A phosphoinositide and RAB switch controls early macropinocytosis

Macropinocytosis is a non-selective endocytic process by which cells take up large amounts of extracellular fluids into giant vesicles known as macropinosomes. This mechanism is used by immune cells to sample the surroundings for antigens and can be exploited by cancer cells for nutrient uptake. What determines the fate of macropinosomes after they have been internalized is largely unknown. Here we investigate the role of the phosphatidylinositol 3-kinase VPS34/PIK3C3 and its product phosphatidylinositol 3-phosphate (PtdIns3P) in macropinosome fate determination. Inhibition of VPS34 led to a decrease in macropinosome survival and fluid phase uptake as well as preventing recruitment of early endosomal factors, including the small GTPase RAB5 and its effectors, to the forming macropinosomes. Instead, forming macropinosomes under VPS34 inhibition accumulated regulators of endocytic recycling, including RAB8A, RAB10, RAB11A, and PtdIns4P, which led to fusion of macropinosomes with the plasma membrane. Whereas RAB5 was critical for macropinosome formation, macropinosome fusion with the plasma membrane depended on RAB8A. Thus, macropinosome maturation is regulated by a PtdIns3P-controlled switch that balances macropinosome fate between the default, endolysosomal maturation and an alternative, secretory route.

Introduction 47 effective and selective VPS34 inhibitor SAR405 [29] to acutely deplete cellular PtdIns3P pools. Treatment 107 of cells with SAR405 blocks de-novo synthesis of PtdIns3P by VPS34. Existing pools are rapidly 108 metabolized, leading to a loss of PtdIns3P from intracellular vesicles within minutes ( Figure 1B). When 109 cells expressing Phafin2-GFP and either mCherry-2xFYVE or mCherry-2xSidM were treated with SAR405,110 we found that localization of mCherry-2xSidM was not affected; rather, this probe showed a slightly longer 111 localization to macropinosome membranes ( Figure 1C). In contrast, localization of mCherry-2xFYVE was 112 completely lost, as expected ( Figure 1C). Phafin2 only localized to nascent macropinosomes, consistent 113 with earlier findings that its localization to mature macropinosomes is dependent on VPS34-generated 114 PtdIns3P [23]. We also found that this localization was slightly longer than in non-treated cells, similar to 115 the localization of mCherry-2xSidM. We conclude that nascent macropinosomes undergo a switch from 116 PtdIns4P to PtdIns3P as they acquire endosomal identity, and that this depends on VPS34. 117 118

PtdIns3P depleted macropinosomes re-fuse with the plasma membrane 119
While tracking macropinosomes in cells treated with VPS34 inhibitor SAR405, we noticed that these 120 vesicles invariably disappeared shortly after their formation ( Figure 1D (asterisk)). Newly-formed 121 macropinosomes still gained the first Phafin2 localization -suggesting that they successfully formed and 122 abscised from the membrane [23], but then rapidly disappeared. High resolution imaging of SAR405 123 treated cells expressing the membrane marker MyrPalm-mCherry and mNeonGreen-2xFYVE showed that 124 these cells could still form macropinosomes -as judged by the appearance of large vesicles -which then 125 shrunk and disappeared ( Figure 1E, movie S1). 126 In order to understand this finding, we used live cell imaging to follow the fate of Phafin2-lablled forming 127 macropinosomes with and without VPS34-inhibitor SAR405 in a cell line stably expressing Phafin2-GFP. In 128 untreated cells, ~65% of all newly-formed macropinosomes matured to the endosomal stage ( Figure 2A). 129 In SAR405-treated cells, we found that macropinosomes showed two distinct fates, depending on their 130 maturation state. Macropinosomes forming after the addition of SAR405 invariably collapsed and 131 disappeared (Figure 2A), whereas preexisting RAB5 positive macropinosomes were largely unaffected by 132 SAR405 ( Figure 2B). We also compared the lifetimes of newly forming macropinosomes in untreated and 133 in VPS34-inhibited conditions and observed a reduction from a mean lifetime of 465s in control cells to 134 135s in SAR405-treated cells ( Figure 2C). Since the survival of nascent macropinosomes was significantly 135 decreased upon VPS34 inhibition, we investigated whether this was also reflected in fluid phase uptake 136 by macropinocytosis. To that end we incubated cells with fluorescently labelled 70 kDa dextran and 137 measured dextran uptake by flow cytometry. Cells treated with the macropinocytosis inhibitor EIPA [30, 138 31] or with SAR405 showed a strong reduction in fluid phase uptake of dextran compared to untreated 139 cells ( Figure 2D). To investigate whether this reduction of fluid phase uptake upon SAR405 treatment is 140 caused by a decreased rate of macropinocytosis, we performed live cell imaging and quantified the 141 number of forming macropinosomes in untreated, DMSO-and SAR405 -treated cells using the first, VPS34-142 independent recruitment of Phafin2-GFP as a marker. The number of forming Phafin2 positive 143 macropinosomes was unaffected in all conditions ( Figure 2E), indicating that the decrease in fluid phase 144 uptake was not due to a defect in macropinosome initiation. 145 We then asked what happened to the newly formed macropinosomes, which disappeared when VPS34 146 was inhibited. When performing 2D imaging, it appeared that macropinosomes shortly after formation 147 started to rapidly shrink until they vanished ( Figure 1E). The morphology and dynamics of these events 148 suggested that they re-fused with the plasma membrane. To visualize this process, we performed high 149 resolution oblique plane light sheet microscopy (OPM), which allows fast volumetric imaging with minimal 150 photobleaching [32]. Using this method, we followed the fate of individual macropinosomes after 151 inhibition of VPS34 with SAR405 in 3D. We observed that macropinosomes shrunk and often, a thin 152 connection to the plasma membrane was visible, further indicating that VPS34 inhibition triggered re-153 fusion of nascent vesicles with the plasma membrane ( Figure 2F, movie S2). 154 Our data indicates that failure to establish PtdIns3P leads to macropinosome refusion with the plasma 155 membrane ( Figure 2G). In contrast, already established macropinosomes that have gained endosomal 156 identity -indicated by RAB5 -are stable and not affected by VPS34 inhibition. 157

VPS34 inhibition blocks recruitment of endocytic factors 159
As we found that vesicles with an established endocytic identity are not affected by VPS34 inhibition,160 whereas newly forming vesicles are highly dependent on VPS34 activity, we reasoned that PtdIns3P might 161 be required to recruit a specific factor, which controls endocytic identity. We therefore tested early 162 endocytic markers and investigated if these are still recruited to the forming macropinosome. 163 One of the first factors recruited to forming macropinosomes is the RAB5 effector protein APPL1 [17]. In 164 addition to RAB5 it binds to the membrane by a PH and a BAR domain but does not interact with PtdIns3P 165 [33,34]. In non-treated control cells, we observed recruitment of mCherry-APPL1 to macropinosomes; 166 this localization occurs after the initial Phafin2 localization ( Figure 3A) [23]. In contrast, VPS34-inihibition 167 completely blocked recruitment of mCherry-APPL1 to forming macropinosomes ( Figure 3A). This was 168 surprising, as APPL1 recruitment is not dependent on PtdIns3P and APPL1 was not displaced from existing 169 vesicles by inhibition of VPS34 (Supplementary Figure S1). 170 Recruitment of APPL1 depends on RAB5, which is one of the key GTPases regulating early endocytosis and 171 is involved in macropinosome formation and closure [35] as well as endosome fusion and maturation [36, 172 37]. RAB5 is further important for recruitment and regulation of the endosomal 37,173 38]. We therefore tested if VPS34 inhibition would influence RAB5 recruitment. Using live cell imaging, 174 we observed that mCherry-RAB5a was recruited to macropinosomes directly after the first transient 175 Phafin2-GFP recruitment ( Figure 3B). In contrast, VPS34 inhibition completely abolished the recruitment 176 of mCherry-RAB5 to newly formed macropinosomes ( Figure 3B). While new vesicles were still initiated 177 and showed the first, transient Phafin2 localization, RAB5 was never recruited to the macropinosome, 178 suggesting that PtdIns3P is critical to establish RAB5 on macropinosome membranes. 179 In line with this, the RAB5 effector and FYVE domain containing protein Rabankyrin5 also failed to be 180 recruited. Rabankyrin5 localizes to macropinosomes and regulates endocytic trafficking [16,23]. Under 181 control conditions, mCherry-Rabankyrin5 localized to macropinosomes directly after the initial Phafin2 182 localization. In line with the absence of both RAB5 and PtdIns3P, VPS34 inhibition completely blocked this 183 recruitment ( Figure 3C). 184 Our findings that VPS34-inhibition completely abolishes recruitment of RAB5 and the RAB5-effectors 185 APPL1 and Rabankyrin5 to newly forming macropinosomes, suggest that VPS34-generated PtdIns3P is 186 essential for newly formed macropinosomes to recruit endosomal markers and establish endosomal 187 identity ( Figure 3D). 188

VPS34 inhibition imparts a recycling identity on macropinosomes 190
Live-cell imaging ( Figure 1E, 2F) suggested that nascent macropinosomes in SAR405-treated cells re-fused 191 with the plasma membrane. As macropinosomes under VPS34-inhibited conditions failed to establish 192 endocytic markers like RAB5, and harbored a PtdIns4P pool, we asked if these vesicles might have a 193 secretory identity. Therefore, we examined if VPS34 inhibition affected recruitment of different RAB 194 GTPases involved in recycling and secretion. To this end, we transfected different RAB GTPases involved 195 in endocytic recycling using a Phafin2-GFP stable line and tracked macropinocytosis by live cell imaging. 196 First, we examined RAB8, which is involved in exocytic trafficking to the plasma membrane but has also 197 been shown to localize to early macropinosomes [39-42]. 198 When imaging Phafin2-GFP and mCherry-RAB8 under untreated conditions, we observed that mCherry-199 RAB8 localized to macropinosomes in the early stages of formation at the plasma membrane. mCherry-200 RAB8 recruitment coincided with the first Phafin2-GFP recruitment. mCherry-RAB8 localized transiently 201 to macropinosomes and then dissociated with the onset of the second Phafin2-GFP recruitment ( Figure  202 4A). Upon VPS34-inhibition, mCherry-RAB8 showed increased and persistent localization to 203 macropinosomes. This localization lasted until the re-fusion of the macropinosome with the plasma 204 membrane ( Figure 4A). Next, we examined RAB11, which mediates slow recycling via the endosomal 205 recycling compartment [43]. Under control conditions, we observed no recruitment of mCherry-RAB11 to 206 forming macropinosomes ( Figure 4B). However, after addition of VPS34-inhibitor, we observed a strong 207 recruitment of RAB11 to forming macropinosomes ( Figure 4B). The recruitment of mCherry-RAB11 upon 208 VPS34 inhibition started with the peak of the Phafin2-GFP recruitment and persisted until the 209 disappearance of macropinosomes. Next, we analyzed the behavior of RAB10, a small GTPase belonging 210 to the same family as RAB8, which is also involved in endocytic recycling [44][45][46]. We observed that 211 mCherry-RAB10 transiently accumulated on forming macropinosomes in untreated cells directly after the 212 first Phafin2 recruitment ( Figure 4C). In SAR405-treated cells, initial mCherry-RAB10 recruitment was 213 unchanged, but mCherry-RAB10 showed a persistent accumulation on macropinosomes until they re-214 fused with the membrane ( Figure 4C). 215 Taken together, we find that VPS34 inhibition increases recruitment of RAB proteins involved in recycling 216 and secretion, like RAB8, RAB11 and RAB10 on newly forming macropinosomes, suggesting that these 217 vesicles gain a secretory identity ( Figure 4D). 218

RAB5 is critical for successful macropinosome formation 220
We next asked if the failure of forming macropinosomes to recruit RAB5 -and thus establish endosomal 221 identity -is the cause for the re-fusion of the vesicle with the plasma membrane. To test this hypothesis, 222 we depleted all three RAB5 isoforms (RAB5A, RAB5B, RAB5C) by siRNA and monitored macropinocytosis 223 in control and knockdown cells. First, we tested if RAB5 depletion affects fluid phase uptake by measuring 224 10 kDa dextran internalization using flow cytometry. Both non-treated cells and siControl treated cells 225 showed similar efficiency in dextran uptake. In contrast, either treatment with SAR405 or depletion of 226 RAB5 (siRAB5ABC) led to a strong reduction in dextran uptake ( Figure 5A,B). 227 The role of RAB5 in macropinosome formation is not completely clear. While some studies place RAB5 228 already at the closure of the macropinosome and the scission from the plasma membrane [35], we see 229 RAB5 only arriving once the macropinosome has been closed and after the initial Phafin2 recruitment [23] 230 . We therefore asked if RAB5 knockdown cells could still initialize macropinosomes, or if already this step 231 was perturbed. To track macropinosome formation, we used plasma-membrane localized mCherry 232 (MyrPalm-mCherry) together with mNeonGreen-2xFYVE to mark the endosomal stage of 233 macropinosomes. Using this assay, we tracked macropinosomes from the formation of cup-shaped 234 membrane ruffles until they entered the endocytic pathway. Cells transfected with siControl showed 235 normal formation of macropinosomes, with membrane ruffles forming cup-shaped, closed vesicles 236 (outlined by MyrPalm-mCherry), which then gradually gained 2xFYVE, indicating that they accumulated 237 PtdIns3P and established endosomal identity ( Figure 5C,D). Likewise, cells depleted for RAB5 (siRAB5ABC) 238 were able to form macropinosomes ( Figure 5C,D). However, these vesicles did not gain 2xFYVE and 239 ultimately re-fused with the plasma membrane. Tracking of individual macropinosomes showed that in 240 control conditions, ~ 45 % of cup-shaped membrane ruffles transitioned to the endosomal stage, whereas 241 less than 5% were able to do so in RAB5ABC knockdown cells ( Figure 5E). 242 243 Thus, the lack of RAB5 shows the same phenotype as the lack of PtdIns3P, suggesting that the failure to 244 recruit RAB5 in the absence of PtdIns3P could be the cause of the observed re-fusion of the 245 macropinosomes with the plasma membrane. 246 247

Forced recruitment of RAB5 does not rescue macropinosome survival upon VPS34 inhibition 248
Recent studies show that RAB5 not only recruits the VPS34 kinase complex, but that PtdIns3P enhances 249 RAB5 membrane recruitment, which likely helps to activate RAB5 [47]. As we observed that RAB5 failed 250 to be recruited to newly formed macropinosomes, we hypothesized that the initial recruitment and 251 activation of RAB5 could be PtdIns3P dependent. In contrast to the dynamic wild-type RAB5, constitutively 252 active RAB5 (RAB5Q79L) is permanently membrane associated once recruited [48,49]. We therefore 253 asked if expression of active RAB5 would be sufficient to overcome the effect of VPS34 inhibition. To this 254 end, we expressed mCherry-tagged RAB5Q79L in cells stably expressing Phafin2-GFP ( Figure 6A). Under 255 control conditions, we observed robust recruitment of RAB5Q79L after the first Phafin2 recruitment, 256 similar to wild-type RAB5 ( Figure 6A). Addition of SAR405 did not affect RAB5Q79L recruitment, and 257 treated cells showed a similar recruitment as non-treated cells ( Figure 6A, movie S2). However, although 258 these macropinosomes successfully recruited RAB5Q79L, this was not sufficient to prevent their re-fusion 259 with the plasma membrane. 260 261 Several studies have placed RAB5 early in macropinocytosis, in some studies as early as during ruffle 262 closure [16,35,50]. As we observed that RAB5Q79L recruitment occurred only after the initial Phafin2 263 recruitment, we wanted to exclude that this recruitment was too late to rescue macropinosome survival 264 in SAR405 treated cells. We therefore exchanged the C-terminal CXC motif of RAB5 with the polybasic 265 region and CAAX box of KRAS. This motif mediates robust plasma membrane recruitment [51]. Live cell 266 imaging showed strong plasma membrane localization of RAB5Q79L-CAAX in both the presence and 267 absence of SAR405 ( Figure 6B). However, even the expression of these constructs was unable to rescue 268 the macropinosome survival phenotype in SAR405-treated cells ( Figure 6B). Quantification of 269 macropinosome survival showed that expression of RAB5Q79L or RAB5Q79L-CAAX did not affect the 270 average survival time of newly-forming vesicles upon SAR405 treatment ( Figure 6C). This was surprising, 271 as we found that vesicles with established endosomal identity were only mildly affected by SAR405 ( Figure  272 2B). 273

274
The transition from RAB8 to RAB5 requires VPS34 activity 275 As we found that SAR405 treatment led to the recruitment of secretory GTPases such as RAB8, RAB10 and 276 RAB11 on newly-formed macropinosomes, we asked if SAR405 would establish a secretory identity on 277 these vesicles even in the presence of RAB5. In non-treated cells expressing wild-type RAB8 and wild-type 278 RAB5, we found a RAB switch, where RAB8 was dissociating, whereas RAB5 was associating with the 279 newly-formed macropinosome ( Figure 6D, movie S3). Intriguingly, in SAR405 treated cells, we observed 280 robust and persistent localization of RAB8, even when the macropinosome in parallel recruited RAB5Q79L 281 ( Figure 6E, movie S4), and the vesicle re-fused with the plasma membrane. This suggests that the presence 282 of active RAB5 is not sufficient to fully impart endocytic identity to macropinosomes in the absence of 283 PtdIns3P, as these vesicles still acquire secretory characteristics. This implies that PtdIns3P could function 284 to either recruit a yet unknown determinant of endosomal identity, or a displacement factor for RAB8 and 285 other secretory GTPases, to stabilize the endocytic character of the newly-formed vesicle. 286 287

Loss of PtdIns3P induces RAB8 and RAB10 mediated recycling of macropinosomes 288
In order to test if RAB8 drives the re-fusion of newly-formed macropinosomes with the plasma membrane, 289 we depleted both isoforms of RAB8 by siRNA and analyzed the formation of macropinosomes. Indeed, in 290 contrast to cells treated with control siRNA, we observed an increased success rate of macropinosome 291 formation in cells depleted for RAB8A and RAB8B ( Figure 7AB). On the other hand, overexpression of RAB8 292 reduced the number of successfully formed macropinosomes ( Figure 7C). We also knocked down the 293 related GTPase RAB10, as well as both RAB8 and RAB10, and assayed fluid phase uptake by flow cytometry 294 ( Figure 7D-F). In both cases, we observed enhanced fluid phase uptake in comparison to control siRNA 295 ( Figure 7D). These results are consistent with a role of RAB8 and RAB10 in recycling of macropinosomes. 296

297
We then proceeded to test if inhibition of RAB8 activity could suppress the effect of VPS34 inhibition and 298 prevent re-fusion of nascent macropinosomes with the plasma membrane. As depletion of RAB8 by siRNA 299 was not complete, we chose to express dominant negative RAB8T22N instead. This mutant is unable to 300 bind GTP, which often results in more robust phenotypes than siRNA-mediated depletion. We transfected 301 cells expressing Phafin2-GFP with mCherry-RAB8T22N, treated them with SAR405 and analyzed 302 macropinosome formation and survival by live cell microscopy ( Figure 7G,H). In control cells, SAR405 303 treatment led to rapid re-fusion of nascent macropinosomes with the plasma membrane. In contrast, 304 expression of RAB8T22N stabilized newly-forming macropinosomes even in the absence of PtdIns3P and 305 suppressed the re-fusion with the plasma membrane ( Figure 7G,H, movie S5). 306 307 Taken together, our results show that VPS34 regulates the balance between macropinsome maturation 308 and exocytosis by controlling the dynamic transition of PtdIns4P to PtdIns3P, and RAB8 to RAB5 (Figure 309 In this study, we provide insight into the mechanisms that establish the membrane identity of newly-313 formed vesicles by using macropinosomes as a model system. Under normal conditions, macropinosome 314 maturation is characterized by a change in PIs from PtdIns4P to PtdIns3P in concert with a transition of 315 RAB8 to RAB5. When VPS34 is inhibited, macropinosomes accumulate RAB8 and other recycling factors, 316 whereas RAB5 recruitment and endocytic maturation is prevented. This leads to re-fusion of the 317 macropinosome with the plasma membrane ( Figure 7I). Since the forced recruitment of RAB5 to SAR405 318 treated macropinosomes failed to rescue macropinosome survival, this indicates that VPS34 has a dual 319 role in macropinosome maturation. In addition to enabling the recruitment of RAB5, VPS34 is also 320 required for the removal of RAB8. Such a mechanism could ensure that forming macropinosomes will 321 mature through the endosomal pathway to deliver their cargo for degradation in lysosomes, rather than 322 being recycled and exocytosed. 323 How can VPS34 control the fate of newly-formed vesicles? Early endocytic vesicles carry RAB5, which 324 recruits effector proteins that define the biochemical properties of the vesicle. In addition, these vesicles 325 gradually gain PtdIns3P, which can then recruit other effector proteins and ultimately controls the 326 maturation of the endocytic vesicles to early and late endosomes [28,37]. RAB5 binds to the VPS34 kinase 327 complex, which generates PtdIns3P and regulates its activity [38,52]. At the same time, PtdIns3P 328 enhances membrane recruitment of RAB5 [47] and could potentially form a positive feedback loop. 329 PtdIns3P also drives the maturation of early to late endosomes and the switch of RAB5 to RAB7 [53][54][55]. 330 We find that PtdIns3P -generated by the VPS34 kinase complex -is not only required for the maturation 331 of early to late endosomes but is also critical for the initial establishment of endosomal identity. Inhibition 332 of VPS34 -and consequently the absence of PtdIns3P -completely blocks the recruitment of RAB5 to 333 newly formed macropinosomes and prevents these vesicles from gaining endosomal identity. 334 RAB5 recruitment and activation is a fine-tuned system that heavily relies on the guanine nucleotide 335 exchange factor (GEF) Rabex5, complexed to the RAB5 effector Rabaptin5 as well as the endosomal 336 VPS34-complex. PtdIns3P could act as an initial recruiter of RAB5 [47,56]. Membrane-associated RAB5 337 could be activated by Rabex5, and active RAB5 would then recruit more Rabex5-Rabaptin complex, thus 338 driving a positive feedback loop [57]. Active RAB5 would also recruit the endosomal VPS34 complex and 339 thereby produce more PtdIns3P, further promoting the recruitment of RAB5. The gradual increase of 340 PtdIns3P at endosomal membranes supports such a positive feedback model. 341 The inhibited recruitment of RAB5 and the RAB5 effectors APPL1 and Rabankyrin5 that we observed in 342 response to VPS34 inhibition could be due to the inability of RAB5 and its activation complex to overcome 343 this initial hurdle to establish active RAB5 domains without active VPS34. This would also fit with our 344 observation that constitutively active RAB5Q79L can rescue RAB5 localization to newly formed 345 macropinosomes under VPS34-inhibited conditions. 346 This model would imply that either an initial pool of PtdIns3P or a stochastic activation of RAB5 is needed 347 to overcome the initial inertia and establish active RAB5 domains on macropinosomes. Indeed, previous 348 studies report a transient, phosphatase derived PtdIns3P pool which could fulfill this role. A phosphatase 349 cascade on macropinosomes can metabolize PtdIns3,4,5P3 via PtdIns3,4P2 to PtdIns3P [14]. It is tempting 350 to speculate that this PtdIns3P pool could act as an initial trigger for RAB5 recruitment. Alternatively, 351 stochastic membrane association and activation of RAB5 might be sufficient to activate VPS34 and thereby 352 initiate recruitment of RAB5. 353 However, active RAB5 is not sufficient to overcome the need for PtdIns3P, since neither constitutive active 354 RAB5Q79L, nor membrane-tethered RAB5Q79L-CAAX, which both are localizing to newly formed 355 macropinosomes, prevent the re-fusion of macropinosomes in the absence of active VPS34. We find that 356 PtdIns3P, produced by VPS34, has a dual function. Apart from the recruitment of RAB5, it is also needed 357 to prevent nascent macropinosomes from acquiring a secretory identity. 358 We show that macropinosomes undergo a shedding of their plasma membrane identity, by undergoing a 359 RAB and phosphoinositide switch from a RAB8 and PtdIns4P positive stage to a RAB5 and PtdIns3P positive 360 stage. This corresponds to a switch from a recycling identity -as PtdIns4P and RAB8 are key regulators of 361 secretion -to an endocytic identity. VPS34 activity appears to be a critical regulator of this switch, as in 362 the absence of PtdIns3P, RAB8 and PtdIns4P remain on the macropinosome and ultimately drive its 363 refusion with the plasma membrane. Similarly, other recycling factors -such as RAB11 and RAB10 -are 364 recruited if VPS34 is inhibited. Recycling and re-fusion with the plasma membrane can be blocked by 365 overexpression of dominant negative RAB8 T22N, suggesting that the prolonged recruitment of RAB8 is 366 one of the major drivers of macropinosome re-fusion. Our findings also suggest that PtdIns3P is required 367 to remove RAB8 and RAB10 from the macropinosome and thus remove this secretory identity. 368 Interestingly, a recent study also described a re-fusion of macropinosomes with the plasma membrane 369 Taken together, we show in this study that forming macropinosomes need to undergo a transition from 396 RAB8 to RAB5 in order to mature along the endo-lysosomal pathway, similar to the transition from early 397 endosomes marked by RAB5 to late endosomes and lysosomes marked by RAB7 [53]. This transition is 398 functionally required for successful macropinosome maturation and is regulated by phosphoinositide 399 transitions and gradients that allow a tightly controlled recruitment of effector proteins needed for the 400 different stages. The default secretory identity of macropinosomes could be a fail-safe mechanism that 401 prevents the spurious uptake of vesicles and requires that they actively gain endocytic activity. This would 402 prevent the accumulation of "vagrant" vesicles lacking a defined identity. At the same time, it could 403 provide a first layer of defense against pathogen invasion, as a pathogen entering cells by induced 404 macropinocytosis would have to actively modulate the vesicle coat or end up in either the endolysosomal 405 pathway or the secretory pathway [59]. 406 We propose the following model for endocytic macropinosome maturation: Newly forming 407 macropinosomes have a secretory identity by default. In order to mature along the endo-lysosomal 408 pathway, they need to be stripped of secretion promoting lipids such as PtdIns4P and RABGTPases, such 409 as RAB8 and RAB10. Maturation along the endo-lysosomal pathway requires both PtdIns3P, RAB5 and 410 RAB5 effectors. Together VPS34 and RAB5 create a stabilizing feedback loop, which allows for the 411 generation of more PtdIns3P and in turn displacement of RAB8 and RAB10 as well as recruitment of RAB5 412 and RAB5 effectors, stably establishing macropinosomes in the endo-lysosomal pathway. Depleting either 413 PtdIns3P or RAB5 inhibits entry into the endo-lysosomal pathway and instead stabilizes recycling identity 414 on the newly formed macropinosomes, by recruitment of RAB8, RAB11 and RAB10 and PtdIns4P. Since 415 entry into the endosomal pathway is blocked, these factors mediate re-fusion of the newly formed 416 macropinosomes with the plasma membrane in order to maintain membrane homeostasis. 417 While our understanding of the exact regulation of the RAB8 to RAB5 transition is not yet complete, our 418 findings elucidate how macropinosome maturation is regulated. This knowledge can potentially be 419 utilized for investigating new treatment approaches for pathogen infections or cancers driven by 420 macropinocytosis. 421

Cell lines 423
Experiments were performed in hTert-RPE1 cells (ATCC ® CRL-4000 ™ ) and HT1080 cells (ATCC ® CCL-121 ™ ). 424 Cells were purchased from ATCC and authenticated by genotyping. Cells were verified to be free of 425 mycoplasma contamination and regularly tested after manipulation. 426 Stable hTert-RPE1 or HT1080 cell lines were lentivirus generated pools using a third-generation system 427 [60]. Tagged versions of proteins of interest were subcloned into a Gateway entry vector with a PGK or 428 CMV promotor by standard molecular biology techniques. Transfer vectors were generated by a 429 Gateway LR recombination with a destination vector of choice. Virus was packaged using a third-430 generation system, by co-transfecting a packaging vectors encoding for GAG/POL, REV and a VSV-G 431 based envelope in addition to the destination vector. Viral supernatant was collected 72h after 432 transfection. Cells were transduced with low viral titers (MOI<1) and stable pools were generated 433 through antibiotic selection. 434 The hTert-RPE1 line stably expressing GFP tagged Phafin2 was generated previously [23] and used as a 435 background for further stable lines. In this study, the following stable cell lines were used: hTert-RPE 436 mNeonGreen-Rab5, hTert-RPE1 mCherry-Rab8, hTert-RPE1 mNeongreen-2xFYVE -MyrPalm-mCherry 437 and HT1080 Phafin2-mNeonGreen-MyPALM-mCherry. 438 Cell Culture 439 hTert-RPE1 cells were maintained in DMEM-F12 medium (Gibco) supplemented with 10% fetal bovine 440 serum (Merck Life Science (Sigma)) and 5 U ml −1 penicillin and 50 μg ml −1 streptomycin (Merck Life 441 Science (Sigma)) at 37 °C and 5% CO 2 . HT1080 cells were maintained in DMEM (Sigma Aldrich) 442 supplemented with 10% fetal bovine serum and 5 U ml −1 penicillin and 50 μg ml −1 streptomycin at 37 °C 443 and 5% CO 2 . 444 Antibodies 445 The following primary antibodies were used: anti-RAB5A, Santa Cruz Biotechnology (sc-46692). 446 Recombinant anti-RAB8A, Abcam (ab188574). Monoclonal anti-Vinculin, Sigma Aldrich/ Merck (V9131). 447 Monoclonal anti-γ-Tubulin Sigma Aldrich/ Merck (T6557). Secondary antibodies were from LI-COR. Technologies). Cells were imaged in two or three colors in conventional mode every 5 seconds between 482 12-25 minutes in total. Acquired images were aligned and deconvolved using softWoRx software 483 (Applied Precision, GE Healthcare) and x-y alignment was checked and when necessary re-calibrated 484 using the "GE Image Registration slide". 485

OPM light sheet microscopy 486
Light sheet microscopy was performed using a custom-built oblique plane microscope. The microscope 487 layout followed the general plan for a stage scanning OPM microscope described in Sapoznik et al [32]. 488 The microscope was built around the ASI modular microscope platform. The inverted microscope 489 consisted of a ASI MIM microscope body, an ASI FTP Z stage and a stage scanning optimized ASI X/Y stage 490 (ASI imaging). Environmental control was provided by an Okolabs stage incubator. A 60x 1.3 NA silicon oil 491 immersion lens (Olympus) served as primary objective. ASI cage elements were used to construct the rest 492 of the optical train, consisting of a 300mm tube lens (TL1), a 357mm tube lens (Tl2) and a 40x 0.95 NA air 493 remote objective (Nikon). The image generated by the remote objective was collected by a bespoke 494 tertiary objective (AMS-AGY v1.0, Special Optics) and focused by a 250 mm tube lens (TL3) on the sensor 495 of a Andor Zyla 4.2 camera, resulting in a final pixelsize of 91 nm. The excitation beam was coupled into 496 the light path by a dichroic mirror placed between TL2 and the remote objective. We used a gaussian light 497 sheet generated by an ASI light sheet generator (ASI Imaging) equipped with a cylindrical lens, coupled to 498 a Toptica laser source (Toptica) with 405, 488, 561 and 647 nm laser lines. Image acquisition was 499 performed by stage scanning. All hardware was synchronized by an ASI Tiger controller (ASI Imaging) and 500 controlled by the ASI diSPIM plugin in MicroManager [63]. Deconvolution and deskewing were performed 501 using the LLSPy software package (https://github.com/tlambert03/LLSpy/), the resulting images were 502 visualized using Imaris. 503 Quantitative real-time PCR of mRNA expression 504 mRNA expression analysis after siRNA transfection was performed as previously described in Pedersen 505 et al. 2020 [64]. The primers used in this paper were obtained from Qiagen and are QT00067767 for 506 RAB10 and QT00000721 for TATA-binding protein (TBP) as a reference housekeeping gene. 507

Image processing and data analysis 508
All live cell imaging files were deconvolved using SoftWorx Software (GE Healthcare) before being 509 analyzed and prepared for presentation. Analysis of life cell imaging was performed using Fiji and 510 custom-made Python scripts (https://github.com/koschink/Phafin2). The intensity tracks were plotted 511 using the Python Seaborn library. Newly forming macropinosomes were tracked manually in Fiji using 512 Phafin2 as a marker for macropinosomes. When tracking macropinosomes in the absence of Phafin2-513 GFP, we used other transfected fluorescently tagged proteins, as well as the exclusion of cytosol from 514 the big vesicular structures to follow the progress of the macropinosomes. The limiting membrane of 515 the tracked vesicle was marked in each frame and added to the ROI manager. From the ROI's the 516 program measured the fluorescence intensity in every channel and created galleries for the tracked 517 macropinosomes. Processing of the fluorescence intensity was performed by a Python script, which 518 aligned the tracks to the first Phafin2-GFP peak. 519

Statistics 520
No statistical methods were used to pre-determine sample size. All statistical calculations were derived 521 from at least 3 biological replicates. For microscopy-based assays, each biological repeat included 522 sufficient cell numbers to ensure that the effects could be robustly measured and was repeated at least 523 3 times. The number of experimental days, cells and vesicles analyzed are indicated in the figure  524 legends. Data was tested for normal distribution using GraphPad Prism Version 8. For comparison of two 525 samples with each other unpaired t-test was used for normally distributed data, while the Mann-526 Whitney test was used for non-normal data sets. For multiple comparisons we used the one-way 527 analysis of variance (ANOVA) or the Kruskal-Wallis test with a suitable post hoc test. One-sample t-test 528 was used when the value of control sample was set to 1. All error bars give the mean values ± 95% CI or 529 as indicated in the figure legends. Samples were not randomized for this study. 530

Data Availability 531
Source data for the figures is provided with the paper. 532 Live imaging of hTert-RPE1 cells stably expressing mCherry-MyrPalm (magenta) and mNeonGreen-777 2xFYVE (green) in the presence of VPS34-inhibitor SAR405 (3 µM). The movie shows the formation of a 778 macropinosome as marked by an arrow and follows its progress after formation (scalebar 2 µm). 779 generated constructs and wrote image and data processing software. H.St. provided funding, co-807 supervised the study, and discussed data. K.O.S. and C.R. conceived and supervised the study and 808 contributed to data validation and visualization. All co-authors reviewed and edited the manuscript. 809

Acknowledgements 810
We thank Ling Wang for the qPCR analysis of RAB10, we thank Simona Migliano for the help with 811 performing an APEX screen and proteomics analysis. We thank Jost Enninga for discussions and help 812 during exploratory proteomics. We thank Ulrikke Dahl-Brinch for help with cloning of constructs. We 813 thank Kia Wee Tan for discussions on the project. 814 K.O.S was supported by a Career grant from the South-Eastern Norway Regional Health Authority 815 (2020038)