Extracellular vesicles released by the human gut symbiont Bacteroides thetaiotaomicron in the mouse intestine are enriched in a selected range of proteins that inuence host cell physiology and metabolism

It is becoming increasingly clear that bacterial extracellular vesicles (BEVs) produced by members of the 2 intestinal microbiota can contribute to microbe-host cell interactions that impact on host health. A major 3 unresolved question is the nature of the cargo packaged into these BEVs and how they can impact on host 4 cell function. Here we have analysed the proteome of BEVs produced by the major human gut symbiont 5 Bacteroides thetaiotaomicron in both in vitro cultures using defined and complex medias, and in vivo in 6 fed or fasted animals to determine the impact of nutrient stress on the BEV proteome, and to identify 7 proteins specifically enriched in BEVs produced in viv o. In contrast to BEVs produced in vitro where 8 limiting nutrient provision resulted in an increase in a large fraction of proteins, the protein content of BEVs 9 extracted from fasted versus fed mice was less affected with similar numbers of proteins showing increased 10 and decreased abundance. We identified 102 proteins exclusively enriched in BEVs in v i vo of which the 11 majority (66/102) were enriched independently of their expression in the parent cells implicating the 12 existence of an active mechanism to drive the selection of a group of proteins for their secretion into BEVs 13 within the intestine. Amongst these abundantly expressed proteins in BEVs in vivo were a bile salt hydrolase 14 and a dipeptidyl peptidase IV that were characterised further and shown to be active and able to degrade 15 host-derived substrates with defined roles in metabolism. Collectively these findings provide additional 16 evidence for the role of BEVs in microbiota-host interactions with their contents playing key roles in the 17 maintenance of intestinal homeostasis, and host metabolism.


Introduction 20
The human gastrointestinal (GI) tract accommodates a microbial community (the microbiota) comprising 21 trillions of cells that carry out vital functions for human health. Increasing our understanding of the basis 22 of this mutualistic relationship and its impact on human health and disease is dependent on defining the 23 pathways and mediators of microbiota-host crosstalk. Many studies have identified the importance of 24 microbe-and host-derived soluble factors in this crosstalk [reviewed in 1,2]. More recently, another 25 pathway of host-microbe crosstalk has been identified that involves bacterial extracellular vesicles (BEVs) 26 [3], which contain various macromolecules with the potential of contributing to interactions with other 27 members of the microbial community but also with host cells [4][5][6][7]. 28 29 BEVs represent a novel secretion system enabling the dissemination of membrane-encapsulated cellular 30 materials including proteins, nucleic acids and metabolites into the extracellular milieu [8,9] and beyond 31 [7]. These include membrane vesicles (MVs) produced by Gram-positive bacteria, and outer membrane 32 vesicles (OMVs) and outer-inner membrane vesicles [10-12] produced by Gram-negative bacteria. BEVs 33 produced by pathogenic bacteria have historically been the most intensively investigated. The animal GI 34 1 mL with a Vivaspin 20 centrifugal concentrator (100 kDa molecular weight cut-off, Sartorius) and filtered 18 through a 0.22 µm PES membrane (Sartorius). Vesicle concentration was determined by Nanoparticle 19 Tracking Analysis (NTA). The volume of the retentate was adjusted to 8.9 mL and the BEV suspension 20 centrifuged (150,000 g at 4 o C or 2 h in a Ti70 rotor (Beckman Instruments)). After centrifugation, the 21 supernatant was removed using a vacuum pump and the vesicle pellets snap frozen in liquid nitrogen and 22 stored at -80°C prior to extraction. 23 24

Proteomics 25
Comparative proteomics was carried out on samples of BEVs produced in BHI versus BDM and from 26 BEVs and EVs isolated from the caecum of fed or fasted mice. For the in vitro experiments vesicles were 27 isolated (above) from 3 independent cultures for each culture medium. One of the samples obtained in BDM 28 was excluded from further analysis as it produced anomalous results. In the comparison of BEVs generated 29 in vivo 5 mice were used for each condition providing 10 datasets including ratios (fasted versus fed) for 30 each protein identified with the level of confidence determined by the false discovery rate (FDR), that were 31 then further analyzed. Parental cells were from BHI cultures or the caecum of Bt colonised mice (3 32 replicates for each condition). 33 34 Samples for proteomics analysis consisted of 100 ug of BEV or cell protein extract prepared and labelled 1 at the Bristol University proteomics facility using TMT reagents (10-Plex format, Isobaric Mass Tagging 2 kit, Thermo Scientific). Labelled samples were pooled and then fractionated using High pH Reverse Phase 3 Liquid Chromatography. The resulting fractions were subjected to nano-LC MSMS using an Orbitrap 4 Fusion Tribrid mass spectrometer with an SPS-MS3 acquisition method. Fragmentation of the isobaric tag 5 released the low molecular mass reporter ions which were used to quantify the peptides. Protein quantitation 6 was based on the median values of multiple peptides identified from the same protein, resulting in highly 7 accurate protein quantitation between samples. The data sets were analysed using the Proteome Discoverer 8 v2.1 software and run against the Bt VPI-5482 or mouse database and filtered with a 5% (1%) FDR cut-9 off. 10 11

Proteomics data curation 12
BEVs versus parent cells produced in vitro and in vivo: from 3092 listed proteins of the raw results to 2047. 13 A hundred contaminant proteins (FALSE) were removed from the data. Using the 99% confidence level 14 (<1% FDR), 213 additional proteins were removed. Proteins that were not found in BEVs (732) were also 15 removed from the list resulting in a total of 2047 identified proteins. For the abundance ratio of BEV 16 proteins (mouse caecum versus BHI) those with a ratio ≥ 15 and a PSMs ≥ 10 were retained, excluding 17 proteins that were not identified in the fasted versus fed animal experiment, resulting in a total of 102 18 proteins. To discriminate between proteins that are enriched in BEVs in vivo, the 36 proteins with an 19 abundance ratio in the cell lysate (mouse caecum versus BHI) ≥ 5 were considered as non-enriched whereas 20 the 66 proteins with an abundance ratio in cell lysates (mouse caecum versus BHI) ≤ 5 were considered 21 enriched in BEVs. 22

Gene ontology analysis 24
The proteins were categorized according to species specific gene ontology (GO) annotations using 25

Electron microscopy 28
Cells were grown in BHI to early stationary phase and visualised by negative staining electron microscopy. 29 2 µL of liquid culture were applied to a 600-mesh copper TEM grid coated with formvar/carbon. The 30 sample was left to settle out for 5 minutes and 2 µL of 2x fixative (5% glutaraldehyde in 200mM sodium 31 cacodylate buffer, pH 7.2) was added and left for 5 minutes. The grid was then immersed for 10 minutes in 32 10 µl of 1x fixative, washed 5 times with100mM sodium cacodylate buffer, pH 7.2 and 5 times with 33 ultrapure water (1 minute each). The grid was air dried before negative staining in 2% aqueous Uranyl 34 acetate-stain was applied and removed immediately. Grids were air dried and viewed in a Jeol 1230 TEM 1 operated at an accelerating voltage of 80kV. Images were recorded on a Gatan One View 16MP digital 2 camera. 3 Pellets of BEVs (including EVs for immunogold staining) were resuspended and fixed by vortex and 4 pipetting in 100 µL 2.5% Glutaraldehyde in 0.1M PIPES buffer. Large aggregates of material still present 5 upon pellet resuspension, were removed by centrifugation for 2 min at low speed (5 g). A 50 µL portion of 6 supernatant was mixed 1:1 with cooled molten 4% low gelling temperature agarose (TypeVII, Sigma), 7 solidified by chilling and cut into approximately 1 mm 3 pieces. The BEV sample pieces were transferred 8 into glass vials for further fixation in 2.5% glutaraldehyde in 0.1M PIPES buffer overnight at 4°C. Fixed 9 BEV sample pieces were washed in 0.1M PIPES buffer (3x) and post-fixed in 1% OsO 4 (Agar Scientific) 10 for 2 h. Following OsO 4 fixation, samples were washed in deionised water (3x), followed by dehydration 11 through an ethanol series (30, 50, 70, 90, 3x 100%). The samples were infiltrated with a 1:1 mix of 100% 12 ethanol to LR White medium grade resin, followed by a 1:2 and a 1:3 mix of 100% ethanol to LR White 13 resin and finally 100% resin, with at least 1 h between changes. The resin was changed twice more with 14 fresh 100% resin with periods of at least 8 h between changes. The sample pieces were each transferred 15 into BEEM embedding capsules with fresh resin and polymerised for 24 h at 60°C. Sections approximately 16 90 nm thick were cut using an ultramicrotome (Ultracut E, Reichert-Jung) with a glass knife, collected on 17 Cu Formvar/carbon grids and stained sequentially with 2% uranyl acetate solution for 1 h at 21 o C, and 0.5% 18 lead citrate solution for 1.5 min at 21 o C. Deionised water washes were performed (5x) following each of 19 the staining steps. Sections were examined and imaged in a FEI Talos F200C transmission electron 20 microscope at 200kV with a "Gatan One View" digital camera. For immunogold staining, a "short" version 21 of the Aurion Immunogold labelling (IGL) protocol 22 (http://www.aurion.nl/the_aurion_method/Post_embedding_conv) was used with 1h antibody incubations 23 and detergent (0.1% TWEEN). The primary antibody (anti-OmpA) was diluted 1/500 and the secondary 24 antibody (GAR-10, Agar Scientific, Stanstead, UK) was diluted 1/50. After antibody labelling, the sections 25 were stained with 2% uranyl acetate for 40 min. The sections were examined and imaged in a FEI Tecnai 26 G2 20 Twin transmission electron microscope at 200 kV. 27 28

Construction of a BT_2086 deletion mutant 29
An 899 bp chromosomal DNA fragment upstream from BT_2086 and including the first 30 nucleotides of 30 its 5ʹ-end region was amplified by PCR using the primer pair f-5ʹbsh1_SpHI, r-5ʹbsh1_SalI. This product 31 was then cloned into the SpHI/SalI sites of the E. coli-Bacteroides suicide shuttle vector pGH014 [22]. A 32 900 bp chromosomal DNA fragment downstream from BT_2086, including the last 44 nucleotides of the 33 3ʹ-end region, was amplified by PCR using the primer pair f-3ʹbsh1_BamHI, r-3ʹbsh1_SacI and was cloned 34 into the BamHI/SacI sites of the pGH014-based plasmid. The resulting plasmid containing the 1 ΔBT_2086::tetQ construct, was mobilized from E. coli strain GC10 into Bt by triparental filter mating [24], 2 using E. coli HB101(pRK2013) as the helper strain. Transconjugants were selected on BHI-haemin agar 3 containing gentamicin (200 mg/L) and tetracycline (1 mg/L). Determination of susceptibility to either 4 tetracycline or erythromycin was done to identify recombinants that were tetracycline resistant and 5 erythromycin susceptible after re-streaking transconjugant bacteria on LB-agar containing tetracycline or 6 both antibiotics. PCR analysis and sequencing were used to confirm allelic exchange. A transconjugant, 7 GH511 containing the ΔBT_2086::tetQ construct inserted into the Bt chromosome was selected for further 8 studies.

Impact of nutrient availability on BEV biophysical characteristics and hydrolytic enzyme content 3
Bt produces large amounts of uniform BEV particles which are released from the bacterial cell surface into 4 the external milieu (Fig. 1). To test whether environmental factors have an impact on BEV structure, 5 production and protein composition, Bt was cultured in either a complex (BHI) or defined and minimal 6 (BDM) media and BEVs isolated from the culture supernatants. BEV concentration harvested from Bt 7 grown in BHI and BDM media were similar while their average size increased from 135 ± 6 nm in BHI to 8 205 ± 3 nm in BDM (Fig. 2a). Electron microscopy imaging confirmed that BEVs from BHI and BDM 9 were similar in appearance and structure although those produced in BDM were larger in size (Fig 2b). The 10 average zeta potential of BEVs from both BHI and BDM media was -25 mV and -22 mV (in PBS, pH 7.2, 11 25°C) respectively, which is similar to what was reported for E. coli-derived BEVs [27]. 12

13
The proteomic profile of BEVs produced in BHI versus BDM cultures were compared by differential 14 proteomic analysis. In general, 1,438 proteins were identified corresponding to approximately 30% of the 15 predicted proteome of parent cells [28]. Of note, the majority of proteins were more abundant in BEVs 16 produced in nutrient-poor, BDM, medium (Fig. 3a). Proteins categorized according to universal gene 17 ontology (GO) annotations showed that many of the proteins displaying an increase abundance (fold change 18 > 3) were hydrolases, and in particular glycoside hydrolases, and proteases in addition to transferases, 19 oxidoreductases, ligases and lyases (Fig. 3b). Whereas the complete predicted proteome of Bt is composed 20 of 64% acidic proteins (pI < 7.4, physiological pH), 79% of the BEV proteins were acidic, confirming the 21 enrichment of acidic proteins in BEVs (data not shown). 22

23
The Bt transcriptome in response to nutrient availability was previously investigated by microarray analysis 24 [29] using probe pairs derived from 4,779 predicted genes to compare transcriptional profiles obtained from 25 Bt grown in rich versus minimal medium (with glucose as the sole carbon source) during early log phase to 26 stationary phase. Accordingly, we selected the 250 most abundant proteins that were more, or less, abundant 27 in the defined medium compared to the rich medium (highest peptide spectrum match PSM) with the level 28 of expression of their corresponding genes determined under the same conditions (BDM-G). The results 29 showed that the fold differences in protein and the corresponding RNA expression correlated with each 30 other ( Fig. 3c) (Spearman correlation coefficient r s = 0.44, p = 2.10 13 ) indicating that BEV protein content 31 reflected RNA levels in the parental cell. 32

33
Our analysis also showed that there was a corresponding increase in the metabolites generated from 1 reactions catalysed by the more abundant proteins present in BEVs produced in BDM (Table 1). 2 Intriguingly, the concentration of the substrates specific for these enzymes [9] was also increased in BEVs 3 indicating that some of the reactions are reversible and/or the diffusion of the substrate from the external 4 milieu into BEVs is facilitated by the presence of higher levels of enzyme. 5 6 Since the BEV protein content was affected by nutrient availability in vitro, we investigated if nutrient 7 deprivation in vivo (in fasting animals) could similarly lead to changes in the BEVs' proteome. 8 9

BEV proteomic profile in vivo 10
To assess whether nutrient deprivation affects the protein composition of BEVs produced by Bt in the GI 11 tract, germfree mice were orally gavaged with Bt with one group of conventionalised mice allowed 12 unrestricted access to food and water with a second group being deprived of food for 16 h. BEVs extracted 13 from the caecum were equivalent in size from both fed or fasted mice with a mean size of approximately 14 190 nm when measured with a NanoSight instrument (Fig. 4a). By contrast, 1.8 times more nanoparticles 15 were recovered from the caecum of fasted mice compared to fed animals (Fig. 4a). The presence and 16 identity of Bt BEVs in mouse caecal preparations was confirmed by immuno-EM using an antiserum 17 specific for the outer membrane protein OmpA (BT_3852) of Bt (Fig. 4b).  (Table S1) 5 and were classified using the Polysaccharide-Utilization Loci DataBase (PULDB) 6 http://www.cazy.org/PULDB/ [32] (Table S1). The starch degrading PUL66 was highly abundant which 7 most likely reflects the high (~34%) starch content of the animal chow. PULs involved in the degradation 8 of rhamnogalacturonan-II (PUL77), pectic galactan (PUL86) and arabinogalactan (PUL65) were also 9 highly abundant. Of note, in fasted mice there was an increased abundance of PULs capable of degrading 10 host glycans and mucins (PULs 6, 19, 35, 37, 80 and 81). 11

A set of proteins is selectively secreted in BEVs in the GI tract 13
To investigate whether proteins are selectively enriched in BEVs in vivo, we first compared the proteome 14 of BEVs harvested from the caecum of Bt mono-colonised germfree mice with that of BEVs generated in 15 vitro in BHI media. A total of 102 proteins were identified based upon the abundance being at least 15-fold 16 higher in in vivo generated BEVs (Table S2,  of the corresponding gene were significantly correlated [Spearman correlation coefficient r s = 0.81 (p < 2 0.0001)] ( Fig. 5a and b, Table S2). These results are consistent with the selective enrichment of a set of 3 (102) protein in BEVs in vivo that can occur in parallel with (36) or independently of (66) changes in protein 4 production in parental cells. 5 6 The enrichment of the 102 proteins in BEVs generated in vivo was independent of food intake and 7 availability since the abundance of these proteins was comparable in BEVs from fasted versus fed mice 8 (fold change  1.0) (Fig. 5a, b). 9 10 We used the SignalP-5.0 Server software programme to predict the presence of known Gram-negative 11 bacteria signal peptides amongst the 102 proteins enriched in BEVs (Fig. 5c). Most of the 36 proteins whose 12 abundance values were 15-fold or higher in BEVs in vivo versus in vitro were predicted to be transported 13 via the bacterial Sec-dependent protein secretion system. By contrast, 31/66 of the proteins displaying a 5-14 fold or less increase in abundance in parent cells were predicted to be secreted independently of known 15 (Sec) bacterial secretion systems. 16 17

Enrichment of bile salt hydrolases and dipeptidyl hydrolases IV in BEVs in vivo 18
Prominent among the 36 proteins enriched in both BEVs and parent cells in vivo (Table S2)  and TCA (Fig. 6). The specificity of the BSH encoded by BT_2086 was established by generating a Bt 23 mutant lacking BT_2086 (BT2086). The mutant was unable to degrade TCA whereas hydrolysis of GCA 24 to produce cholic acid was unaffected most likely reflecting the activity of the other predicted BSH, 25 BT_1259 [28]. In the case of BEVs, levels of bile salt hydrolase activity and CA production were lower 26 than that of parent cells as reflected in higher residual levels of GCA and TCA after incubation with BEVs 27 (Fig. 6). Despite this it was clear that BT2086 generated BEVs produced strikingly less CA compared to 28 BEVs from wild type Bt. These findings indicate that BEVs produced by Bt contain a BSH able to 29 deconjugate tauro-conjugated bile salts. 30

31
The dipeptidyl-peptidase 4 (DPP4)-like protein (DPP6) encoded by BT_1314 was also abundant in BEVs 32 in vivo (Table S2). Human DPP4 or CD26 truncates proteins containing the amino acid proline or alanine 33 in the second position of the N-terminus, and DPP-4-like activity encoded by the intestinal microbiome has 34 been proposed to constitute a novel mechanism to modulate protein digestion and host metabolism [34]. 1 We tested therefore whether intact BEVs could hydrolyse DDP4 specific substrates (H-Ala-Pro-p-2 nitroaniline) [26]. BEVs produced in vitro in BDM exhibited activity of 0.57 nmol/min/mg BEV total 3 protein whereas BEVs isolated from BHI exhibited activity of 0.09 nmol/min/mg. This agrees with the 4 abundance ratios measured for BT_1314 which was 6 times more abundant from BEVs obtained in BDM 5 compared to those obtained in BHI. The same was also true for two other putative DPP4 enzymes detected 6 in BEVs; BT_3254 (3 times more abundant from BDM) and BT_4193 (4 times more abundant from BDM). 7 However, in vivo, BT_1314 was selectively enriched (~20-fold) in BEVs. proteins contributing to multi-organism process and the immune system were also increased in caecal EVs. 31 In comparing the abundance ratio for each protein contained in caecal EVs derived from fasted versus fed 32 animals (Table 3), amongst EVs produced in fasted mice two serine protease inhibitors (A3M and A3K) 33 were more abundant (5-fold and 2.7-fold, respectively). We also observed a 3.5-fold increase in the 1 abundance of the murine specific α-defensin CRISC-2 in EVs produced in fasted mice. 2 3 Discussion 4 Our study provides new insights into microbe-host interactions in the mammalian GI tract and how BEVs 5 can contribute to this crosstalk. Using the ubiquitous human commensal gut bacterium Bt as a model 6 system, we have shown that the profile of proteins it packages into BEVs is influenced by nutrient 7 availability, and provided evidence of the selective and exclusive enrichment of proteins in BEVs in vivo 8 in the mouse GI tract that include enzymes capable of influencing host metabolism. in BEVs compared to their parental cells. In addition, as the levels of these proteins were comparable in fed 29 versus fasted animals the process responsible for the accumulation of these proteins into BEVs functions 30 independently of nutrient (food) supply (Fig. 5a). Thus, local environmental factors other than diet and 31 nutrient supply are involved in the selection and secretion of a set of proteins into BEVs. Furthermore, 32 based upon the known functionality of some of these proteins (e.g. dipeptidyl-peptidase and asparaginase) 33 they are most likely selectively packaged into BEVs by the bacterium with the purpose of influencing host 34 cell physiology and in particular, metabolism. The mechanisms that account for this enrichment of proteins 1 in BEVs is unknown and likely to involve unique processes as nearly half of the proteins enriched in BEVs 2 are not predicted to be secreted by a known bacterial secretion system. 3 4 The dipeptidyl-peptidase encoded by BT_1314 is enriched in BEVs in vivo and has the potential to influence 5 host physiology via its effect on protein and glycan (e.g. gluten) digestion, signal transduction and apoptosis 6 [34]. Based upon its ability to cleave and inactivate various signalling molecules important in metabolism 7 (i.e. incretins), the immune system (i.e. growth factors and cytokines) and CNS (i.e. neuropeptides) [48-50] 8 it is tempting to speculate that upon accessing the systemic circulation [7] BEVs can impact on various 9 aspects of host physiology and behaviour, a possibility that awaits confirmation from further studies. The 10 type II L-asparaginase encoded by BT_2757 was also selectively enriched in BEVs in vivo. Asparaginase 11 activity is required to deamidate asparagine to aspartate, an essential amino acid for proliferating 12 mammalian cells (e.g. cancer cells) and as a neurotransmitter [51,52]. The human asparaginase enzyme 13 (ASPG) exhibits a relatively low affinity for L-asparagine while bacterial enzymes, that are commonly used 14 as anticancer drugs, have a higher affinity for the substrate [53]. Indeed, E. coli-derived asparaginase is 15 in the low nutrient culture medium BDM. Furthermore, we did not observe an increase in host glycan-33 specific and surface-exposed glycohydrolases in BEVs from fasted animals. Indeed, their abundance 1 exhibited a downward trend (up to 2-fold) compared to their levels in BEVs from fed animals. 2 3 It is interesting to note that for all PULs the abundance of the integral membrane oligosaccharide importer 4 SusC is increased by about fifty percent in BEVs produced under fasting conditions whereas for the other 5 proteins belonging to the same PULs, including SusD (nutrient binding accessory protein) they are equally 6 abundant or less abundant under fasting versus fed conditions. The glycosyl hydrolases associated to PUL 7 systems which are preferentially packaged into BEVs [45] can provide substrates to support the growth of 8 other bacteria in animals harbouring a conventional microbiota, conferring a "public good" function to 9 BEVs [13, 61]. 10

11
As part of this study we established the proteome profile of mammalian EVs in the mouse intestine. In 12 comparing the abundance ratio for each protein contained in caecal EVs derived from fasted versus fed 13 animals (Table 3), EVs produced in fasted mice contained two serine protease inhibitors (A3M and A3K 14 serpins) with their abundance increased 5-fold and 2.7-fold, respectively. Moreover, 7 additional serpins 15 (protease inhibition activity, InterPro family IPR000215) were identified with similar abundance in fasted 16 and fed mice. It has been reported that high protease activity measured in the feces of patients suffering 17 from irritable bowel syndrome correlates with a decrease in microbial diversity [62]. It is therefore tempting 18 to speculate that the various serine protease inhibitors detected and identified in EVs produced in mice 19 mono-colonised with Bt (submitted to Vesiclepedia, number) is a consequence of the lack of microbial 20 diversity, and is to counteract the detrimental effect of proteases present in high abundance in the gut lumen 21 [62] of mono-colonised mice. 22

23
We also compared the protein profile of mouse caecal EVs with that obtained from EVs of cultured human 24 primary and metastatic colorectal cancer cells [38]. Human and mouse small intestines share many 25 similarities in their intestinal microbial defence strategies, including production of α-defensins which are 26 also found in EVs [35]. Mice, however, produce a unique antimicrobial peptide and member of the CRS 27 (cryptdin-related sequences)-peptide family, not found in man [63]. We observed a 3.5-fold increase in the 28 abundance of the CRISC-2 α-defensins in EVs produced in fasted mice. Whether a decrease in nutrient 29 availability in the mouse intestine leads to increased expression of CRISC-2, to an increased number of the 30 secretory Paneth cells and/or to CRISC-2 preferentially packaged into EVs still needs to be determined.  Table S1 and Table S2.   (Table S2) are combined in (a), and results of the 36 proteins with a greater than 5-fold change 32 in the parent cells in vivo (Table S2) Figure 1 Release of BEVs from the cell surface of Bt into the external milieu. The cells were grown in BHI to early stationary phase and visualised by negative staining electron microscopy.     (Table S2) are combined in (a), and results of the 36 proteins with a greater than 5-fold change in the parent cells in vivo (Table S2)   at 37oC after which supernatants were spotted onto a silica gel plates. The plate was inserted into a TLC chamber, run for 40 min and stained with phosphomolybdic acid.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. TableS1.pdf TableS2RGS.xlsx