Frankia alni Carbonic Anhydrase Regulates Cytoplasmic pH of Nitrogen-Fixing Vesicles

A phyloprofile of Frankia genomes was carried out to identify those genes present in symbiotic strains of clusters 1, 1c, 2 and 3 and absent in non-infective strains of cluster 4. At a threshold of 50% AA identity, 108 genes were retrieved. Among these were known symbiosis-associated genes such as nif (nitrogenase), and genes which are not know as symbiosis-associated genes such as can (carbonic anhydrase, CAN). The role of CAN, which supplies carbonate ions necessary for carboxylases and acidifies the cytoplasm, was thus analyzed by staining cells with pH-responsive dyes; assaying for CO2 levels in N-fixing propionate-fed cells (that require a propionate-CoA carboxylase to yield succinate-CoA), fumarate-fed cells and N-replete propionate-fed cells; conducting proteomics on N-fixing fumarate and propionate-fed cells and direct measurement of organic acids in nodules and in roots. The interiors of both in vitro and nodular vesicles were found to be at a lower pH than that of hyphae. CO2 levels in N2-fixing propionate-fed cultures were lower than in N-replete ones. Proteomics of propionate-fed cells showed carbamoyl-phosphate synthase (CPS) as the most overabundant enzyme relative to fumarate-fed cells. CPS combines carbonate and ammonium in the first step of the citrulline pathway, something which would help manage acidity and NH4+. Nodules were found to have sizeable amounts of pyruvate and acetate in addition to TCA intermediates. This points to CAN reducing the vesicles’ pH to prevent the escape of NH3 and to control ammonium assimilation by GS and GOGAT, two enzymes that work in different ways in vesicles and hyphae. Genes with related functions (carboxylases, biotin operon and citrulline-aspartate ligase) appear to have undergone decay in non-symbiotic lineages.


Introduction
Actinobacterial genus Frankia contains several symbiotic lineages that fix nitrogen within the root nodules of plants belonging to the Rosales, Fagales and Cucurbitales orders [1]. A phylogenetic analysis showed that these symbiotic Frankia formed three clusters [2]. Conversely there are numerous strains that have lost various symbiotic features; these form a coherent cluster [3]. These non-symbiotic strains fall into two categories: those that are non-infective such as CN3 or DC12 [4,5] and those strains that are infective but non-nitrogen-fixing such as EuI1c or AgB1.9 [6,7]. The relative position of the clusters has been found to fluctuate depending on the marker used or the strains studied, [8,9] but all are robust.
It has been estimated that the symbiotic Frankia strains diverged from other soil actinobacteria around 80-100 MY [2], a time length that corresponds roughly to that of the emergence of the plant lineages in symbiosis with Frankia, predating the emergence of the Fabaceae which is estimated at 60 MY [10]. This is thus a lengthy period of time that could have impacted the Frankia genomes with genes which are non-essential for symbiosis, presumably costly to express and dispensable in free-living lineages, and during which these genes could have been subject to negative evolutionary pressure leading them to be lost in a manner similar to that occurring in non-symbiotic host plant lineages in relation to symbiosis-associated genes [1].
More than 40 Frankia genomes from all clusters have become available since the first ones were published in 2007 [11,12]. These genomes have been compared and have been found to fluctuate in terms of parameters such as size [11], GC% [12], codon usage [13] and Ka/Ks [14] as they have been subjected to different evolutionary pressures in the soil, on the one hand, or in plant tissues, on the other. There has been massive genome erosion between non-sporulating (Sp−) and sporulating strains (Sp+) [15], for instance, in Alnus-infective strains that have lost around 2500 genes, or one third of the total number.
There have also been studies targeted at groups of genes which have shown the presence of canonical nod genes in some cluster 2 strains [16,17] and one cluster 3 strain [18], the absence of nif, sodF and rhb genes in non-symbiotic cluster 4 strains [12], the presence of celA-bcsA genes in symbiotic strains [19] and the loss of saprophytic genes such as gvp in cluster 1c strains and cluster 2 strains that have reduced genomes [11,16]. These genomes are thus a wealth of data that can be analyzed to try to understand the genetic determinants underpinning symbiosis in the absence of a proven genetic transformation protocol. We decided to compare symbiotic and non-symbiotic genomes to determine if the estimated 100 MY since their separation had resulted in the loss of symbiosis-associated determinants in those strains that had lost the ability to establish nodules. One such determinant that emerged was the can genes that code for carbonic anhydrase (CAN), an enzyme that converts CO 2 into carbonic ions and protons mainly for pH homeostasis and the functioning of carboxylases. There are 23 of these genes in Frankia alni, and we investigated them from a physiological viewpoint.

Genome Mining
When a search was made for those proteins conserved at a threshold of 50% identity between symbiotic cluster1 (cl1) strains (ACN14a, QA3, CcI3), cluster 2 (cl2) strains (BMG5.1 and Dg1) and cluster 3 (cl3) strains (EaN1pec and EUN1f) and absent from nonsymbiotic non-infective cluster 4 (cl4) strains (CN3 and DC12), 108 hits were obtained (Table 1). Among those genes retrieved were nif (nitrogenase, FRAAL6800-6814); rhbBCEF (rhizobactin-related metabolite; FRAAL6422-FRAAL6426); sodF (Fe-superoxide dismutase, FRAAL4337); accA (acetyl carboxylase, FRAAL3196); argF (argininosuccinate lyase, FRAAL5202) and two can (carbonic anhydrase; FRAAL1222, FRAAL4889). These two CANs belong to the Beta class prevalent in prokaryotes which is more active than the Gamma CANs [20]. The can genes were found to be present in some genomes belonging to cluster 4, such as Frankia AgW1.1 and AgB1.8 strains, isolated from Alnus and able to reinfect Alnus but unable to fix nitrogen in pure culture or in vitro. When a search for gene annotations was carried out in the MAGE data base using "carbonic anhydrase" as keyword, supplementary hits were obtained in several lineages including Frankia alni, FRAAL2078, that would code for a distant Gamma CAN with homologs in all Frankia strains. Finally, the same keyword search yielded a Beta CAN present in cluster3 and cluster4 strains with a very low ID (<30% AA) with FRAAL1222 and FRAAL4889 CANs.

Cells Growth
Growth, N-fixation and CO 2 levels were monitored over liquid N-fixing and N-replete cultures. Frankia grew rapidly on propionate and ammonium, reaching a plateau after only 3 days, whereas it took 6 days when fixing nitrogen on propionate, and on fumarate it took 13 days ( Figure 1). The CO 2 in the propionate+ammonium-fed culture reached a plateau at day 4, whereas in the propionate-fed nitrogen fixing culture, CO 2 increased steadily over seven days as the cells multiplied. CO 2 in the fumarate-fed culture was low at first and increased slowly thereafter.

Cells Growth
Growth, N-fixation and CO2 levels were monitored over liquid N-fixing and N-replete cultures. Frankia grew rapidly on propionate and ammonium, reaching a plateau after only 3 days, whereas it took 6 days when fixing nitrogen on propionate, and on fumarate it took 13 days ( Figure 1). The CO2 in the propionate+ammonium-fed culture reached a plateau at day 4, whereas in the propionate-fed nitrogen fixing culture, CO2 increased steadily over seven days as the cells multiplied. CO2 in the fumarate-fed culture was low at first and increased slowly thereafter.

Physiology Measurements of Cells
Frankia alni cells grown on an FBM-medium were stained with DND-189 and DND-99 fluorescent probes and observed under fluorescent light with filters as well as with SYTO-9, which showed the vesicles to be fluorescent ( Figure 2) thus indicating their pH is acidic, more so than that of the hyphae that did not fluoresce with the pH-responsive stains. After staining 21 days post infection (dpi), nodule sections showed strong fluorescence with DND-99 and SYTO-9 stains, thus indicating acidity was found inside the cytoplasm of the vesicles (Figure 3).

Physiology Measurements of Cells
Frankia alni cells grown on an FBM-medium were stained with DND-189 and DND-99 fluorescent probes and observed under fluorescent light with filters as well as with SYTO-9, which showed the vesicles to be fluorescent ( Figure 2) thus indicating their pH is acidic, more so than that of the hyphae that did not fluoresce with the pH-responsive stains. After staining 21 days post infection (dpi), nodule sections showed strong fluorescence with DND-99 and SYTO-9 stains, thus indicating acidity was found inside the cytoplasm of the vesicles (Figure 3).

Physiology Measurements of Cells
Frankia alni cells grown on an FBM-medium were stained with DND-189 and DND-99 fluorescent probes and observed under fluorescent light with filters as well as with SYTO-9, which showed the vesicles to be fluorescent ( Figure 2) thus indicating their pH is acidic, more so than that of the hyphae that did not fluoresce with the pH-responsive stains. After staining 21 days post infection (dpi), nodule sections showed strong fluorescence with DND-99 and SYTO-9 stains, thus indicating acidity was found inside the cytoplasm of the vesicles (Figure 3).

Gene Expression
Expression levels of can1 (FRAAL1222), can2 (FRAAL4889), ctp (FRAAL4724) and two carbamoyl phosphate synthases cpsA (FRAAL4684) and cspB (FRAAL4683) are shown in Figure 4. The can1 and ctp genes were highly upregulated under nitrogen starvation in FBM− media and in symbiosis. The second can gene (FRAAL4889) was not differentially expressed under all conditions tested. The two carbamoylphosphate synthase genes, cpsA and cpsB genes, were up-regulated in free-living cultures without nitrogen and downregulated in symbiosis. A low level of expression was observed under growth as free culture in FBM+ with supplied ammonium. starvation in FBM-media and in symbiosis. The second can gene (FRAAL4889) was not differentially expressed under all conditions tested. The two carbamoylphosphate synthase genes, cpsA and cpsB genes, were up-regulated in free-living cultures without nitrogen and down-regulated in symbiosis. A low level of expression was observed under growth as free culture in FBM+ with supplied ammonium.

Proteomics
The proteomic analysis results of fumarate-and propionate-fed cells under nitrogen fixing conditions are shown in Table 2 with the 40 most up-regulated proteins of each condition. Fumarate-fed cells had a high overabundance of acyl-CoA synthase, three dicarboxylate transporters and proteases. Propionate-fed cells had carbamoyl-phosphate synthase, nitrogenase, SHC and lipid synthase as the most overabundant proteins.

Proteomics
The proteomic analysis results of fumarate-and propionate-fed cells under nitrogen fixing conditions are shown in Table 2 with the 40 most up-regulated proteins of each condition. Fumarate-fed cells had a high overabundance of acyl-CoA synthase, three dicarboxylate transporters and proteases. Propionate-fed cells had carbamoyl-phosphate synthase, nitrogenase, SHC and lipid synthase as the most overabundant proteins.

Metabolomics of Organic Acids
The small organic acids present in roots and 21 dpi nodules were formic, acetic, pyruvic, propionic, malonic, malic, fumaric, α-ketoglutaric, succinic, oxalic and citric acids ( Table 3). Among these, citric, malic and oxalic acids were the most abundant, while α-ketoglutaric acid fumaric acid and the monocarboxylates pyruvic acid and acetic acid were those with the highest FCs of 2.1, 2.1, 1.9 and 1.6, respectively, in nodules relative to roots.

Discussion
Frankia symbiotic determinants are still poorly understood mostly due to the absence of an effective genetic transformation protocol. Omics, transcriptomics and proteomics are promising new approaches for the study of plant-bacteria interactions, and have been used to circumvent this drawback in Frankia. A transcriptomic study of 21 dpi symbiotic cells showed upregulation of the genes nif, hup, suf and shc [21]. More recently, it was shown that cellulase and cellulose synthase were upregulated upon early (2 dpi) contact even although Frankia alni feeds neither on glucose nor on cellulose [19], suggesting a role in penetration into the root hair. However, few hints have emerged through omics on the nature of the trophic relations between symbiont and host.
Carbonic anhydrase (CAN) is a zinc-dependent metallo enzyme that catalyzes the reversible reaction of CO 2 with water into carbonic ions and protons, thereby modifying pH and permitting the subsequent action of carboxylases. In animal lungs, carbonic anhydrase converts blood bicarbonate to carbon dioxide that is then exhaled. In the stomach, the CAN of parietal cells produces acid, in the kidneys it helps maintain the blood pH level by forming protons that are then secreted into urine. In plants, CAN helps increase chloroplast CO 2 concentration to facilitate the carboxylation of RuBisCO. Bacteria need to adapt to niches with low pH, for example Helicobacter pylori in the stomach [22] or Mycobacterium tuberculosis in macrophages [23], and CAN has been shown to permit adaptation to these biotopes. An Escherichia coli canmutant can only survive if the atmospheric partial pressure of CO 2 is high or during anaerobic growth in a closed vessel at low pH where copious CO 2 is generated endogenously [24]. Pathogen CANs are also targets of specific enzyme inhibitors and fight antibiotic resistant cells. This mechanism is known to be used against Mycobacterium tuberculosis [23,25], M. leprae [26], Legionella pneumophila [22] and several other bacteria [27]. Thus, CAN activity is a critical feature for these bacteria. Many CAN enzymes have been described with less sequence identity which suggests the convergent evolution of the enzymatic function [28]. Carboxylases play important roles in the cell; for instance phosphoenolpyruvate carboxylase feeds pyruvate into the TCA, propionyl-CoA carboxylase yields methylmalonyl-CoA for feeding succinyl-CoA into the TCA, acetyl-CoA carboxylase yields malonyl-CoA for the lipid synthesis or carbamoyl phosphate synthetase that then permit L-ornithine to transform into L-citrulline.
Propionate and acetate are known antibacterial compounds which are active against a wide range of microbes. Their mechanism of action has been postulated to be a drop in cytoplasmic pH [29] or synergy and potentiation of the anti-bacterial activity of transition metals [30]. They can freely cross membranes when they are not charged (they are neutral at their pKa, 4.8 for acetate, 4.9 for propionate, and much lower at 2.4 for pyruvate) and thus require no dedicated transporter [31] except at low concentrations or at a pH above neutrality [32]. They are used as food preservatives, yet propionate and acetate are the most efficiently used carbon sources in Frankia [33], presumably because Frankia has been selected over millions of years to use them and can counter their toxic properties. One study claimed that one Frankia strain needed a transporter for propionate, but it is probable that the inhibitors used inhibited incorporation into the TCA and the labeled propionate could then simply move freely in and out of the cells [34]. The present study found three expressed dicarboxylate transporters when Frankia was fed the dicarboxylate fumarate but no likely propionate transporter when fed propionate. Furthermore, the onset of nitrogen fixation was much more rapid when Frankia was fed monocarboxylates than with dicarboxylates, suggesting there was no need for the synthesis of a dedicated monocarboxylates transporter [33].
Symbiotic Frankia lives in a sheltered niche, the root nodular cortex, but it nevertheless faces strong physiological constraints. One of those constraints is pH homeostasis as it is modified by several enzymes such as upregulated nitrogenase that requires protons and ATP to reduce dinitrogen to ammonium or to uptake hydrogenase that recycles hydrogen into protons and ATP. Rhizobium, that has similar constraints, has its peribacteroid space made acidic by the host to provide a proton-motive force, presumably to permit uptake of malate and activate proteases [35] and also to trap fixed nitrogen as ammonium and not as gaseous ammonia that would otherwise escape at a pH above 6 [36].
Among Frankia cells, the hyphae and vesicles differ from each other in terms of physiology, with hyphae having a complete ammonia assimilation pathway (GS and GOGAT) and vesicles only having an active GS that would lead to large amounts of glutamine under low O 2 [37]. It has been hypothesized that glutamine could then diffuse toward the plant as a consequence of several peptides which are upregulated in nodules and which increase porosity towards AA [38]. One of the vesicles' main characteristics is to have a multi-lamellate wall with hopanoid lipids providing a passive diffusion barrier to oxygen, yielding an anaerobic environment in which nitrogenase can function [39]. GSII, the primary enzyme responsible for NH 4 + assimilation, was found to have an optimum pH of 6.4 [40], whereas GOGAT and GDH had optimums of 7.6 and 8.6, respectively [37], which is 1 or 2 pH units above that of GSII. This difference in the pH response of the two enzymes could be explained by the more acidic conditions in the vesicles and is consistent with the vesicles only having GSII activity while the hyphae have a complete cycle. In this scenario, the GS would be complemented by carbamoyl-phosphate synthase (CPS) and ornithine carbamoyl transferase (OCT) which use HCO 3 − to incorporate NH 4 + into ornithine and yield citrulline [41]. OCT is also found in the phyloprofile, with only a distant homolog (32% identities) present in non-symbiotic lineages, with its closest relatives being found in the Planctomyces.
Propionate is synthesized through the catabolism of chlorophyll, valine and fatty acids [42]. The synthesis of acetate as acetyl-CoA in plant cells occurs through pyruvate dehydrogenase entering the TCA cycle and yields lipids and flavonoids [43]. Propionate is known to be upregulated as a response to different situations such as drought [44] and high levels of CO 2 . The catabolism of lipids is also known to yield acetate [43]. However, in nodules we see much higher levels of dicarboxylates than of monocarboxylates, suggesting Frankia could be fed a mixture of both types of organic acids, with monocarboxylates also serving to modulate cytoplasmic acidity.
Nodules are known to fix CO 2 at five times the level of adjacent roots [45], with the labeled carbon ending up as amino acids (glutamate, aspartate and citrulline) and organic acids (malate, citrate and succinate) through the action of PEP carboxylase or pyruvate carboxylase, an enzyme that would be active when the carbon source is pyruvate. Frankia strain ArI3 in pure culture was also shown to fix more nitrogen when supplied supplementary CO 2 , but only when fed monocarboxylates [46] and not TCA intermediates such as succinate, malate, etc. Pyruvate is assimilated through a dedicated carboxylase and transformed into oxalacetate; propionate is assimilated through the propionyl CoA carboxylase and transformed into succinate. Both oxalacetate and succinate use biotin, and acetate is assimilated through acetyl-CoA carboxylase into malonyl-CoA and lipids. Another possible pathway for the assimilation of propionate is the methyl citrate cycle that is also present in symbiotic strains and absent in non-symbiotic strains. This pathway does not require carboxylase and could function to complement the methyl-malonyl pathway. It has long been assumed that the host feeds Frankia with dicarboxylates since the expression of AgDCAT, a transporter specific for dicarboxylates, was nodule-specific [47]. However, monocarboxylates could also contribute to the nutrition of Frankia since pyruvate, for instance, is present in the roots and nodules of Casuarina and Alnus [48]. The transport of monocarboxylates appears to be facilitated in Coryebacterium by a specific sodium solute symporter at a high pH, but import can nevertheless occur through simple diffusion at a neutral pH [32]. When Frankia is fed a dicarboxylate, it upregulates three transporters, all of which were not up-regulated in 21 dpi nodules [21].
The most overabundant proteins in propionate-fed in-vitro-grown cells were CPS subunit, proteins that use ammonium, phosphate and HCO 3 − to yield carbamoyl that could then be used to transform L-ornithine into L-citrulline, the compound shown to carry the most labels after 15 N has been fed to A. glutinosa [49]. This could be a pathway for the assimilation of ammonium which is complementary to the GS-GOGAT pathway. Frankia has been shown to be able to use L-ornithine as source of nitrogen without inhibition of nitrogenase. This could also serve to select efficient nitrogen fixers that can provide NH 4 + to quench carbonate and maintain a slightly acidic pH.
The genes encoding CANs are differentially expressed in nodules (FRAAL1222: 1.20; FRAAL2078: 0.86; FRAAL4724: 1.15; FRAAL4889: 0.71, relative to BAP + , [21]), a process which is compatible with the requirement for CO 2 when grown on propionate in vitro and the very high efficiency of the CAN that have a low Km [50]. It is also compatible with the host feeding the symbiont a mixture of dicarboxylates and monocarboxylates. The amount of organic acids measured in this study and by previous authors [33,48] do lead to a certain conclusion in relation to the photosynthates fed to the microsymbiont because of likely compartmentation in the mitochondria, chloroplasts and vacuoles. Nevertheless yet these organic acids are good candidates for the role.
Carbonic anhydrase would thus be added to the list of symbiosis-related genes such as those coding for nitrogenase, hopanoid biosynthesis, iron-sulfur cluster synthesis, hydrogen-uptake [21] and cellulase and cellulose synthase [19]. Similar to hydrogenuptake and cellulase, there are copies that have diverged early in the evolutionary history of the genus. Whenever genetic editing becomes practical, it will be of interest to assay this symbiosis determinant.

Genome Mining
Genome comparisons were made on the Mage platform [51]. Conserved genes present in representative genomes from clusters 1, 2 and 3 were extracted, and those present in non-symbiotic lineage 4 were subtracted using as the threshold 50% of identities over 80% of the length of the shorter AA sequence. For cluster 4, only CN3 [52] and DC12 [5] were used because these are unable to induce nodules in any of the hosts tested other than EuI1c and AgB1.9 that can induce inefficient nodules [53]. A list of the strains and a flow chart of research directions are given in the Supplementary Materials (see Supplementary Materials Table S1 and Figure S1).

Plant Growth
Alnus seeds were surface-sterilized and germinated as described previously [19]. Seedlings were transferred to opaque pots, grown with N-free medium and inoculated with Frankia alni as before [1].

Culture of Cells and Physiology Measurements of Cells
Growth, N-fixation and CO 2 levels were followed in FBM medium [54], both with and without 5 mM ammonium, and in the same medium with 5 mM fumarate instead of propionate. Growth was monitored as previously [33] and CO 2 in the headspace was measured in the flask every day using a gas chromatograph (Micro GC R3000, SRA Instrument, Marcy L'Etoile, France) as described previously [55].

Microscopy
The staining of nodule slices and FBM− grown Frankia was carried out with DND-99 probes (LysoTracker Red) or DND-189 (LysoSensor Green, Molecular Probes, Eugene, OR, USA) in a 50 mm Tris pH = 7 buffer. Excess dye was removed by a short wash with buffer and samples were observed using a confocal microscope (LSM510 META, Carl Zeiss, Le Pecq, France) using a plan-Apo 63×/1.2 NA water immersion objective.

Phylogenomics
The amino acid sequences of the carbonic anhydrase genes from the representative Frankia species and related actinobacteria were retrieved using the BlastP method [26]; and were aligned using the Seaview platform [56] implementing the ClustalW2 v2.1 software [57]. The alignment was then used for a Maximum Likelihood tree reconstruction approach [58] with a bootstrap of 1000 replicates [59] to assess the robustness of the topology obtained.

Gene Expression
Expression analysis of the carbonic anhydrases FRAAL1222, can1 and FRAAL4889, can2; carbonate transporter FRAAL4724, ctp and two carbamoylphosphate synthase genes, FRAAL4883, (cspA) and FRAAL4884, (cspB), was carried out with cells grown on FBM+ medium [54] with 5 mM propionate and with 5 mM ammonium chloride and cells grown on FBM− medium without ammonium chloride and in 21 dpi nodules [21]. This was monitored by qPCR using the primers that are listed in Table 4 and following the procedure described previously [21].

Proteomics
Proteomic analysis of 5 mM propionate-fed and 5 mM fumarate-fed Frankia cells grown in BAP-was carried out as described before [60]. The mass spectrometry proteomics data have been deposited in the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifiers PXD018045 and 10.6019/PXD018045.

Metabolomic of Organic Acids
The organic acids in the nodules and roots were quantified using 22-day-old nodules grown as described in Alloisio et al. [31]. These were extracted from liquid nitrogen ground tissues. Sterile ultrapure water was added to achieve a concentration of 1 g of nodule per mL of water. Samples were centrifuged at 13,000 rpm in Eppendorf centrifuge for 10 min.
After extraction, the supernatants of the nodules or roots were diluted (1:20) in ultrapure water and filtrated on a 0.45 µm PVDF filter prior to being analysed by ionic chromatography and tandem mass spectrometry (IC-MS/MS). IC-MS/MS was used to verify the presence of organic acids already quantified by IC-conductivity and to identify other unknown organic acids.
The IC-MS/MS instrument involved was the ICS-5000 + Ion Chromatography System combined with a TSQ Fortis Triple Quadrupole Mass Spectrometer (all Thermo Scientific, Waltham, MA, USA). The monitoring software was Chromeleon 7.2.10. The IC system was equipped with an AS-AP auto-sampler, a Dual Pump analytical gradient system, a suppressor ADRS 600 (2 mm, Thermo Scientific) and a conductivity detector (P/N 061830). The IC separation was realized using an anion-exchange column IonPac AS11-HC-4 µm (2 × 250 mm 2 , Thermo Scientific, Illkirch, France) preceded by a guard column IonPac AG11-HC-4 µm (2 × 50 mm 2 , Thermo Scientific) and with the same parameters as the IC-conductivity analysis. Concerning the mass spectrometer, the heated electrospray ionization source was used in negative mode with the following parameters: negative ion spray voltage 3.0 kV, sheath gas 52 Arb, auxiliary gas 13 Arb, sweep gas 16 Arb, ion transfer tube temperature 275 • C and vaporizer temperature 375 • C. Acquisitions were realized in full scan mode or in Single Ion Monitoring mode with analytes detected after deprotonation in ionization source.
The flow chart of Materials and Methods is presented in Supplementary Figure S1.

Funding:
The work was funded by the French ANR (BugsInACell ANR-13-BSV7-0013-03) to P.N. and by an MEC postdoctoral fellowship from the Spanish government awarded to L.C. (Programa Nacional de Movilidad de Recursos Humanos del Plan Nacional de I-D+i 2008-2011).

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The original contributions generated for this study are included in the article; further inquiries can be directed to the corresponding authors.