A deep rough type structure in Bordetella bronchiseptica lipopolysaccharide modulates host immune responses


  • Federico Sisti,

    1. VacSal Laboratory, Biotechnology and Molecular Biology Institute, Department of Biological Sciences, Faculty of Sciences, National University of La Plata, National Council of Scientific and Technical Research (CONICET), Calles 47 y 115 (1900) La Plata
    Search for more papers by this author
    • These authors contributed equally to this work and share first author status.

  • Julieta Fernández,

    1. VacSal Laboratory, Biotechnology and Molecular Biology Institute, Department of Biological Sciences, Faculty of Sciences, National University of La Plata, National Council of Scientific and Technical Research (CONICET), Calles 47 y 115 (1900) La Plata
    Search for more papers by this author
    • These authors contributed equally to this work and share first author status.

  • Sarah C. Higgins,

    1. Immune Regulation Research Group, School of Biochemistry and Immunology, Trinity College, Dublin 2, Ireland
    Search for more papers by this author
  • Adriana Casabuono,

    1. Center for Carbohydrate Research, Department of Organic Chemistry, Faculty of Exact and Natural Science, (1428) University of Buenos Aires, Buenos Aires, Argentina
    Search for more papers by this author
  • Alicia Couto,

    1. Center for Carbohydrate Research, Department of Organic Chemistry, Faculty of Exact and Natural Science, (1428) University of Buenos Aires, Buenos Aires, Argentina
    Search for more papers by this author
  • Kingston H. G. Mills,

    1. Immune Regulation Research Group, School of Biochemistry and Immunology, Trinity College, Dublin 2, Ireland
    Search for more papers by this author
  • Daniela Hozbor

    1. VacSal Laboratory, Biotechnology and Molecular Biology Institute, Department of Biological Sciences, Faculty of Sciences, National University of La Plata, National Council of Scientific and Technical Research (CONICET), Calles 47 y 115 (1900) La Plata
    Search for more papers by this author

Daniela Hozbor, Laboratorio VacSal, Instituto de Biotecnología y Biología Molecular, Departamento de Ciencias Biológicas, Facultad de Ciencias Exactas, Universidad Nacional de La Plata, CCT La Plata CONICET, Calles 47 y 115 (1900) La Plata, Argentina. Tel: 54 221 422 9777; fax: 54 221 4229777; email address: hozbor@biol.unlp.edu.ar


The present authors have previously obtained the Bordetella bronchiseptica mutant BbLP39, which contains a deep-rough lipopolysaccharide (LPS) instead the wild type smooth LPS with O antigen. This mutant was found to be altered in the expression of some proteins and in its ability to colonize mouse lungs. Particularly, in BbLP39 the expression of pertactin is decreased. To differentiate the contribution of each bacterial component to the observed phenotype, here mice defective in the LPS sensing receptor TLR4 (TLR4-defective mice) were used. In contrast to wild-type mice, infection of TLR4-defective mice with BbLP39 resulted in lung infection, which persisted for more than 10 days post-challenge. Comparative analysis of the immune responses induced by purified mutant and wild type LPSs showed that the mutant LPS induced significantly higher degrees of expression of TNF-α and IL-10 mRNA than did the wild type. UV matrix-assisted laser desorption/ionization time-of-flight (MALDI TOF) mass spectrometry analysis revealed that both LPSs had the same penta-acylated lipid A structure. However, the lipid A from BbLP39 contained pyrophosphate instead of phosphate at position 1. This structural difference, in addition to the lack of O-antigen in BbLP39, may explain the functional differences between BbLP39 and wild type strains.

List of Abbreviations: 
B. bronchiseptica

Bordetella bronchiseptica

BGA medium

Bordet Gengou supplemented with 10% (v/v) defibrinated fresh sheep's blood


bone marrow-derived immature DC cells


colony forming units


dendritic cells



E. coli

Escherichia coli






matrix-assisted laser desorption/ionization time-of-flight


tumor necrosis factor-α



Lipopolysaccharides are outer membrane molecules that are essential for all Gram-negative bacteria and consist of highly variable as well as conserved segments (1). Conserved regions of LPS assist in the maintenance of a structure that is essential for survival of the bacterium, while the variable regions represent segments that are not essential, allowing for evolutionary variation without significant consequences.

In extremely small amounts, LPS possesses potent bioactivities initiating, for example, the morbidity and mortality associated with Gram-negative sepsis, as well as the modulation of a myriad of host innate inflammatory responses. Interestingly, specific synthetic alterations within conserved segments of LPS have been shown not only to be recognized by host cells, but also to be capable of blocking or down-regulating responses to other LPS forms normally associated with robust innate cell activation (2). In fact, natural heterogeneity of LPS in certain pathogenic bacteria, arising from either altered bacterial growth or environmental conditions, has been shown to result in differential and altered innate host cell responses in vitro (3). LPS heterogeneity confers some evasion advantages, resulting in survival and sustained pathogenicity for the bacterium in vivo (4–6).

Bordetella bronchiseptica, a Bordetella species that can infect a variety of mammals including humans, expresses LPS that plays an essential role in host interactions, being critical for early clearance of the bacteria (7). B. bronchiseptica is the only species of the classical Bordetella group that expresses a hexa-acylated lipid A as a major molecule; some have a ‘4 + 2’ arrangement of acyl chains, although tetra- and penta-acylated species are also present (8, 9).

The use of defective mutants in different portions of the LPS structure has allowed progress in understanding the role of this molecule in bacteria–host interactions. We have previously generated, on a B. bronchiseptica 9.73 (Bb9.73) background, a mutant defective in the expression of waaC gene (B. bronchiseptica LP39, BbLP39), which codes for a heptosyltransferase involved in the biosynthesis of the core region of the LPS (10). This mutation led to a small LPS structure (known as deep rough structure), the smallest so far described in B. bronchiseptica. Although in BbLP39 production of the principal virulence determinants adenylate cyclase-hemolysin, filamentous hemagglutinin and pertactin persisted, the quantity of the two latter factors was diminished, the amounts of pertactin being the most greatly affected. Our previous in vivo experiments demonstrated that the ability of BbLP39 to colonize the respiratory tract of mice containing wild type LPS sensing receptor TLR4 (TLR4-competent mice) is reduced, the bacterium having been effectively cleared from the lungs within 5 days, whereas the parental strain Bb9.73 survived for at least 30 days. Our in vitro experiments demonstrated that, although BbLP39 had impaired adhesion to epithelial cells, it was still able to survive within the host cells as efficiently as the parental strain. These results seem to indicate that the deep rough form of B. bronchiseptica LPS cannot represent a dominant phenotype in the first stage of colonization.

To evaluate in more detail the contribution of the altered LPS in the in vivo behavior of the deep-rough mutant, we performed colonization experiments using mice defective in LPS sensing receptor TLR4 (TLR4-defective mice). We also performed assays using purified LPS from either Bb9.73 or BbLP39. The results presented here clearly demonstrate the role of altered LPS in vivo in TLR4 competent mice.


Bacterial strains and growth conditions

Bordetella bronchiseptica strain 9.73 (Collection de l’Institut Pasteur designation, Bb9.73) was grown on BGA medium (Difco, Sparks, MD, USA) at 36°C for 48 hr. B. bronchiseptica LP39 (BbLP39), a deep rough mutant derived from Bb9.73 (10) was grown in BGA medium supplemented with streptomycin (200 μg/mL) and kanamycin (80  μg/mL). Both strains were replated in the same medium for 24 hr. For LPS extraction, subcultures were grown in Stainer-Scholte liquid medium (11) for 20 hr at 36°C until the optical density measured at 650 nm reached 1.0.

Mouse infection

All mouse procedures were performed in accordance with National Regulations. Three- four-week-old female C3H/HeN and C3H/HeJ mice (originally from Jackson Laboratories, West Grove, PA, USA and maintained at the National Pharmaceutical Company Biological Institute, Buenos Aires, Argentina) were used in our studies. Bacteria were cultured and administered i.n. to mice as previously described (12). The mice were killed by cervical dislocation at 2 hr, 3 and 10 days after infection and the lungs removed and homogenized in 1000 μL of sterile PBS. Appropriate dilutions were plated on BGA blood agar plates and counted after 3 days of incubation at 37°C to determine CFU per lung. A minimum of five mice per group was used in each of the four independent experiments performed.

For immune response analysis, two- three-week-old female BALB/c mice were inoculated with 50 μL containing 500 ng of different LPS onto the tip of the external nares. Two hours later, groups of at least four mice were killed by cervical dislocation. The inferior lobes of their right lungs were immediately processed for total RNA obtaining using Nuclespin RNAII system (Macherey-Nagel, Duren, Germany) according to the manufacturer instructions.

Lipopolysaccharide extraction

Bacterial cultures were centrifuged (10,000 ×g, 15 min, 4°C) and washed twice in distilled water. LPSs were extracted by the hot-phenol-water method of Darveau and Hancock (13). All isolated LPSs were sequentially treated with DNAse and proteinase K (10  μg/mL and 1 mg/mL overnight). The resultant LPS samples were dialyzed and lyophilized. Dry weight measures were used to quantify the LPS obtained. The quality of each sample was checked by SDS-PAGE, Fig. 1a. The isolated LPSs were stored at −20°C until use.

Figure 1.

(a) Silver-stained SDS-PAGE 17.5% wt/vol profiles of phenol-water-extracted LPS samples from Bb9.73 wild-type (line 1) and BbLP39 mutant (line 2) strains. (b) UV-MALDI-TOF structure of the predominant lipid A molecule isolated from Bb9.73. The additional phosphate group boxed in the figure, located in position 1, is only present in BbLP39. (c) UV-MALDI-TOF mass spectrum of BbLP39 LPS in the positive linear ion mode using CMBT as matrix. (d) UV-MALDI-TOF mass spectrum of BbLP39 LPS in the negative linear ion mode using nor-harmane as matrix.

Isolation and purification of lipid A

The LPSs were hydrolyzed with 2% acetic acid for 2 hr at 100°C (14). Precipitated lipid A was recovered by centrifugation at 4°C, 9000 ×g for 60 min. The solution containing the sugar moiety was separated and lyophilized. The solid component was washed with a small volume of water at 4°C, centrifuged as described above and extracted with CHCl3/MeOH/water (12:6:1). The purified lipid A was stored at −20°C.

Ultraviolet matrix-assisted laser desorption/ionization time-of-flight mass spectrometry analysis

Matrices and calibrating chemicals were purchased from Sigma-Aldrich (St Louis, MO, USA). Measurements were performed using an Ultraflex II TOF/TOF mass spectrometer (Bremen, Germany) equipped with a high-performance solid-state laser (λ= 355 nm) and a reflector. The system is operated by the Flexcontrol 2.4 software package (Bruker Daltonics GmbsH, Bremen, Germany). The mass spectra reported are the result of 1000 laser shots. All samples were measured in the linear and reflectron modes and as routine in both positive and negative polarity. The samples were loaded onto a ground steel sample plate (MTP 384 ground steel; Bruker Daltonics GmbsH) using the sandwich method or by the classic dried drop method: a sample/matrix solution mixture 1 μL, 1:1 (v/v) was deposited on the target plate and allowed to dry at room temperature.

Different matrices were assayed to obtain adequate spectra. For lipid A analysis: nor-harmane and 5-chloro-2-mercaptobenzothiazole; for oligosaccharide: nor-harmane and for complete LPS: nor-harmane and 5-chloro-2-mercaptobenzothiazole and gentisic acid gave better spectra.

RNA isolation, reverse transcription and quantitative real-time polymerase chain reaction

RT-qPCR was performed as previously described (15). Reverse transcription was performed on 100 ng of RNA using MMLV-RT (Promega, Madison, WI, USA). The resulting cDNA was amplified in triplicates using SYBR Green PCR assay (Bio Rad Laboratories, Hercules, CA, USA) and the products detected on an ICycler (Bio Rad). PCR samples were incubated for 2 min at 50°C and for 10 min at 95°C, followed by 40 amplification cycles with 1 min annealing/extension at 60°C and 15 s denaturation at 95°C. β-actin expression was used as the normalizer. The specificity of PCR was checked by melting curves. Relative amounts of mRNA were determined by comparing the normalized PCR cycle threshold between cDNA samples of the gene of interest as previously described (15). Specific primers for the different cytokines have already been described (15).

Cytokine/chemokine secretion by bone marrow dendritic cells

Bone marrow-derived immature DC cells were generated by culturing murine bone marrow cells for 7 days in medium with 5–10% granulocyte macrophage colony-stimulating factor cell supernatant, as previously described (16). Ten nanograms per mL of Bb9.73 LPS or BbLP39 LPS were incubated at 37°C with BMDCs cultured in 24-well tissue culture plates at 106 cells/mL. LPS from E. coli (LPS E. coli, 10 ng/mL) was used as control. After 24 hr incubation, the supernatants were separated and the cytokines IL-1β, IL-10, and TNF-α were quantified by two-site ELISA (BD Biosciences, San Jose, CA, USA or R&D Systems, Minneapolis, MN, USA) as previously described (16).

Analysis of dendritic cell maturation by flow cytometry

Expression of surface markers was assessed using biotinylated anti-CD11c with streptavidin PerCP, phycoerythrin-conjugated anti-CD80, fluorescein isothiocyanate-conjugated anti-CD86 and anti-CD40. All antibodies were purchased from BD PharMingen (San Diego, CA, USA). Nonspecific binding was prevented by incubating the cells with 2% normal mouse serum for 30 min on ice. Cells incubated with an isotype-matched directly conjugated antibody with irrelevant specificity were used as a control. After incubation immunofluorescence analysis was performed using a FACScan (BD Biosciences, San Jose, CA, USA) and analyzed using CellQuest software. Twenty thousand cells per sample were analyzed.

Statistical analysis

All the results were compared by analysis of variance followed by the Tukey test. Differences with P < 0.05 were considered significant.

Survival analysis was done by the Kaplan-Meier method. The difference between the survival curves was analyzed by the log-rank test. A value of P < 0.05 was considered significant.


Structure of lipopolysaccharide from Bordetella bronchiseptica LP39

Previously, we reported that the SDS-PAGE LPS profile from BbLP39 has no O-antigen band (10), which we have confirmed here as shown in Figure 1a. Indeed, BbLP39 LPS showed a single band, migrating considerably faster than that corresponding to the Bb9.73 wild-type strain. The wild-type LPS resolved as two bands, A and B, and one region that corresponds to the O-antigen (Fig.  1a).

In addition to the SDS-PAGE profile, we determined the lipid A structures by UV-MALDI-TOF mass spectrometry (Fig.  1b). This methodology revealed that both LPSs contain the same penta-acylated lipid A. Interestingly, a structural difference between both LPSs was detected: MALDI-TOF mass spectra of BbLP39 lipid A (Fig.  1c, d) indicated the presence of a diglucosamine backbone penta-acylated and carrying one phosphate and one pyrophosphate group which was not observed in the wild type LPS (Fig.  1b). Though to our knowledge this is the first time that a pyrophosphate group has been detected in B. bronchiseptica, there are some examples of pyrophosphorylated lipid A in other Gram negative bacteria such as the lipid A of Yersinia pestis, Salmonella typhimurium and E. coli (17–19). We also detected a KDO unit substituted with a phosphoethanolamine group.

Infection of TLR4-defective mice by Bordetella bronchiseptica LP39

To evaluate the contribution of the altered LPS to the in vivo behavior of mutant bacteria we performed B. bronchiseptica colonization assays comparing TLR4-defective mice to TLR4-competent mice. Sublethal doses (5.105 CFU in 50 μL) of either Bb9.73 (wild type) or BbLP39 were used to i.n. inoculate C3H/HeJ TLR4-defective or C3H/HeN parental mice. Whereas TLR4-competent mice survived the infection, it was lethal to the TLR4-deficient strain. However, the mortality rate differed according to the bacterial strain: all mice died within 10 days of inoculation with Bb9.73, compared to only 40% of mice inoculated with BbLP39 (Fig. 2a). We also observed differences in bacterial persistence inside the host. In TLR4-competent mice infection caused by Bb9.73 resolved within 40 days of infection (7), whereas BbLP39 bacteria were cleared within 3 days of challenge (Fig.  2b). Furthermore, infection of TLR4-defective mice with BbLP39 was significantly more protracted than in TLR4-competent mice, the bacteria persisting for more than 10 days. We recovered bacteria from lungs up to day 20 after infection (not shown).

Figure 2.

B. bronchiseptica infection of TLR4-competent C3H/HeN or TLR4 defective C3H/HeJ mice (a) Survival curves of mice infected with 5.105 CFU of Bb9.73 or BbLP39. Statistically significant differences between C3H/HeJ mice infected with Bb9.73 or BbLP39 were observed (log-rank test, P < 0.02). (b) CFU of Bb9.73 wild-type strain or BbLP39 mutant recovered from lungs of infected mice at the indicated post-infection times. No mice inoculated with the wild type strain survived to day 10 post-infection. Results shown are representative of three different experiments.

Immune response induced by Bordetella bronchiseptica LP39 purified LPS

To examine the innate immune response induced in vivo against purified LPS structures, mice were inoculated with either Bb9.73- or BbLP39-purified LPS (500 ng in 50 μL) and the animals killed 2 hr later to evaluate the amounts of TNF-α, IL-1β and IL-10 in their lungs.

TLR4-competent mice inoculated with wild-type LPS had six times higher TNF-α mRNA expression than untreated mice (Fig. 3). Moreover, IL-10 and IL-1β mRNA expression was significantly increased (quadrupled and doubled, respectively). Surprisingly, the deep rough BbLP39 LPS induced 2.6 times greater TNF-α and 3.5 times greater IL-10 mRNA expression than the wild-type LPS. We also observed a slight increase in IL12p40 for BbLP39 LPS in comparison to the wild-type LPS. On the other hand, in TLR4-deficient mice, the wild-type and mutant LPS structures did not induce cytokine expression greater than that in control mice (data not shown).

Figure 3.

Innate immune response to purified B. bronchiseptica LPS in vivo. Mice were inoculated either with wild type Bb9.73 LPS or the deep-rough mutant BbLP39 LPS. Total RNA was extracted at 2 hr post inoculation from the lungs and relative expression of different cytokines evaluated by RT-qPCR. Fold increase values are relative to mock treated lungs. Results show the mean of three individual mice per group. Results shown are representative of three different experiments. *indicates P < 0.02; LPS Bb9.73 versus LPS BbLP39, versus non-treated mice.

Since DCs are intimately involved in the innate response to pathogens and strongly influence adaptive immunity, we examined the effect of LPS on BMDC. We analyzed expression of CD40, CD80, and CD86 following exposure of DC for 24 hr to 10 ng/mL of different purified LPSs. The results shown in Figure  4 demonstrate that stimulation with both B. bronchiseptica LPSs enhances surface expression of CD86, CD80, and CD40. We detected no expression of these markers in untreated DC. Furthermore, when LPS samples were treated with Affi-Prep polymyxin (Bio-Rad, Hercules, CA, USA) no increase in the surface marker expression was observed (data not shown). These results suggest that the mutation in LPS not affect DC maturation.

Figure 4.

B. bronchiseptica LPS induces maturation of BMDC. BMDC were stimulated with Bb9.73 LPS, BbLP39 LPS or medium only. After 24 hr, cells were washed, and immunofluorescence analysis was performed with antibodies specific for CD80, CD86, or CD40 (filled curve), or with isotype-matched control Abs (empty curve). Results are representative of three different experiments.

We then measured BMDC cytokine production in response to stimulation with LPS. Both B. bronchiseptica LPSs induced high concentrations of all cytokines assayed. Moreover, wild type LPS induced eight times higher amounts of IL-10 compared with E. coli LPS (10 ng/mL) used as positive control (Fig. 5). Treatment of BMDC with the deep-rough LPS (10 ng/mL), induced similar concentrations of TNF-α and IL-1β as wild-type LPS. However, the deep-rough LPS induced a significantly higher concentration of IL-10. Treatment with polymyxin reduced the concentrations of cytokines to those observed in untreated cells (data not shown).

Figure 5.

B. bronchiseptica LPS stimulates proinflammatory cytokines and IL-10 production by DC. Bone marrow-derived DCs from mice stimulated with 10 ng/mL of wild-type Bb9.73 LPS, deep rough BbLP39 LPS, or LPS purified from E. coli. Cytokine concentrations were evaluated by immunoassay 24 hr later. *P < 0.002; **P < 0.001 LPS Bb9.73 vs LPS BbLP39.

The difference between the relative results of Bb9.73 versus BbLP39 observed in in vivo and in vitro immune responses (Table 1) could be because in vivo other cells besides dendritic cells respond to LPS. Amongst these cells are macrophages, which are known to have different sensitivity than DC to LPS (20, 21). Therefore it is not surprising that the response to LPS in in vitro assays using a single cell type does not correlate exactly to that observed in in vivo assays, where the response to LPS is the result of the action of different cells with different sensitivities to this molecule.

Table 1.  Immune response to purified B. bronchiseptica LPS
 In vivo Relative mRNA expressionBMDC in vitro Protein in supernatant
TNF-α6.5  ±  0.316.3  ±  1.24707  ±  415 pg/mL5152  ±  170 pg/mL
IL-103.7  ±  0.113.6  ±  1.0534  ±  45 pg/mL1213  ±  200 pg/mL
IL-1β1.5  ±  0.1 0.5  ±  0.1 605  ±  200 pg/mL 596  ±  180 pg/mL
IL-12p401.0  ±  0.1 4.4  ±  0.1106  ±  6 ng/mL 273  ±  9 ng/mL 


Alterations in LPS structure in B. bronchiseptica, and other pathogens lead to changes in the host-bacterium interaction (22–25). Harvill et al. used wlb deletion (Δwlb) mutants to investigate the roles of distal LPS structures in respiratory tract infection induced by Bordetellae (25). They found that, compared to the wild type strain, B. bronchisepticaΔwlb strain is defective in colonization of the respiratory tracts of BALB/c. The mice could resolve infection much faster (1 week post-inoculation) with the mutant than with wild type bacteria. In agreement with these findings, we have previously observed that a complete B. bronchiseptica LPS, but not a deep rough LPS, is necessary to subvert the initial steps of infection (10). A modification in lipid A also appears to have an impact on bacteria-host-interaction. Preston et al. demonstrated that PagP, a palmitoyl transferase that mediates modification of lipid A, is not required for initial colonization of the mouse respiratory tract by B. bronchiseptica, but is necessary for persistence of the bacteria within the host (26).

In this study, we have demonstrated that, in the absence of a TLR4-mediated immune response, B. bronchiseptica mutant carrying a deep-rough LPS structure is able to colonize and persist in mice, bacteria still being detectable in the lungs beyond day 10. In the case of wild type B. bronchiseptica infection, all TLR4-defective mice died before day 10 post-infection. This distinctive effect on survival of mice following infection with the mutant and parental strains may be explained, at least in part, by changes in the expression of certain virulence factors such as pertactin (27, 28).

Infection with the deep-rough mutant B. bronchiseptica strain persisted longer in TLR4-defective than in immune-competent mice. This confirms the importance of the immune response induced by LPSs in the clearance of bacteria from the respiratory tract. To understand the molecular basis of the differential immune responses induced with mutant and wild type bacteria, we performed in vivo experiments with purified LPS. We found that the deep-rough LPS induced significantly higher TNF-α and IL-10 mRNA expression than the wild-type LPS. The enhancement of these innate immune responses by the deep-rough mutant may reflect the here-described structural differences between lipid A and wild-type LPS. Interestingly, MALDI-TOF mass spectra of BbLP39 lipid A indicates that the KDO unit is linked to a phosphoethanolamine group and that there is a diglucosamine backbone penta-acylated carrying one phosphate and one pyrophosphate group. In other pathogens, differences in the phosphorylation pattern of lipid A have been shown to be important for their biological activity. Removal of a phosphate group has been shown to substantially reduce the toxicity of lipid A (1, 29) and its ability to induce IL-1β (3). In contrast, masking of lipid A phosphate groups (e.g., by addition of aminoarabinose) has been shown to affect bacterial resistance to host cationic antimicrobial peptides (2). The biochemical effects of phosphate groups in lipid A have been attributed to their negative charge, which affects recognition by TLR4 and further LPS-induced signaling in the host immune response to bacterial infection (30).

Our findings suggest that changes in the phosphorylation of lipid A lead to differences in the ability of LPS to induce innate immune responses. These differences are characterized by increased expression of TNF-α, which might be partially responsible for the early clearance of bacteria in TLR4-competent mice. The results presented here strengthen knowledge about the functional role of the different LPS structures in B. bronchiseptica-host interactions.


This work was supported by National Agency for the Promotion of Science and Technology (ANCPyT) and Commission for Scientific Investigation Buenos Aires (CICBA) (Argentina) grants to D.F.H., a National Council of Scientific and Technical Research (CONICET) grant to F.S. and A.C. D.F.H. is a member of the Scientific Career of CICBA. A.C., J.F. and F.S. are members of the Scientific Career of CONICET. Kingston Mills's group is supported by Science Foundation Ireland.


The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in, or financial conflict with, the subject matter or materials discussed in the manuscript.