coli DL739 (Table 2). Using the standard polyethylene glycol (PEG) protocol (Dyall-Smith, 2004), non-methylated plasmid DNA isolated from E. coli DL739 was used to transform H. volcanii. To complement ΔartA strains, plasmid pAF9 was transformed into AF103 (derived from H53) to result in AF109, and into AF110 (derived from H98) to result in JZ102. The parent strains H53 and H98 were also transformed with pAF9 to generate strains JZ101 and JZ103, respectively, which thus had two copies of the artA gene under control of distinct promoters.

Motility assay

The motility assay was performed on motility plates (0.3% agar in CA medium). For the defined CA agar media, tryptophan, thymidine, and hypoxanthine were added at final concentrations of 50 μg ml−1, 40 μg ml−1, and 40 μg ml−1 respectively. A toothpick was used to stab inoculate the agar, followed by incubation at 45°C. Halo sizes around the stab-inoculation site were measured after 3–5 days of incubation.

Isolation of ΔartA* motility suppressor strain

Ten colonies of the ΔartA strain were stab-inoculated on the motility plate and incubated for 10 days at 45°C. The colonies that exhibited swimming motility after the incubation period were re-stabbed on the new motility plate to ensure the consistency of the motility suppressor. The halo size of the ΔartA motility suppressor mutants was measured after 4–6 days and compared with the wild type. This ΔartA motility suppressor strain will hereafter be referred to as ΔartA*.

Conjugation assay

The conjugation rate was assayed using a modified version of a protocol previously described (Mevarech and Werczberger, 1985). In brief, equal volumes (3–5 ml) from cultures of two different auxotrophic H. volcanii strains (H53, ΔtrpA, requiring tryptophan and H98, ΔhdrB, requiring thymidine) at OD600 of 0.3–0.5 were mixed. The mixture was filtered through Millipore Swinnex 25 mm filter units equipped with corresponding filters (pore size 0.45 μm, type HA 25 mm filters) using a 5 ml syringe. The filter unit was then disassembled and filter discs were placed cell side up on solid MGM medium supplemented with thymidine, and then incubated at 45°C overnight. After incubation, the filters were removed from the medium, placed in 2 ml Eppendorf tubes containing 1 ml of 18% salt water (Dyall-Smith, 2004), and shaken on a rotator for 1 h. Fifty microlitre of an undiluted sample and samples that were diluted 1:100 were plated on CA agar medium supplemented with uracil, but lacking tryptophan and thymidine. This prevented the growth of the auxotrophic strains H53 or H98, respectively, and selected for conjugants that contained both the trpA and hdrB genes. Twenty microlitre of samples that were diluted 1:100 000 were also plated on MGM agar medium supplemented with thymidine for viable cell counts.

Growth curves

Growth curves were generated using a Biotek Power Wave X2 microplate spectrophotometer. H. volcanii parent strain H53, the ΔartA mutant, and the ΔartA complementation strain were first incubated in 5 ml liquid cultures in CA medium supplemented with tryptophan (final concentration 50 μg ml−1), with continuous shaking at 45°C, until suitable OD600 values (0.2–0.5) were reached. Approximately 6 μl of each culture (adjusted slightly for OD600 differences) were then transferred into 194 μl of fresh CA medium with normal (18%), low (14%), or high (23%) salt concentration and supplemented with tryptophan (final concentration 50 μg ml−1), and grown to the stationary phase, with OD600 recordings taken every 30 min.

Purification of the S-layer glycoprotein

The isolation of the S-layer glycoprotein was performed by taking advantage of the surface structure purification technique as described previously (Fedorov et al., 1994), with modifications described below. H53 wild-type cells or ΔartA cells were inoculated into 5 ml MGM liquid medium. Two litres of MGM medium were inoculated with two 5 ml cultures each and the cultures were harvested at an OD600 of approximately 0.3 by centrifugation at 8700 r.p.m. (JA-10 rotor; Beckman) for 30 min. The supernatant was centrifuged again (8700 r.p.m. for 30 min) and incubated at room temperature with 4% (wt/vol) PEG 6000 for 1 h. The PEG-precipitated proteins were then centrifuged at 16 000 r.p.m. (JLA-16.250 rotor; Beckman) for 50 min at 4°C, and the S-layer purified by cesium chloride (CsCl) density gradient centrifugation (overnight centrifugation at 50 000 r.p.m.) (VTI-65.1 rotor; Beckman). CsCl was dissolved in a 3 M NaCl saline solution to a final density of 1.37 g cm−1. The CsCl gradient fractions were dialysed against water and concentrated using an Ultracel-3K membrane (Millipore). The electrophoresis of the proteins in the fractions was performed with 7% NuPAGE Tris-Acetate (TA) gels (Life Technologies) under denaturing conditions using Tris-Acetate buffer at pH 8.1 to separate the S-layer glycoprotein from other proteins in the fraction. The proteins on the 7% NuPAGE Tris-Acetate gels were stained with Coomassie brilliant blue and the protein band that corresponded to the expected S-layer glycoprotein size was cut from the gel and submitted for mass spectrometry analysis.

Mass spectrometry

The S-layer glycoprotein band from the Coomassie-stained bands from the 7% NuPAGE Tris-Acetate gels was tryptically digested for LC-MS/MS analysis according to published procedures, except that samples were not alkylated (Shevchenko et al., 2006). The previously described capillary Reversed-phase Liquid Chromatography (RPLC) system for peptide separations (Livesay et al., 2008) was slightly modified with regards to the pumping system. Briefly, the High-Performance Liquid Chromatography (HPLC) system consisted of a custom configuration of Agilent 1200 nanoLC pumps (Agilent, Santa Clara, CA), 2-position Valco valves (Valco Instruments, Houston, TX), and a PAL autosampler (Leap Technologies, Carrboro, NC), allowing for fully automated sample analysis across four separate HPLC columns (3 μm Jupiter C18 stationary phase, Phenomenex, Torrance, CA). The mobile phases consisted of 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B). The HPLC system was equilibrated with 100% mobile phase A for 2 min, followed by a 100 min RPLC gradient (0–12% buffer B over 10 min, 12–35% buffer B over 55 min, 35–95% buffer B over 10 min, 95% buffer B for 5 min and 0% buffer A for 20 min). The flow rate during the run was set at 400 nl min−1. ESI using an etched fused-silica tip (Livesay et al., 2008) was employed to interface the RPLC separation to a LTQ Orbitrap Velos mass spectrometer (Thermo Scientific, San Jose, CA). Precursor ion mass spectra (AGC 1 × 106) were collected for 400–2000 m/z range at a resolution of 60 K, followed by data dependent HCD MS/MS (normalized collision energy 35%, AGC 3 × 104) of the eight most abundant ions at a resolution of 15 K. A dynamic exclusion time of 60 s was used to discriminate against previously analysed ions.

Mass spectrometry data analysis

HCD MS/MS data were processed using SEQUEST (Eng et al., 2008) and a H. volcanii proteome database downloaded from GenBank (http://www.ncbi.nlm.nih.gov/protein) (Hartman et al., 2010). The database search against the decoy database, using both forward and reverse sequences, was performed with the following settings: no enzyme rule, methionine oxidation as a variable modification, and precursor mass limit at 5000 Da. A peptide mass tolerance of 3 Da and a fragment mass tolerance of 0.5 Da were allowed. First hits from the SEQUEST analysis were rescored using a MS-Generating function (MSGF) for spectral probabilities (Kim et al., 2008). Peptide identifications were filtered based on a 1% false discovery rate against the decoy database (Elias and Gygi, 2007).

Peptide quantitation was performed using the identified peptide signal intensity in precursor spectra eliminating redundant peptides with weak signal intensity. ICR2LS software (http://omics.pnl.gov) was used to determine the most abundant isotope peaks of the expected peptides. Matched peaks were filtered using 10 ppm mass accuracy and 0.2 tolerance LC window in normalized elution time. Maximum signal intensity was extracted from each dataset (no normalization). Results of this analysis are presented in Table S2.

Light microscopy and cell measurements

The H. volcanii H53 wild type, ΔartA, ΔartA*, ΔartA+artA, and wt + artA strains were grown in 5 ml liquid cultures to an OD600 of ∼ 0.1. One ml of cells was concentrated by centrifuging at 6800 r.p.m. for 5 min and pellets were resuspended in 20 μl of MGM liquid media. The concentrated cells were transferred on to a microscope slide and observed under the light microscope. Light microscopy was performed using an Eclipse TE2000-U inverted microscope (Nikon) equipped with a Plan Apo 100 × 1.0 NA objective and Cascade 512B CCD camera (Photometrics) driven by Metamorph imaging software (Molecular Devices). ImageJ software was used to measure the lengths and widths of each cell, and to calculate the average cell perimeter. At least 20 cells were used to measure each parameter. Cells were counted using a Petroff-Hausser Counting Chamber.

Electron microscopy

The H. volcanii H53 wild-type and ΔartA cells were cultured in 5 ml liquid MGM until they reached OD600 of 0.3. Cells were concentrated by centrifugation and drawn into 200 μm diameter Spectrapor® microdialysis tubing (Spectrum Laboratories). Segments of 2.5 mm tubing containing cells in culture media were frozen at high pressure (4350 psi) using an Abra model HPM010. Cellular water was replaced with acetone, 2% OsO4, and 0.1% UA (uranyl acetate) at −90°C for 72 h. Cells were washed three times in acetone at room temperature and incubated overnight in 1:1 acetone : EPON followed by two exchanges of EPON, before polymerization at 60°C for 48 h. Thin sections were cut at 60 nm thickness. Sections were counterstained in 2% UA and Reynolds lead citrate solution before observation at 80 KeV on an FEI Tecnai12 electron microscope. Images were recorded digitally using a Gatan Ultrascan 1000 4-megapixel camera at the indicated magnifications.

Acknowledgements

J.Z. was supported by National Science Foundation grant MCB02. M.P., A.F., and M.F.A.H. were supported by National Aeronautics and Space Administration grant NNX10AR84G. We thank Dewight Williams and Tatyana Svitkina for invaluable advice on microscopy and Fevzi Daldal for helpful discussions. Mass Spec analysis was performed in the W.R. Wiley Environmental Molecular Science Laboratory (EMSL), a national scientific user facility sponsored by the U.S. Department of Energy's Office of Biological and Environmental Research and located at Pacific Northwest National Laboratory. Pacific Northwest National Laboratory is operated by Battelle Memorial Institute for the U.S. Department of Energy under contract DE-AC05-76RLO-1830. S.H.P. is supported by the U. S. Department of Energy (DOE) Early Career Research Award.

Ancillary