SHIP164 is a Chorein Motif Containing Lipid Transport Protein that Controls Membrane Dynamics and Traffic at the Endosome-Golgi Interface

Cellular membranes differ in protein and lipid composition as well as in the protein-lipid ratio. Thus, progression of membranous organelles along traffic routes requires mechanisms to control bilayer lipid chemistry and their abundance relative to proteins. The recent structural and functional characterization of VPS13-family proteins has suggested a mechanism through which lipids can be transferred in bulk from one membrane to another at membrane contact sites, and thus independently of vesicular traffic. Here we show that SHIP164 (UHRF1BP1L) shares structural and lipid transfer properties with these proteins and is localized on a subpopulation of vesicle clusters in the early endocytic pathway whose membrane cargo includes the cation-independent mannose-6-phosphate receptor (MPR) and ATG9. Loss of SHIP164 disrupts retrograde traffic of these organelles to the Golgi complex. Our findings raise the possibility that bulk transfer of lipids to endocytic membranes may play a role in their traffic.


44
The homeostasis of intracellular membranes and their adaptation to changes in the functional 45 state of the cell requires the coordination of protein and lipid transport. While much has been 46 learned about protein traffic, less is known about the dynamics and transport of bilayer lipids. A 47 significant fraction of such lipids move between organelles as part of the membranes of vesicular 48 carriers. However, lipids also move via transport proteins that harbor them in hydrophobic 49 cavities as they travel through the aqueous environment of the cytosol. This mode of transport 50 has been known for decades, but has been increasingly appreciated over the last several years with 51 the discovery of many new lipid transport proteins. Moreover, it has also become clear that many 52 such proteins function at membrane contact sites, thus facilitating specificity and speed of lipid 53 transport. Typically, these proteins contain modules or motifs that tether them to the two apposed 54 membranes while lipid transfer modules move back and forth between them to extract and deliver 55 lipids by a shuttling mechanism 1-5 .

57
Recently, the characterization of VPS13 and its distant paralogue ATG2 has suggested a new mode 58 of transport involving the flow of lipids along a protein bridge that connects the two membranes 59 in eukaryotic cells [6][7][8][9][10] . A defining feature of these proteins is the presence of a conserved N-60 terminal region, ~125 residues long, referred to as the chorein-N motif 6,11-13 . In VPS13 and ATG2, 61 this motif caps one end of an elongated rod (extended chorein domain). The rod comprises an 62 extended ß-sheet that is highly curved to resemble a taco shell harboring a groove along its 63 length 7,8 . A hydrophobic cavity in the chorein motif is continuous with the groove, whose floor is 64 lined by hydrophobic amino acids, and so suited to accommodate many lipids at once and to allow scale whole exome sequencing (WES) study 25 . To date, Nothing is known about the cell biology of leaflet. If SHIP164 had induced liposome fusion, the residual fluorescence after addition of 144 SHIP164 would be higher than in its absence, as lipids in the internal leaflet would have been 145 diluted 26 . In contrast, we found that the residual fluorescence after addition of dithionite was the 146 same with and without addition of SHIP164 reflecting the fluorescence of lipids in the internal 147 leaflet of donor liposomes that did not undergo dilution in the absence of membrane fusion 148 ( Figure 1F). These experiments support that SHIP164 can solubilize lipids and transfer them 149 between membranes.

153
Comparison of SHIP164 with a MBP-SHIP164 fusion indicates that this rod represents a tail-to-154 tail dimer, as there are two densities corresponding to MBP, one at each end. The MBP-tag locates 155 the SHIP164 N-terminus at the rod end. Whether this dimerization is physiologically relevant is 156 unclear; for example, ATG2 similarily dimerizes in vitro, at high concentration, and in the absence 157 of binding partners, but is thought to be monomeric in vivo 7 . Further characterization of 3XFLAG-158 SHIP164Δ901-1099 using single particle cryo-electron microscopy techniques yielded a 159 reconstruction at an estimated resolution of ~8.3 Å (Supplemental Figure 1B&C). As 160 suggested by fold prediction algorithms (ref. 18,19 )(see Figure 1Ac-d), a cavity runs along the 161 entire length of the rod, as in VPS13 and ATG2 (Figure 1 H&I). The algorithms also predict that 162 this cavity is lined entirely with hydrophobic residues (Figure 1Ac-d). Thus, the cavity can 163 accommodate the multiple lipids bound by SHIP164, and we propose that it could be a conduit 164 for lipids to transit between membranes.

166
Collectively, these findings are consistent with the hypothesis that SHIP164 is a lipid transport 167 protein. We next investigated its site of action within cells.

169
Localization of exogenous SHIP164 points to a role on endocytic organelles

171
Antibodies directed against SHIP164 did not yield a consistently reliable signal when tested by 172 immunofluorescence. Thus, we examined the localization of exogenously tagged human SHIP164.

173
In agreement with the study by Otto et al 20 , SHIP164 tagged at the N-terminus (GFP-SHIP164) 174 had a diffuse cytosolic localization when expressed in mammalian cell lines (e.g. HeLa, COS-7, 175 RPE1) (Supplemental Figure 2A). As studies of VPS13 had shown that tags appended to the 176 N-terminus (i.e the N-terminus of the chorein motif) interfere with its physiological function 6,27 , 177 we engineered tags at the C-terminus of SHIP164 (SHIP164-Halo) or at an internal site within the 178 predicted disordered region (after amino acid residue 915; SHIP164^mScarlet-I) (Figure 2A) 179 (ref. 6,27 ). When either of these tagged SHIP164 constructs was expressed alone, punctate 180 structures of varying size were observed throughout the cytoplasm, with the larger and brighter 181 spots being localized in the central region of the cell, in proximity of the Golgi complex area on foci juxtaposed to them (Figure 2C-E). This was even more clear by observing 195 macropinosomes (both naturally occurring or induced by expressing constitutively active 196 Ras G12V ), given the large size of these vesicles 29 (Supplemental Figure 2B). As shown by live-197 cell imaging, SHIP164 foci not only were variable in size as well as in fluorescence intensity, but 198 were also highly dynamic structures. They changed shape, often underwent fission, or coalesced 199 into larger spots although they tended to remain tethered to endosomes and micropinosomes 200 ( Figure 2F; Supplemental Movie 1). In some cases, they were localized at sites where the 201 endosomes appeared to be close to the ER (Figure 2E), but were not restricted to the space 202 between these two organelles.

205
confirming previous results 20,24 ( Figure 2D). In this case a pool of SHIP164 colocalized with Rab5 206 along the entire profile of vacuoles and, conversely, Rab5 robustly colocalized with the bright 207 SHIP164 foci closely apposed to large vacuoles ( Figure 2D). This very strong colocalization 208 supports the hypothesis that SHIP164 is a Rab5 effector. However, expression of a dominant 209 negative Rab5 mutant (GFP-Rab5a S34N ) did not abolish bright SHIP164 (SHIP164-Halo) foci 210 suggesting that Rab5 is not necessary for their formation (Supplemental Figure 2C).

226
This difference is in line with the known traffic of MPR and sortilin versus TGN46 in different 227 vesicular carriers 33,34 (see also below).

229
Foci of exogenous SHIP164 reflect accumulations of small vesicles

231
To gain further insight into the precise nature of foci of over-expressed SHIP164, we performed 232 correlative light-electron microscopy (CLEM) of COS-7 cells co-expressing the early endosome 233 marker GFP-WDFY2 and SHIP164^mScarlet-I ( Figure 3A). Both conventional EM and FIB-234 SEM were performed. We again employed Ras G12V to induce the formation of macropinosomes 35 "phases" often involve low affinity interaction of disordered protein regions 39 , we considered the 246 possibility that the "disordered region" of SHIP164 (see Figure 1Ac) may be responsible for their 247 formation. However, a truncated form of SHIP164 (SHIP1641-873-mRFP) that lacks the disordered 248 region also induced the formation of large fluorescence puncta closely juxtaposed to endosomes 249 (Supplemental Figure 3). Thus, the mechanism responsible for the clustering of SHIP164 250 positive vesicles remains unclear. Also unclear is how such clusters are anchored to the large 251 vacuoles.

253
As these vesicle accumulations, which to our knowledge were never observed in WT cells,

254
suggested an abnormal effect of over-expressed SHIP164, we next tagged SHIP164 at the 255 endogenous locus to assess its localization when expressed at the endogenous level.

257
Endogenous SHIP164 localizes to small clusters of vesicles near the cell edge

259
We tagged SHIP164 at the endogenous locus in HeLa cells. Specifically, we engineered a

267
Like exogenous SHIP164, eSHIP164^mNG had a punctate appearance. However, 268 eSHIP164^mNG puncta were very much smaller and, in contrast to foci of over-expressed 269 SHIP164 (Figure 2A), were primarily localized at the cell edge ( Figure 4D). On the other hand,

276
To determine whether puncta of eSHIP164^mNG represent MPR positive vesicles, we generated

291
To gain further insight into the nature of endogenous SHIP164 puncta, we again employed CLEM 292 in eSHIP164^mNG HeLa cells also expressing a marker of mitochondria (mito-BFP) to aid in analysis revealed that, similar to the large fluorescence puncta of exogenously expressed SHIP164, 297 spots of eSHIP164^mNG fluorescence reflected tightly packed clusters of small vesicles and short 298 tubules ( Figure 4F&G). However, these clusters were much smaller (10's instead of 100's of 299 vesicular structures). As revealed by 3D reconstruction of FIB-SEM volumes ( Figure 4G) and by 300 fluorescence of cells also expressing the ER marker RFP-Sec61 (Figure 4H), ER tubules were 301 localized in close proximity of these clusters but vesicles were preferentially associated with 302 themselves rather than with the ER.

304
Collectively, these findings indicate that SHIP164, when expressed at endogenous levels, is

314
We confirmed the previously reported colocalization of exogenous SHIP164 with over-expressed  Figure 2A). This indicates that binding to Stx6 does not play a major role in 319 the localizations of SHIP164 described above.

321
The location of the Stx6 binding site within the SHIP164 protein was not known. A short

332
To identify additional SHIP164 interacting partners that may impact its localization and/or 333 function, we performed pull-down experiments from detergent solubilized mouse brain lysates

370
To assess the impact of the absence of SHIP164 on cell physiology we chose RPE1 cells as a model 371 system due to their reliable genetic editing and flat morphology optimal for organelle imaging.

372
Using CRISPR/Cas9 methodology, two independent SHIP164 knockout (KO) clones were 373 generated where indels led to a frameshift of the reading frame that resulted in premature stop 374 codons as shown by sequencing (Supplemental Figure 6A). In both clones, identical indels 375 were observed on both alleles (but different between the two clones) and absence of the SHIP164 376 protein was confirmed by western blotting (Figure 6A). No obvious changes were observed in 377 cell shape, in mitochondria, ER, or Golgi morphology as assessed by the cis-Golgi marker 378 (GM130) (see below). Two differences, however, were noted.

380
One was a difference in the size of endosomes labeled by EEA1 (visualized by immunostaining),

384
The other difference was a change in the steady state localization of MPR and other proteins that

396
To validate the dependence of these phenotypes from the lack of SHIP164, we attempted to rescue 397 them by expressing exogenous SHIP164 in the KO cells. To this aim we used the polycistronic construct described above encoding both untagged SHIP164, in order to avoid any artifact due to 399 the tag, and soluble RFP, to identify transfected cells. In both KO clones, transfection of this 400 construct (likely resulting in levels of SHIP164 higher than in WT cells), but not of a construct 401 encoding only the RFP reporter, resulted in the rescue of the vesicles-to-Golgi ratio of TGN46 402 immunoreactivity ( Figure 6G; *, p<0.05; **, p<0.01). However, in the case of MPR and sortilin, 403 expression of the SHIP164 construct did not restore their WT localization, but resulted in their 404 robust accumulation into large foci as shown above upon SHIP164 over-expression in wild type 405 cells (Supplemental Figure 6D). This result is most likely explained by the disrupting effect of 406 SHIP164 over-expression, which, as shown above, does not impact TGN46 localization 407 (Supplemental Figure 2G&H). Likewise, SHIP164 over-expression did not seem to rescue the

417
We conclude that loss of SHIP164 function may globally impact membrane traffic between the 418 cell periphery and the Golgi complex, while over-expression of SHIP164 disrupts selectively the 419 retrograde traffic of a subset of the vesicles with which SHIP164 is associated.

421
Presence of ATG9A in clusters of SHIP164-positive vesicles

423
The structural similarity of SHIP164 to proteins thought to mediate bulk lipid transport-

459
However, in contrast to SHIP164 foci, which were primarily associated with early endosomes, 460 UHRF1BP1 foci were predominantly associated with LAMP1-GFP-positive organelles, in 461 agreement with previous work suggesting that UHRF1BP1 is a Rab7 effector, rather then a Rab5 462 effector 24 (Supplemental Figure 7). We conclude that SHIP164 and UHRF1BP1 likely function

494
The fourth is Rab45, an unconventional Rab that anchors endocytic vesicles to dynein, creating expressed at endogenous levels is predominantly associated with peripheral vesicles, suggesting

513
A key question is how the structure and molecular properties of SHIP164, which predict a lipid

552
Finally, in view of the link between the function of SHIP164 and GARP suggested by our study 553 and a previous study 20 , it is of interest that defects in retrograde traffic from the cell periphery to 554 the Golgi complex produced by SHIP164 loss-of-function have similarities to those observed upon 555 loss of function of the GARP complex 20,58 . GARP has also been linked to lipid metabolism. In

1314
(ThermoFisher) for 48 hours before fixation for immunocytochemistry experiments or collected cells using the transfection protocol described above. Cells were split once to keep confluency