The large GTPase Sey1/atlastin mediates lipid droplet- and FadL-dependent intracellular fatty acid metabolism of Legionella pneumophila

The facultative intracellular bacterium Legionella pneumophila employs the Icm/Dot type IV secretion system (T4SS) to replicate in a unique membrane-bound compartment, the Legionella-containing vacuole (LCV). The endoplasmic reticulum (ER)-resident large fusion GTPase Sey1/atlastin promotes remodeling and expansion of LCVs, and the GTPase is also implicated in the formation of ER-derived lipid droplets (LDs). Here we show that LCVs intimately interact with palmitate-induced LDs in Dictyostelium discoideum amoeba. Comparative proteomics of LDs isolated from the D. discoideum parental strain Ax3 or ⊗sey1 revealed 144 differentially produced proteins, of which 7 or 22 were exclusively detected in LDs isolated from strain Ax3 or ⊗sey1, respectively. Using dually fluorescence-labeled amoeba producing the LCV marker P4C-GFP or AmtA-GFP and the LD marker mCherry-perilipin, we discovered that Sey1 and the L. pneumophila Icm/Dot T4SS as well as the effector LegG1 promote LCV-LD interactions. In vitro reconstitution of the LCV-LD interactions using purified LCVs and LDs from D. discoideum Ax3 or ⊗sey1 revealed that Sey1 and GTP promote this process. The LCV-LD interactions were impaired for ⊗sey1-derived LDs, suggesting that Sey1 regulates LD composition. Palmitate promoted the growth of (i) L. pneumophila wild-type in D. discoideum Ax3 but not in ⊗sey1 mutant amoeba and (ii) L. pneumophila wild-type but not ⊗fadL mutant bacteria lacking a homologue of the E. coli fatty acid transporter FadL. Finally, isotopologue profiling indicated that intracellular L. pneumophila metabolizes 13C-palmitate, and its catabolism was reduced in D. discoideum ⊗sey1 and L. pneumophila ⊗fadL. Taken together, our results reveal that Sey1 mediates LD- and FadL-dependent fatty acid metabolism of intracellular L. pneumophila.


Introduction
protein disulfide isomerase, PDI; lipid droplet membrane protein, LdpA; and 15 lipid metabolism 126 enzymes), the latter reflecting their cellular organelle origin (Du et al., 2013). LDs are 127 transported along microtubules and actin filaments or moved by actin polymerization (Welte,  isotopologue profiling studies indicated that extracellular L. pneumophila also efficiently 142 catabolizes exogenous [1,2,3,4-13 C 4 ]palmitic acid, yielding 13 C 2 -acetyl-CoA, which is used to 143 synthesize the storage compound poly-hydroxybutyrate (PHB) (Häuslein et al., 2017). 144 It is unknown how fatty acids are taken up by L. pneumophila. In E. coli, the long-chain fatty 145 acid transporter FadL localizes to the outer membrane, where the monomeric protein adopts a 146 7 14-stranded, anti-parallel β barrel structure (van den Berg et al., 2004). The N-terminal 42 amino 147 acid residues of FadL form a small 'hatch' domain that plugs the barrel, and the hydrophobic 148 substrate leaves the transporter by lateral diffusion into the outer membrane (Hearn et al., 2009). 149 L. pneumophila encodes a homolog of E. coli FadL, Lpg1810, which was identified as a surface-150 associated protein by fluorescence-labeling and subsequent mass spectrometry (MS), confirming 151 its presence in the outer membrane (Khemiri et al., 2008). 152 Given the role of LDs as lipid storage organelles regulated by atlastins, we set out to analyze  (Fig. S1A). To assess whether the growth-promoting effect of palmitate might involve 165 LDs, we thought to visualize the possible interactions between LCVs and LDs. To this end, D. 166 discoideum strain Ax3 was fed with palmitate, infected with the L. pneumophila wild-type strain 167 JR32 and subjected to cryo-electron tomography (cryoET). The obtained cryotomograms clearly 168 show an extensive interaction between LCVs and LDs (Fig. 1A). Upon contact with the LCV, acetoacetyl-CoA hydrolase) (Fig. 1B, Table S1). Among the differentially produced proteins, 7 185 or 22 were exclusively detected in LDs isolated from strain Ax3 or Δsey1, respectively. Sey1 186 was identified on LDs isolated from strain Ax3, but as expected not on LDs isolated from Δsey1 187 mutant amoeba. Contrarily, the phospholipase PldA, the ER protein calnexin (CnxA) and the 188 protein SCFD1/SLY1 implicated in ER to Golgi transport were identified only on LDs isolated 189 from the Δsey1 strain (Table S1). The 50 most highly abundant proteins, which were not 190 significantly different on LDs isolated from Ax3 or Δsey1 mutant amoeba, included perilipin 191 (PlnA), which is involved in the formation and maintenance of LDs (Du et al., 2013), as well as 192 the small GTPase RanA (Du et al., 2013) and its effector RanBP1 (Table S1). RanA is activated 193 in L. pneumophila-infected cells and implicated in microtubule stabilization and LCV motility 194 (Rothmeier et al., 2013;Swart et al., 2020b). 195 To assess the localization of Sey1 with regard to LDs, we used dually labeled D. discoideum 196 Ax3 producing GFP-Sey1 as well as the LD marker mCherry-Plin, and further stained LDs with 197 LipidTOX Deep Red (Fig. 1C). Under the conditions used, GFP-Sey1 accumulated in the 198 vicinity of LDs in intact cells as well as in cell homogenates, but apparently did not co-localize 199 with LDs. This staining pattern suggests that Sey1 localizes only in very low amounts to LDs, or with LDs were recorded for 60 s each at different time points ( Fig. 2A, Fig. S2B). As the LCVs 217 matured over the course of 1 h post infection (p.i.), the overall LCV-LD contacts gradually 218 increased in D. discoideum Ax3, while they remained lower in Δsey1 mutant amoeba (Fig. 2B). 219 Moreover, the retention time of individual LDs on LCVs was also signficantly higher in strain 220 Ax3 than in Δsey1 mutant amoeba (Fig. 2C). Taken together, these real-time data indicate that 221 Sey1 promotes the dynamics of LCV-LD interactions during the course of LCV maturation.  (Table S1), we tested the hypothesis that LegG1 is implicated in LCV-LD 227 dynamics. To this end, we infected palmitate-fed D. discoideum Ax3 or Δsey1 producing P4C-  Sey1 promotes LD recruitment to LCVs in D. discoideum 239 To validate that Sey1 promotes LCV-LD interactions in palmitate-fed, fixed D. discoideum and 240 to test whether the process is dependent on the L. pneumophila Icm/Dot T4SS, we used dually 241 fluorescence-labeled amoeba producing mCherry-Plin and AmtA-GFP, a probe localizing to 242 vacuoles containing either wild-type L. pneumophila or ΔicmT mutant bacteria lacking a 243 functional T4SS (Fig. 3A). This approach indicated that the mean number of LDs localizing to 244 LCVs harboring wild-type L. pneumophila was more than twice as high in D. discoideum Ax3 as 245 compared to the Δsey1 mutant amoeba (Fig. 3B), and the effect was of similar magnitude, when 246 the number of LDs per LCV area was calculated (Fig. 3C). Contrarily, Sey1 did not promote the 247 interaction of vacuoles harboring ΔicmT mutant bacteria with LDs, and overall, significantly 248 fewer LDs associated with these vacuoles (Fig. 3BC). Taken together, these studies using fixed 249 D. discoideum amoeba reveal that Sey1 promotes LCV-LD interactions and the Icm/Dot T4SS is 250 required for LD accumulation on LCVs.  253 Since Sey1 is a large fusion GTPase, we thought to test the nucleotide requirement of the LCV-254 LD interactions. To this end, we purified LCVs from D. discoideum Ax3 producing P4C-GFP 255 infected with mCerulean-producing L. pneumophila JR32, mixed the pathogen vacuoles with 256 purified LDs from palmitate-fed strain Ax3 producing mCherry-Plin and added 5 mM of 257 different nucleotides (Fig. 4A). Using this in vitro reconstitution approach, the addition of GTP 258 resulted in a ca. 2.5-fold higher number of LDs per LCV as compared to the addition of GDP, 259 Gpp(NH)p or GTPγS (Fig. 4B). 260 In an analogous approach, we thought to test whether Sey1 on the purified LCV or LD 261 fraction promotes the interaction between the two compartments. We mixed LCVs purified from 262 D. discoideum Ax3 or Δsey1 producing P4C-GFP infected with mCerulean-producing L. 263 pneumophila JR32 with purified LDs from palmitate-fed strain Ax3 or Δsey1 producing 264 mCherry-Plin in presence of 5 mM GTP or GDP (Fig. 4C). The mean number of LDs per LCV 265 was highest for both compartments isolated from D. discoideum Ax3, followed by LDs purified 266 from strain Ax3 and LCVs from Δsey1 mutant amoeba (Fig. 4D). Contrarily, the LCV-LD 267 interactions were impaired for Δsey1-derived LDs, suggesting that Sey1 regulates LD traits   pneumophila JR32, we quantified the portion of intra-LCV LDs without perilipin coat (Fig. 5A). 280 While ca. 90% of the LDs adhering to LCVs from the cytoplasmic side were decorated with 281 perilipin, less than 10% of the LDs in the LCV lumen were decorated with perilipin, regardless 282 of whether the amoeba produced Sey1 or not (Fig. 5B). Accordingly, most LDs had shed the 283 13 perilipin coat on their way from the host cell cytoplasm to the lumen of the pathogen vacuole, 284 and this process was independent of the large fusion GTPase Sey1.

285
To further analyze the shedding process, we mixed LCVs isolated from D. discoideum Ax3 286 producing P4C-GFP with LDs purified from palmitate-fed Ax3 producing mCherry-Plin in 287 presence of 5 mM GTP or GDP (Fig. 5C). Again, ca. 90% of the LDs adhering externally to 288 LCVs were decorated with perilipin, and less than 10% of the LDs in the LCV lumen were 289 decorated with perilipin, regardless of whether GTP or GDP was added (Fig. 5D). In summary, 290 these results indicate that upon LCV membrane crossing, LDs shed their perilipin coat in a 291 process that does not involve Sey1 or GTP. identified. 308 We first tested whether fadL (lpg1810) is expressed and how expression is controlled. To this 309 end, we constructed a P fadL -gfp transcriptional fusion and assessed GFP production in L. 310 pneumophila JR32 and ΔlqsA, ΔlqsR, or ΔlvbR mutant strains lacking components of the Lqs 311 quorum sensing system or the pleiotropic transcription factor LvbR, respectively (Fig. S4AB). 312 Compared to the parental strain JR32, P fadL -gfp was expressed less strongly and with some delay  To assess the role of fadL for the growth of L. pneumophila, we constructed a ΔfadL deletion 317 mutant strain by double homologous recombination. The ΔfadL mutant strain grew like the 318 parental strain JR32 in AYE broth and MDM minimal medium (Fig. S4C). However, the ΔfadL 319 mutant strain was impaired for intracellular growth in D. discoideum, and the growth defect was 320 complemented by inserting the fadL gene back into the L. pneumophila genome (Fig. 6A). In 321 agreement with a role of fadL for intracellular growth of L. pneumophila, the P fadL reporter 322 construct was expressed in D. discoideum throughout an infection (Fig. S5). 323 Next, we tested the effects of overnight palmitate feeding of D. discoideum Ax3 or Δsey1 on 324 intracellular growth of L. pneumophila JR32 or ΔfadL. Feeding with palmitate augmented the 325 growth of L. pneumophila JR32 in D. discoideum Ax3 but did not significantly affect the growth 326 of the ΔfadL strain in either D. discoideum Ax3 or Δsey1 (Fig. 6B). These results revealed that  In this study we assessed the role of LDs, Sey1 and FadL for intracellular growth of L. 349 pneumophila. We reveal that LCVs intimately interact with palmitate-induced and Sey1-350 containing LDs in D. discoideum (Fig. 1). Moreover, using dually fluorescence-labeled amoeba, 351 we demonstrated in vivo by live-cell microscopy ( Fig. 2) and fixed samples (Fig. 3), or in vitro 352 using reconstituted purified LCVs and LDs (Fig. 4) that Sey1 and GTP promote LCV-LD 353 interactions. Palmitate was catabolized and promoted the intracellular replication of wild-type L. 354 pneumophila in D. discoideum Ax3, but not of ΔfadL mutant bacteria in Δsey1 mutant amoeba 355 (Fig. 6). Taken together, our results indicate that Sey1-dependent recruitment of LDs, LCV-LD 356 interactions, LD transfer to the LCV lumen and FadL-dependent catabolism of fatty acids 357 promote intracellular growth of L. pneumophila (Fig. 7). 358 The LCV-LD interactions are controlled by the L. pneumophila Icm/Dot T4SS (Fig. 3), and 359 the Icm/Dot substrate LegG1 (Fig. 2). LegG1 belongs to a family of RCC1 repeat domain-

398
L. pneumophila FadL promotes the catabolism of palmitate (Fig. 6), but overall, the 399 degradation of lipids by Legionella is not well understood. Analogously to M. marinum (Barisch 400 and Soldati, 2017b), L. pneumophila might hydrolyse triacylglycerols; however, rather than 401 building up ILIs, L. pneumophila accumulates the storage compound PHB (Fig. 6). L.   Sey1 was identified by proteomics on LDs purified from the D. discoideum Ax3 parental strain, 423 but not from the Δsey1 mutant strain (Fig. 1B). Accordingly, Sey1 might directly participate in 424 the LD-LCV fusion process. However, we cannot rule out that the ER-residing fusion GTPase 425 Sey1 affects the formation and composition of ER-derived LDs, and thus, indirectly affects LD-426 LCV fusion. Since Sey1 does not seem to significantly affect the number and size of LDs in D. 427 discoideum (Fig. S1BC), an indirect role of Sey1 for LD-LCV interactions seems less likely.

428
The LD protein perilipin was shed upon uptake of LDs into LCVs (Fig. 5). Perilipin    The plasmids used in this study are listed in Table S2. Cloning was performed according to 465 standard protocols and plasmids were isolated using the NucleoSpin Plasmid kit (Macherey-466 Nagel). DNA fragments were amplified using Phusion High Fidelity DNA polymerase (NEB) 467 and the oligonucleotides listed in Table S3. FastDigest restriction enzymes (Thermo-Fisher) 468 were used for plasmid digestion and Gibson assembly was performed using the NEBuilder HiFi 469 DNA assembly kit (NEB). All constructs were verified by DNA sequencing.

470
To construct the plasmid pLS187, the gene region of ranBP1 (P ranBP1 ) was amplified from with BglII and SpeI. To construct pLS221 and pLS222, the gene region of ranA (P ranA ) and 474 ranBP1 (P ranBP1 ) was amplified from pSU17 and pLS187, respectively, using the primer pairs 475 oLS333/oLS334 and oLS296/oLS297. The PCR products were purified as described above and    to be the most effective in increasing intracellular replication of L. pneumophila (Fig. S1A). The   614 For infection assays for confocal microscopy, D. discoideum strains producing the desired  accumulation equal 2 were used to capture still images (Fig. 1, Fig. 3-5, Fig. S2A, Fig. S4-S5). 640 Resonant scanner (scanning speed: 8000 Hz) and line average equal 4 were used to capture 641 movies (Fig. 2, Fig. S2B, Supplementary Movies S1-S4). 642 For image processing, all images and movies were deconvolved with Huygens professional  Infection assays for electron microscopy were performed with D. discoideum Ax3 grown directly 651 on grids as previously described (Medeiros et al., 2018). Briefly, the amoebae (5×10 5 per well) 652 were seeded onto EM gold finder grids (Au NH2 R2/2, Quantifoil) and incubated for 1 h, to 653 allow the cells to attach to the grids. Cells were infected at an MOI of 100 and were vitrified at  Cryo-focused ion beam (cryoFIB) milling was used to prepare samples of plunge-frozen 661 infected amoebae for imaging by cryo-electron tomography (cryoET) (Marko et al., 2007).

662
Frozen grids with infected cells were processed as previously described (Medeiros et al., 2018) 663 using a Helios NanoLab600i dual beam FIB/SEM instrument (Thermo-Fisher). Briefly, lamellae 664 with ~2 µm thickness were generated first (at 30 kV and ~400 pA). The thickness of the lamellae 665 (final ~200 nm) was then gradually reduced using decreasing ion beam currents (final ~25 pA).  Results were filtered for proteins quantified in at least two out of three biological replicates 712 before statistical analysis. Here, both strains were compared by a student's t-test applying a 713 threshold p-value of 0.01, which was based on all possible permutations. Proteins were 714 considered to be differentially abundant if the log2LFQ-fold change was greater than |0.8|. The 715 dataset was also filtered for so-called "on-off" proteins. These proteins are interesting candidates 716 as their changes in abundance might be so drastic that their abundance is below the limit of 717 detection in the "off" condition. To robustly filter for these proteins, "on" proteins were defined 718 as being quantified in all three biological replicates of the Δsey1 mutant setup and in none of the 719 replicates of the Ax3 wild-type setup, whereas "off" proteins were quantified in none of the three 720 biological replicates of the mutant setup but in all replicates of the wild-type setup. In addition, 721 the dataset was also filtered for the 50 most abundant proteins in each sample (based on the 722 corresponding iBAQ value). After filtering, we ended with a list of 192 proteins that were either 723 differentially abundant on LDs from D. discoideum Ax3 (39 proteins) or Δsey1 (76 proteins), or 724 that belonged to the "on" (7 proteins) / "off" (22 proteins) categories or to the 50 most abundant 725 proteins (48 additional proteins) ( Table S1). 726 We then functionally mapped these proteins manually according to dictyBase classification,