Extracellular vesicle formation in Euryarchaeota is driven by a small GTPase

Significance Extracellular vesicles (EVs) play important roles in intercellular communication by transferring proteins, nucleic acids, lipids, and metabolites between cells. Few studies have investigated their role in Archaea. Here, we show that EVs of halophilic Archaea (haloarchaea), members of the Euryarchaeota, transfer an RNA cargo enriched in noncoding RNAs (ncRNAs), likely contributing to intercellular communication. We show that EV formation in haloarchaea is driven by a small guanosine triphosphatase (GTPase), ArvA, that is also conserved across other archaeal lineages, along with two genes closely associated with arvA that are also involved in vesicle production. Our work provides important insights into small GTPase-driven vesicle formation and a basis for further studies into the evolutionary relationships between prokaryotic and eukaryotic vesicle formation.


EV-associated RNA is best analyzed when using small RNA libraries and normalizing EV RNA content with host cell RNA content
To determine the nature of the EV enclosed RNA, total RNA and small RNA (enriching for transcripts below 150 nt in length) libraries were prepared from EV-extracted RNA.When comparing sequencing results from both libraries, we observed a drastically different transcriptional profile (Supplementary Table 3).Around 95% of reads from total RNAseq mapped to ribosomal RNA (rRNA) and only 17 transcripts recruited enough reads to reach the threshold (TPM > 10).In contrast, the small RNA library revealed a more diverse array of transcripts with transfer RNA (tRNA) being the most dominant RNA species (around 85% of reads) and 264 transcripts identified within the threshold.Further, over 2000 transcripts were only identified in the small RNA library and not in the total RNA library, the majority of them being tRNAs and non-coding RNAs (ncRNA), indicating that the total RNA library excludes important smaller transcripts.
Therefore, we decided to use small RNA libraries for further analysis of EV-associated RNAs, as this seems to yield a more accurate picture of the RNA composition of EVs.
We also compared transcripts from EVs in the upper and lower bands in density gradients to determine whether the bands represented different subpopulations of EVs with respect to RNA content (Supplementary Table 4).Indeed, we identified transcripts that were only present in the upper band (app.200) or only in the lower band (80).However, the abundance of these transcripts was below the threshold (TPM > 10) and they were disregarded.Overall, the RNA composition between both bands was mostly identical, with few outliers.We concluded that the RNA composition alone is not the differentiating factor between the two subpopulations of EVs in different density gradient bands, and pooling the bands for further sequencing analysis is acceptable.
Since vesicle production could be linked to UV exposure [1], we aimed to determine whether subjecting cultures to UV radiation would alter the RNA composition of EVs.Indeed, 145 transcripts appeared to be present in a higher abundance (log2 fold change > 1) in the UV-treated sample, and 32 transcripts were present in a higher abundance (log 2 fold change < -1) in the untreated sample (Supplementary Table 5).This population of EV-associated RNA from UV-treated cultures included all forms of RNA, including mRNAs, tRNAs, rRNAs and ncRNAs.Nevertheless, we realized that without determining transcriptional changes within the cells, it is difficult to distinguish transcripts that are associated with EVs as a response to UV exposure and transcripts that are present in EVs simply due to changes in intracellular levels.In order to differentiate between random packaging and potentially selective packaging of RNA into EVs, it became clear that sequencing intracellular RNA at the time of EV harvesting was imperative for any kind of analysis.

Analysis of EV-associated RNA under infection with a virus reveals viral transcripts associated with EVs
Direct interactions between EVs and viruses have been documented, demonstrating the capacity for EVs to act as a viral defense mechanism [2] or to facilitate viral propagation [3].While we detected only minor changes to EV quantities under infection with the virus HFPV-1 (Figure 2B), we wanted to test whether infection with HFPV-1 would influence the RNA composition of EVs and thereby possibly indirectly influence virus-host interactions.
While it was shown that infection with HFPV-1 drastically altered the transcriptomic landscape of the cell during exponential and early stationary growth [4], sRNAseq in late exponential growth revealed a nearly identical transcriptional profile when comparing infected and uninfected cells (Supplementary Figure 6A).
Only two genes showed a significant upregulation (log2 > 1) in the infected cells, HVO_2657 and HVO_0272; however, both are in general rather weakly expressed (TPM < 15).When comparing the RNA content of EVs between infected and uninfected cells, two transcripts were found to be significantly higher in abundance in EVs of infected cells: HVO_A0466 and HVO_0272 (Supplementary Figure 6A).While HVO_0272 mRNA was about 4-fold upregulated in infected cells (log2 ~ 2), it was about 10-fold upregulated (log2 ~ 5) in EVs of infected cells (Supplementary Figure 6B), indicating that the packaging of this transcript into EVs increases significantly upon infection.Surprisingly, it appeared that the majority of reads mapping to HVO_0272 only map to two short regions of about 30 nt within the coding region of the gene that are identical to a region on the viral genome.Therefore, we conclude that the upregulation of HVO_0272 is due to viral transcripts mapping to the host genome.
Subsequently, when mapping reads to the virus genome, we detected a significant amount of viral transcripts in EVs.While only 1.7 ± 0.07% of intracellular RNA mapped to the HFPV-1 genome, 4.0 ± 0.10% of EV-associated RNA mapped to the viral genome, suggesting a slight enrichment of virus-derived transcripts in EVs.Both cellular and EV-associated RNA mapped the entire HFPV-1 genome, and no enrichment of particular viral RNAs could be detected in EVs (Supplementary Figure 7).However, the detection of viral transcripts within EVs shows that they are also exported in EVs together with host RNA.

Analysis of proteins in EVs from UV-treated cultures did not reveal significant differences
We also analyzed the protein composition of EVs from UV-treated cells and compared them with membrane-associated proteins isolated from their respective cells, to determine whether UV treatment would alter protein composition of the EVs.We identified 377 proteins associated with EVs and 668 proteins associated with their respective cell membranes.We identified 11 proteins to be enriched in EVs from UV-treated cells (Supplementary Table 9, Supplementary Figure 10A).All proteins identified as enriched in EVs from untreated cells were also identified as enriched in EVs from UV-treated cells, except for the small GTPase, HVO_3014, that was calculated as equally enriched but did not pass the p-value threshold.Instead, one additional ABC transport protein (HVO_2399) was identified as enriched.In comparing EV-associated proteins between untreated and UV-treated cultures, we did not identify major differences (Supplementary Figure 10B).Only one protein (HVO_B0027) was identified to be more enriched in EVs from untreated cultures, while two proteins (HVO_1751 and HVO_2529) were identified to be more enriched in EVs from UV-treated cultures.None of the enriched proteins from either preparation held functions that appear significant to EV production.

Testing other knockout mutants provides further insight into the mechanisms of EV formation
CetZ1 and CetZ2 were amongst the most abundant proteins in EVs; however, we were able to isolate EVs from the supernatant of both CetZ1 and CetZ2 knockout strains (Supplementary Figure 18).While quantification of EVs by the immunodetection-based method was not possible for the CetZ1 knockout strain, we did not have any indication when purifying EVs that EV production was drastically altered for CetZ1 and CetZ2 knockout strains (Supplementary Figure 18).RNA could also be isolated from EVs of this strain, and the size distribution of EV-associated RNA was nearly identical when compared to the parental strain.
Previous studies in Bacteria have shown that destabilization of the cell envelope results in a 'hypervesiculation' phenotype [5,6].To investigate whether changes in cell envelope stability would similarly affect EV production in H. volcanii, we assessed EV production in an aglB knockout strain.Cells lacking AglB are unable to N-glycosylate the S-layer glycoprotein and absence of AglB results in enhanced release of the S-layer glycoprotein [7].Thus, deletion of this protein causes a destabilization of the structural integrity of the cell envelope.Indeed, we observed a noticeable increase in EV production from the aglB knockout strain during the purification process as well as by TEM (Supplementary Figure 19A-C).While we could not confirm this result when using the immunodetection-based assay for quantifying EVs (Supplementary Figure 19D and E), EV quantification by fluorescence staining revealed a 1.52-fold (p = 0.003) increase in EV production (Supplementary Figure 19F), indicating that CetZ1 incorporation into EVs is altered in this mutant.Interestingly, we observed a drastic change to the morphology of EVs in the mutant.The surface of EVs isolated from the AglB knockout strain was significantly different from EVs of the parental strain, appearing very fuzzy (Supplementary Figure 19B and C), likely due to the instability of the S-layer.Further, while we isolated a significantly larger amount of EVs from the mutant, the RNA yield remained the same (Supplementary Figure 19G), indicating that RNA distribution in EVs is altered.

Lipid analysis reveals differences in the relative abundance of distinct lipids between cells and EVs
To determine whether EVs selectively enclose particular lipids, we analyzed the lipid content of EVs and compared the relative abundances of different lipid compounds in EVs to that of cell membranes and total cells of H. volcanii.We only detected minimal differences in proportions of lipid types between EVs deriving from upper and lower bands of a density gradient (Supplementary Figure 20B), indicating that the lipid content alone is not the differentiating factor between the two subpopulations.We therefore chose to pool samples from different bands of each replicate for comparison.
Lipids with a neutral headgroup, such as diglycosyl-archaeol (2G-AR) or no head group, such as corearchaeol (C-AR), were detected in all fractions but showed higher relative abundances in the EV samples (2.57± 0.19% and 4.1 ± 1%) compared to lipid extracts from cells and cell membranes (<1.2 ± 0.2%).A notable difference was also observed for dimeric phospholipids (or cardiolipins, CL).While they contributed 20.1 ± 9.7% and 9.04 ± 8.1% of the total lipids in whole cells and cell membrane samples respectively, they were almost undetectable in the EV samples (0.91 ± 0.41%).Interestingly, we were not able to detect any extended archaeol lipids (C25 instead of C20 isoprenoidal chains) with relevant concentrations in any of the samples, despite how common they are among many haloarchaea [8].
We could not detect any lipid compounds which were only present in the vesicular fraction but not in cells or cell membranes.However, the lipid composition of EVs differed significantly to that of cells and cell membranes when comparing the relative abundance patterns of different lipid groups (Supplementary Figure 20B, Supplementary Figure 21).The distribution between unsaturated and saturated compounds shifts towards saturated lipids from 67.5 ± 2.7% and 68.7 ± 1.7% in whole cell and cell membrane extracts, to 84.4 ± 4.7% in EVs (Supplementary Table 11).In the vesicle fraction this is likely attributable to the absence of cardiolipins and the lower abundance of unsaturated Me-PGP-ARs.

Supplementary Discussion
We were able to detect the major bilayer forming lipids PG-AR, Me-PGP-AR, S-2G-AR, C-AR, 2G-AR and cardiolipins, that were previously described for H. volcanii [9,10] in all samples, albeit in different relative amounts.Me-PGP-AR and PG-AR were the two most abundant lipid species across all samples, while the cardiolipins (CL) contributed to a notable portion of the intact polar lipids (IPLs) in cells and cell membranes and were surprisingly only detected in low abundances in EVs.CLs are considered to be important for membrane curvature [11], therefore, we expected them to be essential in EVs due to the high degree of bilayer curvature in the vesicles.However, Kellermann et al [9] observed that changing extracellular Mg 2+ levels influence CL and Me-PGP ratios in H. volcanii and proposed that changes to the ratio of the two compounds are used to control membrane permeability in neutrophilic haloarchaea, in response to extracellular Mg 2+ levels.As we cultivated H. volcanii in medium with a constant high Mg 2+ concentration (174 mM) it is not surprising that Me-PGP-AR was the most prominent phospholipid species across all samples.This could also explain the absence of CLs in EVs, as Me-PGP-AR may be sufficient to ensure membrane stability in the smaller-sized EVs under high Mg 2+ concentrations.C-ARs and 2G-AR showed the opposite trend to cardiolipins, with an increase in their relative abundance in EVs compared to the cellular fraction.

Isolation and purification of EVs
For isolation of EVs from H. volcanii, cultures were grown at 45 °C in minimal media with serial dilution (two times in exponential growth to OD600 = 0.05) before being transferred into nutrient rich media and grown at 28 °C (unless otherwise specified).EVs were isolated and purified as described in [12].Briefly, cells were removed at late stationary (~ 144 hours growth) by centrifugation (4,500 x g, 40 min), and EVs were precipitated with the addition of polyethylene glycol (PEG) 6000 and incubation at 4 °C.EVs were subsequently pelleted by centrifugation (14,000 x g¸ 50 min, 4 °C) and after resuspending the pellet, remaining cell contaminations were removed by an additional centrifugation (14,000 x g, 10 min) and filtration (1 x 0.45 µm filter, 1 x 0.2 µm filter).Extracellular nucleic acids were removed with DNase I (New England Biolabs, 20 U/mL) and RNase A (New England Biolabs, 20 U/mL) [13].The samples were further purified through an OptiPrep™ density gradient, yielding two bands containing EVs.

EV quantification
Two different quantification methods were used, because each of them proved unsuitable for some conditions tested.We assume that enclosing CetZ1 into EVs can be influenced by particular conditions.
Using CetZ1 [14] as a reporter gene for detection of EVs in culture supernatants (immunodetection) was unsuitable when testing temperatures dependencies (Supplementary Figure 2A and B) and did also not reflect results that we obtained for EVs from the aglB knockout strain (Supplementary Figure 16D and E).
The fluorescence-based method proved unsuitable for quantification of EVs in virus infected cultures, because viral particles also appeared to be stained with the fluorescent dye (Supplementary Figure 2F).Pvalues are calculated by unpaired, two-tailed t-test.

RNA extraction and transcriptomic analysis
RNA was extracted from cell pellets or EV pellets using TRIzol™ (Thermo Fischer Scientific). 1 mL TRIzol™ reagent was added to the pellet, homogenized by pipetting, and incubated at room temperature for 5 min.0.2 mL chloroform was added to the sample, gently mixed via inversion, and incubated at room temperature for 10 min.The sample was then centrifuged at 4 °C for 10 min at maximal speed (~20,000 x g).Upper phase was transferred to a new tube, and 500 mL isopropanol was added, mixed gently by inversion, and incubated at room temperature for 10 min.The sample was then centrifuged at 4 °C for 15 min at maximal speed.The supernatant was removed and pellets washed twice with ice-cold 75% ethanol.
The remaining liquid was removed and the pellet was air-dried for 10 min.Pellets were resuspended in RNase/DNase free water.

Northern blot
The Northern blotting protocol was adapted from [15].Briefly, RNA was extracted as described above and separated on formaldehyde-MOPS agarose gels, with a final concentration of 2% formaldehyde and 2% NuSieve 3:1 agarose (Lonza).5 µg RNA was denatured for 10 min at 70 °C with 1 X MOPS buffer (20 mM MOPS, 5 mM NaOAc, 1 mM EDTA, pH 7.0), 3.7% formaldehyde and loading dye (67 nM EDTA pH 8, bromophenol blue and xylene cyanol in deionized formamide).Samples were heat denatured for 10 min at 70 °C then placed on ice for 3 min before loading onto gel.The gel was run at 125 V for 3 to 4 hours and the RNA was then transferred to a Zeta-Probe GT membrane (Bio-Rad) by capillary action with 20 x SSC buffer (3 M NaCl, 300 mM Sodium Citrate, pH 7.0) and 2 x SSC buffer.The oligonucleotide probe is listed in Supplementary Table 2, and was labelled with [γ-32 P] ATP using polynucleotide kinase (Thermo Fisher).

Plasmid construction and expression of OapA
The plasmid, pTA1852 (provided by Thorsten Allers), is derived from pTA1392 [16] with a replacement of the 112 bp NdeI and NotI region containing an N-terminal 6 x His tag and a C-terminal 1 x StrepII tag with an N-terminal 7 x His tag and 2 x StrepII tag.Expression of tagged OapA (OapAt) on pTA1852 is controlled by tryptophan-inducible promoter, p.tnaA.
For Expression of OapAt cultures were grown in Hv-YPC supplemented with 200 µg/mL tryptophan at 28 °C until OD600 of approximately 1. Cultures were then supplemented with tryptophan by adding 18% BSW containing 5 mg/mL tryptophan (final concentration of 450 µg/mL tryptophan).Cultures were grown for 2 hrs at 28 °C before EVs were quantified as described.
Cells were lysed by sonication (6 x 30 seconds at 35% amplitude) on ice, and treated with 20 µL DNase I (New England Biolabs, 20 U/mL) for 1 hr at 28 °C.Lysates were centrifuged (20,000 x g, 15 min, 4 °C) and filtered through 0.8 µm, 0.45 µm and 0.22 µm pore-size filters.The remaining flow through was incubated overnight with 1 mL Strep-Tactin® Sepharose® beads (iba-lifesciences) equilibrated with Binding Buffer and applied to a Poly-Prep chromatography column (Bio-Rad).Flow through was run twice on the column, and the column was then washed 5 x with Binding Buffer.The column was then incubated with 3 mL Elution Buffer (Binding buffer with 5 mM D-desthiobiotin) for 30 min, and flow through was concentrated using Vivaspin® 500 centrifugal concentrator (10,000 MWCO, Sartorius).Expression of OapAt was then confirmed using Western blot (Supplementary Figure 11C).

Lipid extraction and analysis
For the total cell and cell membrane fraction, three biological replicates of cells pellets from H. volcanii cultures were dissolved in DBCM2 salt solution [18] and half of each replicate was used for total cell analysis and half for cell membrane extraction.For extraction of cell membranes, the dissolved cell pellet was sonicated with a microtip sonicator (MS 73 Sonoplus, Bandelin electronic, Germany) 3 times for 30 s on ice at 35% output.The lysate was treated with DNase I (30 min at 28°C, 10 µL per mL), and spun down (8,000 x g, 30 min, 4 °C) to remove cell debris.Cell membranes were pelleted from the supernatant by ultracentrifugation (248,000 x g for 15 min) and dissolved in DBCM2 salt solution.EVs from three biological replicates (200 mL cultures) were treated with DNase and RNase and purified with an Optiprep™ gradient (4 hr at 150,920 x g).The resulting EV bands were extracted separately from gradients.The samples were concentrated (Vivaspin 6, 100,000 MWCO PES, Sartorius, Germany) at 4 °C and washed twice with DBCM2 salt solution.Each gradient band was concentrated to 900 µL, from which 3 x 300 µL technical replicates were aliquoted.
For lipid extraction, samples in DBCM2 salt solution were sonicated for 1 h in an ice-cooled ultrasonication bath and treated with a protocol based on [19].Phase separation after the final centrifugation resulted in an upper lipid-containing organic phase, a lower metabolite-containing aqueous phase and a proteincontaining pellet.The separate phases were isolated into combusted glass LC-MS vials, dried under constant N2 flow and stored at -20 °C until further analysis.Three 300 µL aliquots of sterile DBCM2 salt solution were treated with the same protocol as negative controls.
For ultrahigh performance liquid chromatography (UHPLC) coupled to mass spectrometry (MS) analysis, dried samples were resuspended in a solvent mixture of dichloromethane:methanol (1:9).Measurements were performed on a Dionex Ultimate 3000 RS UHPLC system coupled to a maXis ultrahigh-resolution quadrupole time of flight tandem mass spectrometer (Q-TOF MS, Bruker Daltonics).Separation of archaeal lipids was achieved on a Waters Acquity UHPLC BEH C18 column (1.7 µm, 2.1 x 150 mm) at 65 °C using reverse phase chromatography [20].Briefly, a 26 min gradient was run at a flow rate of 400 µL min -1 beginning with 100% A (held for 2 min), followed by an increase to 15% B within 0.1 min and ramping to 85% B in 19 min, followed by 8 min re-equilibration with eluent B. Eluent A was MeOH:H2O (85:15) and eluent B was IPA:MeOH (50:50), both with addition of 0.04% HCO2H and 0.1% NH3.Analysis was performed in positive ionization mode, scanning from m/z 100 to 2000.MS2 scans were obtained in data dependent mode.
Several technical replicates were measured for each sample type, of which representative replicates were selected for each biological replicate.Since the EV samples showed minimal differences in lipid distribution between bands after ultracentrifugation (Supplementary Figure 20B), gradient bands were pooled for each biological replicate for further analysis.The samples were compared with respect to their relative abundance distributions without absolute quantification.
The relative abundances were normalized per replicate and averages for each fraction were calculated from three biological replicates (total cells and cell membrane fraction) and from the upper bands after density gradient centrifugation from three biological replicates (EV fraction).Figures were created in R Statistical Software (v4.1.2;R Core Team 2021) with the ggplot2 [23], plyr [24] and dplyr packages [25].

Identification of arvB and arvC homologs
For analysis of the distribution of arvB and arvC we examined the proteins encoded by the first and second gene downstream of the 1,666 arvA genes.We retrieved all proteins that were annotated as COG3365 and COG3364 for the gene at the first and second position downstream of arvA, respectively.COG3365 are both predominantly found in Archaea.This selection resulted in a set of 1,374 proteins for the first gene downstream COG3365, and 948 proteins for the second gene downstream (COG3365).To assess whether there were other homologs of these proteins in archaeal and bacterial genomes, and to determine whether we had missed homologs downstream of arvA by relying on COG annotation, we then used these protein sets as reference databases in a DIAMOND search as described for ArvA.For ArvB (first gene downstream, COG3365) this search resulted in 1,749 hits that were inspected using an alignment score ratio approach as described for ArvA.All hits met the selection criteria, and were used for further analysis.For ArvC (second gene downstream, COG3364) this search resulted in 14,474 hits.After filtering of the hits using an alignment score ratio approach as described for ArvA, 1,650 sequences were selected for further analysis.
For both arvB and arvC the presence or absence in genomes containing arvA was then determined, as well as the genomic location relative to arvA.