WadD, a New Brucella Lipopolysaccharide Core Glycosyltransferase Identified by Genomic Search and Phenotypic Characterization

Brucellosis, an infectious disease caused by Brucella, is one of the most extended bacterial zoonosis in the world and an important cause of economic losses and human suffering. The lipopolysaccharide (LPS) of Brucella plays a major role in virulence as it impairs normal recognition by the innate immune system and delays the immune response. The LPS core is a branched structure involved in resistance to complement and polycationic peptides, and mutants in glycosyltransferases required for the synthesis of the lateral branch not linked to the O-polysaccharide (O-PS) are attenuated and have been proposed as vaccine candidates. For this reason, the complete understanding of the genes involved in the synthesis of this LPS section is of particular interest. The chemical structure of the Brucella LPS core suggests that, in addition to the already identified WadB and WadC glycosyltransferases, others could be implicated in the synthesis of this lateral branch. To clarify this point, we identified and constructed mutants in 11 ORFs encoding putative glycosyltransferases in B. abortus. Four of these ORFs, regulated by the virulence regulator MucR (involved in LPS synthesis) or the BvrR/BvrS system (implicated in the synthesis of surface components), were not required for the synthesis of a complete LPS neither for virulence or interaction with polycationic peptides and/or complement. Among the other seven ORFs, six seemed not to be required for the synthesis of the core LPS since the corresponding mutants kept the O-PS and reacted as the wild type with polyclonal sera. Interestingly, mutant in ORF BAB1_0953 (renamed wadD) lost reactivity against antibodies that recognize the core section while kept the O-PS. This suggests that WadD is a new glycosyltransferase adding one or more sugars to the core lateral branch. WadD mutants were more sensitive than the parental strain to components of the innate immune system and played a role in chronic stages of infection. These results corroborate and extend previous work indicating that the Brucella LPS core is a branched structure that constitutes a steric impairment preventing the elements of the innate immune system to fight against Brucella.

Brucellosis, an infectious disease caused by Brucella, is one of the most extended bacterial zoonosis in the world and an important cause of economic losses and human suffering. The lipopolysaccharide (LPS) of Brucella plays a major role in virulence as it impairs normal recognition by the innate immune system and delays the immune response. The LPS core is a branched structure involved in resistance to complement and polycationic peptides, and mutants in glycosyltransferases required for the synthesis of the lateral branch not linked to the O-polysaccharide (O-PS) are attenuated and have been proposed as vaccine candidates. For this reason, the complete understanding of the genes involved in the synthesis of this LPS section is of particular interest. The chemical structure of the Brucella LPS core suggests that, in addition to the already identified WadB and WadC glycosyltransferases, others could be implicated in the synthesis of this lateral branch. To clarify this point, we identified and constructed mutants in 11 ORFs encoding putative glycosyltransferases in B. abortus. Four of these ORFs, regulated by the virulence regulator MucR (involved in LPS synthesis) or the BvrR/BvrS system (implicated in the synthesis of surface components), were not required for the synthesis of a complete LPS neither for virulence or interaction with polycationic peptides and/or complement. Among the other seven ORFs, six seemed not to be required for the synthesis of the core LPS since the corresponding mutants kept the O-PS and reacted as the wild type with polyclonal sera. Interestingly, mutant in ORF BAB1_0953 (renamed wadD) lost reactivity against antibodies that recognize the core section while kept the O-PS. This suggests that WadD is a new glycosyltransferase adding one or more sugars to the core lateral branch. WadD mutants were more sensitive than the parental strain to components

INTRODUCTION
Members of the genus Brucella are the etiologic agents of brucellosis, a worldwide spread zoonosis that affects ruminants, camelids, swine, dogs, and several forms of marine and terrestrial wildlife and causes abortions, infertility, and the subsequent economic losses in livestock. Humans become infected via direct contact with affected animals and through consumption of unpasteurized dairy products, and develop a chronic and debilitating condition that requires prolonged antibiotic treatment, being lethal in 1-5% of untreated cases (Ariza, 1999). Because of its impact on animal production and Public Health, it is estimated that brucellosis imposes a heavy burden in the developing world (McDermott et al., 2013).
The genus includes several nominal species that show host preferences 1 . Those that have been known for a long time (often referred to as "classical" Brucella species) include B. abortus and B. melitensis (the brucellae that infect domestic ruminants), B. suis (infecting swine, reindeer, hares, and several species of wild rodents), B. canis (infecting dogs), B. ovis (not zoonotic and restricted to sheep), and B. neotomae (infecting the desert woodrat). Because of their early identification and their economic and public health importance, B. abortus, B. melitensis, and B. suis are the best-characterized members of the genus, and all of them produce smooth (S) glossy colonies, a morphology that reflects the existence of a lipopolysaccharide (LPS) carrying an O-polysaccharide (O-PS) linked to the corelipid A section that anchors the molecule to the outer membrane (OM). These Brucella spp. behave as facultative intracellular parasites of professional and non-professional phagocytes, an ability that depends on a number of virulence factors, chiefly a type IV secretion system and a peculiar OM structure. Critical OM components such as the S-LPS, lipoproteins, and ornithine lipids differ in relevant molecular details from the homologous molecules that in other bacteria bear the pathogen-associated molecular patterns (PAMP) readily detected by innate immunity pattern recognition receptors (PRRs). Consequently, these brucellae induce comparatively low and delayed proinflammatory responses, which create a time window allowing the pathogen to traffic intracellularly in dendritic cells and macrophages to reach a safe niche before effective phagocyte activation takes place (Lapaque et al., 2005;Barquero-Calvo et al., 2007;Palacios-Chaves et al., 2011). In this regard, the Brucella S-LPS carries the most significant PAMP modifications and is thus a major virulence factor (Lapaque et al., 2005). 1 http://www.bacterio.net /-allnamesac.html Whereas the structure of the O-PS (a N-formylperosamine homopolymer) and its role in virulence in animal models and in the natural host have been known for a long time, the importance of the core and its structure have only recently been established.
Accordingly, the reactivity with R-LPS-specific MoAbs strongly suggests that the structure elucidated for B. melitensis is conserved in the classical species (Bowden et al., 1995;Zygmunt et al., 2012). Moreover, availability of the corresponding structure of several mutants has also allowed assigning genes that upon mutation generate LPSs that lack [i.e., rough (R) LPS] or carry O-PS (Figure 1). Gene wadA corresponds to the enzyme linking KdoII and glucose, wadC to the mannosyltransferase acting on KdoI, and wadB to a glucosaminyltransferase involved in the assembly of the GlcN branch (González et al., 2008;Conde-Álvarez et al., 2012;Gil-Ramírez et al., 2014). These genes are highly conserved in the classical Brucella species (Monreal et al., 2003;Iriarte et al., 2004;González et al., 2008;Conde-Álvarez et al., 2012;Gil-Ramírez et al., 2014;Soler-Lloréns et al., 2016) and as expected, all Brucella genomes also carry a waaA homolog, the essential gene coding for the Kdo transferase of Gram-negative bacteria (Raetz and Whitfield, 2002;Iriarte et al., 2004). However, since most but not all glycosyltransferases involved in LPS synthesis are monofunctional (Raetz and Whitfield, 2002), it remains to be determined whether glucosaminyltransferases other than WadB are required for the synthesis of the GlcN branch.
Based on the complete structure of the core and the phenotype of mutants in wadB and wadC, it is postulated that the lack of acidic groups other than the two Kdo and lipid A phosphates and the mannose-GlcN branch account for the role of Brucella core in virulence both in cellular and animal models. By virtue FIGURE 1 | Schematic representation of Brucella melitensis LPS core. The site of action of the glycosyltransferases identified thus far is indicated with arrows. Kdo (3-deoxy-D-manno-octulosonic acid), Glc (glucose), Man (mannose), GlcN (glucosamine), and Quin (quinovosamine) [adapted from Fontana et al. (2016) and Gil-Ramírez et al. (2014)].
of the density of amino groups and close position to the inner core and lipid A, the GlcN tetrasaccharide both neutralizes and sterically protects those inner anionic groups, thereby hampering binding of bactericidal peptides and PRRs such as the activators of the antibody-independent classical complement pathway and MD2, the TLR4 co-receptor. Accordingly, core defects bolster proinflammatory responses causing an activation of innate immunity earlier than that of the wild type, thereby generating attenuation (Conde-Álvarez et al., 2012;Gil-Ramírez et al., 2014;Soler-Lloréns et al., 2014;Fontana et al., 2016). Also, although both wadB and wadC mutants maintain an intact O-PS, attenuation in mice is more severe for the latter (Conde-Álvarez et al., 2012;Gil-Ramírez et al., 2014;Fontana et al., 2016) strongly suggesting a correlation between the extent of core damage and the intensity of the immunoactivation that brings about attenuation. A complete elucidation of the genetics of Brucella LPS core could confirm such a correlation and, since LPS core mutants represent a tool for developing a new generation of brucellosis vaccines (Conde-Álvarez et al., 2013;Zhao et al., 2017), also provide a graded array of possibilities.
With these possibilities in mind, we investigated B. abortus 2308 genes annotated as glycosyltransferases for their possible involvement in LPS core synthesis and relevant biological effects.

Bacterial Strains and Growth Conditions
The bacterial strains and plasmids used in this study are listed in Supplementary Table S1. All bacteria were grown either on tryptic soy agar (TSA, Pronadisa) plates or in tryptic soy broth (TSB, Scharlau) or Mueller-Hinton broth (Becton Dickinson, Difco) at 37 • C. Where indicated, growth media were supplemented with kanamycin (Km) at 50 mg/ml, nalidixic acid (Nal) at 25 mg/ml, ampicillin (Amp) at 100 mg/ml, and/or 5% sucrose. Bacterial growth rates were determined at 37 • C in Mueller-Hinton broth (Becton Dickinson, Difco), using a Bioscreen C apparatus (Lab Systems). All strains were stored in skim milk at −80 • C. Work with Brucella was performed at the Biosafety Level 3 (BSL-3) laboratory facilities of the "Centro de Investigación Médica Aplicada de la Universidad de Navarra" (CIMA) and "Centro de Investigación y Tecnología Agroalimentaria de Aragón" (CITA), Spain.

DNA Manipulations and Analyses
Sequence data were obtained from Kyoto Encyclopedia of Genes and Genomes (KEGG 2 ). Searches for DNA and protein homologies between Brucella species and other α-proteobacteria such as Ochrobactrum, Rhizobium, or Agrobacterium were carried out using KEGG, Basic Local Alignment Sequence Tool (BLAST 3 ), and Clustal Omega 4 from the European Molecular Biology Laboratory -European Bioinformatics Institute (EMBL-EBI 5 ). New glycosyltransferase identification, using B. abortus 2308 was supported by Carbohydrate-Active enZymes database (CAZy 6 ). Primers were designed using Primer 3 input 7 and synthesized by Sigma-Aldrich. Plasmid DNA was extracted with Qiaprep spin Miniprep (Qiagen GmbH). When needed, DNA was purified from agarose gels using Qiack Gel extraction kit (Qiagen) and sequenced by the Servicio de Secuenciación of CIMA.
For the construction of Ba BAB2_0133 mutant, we first generated two PCR fragments: oligonucleotides BAB2_0133-F1 (5 -GCGTTGGACAAGTTGAGGTT-3 ) and BAB2_0133-R2 (5 -CATAGCGGTCGGTTAAATGC-3 ) were used to amplify a 572 base pairs (bp) fragment including codons 1-38 of BAB2_0133, as well as 458 bp upstream of the BAB2_0133 start codon. Oligonucleotides BAB2_0133-F3 (5 -GTATCGCCAGCCAATTTACGTCCGTATTGGAAGCCAAGA A-3 ) and BAB2_0133-R4 (5 -CAGTAACAAAAGGCCG CTAT-3 ) were used to amplify a 442 bp fragment including codons 299-326 of BAB2_0133 and 355 bp downstream of the BAB2_0133 stop codon. Both fragments were ligated by overlapping PCR using oligonucleotides F1 and R4 for amplification, and the complementary regions between R2 and F3 for overlapping. The resulting fragment, containing the BAB2_0133 deletion allele, was cloned into pCR2.1 (Invitrogen), to generate plasmid pMSB-01, sequenced to ensure the maintenance of the reading frame, subsequently subcloned into the BamHI and the XbaI sites of the suicide plasmid pJQK (Scupham and Triplett, 1997) and transformed into competent E. coli S17 λpir (Simon et al., 1983). The resulting suicide pJQK-derived plasmid was introduced into B. abortus 2308 by conjugation. The first recombination event (integration of the suicide vector in the chromosome) was selected by Nal and Km resistance, and the second recombination (excision of the mutator plasmid leading to construction of the mutant by allelic exchange) was selected by Nal and sucrose resistance and Km sensitivity. The resulting colonies were screened by PCR with primers F1 and R4 which amplified a fragment of 1014 bp in the mutant and 1794 bp in the sibling strain that keeps the wild-type gene. Primers BAB2_0133-F1 and BAB2_0133-R5 (5 -AAGACCCAGTAGTTAGCACT-3 ) amplified a fragment of 919 bp only in the wild-type strain. The mutation generated results in the loss of the 80% of the ORF. Ba BAB2_0135 mutant was constructed following the same procedure and using oligonucleotides BAB2_0135-F1 (5 -TGGCGGCCGCTCTAGAACACCGGACTGCCTGATAA-3 ) and BAB2_0135-R2 (5 -CGGGCAATTTCGGCATAG-3 ) that amplified a 240 bp fragment including codons 1-40 of BAB2_0135, as well as 120 bp upstream of the BAB2_0135 start codon, and oligonucleotides BAB2_0135-F3 (5 -CT ATGCCGAAATTGCCCGCCGGTTTGGAAATGCGGTCAA-3 ) and BAB2_0135-R4 (5 -ATCCACTAGTTCTAGTT ATGTAGCCGCCACCGTTT-3 ) that amplified a 232 bp fragment including codons 441-478 of BAB2_0135 and 115 bp downstream of the BAB2_0135 stop codon. The resulting colonies were screened by PCR with primers F1 and R4 that amplified a fragment of 472 bp in the mutant and 1672 bp in the sibling strain that keeps the wild-type gene. Primers BAB2_0135-F1 and BAB2_0135-R5 (5 -CGATTGCCAGTCCCAGAAAG-3 ) amplified a fragment of 628 bp only in the wild-type strain. The mutation generated results in the loss of the 84% of the ORF.
The rest of the ORFs were mutagenized by recombination and gene disruption using as suicide vectors pJQK or pSKoriT (Tibor et al., 2002) carrying an internal fragment of the ORF.
We also constructed a non-polar wadD mutant (Ba wadD) by in frame deletion. Oligonucleotides wadD-F1 (5 -TCTATAATGAGAGGCGGCTTTT-3 ) and wadD-R2 (5 -A GAAGTGCTGGTCCTGTTGT-3 ) were used to amplify a 304 (bp) fragment including codons 1-50 of BAB1_0953, as well as 154 bp upstream of the BAB1_0953 start codon. Oligonucleotides wadD-F3 (5 -ACAACAGGACCAGCACTT CTATCCTCACCCTGCCATTCAA-3 ) and wadD-R4 (5 -CTGGTACTAGACGCCCTGTT-3 ) were used to amplify a 175 bp fragment including codons 281-324 of BAB1_0953 and 43 bp downstream of the BAB1_0953 stop codon. Both fragments were ligated by overlapping PCR using oligonucleotides F1 and R4 for amplification, and the complementary regions between R2 and F3 for overlapping. The resulting fragment, containing the BAB1_0953 deletion allele, was cloned directly into pJQK by the InFusion cloning system (Clontech) to generate pMSB-34. This suicide vector was sequenced to ensure the maintenance of the reading frame and transferred into B. abortus 2308 by conjugation. The resulting colonies were screened by PCR with primers F1 and R4 that amplified a fragment of 479 bp in the mutant and 1169 bp in the sibling strain which keeps the wild-type gene. Primers wadD-F1 and wadD-R5 (5 -AGGCACGGTTTCATAGACGA-3 ) amplified a fragment of 844 bp only in the wild-type strain. The mutation generated results in the loss of the 71% of the ORF and the mutant was called Ba wadD.

Crystal Violet Exclusion Test
To study if the mutants had smooth (complete LPS) or rough (O-PS-lacking LPS) phenotype, 5 ml of a crystal violet solution at 0.1 mg/ml in distilled water were used to cover isolated colonies on TSA plates for 20 s. Smooth colonies excluded crystal violet and looked white, whereas rough colonies captured crystal violet and looked violet.

LPS Extraction and Characterization
LPS was extracted by the proteinase-K sodium dodecyl sulfate (SDS) protocol (Dubray and Limet, 1987;Garin-Bastuji et al., 1990) with some modifications. Bacteria grown overnight in 10 ml of TSB were killed with 0.5% phenol during 3 days in agitation at 37 • C. After that, samples were weighed and pipetted into small polycarbonate cap tubes and then suspended by ultrasounds in 2% SDS-60 mM Tris-HCl buffer (pH 6.8) at a concentration of 0.5 g (wet weight) of bacteria per 10 ml of buffer. Samples were then heated at 100 • C for 10 min, and lysates were cooled to 55 • C. This treatment was followed by digestion with 60 µl of proteinase-K at 2.5 mg/ml in HCl-Tris per ml of sample (Merck KGaA) for 3 h at 55 • C, and overnight incubation at 20 • C. Afterward, they were centrifuged at 20,000 × g for 30 min at room temperature, and the LPS was precipitated from the supernatant by addition of 3 volumes of methanol containing 1% sodium acetate-saturated methanol at −20 • C. After 60 min, the precipitate was harvested by centrifugation at 5,000 × g for 15 min at 4 • C and resuspended by sonication in 10 ml of distilled water. After a second methanol precipitation and centrifugation, the pellets were resuspended by sonication in 2-3 ml of 60 mM HCl-Tris (pH 6.8) and left at 37 • C. Then samples were treated with 20 µl/ml of RNase and DNase stock solutions at 0.5 mg/ml in HCl-Tris (MP Biomedicals and Sigma-Aldrich, respectively) at 37 • C for 30 min. Subsequently, the LPS was treated again with 5 µl/ml of proteinase K at 2.5 mg/ml in HCl-Tris, at 55 • C for 3 h and then, at room temperature overnight. After a third methanol precipitation in the same conditions described above, the pellet containing LPS was recovered in 1 ml of distilled water and frozen at −20 • C.

Sensitivity to Polycationic Bactericidal Peptides
The minimal inhibitory concentration (MIC) of polymyxin B and poly-L-ornithine (both from Sigma-Aldrich) were determined in Mueller-Hinton medium. Exponentially growing bacteria were adjusted to an OD equivalent to 1 of the McFarland scale, and exposed to serial dilutions of the bactericidal peptides. MICs were determined by technical duplicates after 2 days of incubation at 37 • C. Experiments were performed in triplicate.

Sensitivity to the Bactericidal Action of Non-immune Serum
Exponentially growing bacteria were adjusted to 10 4 Colony Forming Units (CFU)/ml in saline and dispensed in duplicate in microtiter plates (30 µl/well) containing 60 µl of newborn bovine serum. After 90 min of incubation at 37 • C with gentle agitation, complement action was blocked by adding brain heart infusion (BHI) broth (150 µl/well). After mixing the BHI broth with the bacterial suspension, 75 µl were plated by triplicate on TSA plates. Five days after incubation at 37 • C, results were expressed as the percentage of CFU recovered with respect to control samples where new-born bovine serum was substituted by PBS. The experiment was repeated three different times.

Virulence in Mice
Seven-week-old female BALB/c mice (ENVIGO, Harlan) were lodged in cages in BSL-3 facilities with water and food ad libitum for 2, 8, or 12 weeks. Six groups of five mice each were inoculated with Ba wadD or Ba-parental. Inocula were prepared in sterile PBS and each mouse was administered intraperitoneally approximately with 5 × 10 4 CFU in 0.1 ml. To assess the exact dose retrospectively, dilutions of each inoculum were plated by triplicate on TSA plates. Spleen CFUs in infected mice were counted at 2, 8, and 12 weeks after inoculation. The CFU counts were normalized by logarithmic transformation and the mean log CFU/spleen values and the standard deviations were calculated. The spleens were weighed and homogenized in 9 volumes of PBS and serial 10-fold dilutions were accomplished and plated by triplicate on TSA plates. After 5 days of incubation at 37 • C the colonies were checked by crystal violet exclusion test and PCR.

Statistical Analysis
Statistical significance for sensitivity to normal serum was evaluated with one-way ANOVA followed by Dunnett's multiple comparisons test ( * * * * p < 0.0001). For virulence analysis, statistical significance between the parental strain and the wadD mutant was evaluated using Student's t-independent-samples test ( * p < 0.05).

Screening for Putative LPS Core Glycosyltransferases
A bioinformatic search in the Carbohydrate-Active Enzymes database CAZy 8 revealed 23 ORFs in the genome of B. abortus 2308 (Supplementary Table S1) that could code for glycosyltransferases. We excluded from further analysis BAB1_0108-cgs, which is involved in cyclic glucan synthesis (Briones et al., 2001), BAB1_1786-mtgA and BAB1_1450-murG, both related to peptidoglycan synthesis, BAB1_1171-lpxB, probably implicated in lipid A formation (Iriarte et al., 2004), and BAB1_0553-wbkA, BAB1_0563-wbkE; BAB1_1000-wboA and BAB1_1000-wboB, four genes that belong to the O-PS synthesis route (McQuiston et al., 1999;Godfroid et al., 2000;González et al., 2008). Similarly, four ORFs correspond to those glycosyltransferases already known to be involved in the synthesis of the LPS core: BAB1_0639-wadA (Monreal et al., 2003) Table 1, and data on their presence in other Brucella spp. and genetic location are in the Supplementary Material (Supplementary Table S2 and Supplementary Figure S1). Of these, seven (BAB1_0953, BAB2_0105, BAB2_0133, BAB2_0135, BAB1_1620, BAB1_0607, and BAB1_0932) were highly conserved in all Brucella spp., but four (BAB2_0693, BAB1_0417, BAB1_0114, and BAB1_0326) presented significant differences when compared to B. abortus sequences, mainly due to frameshifts generating shorter proteins (Supplementary Table S2). Perusal of the literature revealed some information on 4 of those 11 putative glycosyltransferases. Expression of BAB1_0326, BAB2_0133, and BAB2_0135 has been shown to be controlled by MucR, a general virulence regulator in Brucella (Caswell et al., 2013). However, although it has been reported that B. abortus and B. melitensis mucR mutants have a defective LPS core, the glycosyltransferases involved have not been identified (Caswell et al., 2013;Mirabella et al., 2013). Also, BAB1_1620 expression has been reported to be controlled by BvrR/BvrS, a master regulator of Brucella virulence that modulates OM homeostasis and undetermined aspects of LPS structure (Manterola et al., 2005;Viadas et al., 2010).
We first analyzed the MucR and BvrR/BvrS controlled ORFs for involvement in LPS synthesis. To this end, we constructed an insertion mutant in BAB1_0326 (since the downstream ORF is oriented in the opposite direction) as well as nonpolar deletion mutants in BAB2_0133 and BAB2_0135 (both part of an operon), and BAB1_1620 (which, although isolated, is surrounded by genes implicated in the cell cycle). These four mutants maintained the S phenotype in the crystal violet assay, suggesting that they kept an intact O-PS. Then, SDSproteinase-K LPSs were analyzed by SDS-PAGE and Western blot with both a polyclonal serum against S brucellae and anticore MoAbs A68/24G12/A08 and A68/24D08/G09, using as controls LPS from B. abortus 2308W (Ba-parental), a B. abortus wadC mutant (Ba wadC), and a R per mutant (Ba per) (Martínez-Gómez et al., 2018). The LPS of the four mutants presented S and R fractions with migration profiles identical to those of Ba-parental LPS and reacted similarly with the serum and MoAbs (Supplementary Figure S2) strongly suggesting that the corresponding ORFs are not required for normal LPS synthesis. When we complemented these observations by inoculating BALB/c mice with BAB2_0133, BAB2_0135, and BAB1_1620 mutants, they produced CFU/spleen that did not differ from those of Ba-parental at weeks 2 (p = 0.99; 0.75 and 0.45, respectively) and 8 (p = 0.95; 0.99 and 0.99) after infection. Moreover, mutants in BAB2_0133 and BAB1_1620 behaved similarly to Ba-parental in polycationic peptide resistance, and the former also performed as Ba-parental in sensitivity to normal serum (Supplementary Figure S3). These results are consistent with the idea that the putative glycosyltransferases regulated by MucR or BvrR/BvrS are not involved in the synthesis of LPS or of other components implicated in virulence, at least under the conditions used in this study.
To investigate whether the remaining seven putative glycosyltransferases (BAB1_0953, BAB2_0105, BAB2_0693, BAB1_0607, BAB1_0114, BAB1_0932, and BAB1_0417) were required for LPS synthesis, we constructed B. abortus 2308W insertion mutants in each of them. All mutants were S by the crystal violet assay and the analysis of the extracted LPS showed S fractions with a migration profile like that of Ba-parental and reacted similarly with the anti S-Brucella polyclonal serum (Figure 2 and Supplementary Figure S4). Interestingly, although keeping the S fraction, mutant in BAB1_0953 lost reactivity in the R fraction, suggesting a  defect in the core and/or lipid A epitope(s) recognized by polyclonal sera of infected animals (Rojas et al., 2001). Since this was not observed for the other mutants, we investigated further BAB1_0953 and the phenotype associated with its mutation.

BAB1_0953 Encodes WadD, a Previously Unidentified Glycosyltransferase Involved in the Synthesis of the LPS Core Lateral Branch
BAB1_0953 is an isolated gene and the adjacent ORFs are encoded in the complementary strand. Thus, it was very unlikely that a polar effect caused the LPS phenotype of the insertion mutant. However, to rule out such a possibility, we constructed a non-polar deletion mutant, hereafter named Ba wadD following the nomenclature previously established for Brucella LPS core genes (Reeves et al., 1996;Gil-Ramírez et al., 2014). Ba wadD LPS showed a migration profile similar to that of Ba-parental in the high molecular weight S-LPS fraction and an increased mobility in the R-LPS one. Western-blot analysis with a polyclonal serum showed that, while the former fraction kept the reactivity with this serum, the latter failed to react indicating a significant alteration of the core-lipid A epitopes. To assign the defect to the core oligosaccharide, we probed the LPS with MoAbs A68/24G12/A08 and A68/24D08/G09, the binding of which to the R-LPS requires an intact mannose-GlcN tetrasaccharide (Conde-Álvarez et al., 2012;Fontana et al., 2016). Both antibodies failed to react with the R-LPS fraction and this failure was reverted upon insertion of a complete wadD gene in the bacterial chromosome of the deletion mutant (Ba wadD::Tn7-PwadD). Moreover, this complementation restored both the migration pattern of the R-LPS fraction to the level of the Ba-parental LPS and the reactivity with the polyclonal serum (Figure 3). An ELISA with several anti-core MoAbs and whole bacteria confirmed the core defect (Figure 4, upper panel).
We only observed small but constant differences in reactivity of Ba-parental and Ba wadD with anti-Outer Membrane Proteins (OMP) antibodies (Figure 4, lower panel). Although considering the limitations of the method used, this suggests that the presence of the O-PS and the defect in the core LPS could generate a steric hindrance that would allow the access of antibodies to the OMPs (Bowden et al., 1995) or the possibility that wadD mutation could affect the amount of LPS in the OM or how it is inserted in the bacterial surface.
Although as signaled above, the final confirmation would require a complete chemical analysis, all these results FIGURE 3 | Mutation of wadD generates an LPS core defect. Left panel, SDS-PAGE electrophoresis and silver staining of SDS-proteinase K extracts; central panel, Western blot analysis of SDS-proteinase K extracts with a polyclonal serum of a B. melitensis-infected rabbit; right panel, Western blot analysis of SDS-proteinase K extracts with monoclonal anti-core antibody A68/24G12/A08. strongly suggest that wadD encodes a previously unidentified glycosyltransferase involved in the synthesis of the core lateral branch.

WadD Orthologs Are Present in all Brucella spp. but in a Recently Characterized Isolate From Amphibians
In silico analysis (Supplementary Figure S5) showed that wadD was highly conserved in the core brucellae including the "classical" spp. B. melitensis, B. suis (smooth LPS), and B. ovis and B. canis (rough LPS) and also in other "nonclassical" smooth Brucella spp.: B. pinnipedialis, B. microti, B. ceti, and B. vulpis. We also analyzed the presence of wadD in the group of early-diverging brucellae that depart from the classical spp. and includes B. inopinata strain BO1, B. inopinata-like strain BO2, an isolated strain from native rodents in Australia (NF2653), and the Brucella spp. recently isolated from amphibians. These early-diverging Brucella produce an atypical LPS (Scholz et al., 2010;Tiller et al., 2010;Soler-Lloréns et al., 2016;Al Dahouk et al., 2017). Unexpectedly, wadD was present in all of them but absent in Brucella spp. B13-0095, one of the four Brucella strains isolated from frogs that have been completely sequenced (Soler-Lloréns et al., 2016). In contrast, this strain conserves wadB and wadC. Finally, WadD was 72 and 71% homologous to Ochrobactrum anthropi and O. intermedium orthologs, respectively, two species that also belong to the α-2 Proteobacteria subclass and are the closest genetic neighbors of Brucella.

Dysfunction of wadD Generates Increased Sensitivity to Cationic Peptides and Normal Serum
To test if the core defect displayed by Ba wadD affected resistance to polycationic peptides, we used poly-L-ornithine, a mildly bactericidal cationic peptide and the two known core mutants Ba wadB and Ba wadC plus the parental strain as controls. The results ( Figure 5A) showed that wadD dysfunction brought about a sensitivity similar to that of Ba wadB but inferior to that of Ba wadC. These differences in sensitivity were not due to growth defects because Ba wadD had a growth rate similar to that of Ba-parental, and experiments with the highly bactericidal lipopeptide polymyxin B confirmed the role of wadD (Ba wadD MIC = 0.094 µg/ml versus MIC = 2 µg/ml for both the mutant complemented with wild-type wadD and Ba-parental).
S brucellae are resistant to the bactericidal action of normal serum, a property associated with both the O-PS hindrance to inner OM targets such as OMPs and the PAMP modifications of the core that reduce binding of the complement activators of the antibody-independent classical pathway (Conde-Álvarez et al., 2012;Gil-Ramírez et al., 2014;Fontana et al., 2016). We compared the sensitivity to newborn bovine and ovine serum of Ba-parental and wadB, wadC, or wadD mutants and observed that the three core mutants were more sensitive than Ba-parental. The effect was more remarkable for mutant wadC than for wadD or wadB mutants ( Figure 5B).

Dysfunction of wadD Generates Attenuation Detectable in the Chronic Phase in the Mouse Model
To analyze the role of wadD in virulence, we infected BALB/c mice (n = 5) with Ba wadD or Ba-parental and compared the CFU/spleen at weeks 2, 8, and 12 (Figure 6). At weeks 8 and 12 post-infection, the CFU numbers of Ba wadD were significantly lower than those of Ba-parental (p = 0.0003 and p = 0.0073, respectively), showing that wadD is required for full Brucella virulence in mice. This result is in line with previous observations with wadB and wadC mutants (Conde-Álvarez et al., 2012;Gil-Ramírez et al., 2014) and further confirms that an intact LPS core is necessary for virulence.
Our results indicate that mutants in these ORFs react similarly to the parental strain in the S and R-LPS fractions and suggest that, in the studied conditions and with the available techniques, they seem not to be required for the synthesis of a complete LPS. Interestingly, the last three ORFs have been shown to be controlled by mucR, a regulator of Brucella virulence. Although it has been reported that B. abortus and B. melitensis mucR mutants have a defect in the core LPS (Caswell et al., 2013;Mirabella et al., 2013), the glycosyltransferases responsible for this defect have not been identified. In this work we have shown that mutation of the mucR-regulated putative glycosyltransferases BAB2_0133, BAB2_0135, and BAB1_0326 (Caswell et al., 2013) does not affect the synthesis of the core, at least in the growth conditions tested. Nevertheless, since the expression of these genes seems to be repressed by mucR (Caswell et al., 2013), a single mutation in the ORF could not be sufficient for the complete clarification of their role in LPS synthesis and further work would be required.
We have also analyzed in detail the role of the hypothetical glycosyltransferase BAB1_1620, as it is regulated by the master two-component regulator BvrR/BvrS that controls Brucella FIGURE 6 | Mutants in wadD showed slight attenuation at late stage of infection. Spleen CFUs in infected BALB/c mice were counted at 2, 8, and 12 weeks after inoculation of 5 × 10 4 CFU. Means were compared by Student's t independent-samples test ( * p < 0.05). For mice infected with the parental strain the counts were 6.25 ± 0.38, 6.98 ± 0.24, and 5.85 ± 0.32, respectively, and for mice infected with the wadD mutant 6.56 ± 0.22, 6.28 ± 0.54, and 5.17 ± 0.27.
virulence and the expression of surface components. According to our results, this ORF is not required for the synthesis of a complete LPS and is not implicated in surface-dependent characteristics that confer resistance to polycationic peptides or in virulence in the mouse model.
More interestingly, we report the identification of wadD, a gene encoding a previously unidentified glycosyltransferase involved in the synthesis of the core section not linked to the O-PS and thus, corroborate and extend previous work indicating that the Brucella LPS core is a branched structure that constitutes a steric impairment preventing the elements of the innate immune system to fight against Brucella (Conde-Álvarez et al., 2012;Kubler-Kielb and Vinogradov, 2013;Gil-Ramírez et al., 2014;Fontana et al., 2016), and contribute to Brucella virulence.
The discovery of genes wadC and wadB, involved in the synthesis of the lateral branch not linked to the O-PS, was critical for the understanding of the structure and the role of the core section in virulence. It has been clearly demostrated that in a wadC mutant, the complete core lateral branch is absent because this mutant cannot incorporate the mannose residue that is the depart of the lateral branch and it links to the lipid A-core section (Conde-Álvarez et al., 2012;Fontana et al., 2016). In accordance, deletion of wadC results in higher sensitivity to polycationic peptides and complement, better recognition by the CD14-MD2-TLR4 receptor complex, maturation of dendritic cells, secretion of pro-inflammatory cytokines (including Th1-type cytokines IL-12 and IFN-γ), and attenuation in mice (Conde-Álvarez et al., 2012;Fontana et al., 2016). A wadB mutant is also more sensitive to elements of the innate immune system and shows attenuation in mice, although not to the levels of the wadC mutant . As we show here, disruption of wadD in B. abortus leads to a S strain with a core defect less severe than that of the wadC mutant, more sensitive to polycationic peptides and normal serum than the parental strain and attenuated in the murine model. Interestingly, its sensitivity to polycationic peptides is similar to that of the wadB mutant  and not as strong as that of the wadC mutant (Conde-Álvarez et al., 2012), that has lost the complete branch, and its role in virulence became apparent already at 8 weeks post-infection.
The resistance to polycationic peptides and the bactericidal action of normal serum of mutant wadD strongly suggest a role in thwarting some effectors of innate immunity and that this could be manifested in the early stages of infection. However, to observe these effects would depend on the virulence model used. Indeed, the results obtained in an in vitro test, where the bacteria are put directly in contact with the polycationic peptides or with the serum can not be completely extrapolated to all situations in the in vivo model where other factors apart from polycationic peptides and complement are clearly taking part during the infection process. Still, the mouse model is the only well-characterized laboratory model for Brucella virulence studies and requires that mice are inoculated by the IP route. By this route, bacteria are not in contact with polycationic peptides or serum proteins at the beginning of the infection process as they are taken up and transported to the spleen rapidly. Nevertheless, the attenuation observed for Brucella LPS core mutants is caused by an early activation of innate immunity, as we have proved before (Conde-Álvarez et al., 2012;Fontana et al., 2016).
Our results would be compatible with the loss of one or few glucosamine residues in the lateral branch of the wadD mutant, and with the fact that removal of these residues would cause an increase in overall negative charge of the remaining LPS inner section that will facilitate the binding of polycationic peptides.
According to chemical studies performed in B. melitensis, the core lateral branch contains a mannose and four glucosamines residues assambled as follows: β-D-GlcpN-(1 → 6)-β-D-GlcpN-(1 → 4)-[β-D-GlcpN-(1 → 6)]-β-D-GlcpN-(1 → 3)-α-D-Manp-(1 → 5) (Figure 1). Taking into account that wadC, wadB, and wadD are perfectly conserved in B. melitensis and B. abortus, and since WadC adds the mannose (Conde-Álvarez et al., 2012;Fontana et al., 2016), in all likelihood the four glucosamines should be added by WadB  and WadD. These glucosamines are bound to each other by β-(1 → 6), or β-(1 → 4) links, and the one bound to mannose by β-(1 → 3) is also linked to two glucosamine residues, both in β-(1 → 6) and β-(1 → 4). If WadB and WadD are the only glycosyltransferases involved in the assembly of the glucosamine tetrasaccharide and its binding to the mannose residue, one of them (or both) could be multi-fiunctional and thus able to add sugars in different linkage. Most glycosyltransferase enzymes involved in lipooligosaccharide (LOS) or LPS core biosynthesis are responsible for one type of sugar addition onto the growing chain (Raetz and Whitfield, 2002). However, some bacterial glycosyltransferase enzymes of the GT-2 family, to which WadD belongs, can be multi-functional and are characterized by the presence of tandems of two active domains (DXD) on one polypeptide, as is the case of Lgt3, responsible for the addition of three glucoses with different linkages [β-(1-3), β-(1-4), and β-(1-6)] onto the inner core of Moraxella catarrhalis LOS (Coutinho et al., 2003;Luke-Marshall et al., 2013). Interestingly, WadD from B. melitensis, B. abortus, and all the orthologs in the other Brucella spp. conserves two DXD domains, opening the door to the possibility of a bi-functional role for this glycosyltransferase (Supplementary Figure S5). This DXD domain is not present in WadB. Nevertheless, the understanding of the particular role of each glycosyltransferase in the linkage of the different glucosamines to form the pentasaccharide (glucosamine tetrasaccharide bound to mannose) would require the elucidation of the core chemical structure of wadB and wadD mutants.
Contrary to most of the genes encoding glycosyltransferases implicated in the synthesis of the LPS, that are clustered in the same or related regions of the Brucella genome (Vemulapalli et al., 2000;Monreal et al., 2003;Rajashekara et al., 2004Rajashekara et al., , 2008González et al., 2008), wadC, wadB, and wadD (BAB1_1522, BAB1_0351, and BAB1_0953, respectively), although all situated in chromosome I, are isolated and surrounded by other ORFs apparently not related to LPS synthesis. This, and the fact that some other genes involved in the synthesis of the core (manBcore and manCcore) are situated in chromosome II (Monreal et al., 2003;González et al., 2008), makes even more intriguing the identification of genes needed for the synthesis of Brucella core LPS and its lateral branch. Thus, although we think all glycosyltransferases have been identified, we can not rule out that other glycosyltransferases could be required for the assembly of the pentasaccharide that forms the core lateral branch.
In a chemical characterization of the core LPS previously performed in a B. melitensis strain different from the one used in our studies, a glucose residue was found linked to the mannose that is the depart of the lateral branch, and, if this were the case, a new glycosyltransferase could be needed (Kubler-Kielb and Vinogradov, 2013). However, it should be taken into account that the LPS extraction method and the B. melitensis biovar used for the determination of the core structure in this experiement were different from those used in our genetic and biochemical studies (Figure 1). It is important to notice that the chemical structure we discuss in Figure 1 has been elucidated in the same B. melitensis strain where the wadC gene was mutated (Fontana et al., 2016) and, in this case, no glucose residues were detected. The fact that, as discussed above, wadC, wadB, and wadD are perfectly conserved in B. melitensis, reinforces the interpretation of our results, and the idea that the glycosyltransferases encoded by the last two genes would be involved in the assembly of glucosamine residues.
Nevertheless, we should consider that, although the phenotype of wadC mutant in B. melitensis and B. abortus is similar (Conde-Álvarez et al., 2012;Fontana et al., 2016), previous results suggest that there could exist differences in the structure of the core in these two Brucella spp., since they react differently with MoAbs against core epitopes (González et al., 2008). Moreover, we have already seen that some of the studied ORFs (and discarded in our first screening since they reacted as the parental strain in the rough and smooth LPS fractions) present differences between B. abortus and B. melitensis (Supplementary Table S2). Thus, we could not discard them as the responsible for these differences. To understand the final role of wadB, wadC, and wadD, it would be necessary to analyze and compare the chemical structure of the core section in mutants in these genes in both spp.

ETHICS STATEMENT
Female BALB/c mice (ENVIGO, Harlan) were kept in cages with water and food ad libitum under P3 biosafety conditions in the facilities of CIMA (registration code ES31 2010000132) or CITA (registration code ES502970012005) 2 weeks before and during the experiments. The procedures were in accordance with the current European (directive 86/609/EEC) and Spanish (RD 53/2013) legislations, supervised by the Animal Welfare Committee of the University of Navarra or CITA, and authorized by "Gobierno de Navarra" (protocol number 134-14) or "Gobierno de Aragón" (protocol number R108/2009).