Nonulosonic acid biosynthetic gene clusters are present among pathogenic and some intermediately pathogenic Leptospira species

The genome sequences of L. interrogans serovar Copenhageni strain L1-130 and L. interrogans serovar Lai strain 55601 contain genes predicted to synthesize sialic acids or related molecules (Figure 1A). Using PCR and Southern blotting, we evaluated the presence of this gene cluster in other isolates of Leptospira, including pathogenic, saprophytic, and intermediate strains. Polymerase chain reactions using primers designed from the genome strains amplified genes in the pathogenic strains L. interrogans serovar Copenhageni and Lai but not in the saprophyte L. biflexa (Figure 1B). Interestingly, one of the intermediate strains, L. licerasiae, gave a negative result, while the other, L. fainei, gave a faint positive. Control reactions using primers designed from 16S rRNA gene showed amplification in all the samples, verifying DNA integrity. A probe based on the neuA2 gene of L. interrogans was used for southern blotting of genomic DNA from a number of Leptospira reference strains and isolates. These experiments confirm and extend the PCR data. Of particular interest is a pair of wild rodent isolates of Leptospira in lanes 6 and 7 (MMD4847 identifies as L. licerasiae and MMD3731 identified as L. interrogans serovar Copenhageni). Whereas the intermediately pathogenic L. licerasiae strain did not give a positive result, the pathogenic serovar Copenhageni isolate gave a strong positive band. Also, the intermediate strain L. fainei gave a positive result in southern blotting, further confirming the faint positive result observed by PCR for this strain. Since low sequence identity between primers or probes and the target sequences from less closely related species could produce a negative result in these experiments, other more functional assays were utilized next.

DMB-derivatization and HPLC-MS analysis reveals multiple varieties of nonulosonic acids expressed by Leptospira

Strains were evaluated biochemically to determine whether nonulosonic acid biosynthetic pathways were functional in different species and strains of Leptospira. Bacteria were hydrolyzed with mild acetic acid to release nonulosonic acid species, and low molecular weight fractions were fluorescently derivatized with 1,2-diamino-4,5-methylene dioxybenzene (DMB), a molecule that specifically reacts with alpha keto acids, including NulOs. DMB-derivatized reaction products were separated by high performance liquid chromatography (HPLC) with a tandem electrospray ionization mass spectrometer. As expected by the Gram-negative-like structure of Leptospira, all samples displayed an early-eluting HPLC peak corresponding to the retention time and mass of 2-keto-3-deoxy-D-manno-octulosonic acid (Kdo). Kdo is an 8-carbon α-keto acid present in the core region of lipopolysaccharide in most Gram-negative bacteria. It serves as an internal positive control in these assays (Figure 2 peak b, m/z 355) and allowed comparison between different HPLC runs. Masses of some DMB-derivatized peaks did not readily correspond to masses of known varieties of nonulosonic acids (for example Figure 2 peak a, 407 and peak d, 440). It is not known whether these masses represent nonulosonic acids. In contrast, a consistent m/z of 433 (peak c) indicates the presence of di-N-acetylated nonulosonic acids and was found in pathogenic L. interrogans serovar Lai and L. alexanderi, and intermediate strain L. fainei. In all cases, the DMB-derivatized di-N-acetylated masses were accompanied with characteristic masses corresponding to the hydrated and hydrated sodium salt (m/z 451 and 473 respectively). These biochemical data show that pathogenic and intermediately pathogenic strains of Leptospira are capable of expressing di-N-acetylated nonulosonic acids. However, in contrast, the pathogenic strain L. santarosai was not found to synthesize identifiable nonulosonic acid species at detectable levels (Figure 2). We also performed analyses on L. biflexa serovar Patoc. In this case, we observed the presence of Kdo by HPLC and mass spectrometry, but identifiable NulO molecules were not present at detectable levels (not shown).

Interestingly, HPLC analysis of the two different genome strains of L. interrogans (serovar Copenhageni strain L1-130 and serovar Lai strain 56601) gave distinct results. While L. interrogans serovar Lai expresses di-N-acetylated nonulosonic acid (Figure 2, m/z 433), strain L1-130 (serovar Copenhagenii) exhibited a peak with mass and retention time consistent with Neu5Ac (m/z 408, hydrated 426, and hydrated sodium salt 448) (Figure 3A-B). Additional MS² analysis consistently reduced this trio of masses almost exclusively to the parent mass of 408 (Figure 3B), as expected based on the behavior of standard Neu5Ac derivatized in parallel (Figure 3C). Since the common animal sialic acids Neu5Ac and Neu5Gc were found in the standard culture media used for Leptospira (EMJH, Figure 4A), experiments were designed to exclude the possibility that L. interrogans strain L1-130 may incorporate exogenous sialic acid from the culture media. Unfortunately, the lack of a readily available genetic system for Leptospira rules out gene deletion as an approach to demonstrate endogenous synthesis. However, leptospires grown in defined serum-free media without sialic acids (as confirmed by HPLC) still produced a Neu5Ac peak, confirming that L. interrogans strain L1-130 synthesizes Neu5Ac and this sugar is not acquired from growth media (Figure 4B).

Composition and phylogenetic analysis of NulO biosynthetic gene clusters and enzymes

Next we performed analysis of the composition and phylogeny of the putative NulO biosynthetic gene clusters and the enzymes they encode in L. interrogans serovars Lai (strain 56601) and Copenhageni (strain L1-130). Consistent with the biochemical analysis of L. interrogans, genomic analysis of the NulO gene cluster reveals that the organism encodes a complete pathway for di-N-acetylated nonulosonic acid biosynthesis (see Table 1 in comparison with Figure 5). There are multiple distinct open reading frames encoding synthesis of aminotransferases, NulO synthases, and CMP-NulO synthetases (see Table 1 and Figure 5), suggesting that L. interrogans may express multiple nonulosonic acid species, a conclusion supported by our biochemical investigations (Figure 2 and Figure 3).

Campylobacter enzymes for legionaminic acid biosynthesis[14, 17–21]	C. jejuni Pathway number (Figure 5)	L. interrogans L1-130 & 56601 NCBI accession numbers	Predicted L. interrogans Pathway number (Figure 5)	Predicted enzymatic Function
PmtE (cj1329)	1	YP_002106	1	Glc-1-P guanyltransferase
		NP_711792
GlmU	2	YP_000413	2	(housekeeping)
		NP_714003		N-acetyltransferase
LegB (cj 1319)	3	YP_002111	3	4,6-dehydratase
		NP_711787
LegC (cj1320)	4	YP_002110	4	Aminotransferase in legionaminic acid synthesis (Figure 6A)
		NP_711788
		YP_002103	4, 13, or ?	Aminotransferase
		NP_711795
LegH (cj1298)	5	YP_002109	5	N-acetyltransferase
		NP_711789
LegG (cj1328)	6	YP_002107	6	2-epimerase/NDP sugar hydrolase in legionamimic acid synthesis
		NP_711791
LegI (cj1327)	7	YP_002108	7	Legionaminic acid synthase (Figure 6B)
		NP_711790
		YP_002104	10	Legionaminic or neuraminic acid synthase (Figures 6B & 7)
		NP_711794
LegF (cj1331)	8	YP_002102	8 or 11	CMP-Legionaminic acid or neuraminic acid synthetases (Figure 6C)
		NP_711796
		YP_002112	8 or 11
		NP_711786

Phylogenetic comparisons were performed to provide additional insights into the potential functions of Leptospira nonulosonic acid biosynthesis enzymes. We included in the phylogenetic analysis the well-characterized enzymes of Campylobacter jejuni that participate in parallel pathways of legionamimic, pseudaminic, and neuraminic acid synthesis [14, 17–21]. A schematic of these biosynthetic pathways is shown in Figure 5, noting the structural differences between neuraminic (sialic), legionamimic, and pseudaminic acids. These different NulOs are used by C. jejuni to modify a variety of different surface structures including the O-antigen of lipooligosaccharides, flagellin, and other surface proteins. To add further resolution to our phylogenetic analysis, we also included NulO biosynthetic enzymes from two Photobacterium profundum genome strains (3TCK and SS9), previously demonstrated to synthesize legionamimic and pseudaminic acids respectively [16]. In addition, homologous enzymes from other Leptospira genomes (L. noguchii str. 2006001870, L. biflexa serovar Patoc, L. santarosai str. 2000030832, L. borgpetersenii serovar Hardjo-bovis str. L550) were included in the phylogenetic analysis to better place the L. interrogans NulO enzymes into context with other putative leptospiral NulO biosynthetic enzymes.

The phylogenetic analysis of L. interrogans NulO biosynthetic enzymes demonstrates that a subset of these enzymes is more closely related to the C. jejuni legionaminic acid biosynthetic enzymes and more distantly related to the pseudaminic acid biosynthetic enzymes (Figure 6). Specifically, the aminotransferases YP_002110 and NP_711788 and the NulO synthetases YP_002108 and NP_711790 in L. interrogans serovars Copenhageni and Lai respectively, are more closely related to legionaminic acid synthesis enzymes and more distantly related to C. jejuni and P. profundum pseudaminic acid synthesis enzymes (Figure 6A-B, note green and pink shading indicates legionaminic acid pseudaminic acid pathways respectively). A similar relationship was found for the predicted epimerase/NDP-sugar hydrolases YP_002107 and NP_711791(not shown). Moreover, we find that both homologs of the putative CMP-NulO synthetases in L. interrogans (YP_002102 and YP_002112 in L1-130 and NP_711786 and NP_711796 in 56601) are more closely related to legionaminic acid and neuraminic acid synthetases than to CMP-pseudaminic acid synthetases (Figure 6C). Note in this figure that CMP-Kdo synthases were included to provide contrast and distinguish between enzymes that likely participate in CMP activation of eight carbon sugars (i.e. Kdo) and nine carbon sugars (i.e. NulOs). The sequences retrieved by BLAST of these Leptospira genomes, together with their phylogeny, suggest that a number of leptospires do not encode homologs of CMP-NulO synthetases. In contrast, some leptospires encode putative NulO biosynthesis enzymes that are more closely related to the C. jejuni and P. profundum pseudaminic acid biosynthesis enzymes and more distantly related to the legionaminic acid enzymes (e.g. L. noguchii Figure 6A-B).

In contrast to bacterial NulO biosynthetic pathways that synthesize Neu5Ac from ManNAc (N-acetyl mannosamine), the mammalian pathway relies on a NulO synthase with unique specificity for 6-phosphate-modified ManNAc, to generate 9-phosphate-modified Neu5Ac [22]. A set of adapter enzymes precede (kinase) and follow (phosphatase) the NulO synthase in the animal pathway (see Figure 7). In some cases, ‘adapter’ enzymes have become fused into the same open reading frame with one of the other nonulosonic acid biosynthesis genes. One example is the mammalian UDP-GlcNAc-2-epimerase, which is fused to a kinase that phosphorylates ManNAc to generate the substrate for the next step of the pathway, ManNAc-6-P. Interestingly, when performing analyses of L. interrogans NulO biosynthetic pathway, we noted that one of the NulO synthases encoded by L. interrogans (YP_002104 in serovar Copenhageni and NP_711794 in serovar Lai) has a unique C-terminal domain that is homologous to endonucleases that cleave phosphodiester bonds. By inference, we suggest that the route for N-acetylneuraminic acid biosynthesis in L. interrogans may be very similar to the animal pathway, condensing phosphoenolpyruvate with a phosphorylated 6-carbon intermediate to generate a phosphorylated 9-carbon sugar, followed by dephosphorylation catalyzed by the fused C-terminal phosphatase domain (Figure 7). This enzyme is distantly related to other NulO synthases and did not cluster with animal neuraminic acid synthases when these enzymes were included in the analysis (not shown), suggesting that this biosynthetic route may be ancestral. This conclusion is supported by previous evolutionary analyses of NulO pathways [16].

Nonulosonic acids are elaborated on Leptospira surface lipoproteins

Finally, efforts were made to identify the type of molecule(s) modified with sialic acids in L. interrogans strain L1-130. Immobilized sialic acid-binding lectins from Sambucus nigra agglutinin (SNA) and Maackia amurensis lectin (MAL), which recognize sialic acids in α2-6 and α2-3-linked sialic acids respectively, were used to affinity purify sialic acid-modified molecules in lysates of the L1-130 strain. Wheat germ agglutinin (WGA) also recognizes sialic acids, but is less specific, and also recognizes N-acetylglucosamine residues. As a control, buffers used in the solid phase assay were analyzed in parallel lanes of the gel, revealing that the faint bands present at ~60 kDa were part of the supplied buffers and not specific for sialylated L. interrogans molecules. Silver staining after SDS-Page gel electrophoresis of the eluted material from the affinity columns shows clear bands at ~21 kDa and ~25 kDa that are present at similar intensities in the MAL and SNA lanes (Figure 8A). Other bands appear to be enriched by affinity purification using one or the other lectin. For example, a faint band at ~43 kDa is apparent in the material isolated by MAL, but not by SNA. Alternatively, bands at ~15, ~37, and ~41 kDa are much stronger in the SNA-purified sample. These finding suggests that L. interrogans may modify surface structures with both α2-3- and α2-6-linked nonulosonic acids (Figure 8A). However, future studies should further investigate the molecule(s) modified by nonulosonic acids in leptospires, as well as their exact context and importance.

The affinity-purified material was subjected to DMB-derivatization and HPLC analysis, which showed the Neu5Ac peak, but not the Kdo peak (data not shown), strongly suggesting that this material was free of LPS-components. This does not rule-out that LPS may be modified with NulOs, just that LPS was not present in this affinity-purified preparation. We performed mass spectrometry to identify protein components in the affinity-purified material. Three proteins were identified by mass spectrometry (Figure 8B): Loa22, LipL32, and LipL41, all of which have been described in previous publications as surface-exposed lipid-linked outer membrane proteins of L. interrogans[23–27]. Indeed, Loa22 and LipL31 are among the most abundant proteins expressed on the Leptospira cell surface [28]. Loa22 was identified with the highest number of peptide matches. Loa22 is an outer membrane protein encoded within all Leptospira genomes sequenced to date. It has been observed to be upregulated in vivo[27] and it is one of very few leptospiral proteins so far that has been shown to be necessary for virulence [3]. Additional studies are needed to define the precise context of NulO expression on L. interrogans and understand its potential significance in virulence.

Strains and culture conditions

Intermediately pathogenic strains L. licerasiae serovar Varillal strains MMD3731, MMD4847 and CEH008 (isolated from rodents in Peru), L. fainei serovar Hurstbridge strain BUT 6^T and the saprophyte L. biflexa serovar Patoc were used for these experiments. Pathogenic Leptospira used in this study included L. interrogans serovar Copenhageni strain L1-130, L. interrogans serovar Lai strain 55601, and L. interrogans serovar Icterohaemorrhagiae wild rodent isolate MMD 3731 that were passaged fewer than 5 times in vitro after re-isolation from hamster liver to maintain virulence. L. santarosai and L. alexanderi serovar Manhao were originally isolated from clinical cases of leptospirosis and now serve as reference laboratory strains. Generally, Leptospira were cultivated at 30°C in Ellinghausen-McCullough-Johnson-Harris (EMJH) medium (catalog #279510, Becton Dickinson, Sparks, Maryland). Chemically-defined, sialic acid-free medium, prepared as previously described and verified by HPLC to be sialic acid free, was used to cultivate Leptospira in experiments where the lack of exogenous sialic acids was a necessary condition [38].

PCR of sialic acid cluster genes

Primers based on the genome of L. interrogans L1-130 were designed for the detection of genes in the sialic acid cluster as follows: sasfrontF (5′- TCC GGA AAT GCG AAT GAT G-3′), sasfrontR (5′- CAC CGG GCA AAA GAC TAA CCT - 3′), sasendF (5′- CGG ATA TAG CGG ACG ATG TAA - 3′), sasendR (5′- CGC CAA AAA GCC AAG GAA - 3′), neuA2F (5′- TGA AGC GGC AAA AAG AGC - 3′), neuA2R (5′- TGA AAT AAC ATG CCG ACA AAT A - 3′), neuCfrontF (5′- CGC TAC GGG AAT GCA TCT GTC TC - 3′), neuCfrontR (5′- CCC ATT CCC CCA ACC AAA AA - 3′), neuCendF (5′- GGC GAG GAT CCT TCT AAT GTT TTT - 3′) and neuCendR (5′- ACT CGC TCC GCC TTC ACC A - 3′). PCR reactions were prepared using 0.2 mM of each primer in a 20 μL reaction with DNA from the pathogens L. interrogans Lai, L. interrogans L1-130, the intermediates L. licerasiae and L. fainei and the saprophyte L. biflexa serovar Patoc. NeuA2 and neuBfront reactions used an annealing temperature of 52°C. NeuCfront, neuCend, sasfront and sasend were run using an annealing temperature of 58°C. A 16 S gene PCR reaction using previously published primers fLIP and rLIP1 was used as a control for integrity of DNA.

NeuA2 southern blot

Genomic DNA samples of Salmonella enterica, L. interrogans serovar Lai str. 56601, L. interrogans serovar Copenhageni str. L1-130, L. biflexa serovar Patoc, L. licerasiae strains CEH008 and MMD4847, L. interrogans serovar Icterohaemorrhagiae str.MMD3731 and L. fainei serovar Hurstbridge were prepared into plugs using 1 % agarose and 0.5x TBE. These were subjected to depurination and denaturing conditions. DNA was then transferred to a positively charged membrane via overnight capillary transfer with 20x SSC. Finally the DNA was cross-linked to the membrane using short wave DNA for 15 min. 10 mL of pre-hybridization solution (QuikHyb, Stratagene) were warmed to 40°C prior to hybridization. Hybridization was done overnight at 40-42°C using the same solution and adding 10 mL of DIG-labeled PCR product of primer neuA2F (5′ - TGA AGC GGC AAA AAG AGC - 3′) and neuA2R (5 ′- TGA AAT AAC ATG CCG ACA AAT A - 3′). 2xSSC at room temperature and 1x SSC at 68°C were used for stringency washes. A chemiluminescent substrate and an alkaline phosphatase conjugated anti-DIG antibody were used to demonstrate binding of the probe.

Mild acid hydrolysis and DMB-derivatization of nonulosonic acids

Mild (2 N) acetic acid hydrolysis was performed to release surface nonulosonic acids from Leptospira. 4 N acetic acid was added to an equal volume of extensively washed and resuspended pellets followed by 3 h of incubation at 80°C. The resulting soluble fraction was filtered to remove large molecular weight components and derivatized with DMB (1,2-diamino-4,5-methylene dioxybenzene), a reagent that reacts with the α-keto acid portion of nonulosonic acids. Final reaction conditions were 7 mM DMB, 18 mM sodium hydrosulfite, 1.4 M acetic acid, and 0.7 M 2-mercaptoethanol. Derivatization was carried out for 2 hours at 50°C in the dark.

High performance liquid chromatography and mass spectrometry

DMB-NulO derivatives were resolved by HPLC using a reverse phase C18 column (Varian) eluted isocratically at a rate of 0.9 ml/min over 50 minutes using 85 % MQ-water, 7 % methanol, 8 % acetonitrile as previously described [16, 39, 40]. In some experiments, HPLC was performed without online mass spectrometry and detection of fluorescently labeled NulO sugars was achieved using an online fluorescence detection using excitation and emission wavelengths of 373 nm and 448 nm respectively. In other experiments HPLC was combined with online mass spectrometry using a Thermo-Finnigan model LCQ ion trap mass spectrometer system. When mass spectrometry was performed, the mobile phase also included 0.1 % formic acid, and online UV detection of DMB-NulO molecules preceded mass spectrometric analysis. We note that similar HPLC-MS analyses have been described previously DMB-derivatized α-keto acids [39–41].

Phylogenetic analysis

We performed BLAST searches (blastp) against the NCBI genome database using as seeds the sequences of 1) proteins encoded by Campylobacter jejuni pseudaminic, legionaminic, and neuraminic acid biosynthetic pathways or 2) enzymes encoded in the Leptospira interrogans NulO biosynthetic gene cluster (Figure 1A). NCBI accession numbers are provided in Table 1 and a schematic of the biosynthetic pathways is illustrated in Figure 5. Complete protein sequences of homologous amino acids were aligned using ClustalW in MacVector 11.1.1 software and alignments were checked manually. The Neighbor Joining (NJ) method was utilized for phylogenetic tree construction using MacVector 11.1.1 software, including 1000 Bootstrap replications to obtain confidence values for branches of the NJ trees.

Solid-phase lectin binding

Whole cell lysates were prepared using three cycles of freeze-thawing of PBS washed L. interrogans serovar Copenhageni strain L1-130. In order to probe the abundance and nature of the sialylated molecules on L. interrogans, these lysates were fractionated using a lectin-based solid phase assay (Q Proteome Sialic Acid kit, Qiagen) using three immobilized sialic acid binding lectins: wheat germ agglutinin (WGA), Sambucus nigra agglutinin (SNA), and Maackia amurensis lectin (MAL), according to manufacturer’s instructions. Molecules captured by each of these lectins were eluted according to the manufacturers instructions. then analyzed by SDS-PAGE followed by silver staining (SilverQuest Silver Staining Kit, Invitrogen).

Mass spectrometry

To determine whether L. interrogans uses nonulosonic acids for post-translational modification of proteins, pooled affinity-purified material from above mentioned experiment was subjected to denaturation, reduction, and alkylation, followed by trypsin digestion and MS/MS analysis using a nano-flow LC- tandem mass spectrometer. Peptide mass fingerprints identified in the affinity-purified material were used to identify L. interrogans proteins by searching against the NCBInr bacterial genome database.