DRAM1 requires PI(3,5)P2 generation by PIKfyve to deliver vesicles and their cargo to endolysosomes

Endolysosomal vesicle trafficking and autophagy are crucial degradative pathways in maintenance of cellular homeostasis. The transmembrane protein DRAM1 is a potential therapeutic target that primarily localises to endolysosomal vesicles and promotes autophagy and vesicle fusion with lysosomes. However, the molecular mechanisms underlying DRAM1-mediated vesicle fusion events remain unclear. Using high-resolution confocal microscopy in the zebrafish model, we show that mCherry-Dram1 labelled vesicles interact and fuse with early endosomes marked by PI(3)P. Following these fusion events, early endosomes mature into late endosomes in a process dependent on the conversion of PI(3)P into PI(3,5)P2 by the lipid kinase PIKfyve. Chemical inhibition of PIKfyve reduces the targeting of Dram1 to acidic endolysosomal vesicles, arresting Dram1 in multivesicular bodies, early endosomes, or non-acidified vesicles halted in their fusion with early endosomes. In conclusion, Dram1-mediated vesicle fusion requires the formation of PI(3,5)P2 to deliver vesicles and their cargo to the degradative environment of the lysosome.


22
Endocytic processes are specialised in the uptake of substances from the microenvironment of the 23 cell. Although most of the endocytic cargo is used for cellular sustenance or recycled back to the 24 plasma membrane, a proportion of endocytosed material (e.g. pathogens and remnants of dead cells) 25 is routed towards acidic and hydrolytic lysosomes for its degradation (Huotari and Helenius, 2011). 26 While the endolysosomal system is responsible for degradation of unwanted extracellular material, 27 autophagy performs a similar housekeeping function for the removal of intracellular material. During 28 autophagy, cytoplasmic content is captured in double membraned vesicles and delivered to the 29 endolysosomal system for degradation and recycling (Glick et al., 2010). In this way, autophagy 30 replenishes nutrient levels in times of cellular starvation, and clears the cytoplasm of unwanted 31 elements like protein aggregates, malfunctioning organelles, and intracellular pathogens (Saha et  Tzeng and Wang, 2016). 37 DNA damage regulated autophagy modulator 1 (DRAM1) regulates autophagy and endolysosomal 38 fusion events. DRAM1 was first identified as a cellular stress-induced regulator of autophagy and cell 39 death downstream of tumour suppressor protein p53 (Crighton et al., 2006). DRAM1 primarily 40 localises to lysosomes but can also be detected on other organelles of the vesicular trafficking system, 41 including endosomes, autophagosomes, autolysosomes, the Golgi apparatus, and the endoplasmic 42 reticulum (Crighton et al., 2006;Mah et al., 2012). Furthermore, DRAM1 was found to regulate fusion 43 between autophagosomes and lysosomes, a process called autophagic flux (Zhang et al., 2013). When 44 host cells detect pathogenic mycobacteria that cause tuberculosis, DRAM1 expression is activated 45 downstream of the immunity regulating transcription factor NFκB (van der Vaart et al., 2014). 46 Knockdown or knockout of dram1 increased susceptibility to mycobacterial infection in zebrafish, 47 identifying Dram1 as a host resistance factor (van der Vaart et al., 2014; Zhang et al., 2020). In support 48 of this finding, overexpression of dram1 was protective against mycobacterial infection by enhancing 49 autophagic defences against intracellular bacteria and stimulating vesicle fusion events with 50 lysosomes (van der Vaart et al., 2014). Although the effects of DRAM1 activation on autophagy and 51 endolysosomal fusion events are described for several situations, its underlying molecular function in 52 these processes remains unknown. 53 Endocytic cargo is sorted in early endosomes marked by the GTPase Rab5 (Zerial and McBride, 2001). 54 Early endosomes containing cargo destined for degradation gradually replace Rab5 on their 55 membrane for Rab7, while lowering their luminal pH from values above pH 6 to pH 6.0-4.9 to become 56 late endosomes (Maxfield and Yamashiro, 1987; Zerial and McBride, 2001). During this phase of the 57 maturation process, the outer endosomal membrane starts budding inwards to form intraluminal 58 vesicles (Raiborg et al., 2002;Sachse et al., 2002). The resulting multivesicular bodies are a type of late 59 endosome that also receive cargo destined for degradation by fusing with autophagosomes (Fader 60 and phosphatases, generating a total of 7 different PIs in animals (Banerjee and Kane, 2020). Typically, 70 early endosomes are defined by the presence of PI(3)P in their membrane, which is converted into 71 PI(3,5)P 2 by the lipid kinase PIKfyve (Fab1p in yeast) during maturation into late endosomes (Wallroth 72 and Haucke, 2018 resource to predict functional sites in the human DRAM1 protein ( Figure 1A). This analysis confirmed 94 the previously reported presence of 6 transmembrane domains (Crighton et al., 2006), suggesting that 95 DRAM1 is embedded in cellular membranes with parts of the protein exposed to opposite sides of this 96 membrane. Amongst the predicted protein domains, we identified two domains that support a 97 function for DRAM1 in vesicle trafficking. Eps15 homology (EH) domains are generally present in 98 proteins that regulate endocytosis or vesicle trafficking processes (Naslavsky and Caplan, 2005). The 99 autophagy-related protein Atg8 and its homologs LC3 and GABARAP are markers of autophagosomes 100 (Glick et al., 2010). The presence of Atg8 interacting domains therefore suggests that DRAM1 can 101 interact with the autophagy-machinery. 102 We aimed to study the dynamic localisation of DRAM1 during endolysosomal maturation processes. 103 For this purpose, we used a previously described mCherry-Dram1 construct under control of the 104 ubiquitous beta actin promoter to generate a transgenic zebrafish line fluorescently reporting the 105 subcellular localisation of Dram1 (van der Vaart et al., 2014), named Tg(bactin:mCherry-dram1). We 106 could readily trace mCherry-Dram1 over time by confocal imaging epithelial cells in the thin tissue of 107 the tail fin of 3 days post fertilisation (dpf) zebrafish larvae ( Figure 1B and C). Time-lapse imaging 108 revealed that mCherry-Dram1 labels motile and morphologically diverse globular and tubular vesicles 109 ( Figure 1C, Supplementary movie 1). We used a LysoTracker probe that accumulates and fluoresces in 110 endolysosomal compartments with low luminal pH to confirm that mCherry-Dram1 mainly localises 111 to these acidic organelles ( Figure 1D). increased the number of autophagosomes observed per cell ( Figure S1A and B). We can therefore use 119 artificial expression of mCherry-Dram1 as a gain-of-function approach to study the role of Dram1 in 120 vesicle trafficking. 121 To visualise endosomal vesicles that Dram1 interacts with, we used a transgenic line that fluorescently 122 reports early endosomes in basal cell layer epithelial cells of the zebrafish epidermis: 123 TgBAC(ΔNp63:Gal4FF) la213 ; Tg(4xUAS:EGFP-2xFYVE) la214 , hereafter referred to as GFP-2xFYVE 124 (Rasmussen et al., 2015). The GFP-2xFYVE probe incorporates specifically in membranes containing 125 PI(3)P via its FYVE domains, thereby labelling early endosomes. However, a specific pool of PI(3)P also 126 labels (nascent) autophagosomes (Nascimbeni et al., 2017). We therefore first tested the specificity 127 of the GFP-2xFYVE probe by combining it with a Tg(bactin:mCherry-Lc3) line that marks 128 autophagosomes, hereafter referred to as mCherry-Lc3. We found that GFP-2xFYVE and mCherry-Lc3 129 labelled autophagosomes rarely colocalise, but label distinct vesicles that occasionally are found in 130 close proximity of each other ( Figure S1C). Since the GFP-2xFYVE probe does not label 131 autophagosomes, we therefore refer to vesicles labelled by GFP-2xFYVE in their membrane as 'early 132 endosomes'. 133 Confocal imaging of the GFP-2xFYVE and mCherry-Dram1 transgenes in the accessible zebrafish tail 134 fin tissue allowed us to study endosomal dynamics in great detail. Time-lapse imaging demonstrated 135 that globular mCherry-Dram1 labelled vesicles frequently interact with the PI(3)P-containing 136 membrane of early endosomes ( Figure 1E). We could also observe mCherry-Dram1 vesicles forming 137 tethers between two distant early endosomes that are subsequently brought together ( Figure 1E). 138 Ultimately, mCherry-Dram1 labelled vesicles fuse with early endosomes and localise to their lumen. 139 Early endosomes that have undergone such fusion events gradually lose the GFP-2xFYVE labelling of 140 their membrane, representing a reduction of PI(3)P lipids present in these membranes. Taken  141 together, ectopically expressed mCherry-Dram1 labels acidic and morphologically diverse vesicles that 142 interact and fuse with early endosomes. Subsequently, these early endosomes alter the PI lipid 143 composition of their membrane. 144 Inhibiting PIKfyve and PI(3,5)P2 formation affects mCherry-Dram1 labelled vesicles 145 Early endosomes that have fused with mCherry-Dram1 labelled vesicles lose the GFP-2xFYVE labelling 146 of their membrane. We hypothesised that the enzymatic activity of the 1-phosphatidylinositol 3-147 phosphate 5-kinase PIKfyve was responsible for the conversion of PI(3)P into PI(3,5)P2 in this process. 148 To test this, we used YM201636 and apilimod to selectively inhibit the kinase activity of PIKfyve ( vesicles marked by PI(3)P in their membranes ( Figure S2A). As the more potent and selective of the 154 two inhibitors (Cai et al., 2013), we tested a range of treatment durations for apilimod and found that 155 a relatively short incubation of 2 hours robustly enlarged GFP-2xFYVE labelled vesicles ( Figure S2B). 156 We selected this treatment window for further experiments in which we exposed zebrafish larvae 157 expressing both the GFP-2xFYVE and mCherry-Dram1 constructs to either apilimod or DMSO as a 158 solvent control. We used confocal microscopy to image epithelial cells in the zebrafish tail fin and 159 analysed the number and morphology of GFP-2xFYVE and mCherry-Dram1 vesicles per cell ( Figure S3). 160 Inhibition of the enzymatic activity of PIKfyve resulted in enlarged early endosomes and mCherry-161 Dram1 labelled vesicles, while the number of both types of vesicles per cell was reduced ( Figure 2A mCherry-Dram1 in epithelial cells in the zebrafish tail fin, we could categorise mCherry-Dram1 signal 175 into four groups: 1) mCherry-Dram1 signal that is distant from early endosomes; 2) mCherry-Dram1 176 signal that is in close proximity or directly adjacent to early endosomes; 3) mCherry-Dram1 signal that 177 overlaps with the membrane of early endosomes; and 4) mCherry-Dram1 signal that is contained 178 within early endosomes ( Figure 3A). We then used Fiji/ImageJ to analyse the localisation of mCherry-179 Dram1 in respect to early endosomes according to these four categories ( Figure S3). We found that 180 PIKfyve inhibition reduced the number of times mCherry-Dram1 was localised distant from, adjacent 181 to, or overlapping with early endosomal membranes, while it increased the number of times that 182 mCherry-Dram1 was contained within early endosomes ( Figure 3B). However, since inhibition of 183 PIKfyve reduced the total number of mCherry-Dram1 vesicles per cell ( Figure 2C), we also analysed 184 the categories as a percentage of the total mCherry-Dram1 labelled vesicles present in each cell. We 185 found that PIKfyve inhibition increased the percentage of mCherry-Dram1 signal that is contained 186 within early endosomes or overlaps with early endosomal membranes, at the expense of the 187 percentage of mCherry-Dram1 signal that is localised distant from or adjacent to early endosomes 188 ( Figure 3C). In conclusion, mCherry-Dram1 accumulates in the lumen of early endosomes and on their 189 membranes when the conversion of PI(3)P into PI(3,5)P2 is inhibited. 190

Inhibition of PI(3,5)P2 formation reduces the dynamic interactions between mCherry-Dram1 and 191
early endosomes 192 The observation that mCherry-Dram1 labelled vesicles accumulated in and on early endosomes upon 193 inhibition of PIKfyve means that the dynamic interaction between these two types of vesicles was 194 altered. Either mCherry-Dram1 labelled vesicles interacted more frequently with early endosomes 195 upon inhibition of PIKfyve, or subsequent processes were inhibited that caused their accumulation. 196 We performed time-lapse imaging of the interactions between mCherry-Dram1 and early endosomes 197 to study these possible explanations. We exposed zebrafish larvae expressing both the GFP-2xFYVE 198 and mCherry-Dram1 constructs to either apilimod or DMSO as a solvent control and imaged epithelial 199 cells in the zebrafish tailfin after two hours of drug treatment using confocal microscopy ( Figure 4A). 200 In the control group, we observed many interactions between mCherry-Dram1 and early endosomes 201 over time. This included temporary 'kiss-and-run' interactions, as well as long term contact between 202 two or more vesicles which frequently ended in mCherry-Dram1 fusing into early endosomes 203 (Supplementary movie 2). In contrast, inhibition of PIKfyve greatly reduced the motility of both types 204 of vesicles, with interactions taking place infrequently and novel fusion events between mCherry-205 Dram1 and early endosomes only occurring rarely (Supplementary movie 3). Analysis of the number 206 of interactions that took place with mCherry-Dram1 per early endosome confirmed our observations, 207 as these were significantly reduced upon inhibition of PIKfyve ( Figure 4B). 208 209 The anticipated effect of PIKfyve inhibition is that PI(3)P in the membrane of early endosomes can no 210 longer be converted into PI(3,5)P 2. As observed before ( Figure 1E), early endosomes in the control 211 group gradually lost the PI(3)P lipids marked by GFP-2xFYVE in their membrane following fusion events 212 with mCherry-Dram1 ( Figure 4A, Supplementary movie 2). Upon inhibition of PIKfyve, early 213 endosomes that had already fused with mCherry-Dram1, or underwent novel fusion events on rare 214 occasions, no longer lost the GFP-2xFYVE labelling of their membranes (Supplementary movie 3). By 215 quantifying this process for multiple time-lapse recordings, we confirmed that the duration for which 216 GFP-2xFYVE labelling of early endosomal membranes remained detectable following fusion with 217 mCherry-Dram1 vesicles was significantly increased upon PIKfyve inhibition ( Figure 4C). Not all time-218 lapse recordings of epithelial cells in zebrafish tail fins were of equal length due to technical difficulties 219 associated with this type of imaging in live animals (e.g. samples drifting out of focus). We therefore 220 also plotted the duration for which a GFP-2xFYVE ring containing mCherry-Dram1 signal remained 221 detectable in relation to the total duration for which the cell in which the fusion event occurred could 222 be followed ( Figure 4D). This visualisation clearly illustrates the difference between the control group 223 in which mCherry-Dram1 frequently fused with early endosomes that subsequently lost the GFP-224 2xFYVE labelling of their membrane, and the apilimod treated group in which the majority of early 225 endosomes containing mCherry-Dram1 signal retain the GFP-2xFYVE labelling of their membrane for 226 the entire duration of the time-lapse. Taken together, inhibition of PI(3,5)P 2 formation reduced the 227 dynamic interactions between mCherry-Dram1 and early endosomes, and caused mCherry-Dram1 to 228 accumulate in early endosomes by halting processes that normally follow upon vesicle fusion. 229

230
Acidification of mCherry-Dram1 vesicles is reduced upon inhibition of PI(3,5)P2 formation 231 We have thus shown that early endosomes that have fused with mCherry-Dram1 labelled vesicles lose 232 the GFP-2xFYVE labelling of their membrane in a process dependent on the kinase activity of PIKfyve. 233 This loss of signal suggests that an endosomal maturation process takes place in which PI(3)P is 234 converted into PI(3,5)P2 present in late endosomal membranes. The maturation of early into late 235 endosomes is associated with a decrease in luminal pH (Maxfield and Yamashiro, 1987). This prompted 236 us to investigate how inhibition of PIKfyve affected the acidification of early endosomes and mCherry-237 Dram1 labelled vesicles. We therefore imaged GFP-2xFYVE and mCherry-Dram1 in epithelial cells in 238 the zebrafish tail fin, combined with LysoTracker staining to label acidic vesicles. In the control group, 239 we observed that the majority of mCherry-Dram1 labelled vesicles are acidic ( Figure 5A), confirming 240 our earlier findings ( Figure 1D and (van der Vaart et al., 2014)). GFP-2xFYVE labelled vesicles in the 241 control group varied in the extent of their acidity, ranging from (almost) no detectable LysoTracker 242 staining to clear staining of their lumen ( Figure 5A). This variation in acidity for PI(3)P labelled vesicles 243 likely reflects the gradual acidification of early endosomes that takes place as they mature. Upon 244 inhibition of PIKfyve by apilimod treatment, early endosomes continue to display this range of luminal 245 acidification, with smaller PI(3)P labelled vesicles frequently not or dimly stained by LysoTracker and 246 larger vesicles typically stained intensely ( Figure 5B). In contrast, mCherry-Dram1 labelled vesicles 247 appeared to be less frequently and less intensely stained by LysoTracker when PIKfyve was inhibited 248 ( Figure 5B). We used Fiji/ImageJ to analyse the spatial overlap (colocalisation) between mCherry-249 Dram1 and LysoTracker staining and found that the correlation between these two fluorescent signals 250 decreased significantly upon inhibition of PIKfyve ( Figure 5C). We therefore conclude that the 251 acidification of mCherry-Dram1 vesicles is at least partially dependent on the formation of PI(3,5)P 2 252 by PIKfyve. 253 254 While analysing the colocalisation between mCherry-Dram1 and LysoTracker, we encountered 255 multiple large mCherry-Dram1 labelled vesicles that contained acidic (Lysotracker stained) and non-256 acidic GFP-2xFYVE labelled vesicles ( Figure 5D). These intraluminal vesicles appeared to accumulate 257 within the mCherry-Dram1 labelled compartments, forming what resembles a multivesicular body. To 258 visualise the dynamics of these events, we performed time-lapse imaging of GFP-2xFYVE and mCherry-259 Dram1 combined with LysoTracker staining. In the control situation, a mCherry-Dram1 + /LysoTracker + 260 vesicle formed a tether between two early endosomes with dim LysoTracker staining, causing the two 261 early endosomes to fuse together ( Figure 5E and Supplementary movie 4). The mCherry-262 Dram1 + /LysoTracker + vesicle continued to interact with this newly formed endosome and ultimately 263 fused with it. Following this fusion event, the early endosome displayed more intense luminal 264 LysoTracker staining and lost the GFP-2xFYVE labelling of its membrane over time. This maturation 265 process forms a stark contrast to what occurred upon inhibition of PIKfyve. As described before ( Figure  266 4A), mCherry-Dram1 and GFP-2xFYVE labelled vesicles rarely interacted nor altered their existing 267 associations ( Figure 5E and Supplementary movie 5). Large mCherry-Dram1 labelled vesicles varied in 268 their acidity, ranging from no or dim LysoTracker staining to intense LysoTracker staining. Inside the 269 lumen of non-acidified mCherry-Dram1 labelled compartments, we regularly observed small acidic 270 vesicles that moved around in a seemingly random pattern ( Figure 5E and Supplementary movie 5). 271 These acidic intraluminal vesicles persisted over time, with no indication of releasing their content 272 into the lumen in which they reside. In conclusion, the kinase activity of PIKfyve is required for 273 mCherry-Dram1 labelled vesicles to tether early endosomes and fuse with them to kickstart a 274 maturation process in which their signature PI (3) Representative images were deconvoluted using the Iterative Deconvolution 3D plugin in Fiji/ImageJ 406 (Dougherty, 2005). 407

Image analysis 408
Raw imaging data was analysed in Fiji/ImageJ to obtain measurements for vesicle morphology, 409 interactions between vesicles, and colocalisation of fluorescent signals. For measurements of vesicle 410 morphology, a maximum intensity Z-projection was generated for a single layer of epithelial cells 411 imaged in the zebrafish tailfin tissue. Individual cells were selected and stored as regions of interest 412 (ROIs) using the Polygon selection tool. The Phansalkar Auto-Local Threshold method was used for 413 segmentation of vesicles. Segmented vesicles that were directly adjacent to each other were separated 414 using a Watershed function. The resulting individual vesicles were measured per cell using the Analyze 415 Particles function. 416 To measure interactions between vesicles, the same method as described above was used to segment 417 individual vesicles per cell. Vesicles labelled by their respective fluorescent signal were stored as ROIs. 418 Subsequently, the distance between each mCherry-Dram1 ROI and the nearest GFP-2xFYVE labelled 419 ROI was determined. Based on this measurement, mCherry-Dram1 ROIs were categorised into four 420 groups: 1) mCherry-Dram1 ROI that is distant from a GFP-2xFYVE ROI (distance ≥ 5 pixels); 2) mCherry-421 Dram1 ROI that is in close proximity or directly adjacent to a GFP-2xFYVE ROI (distance < 5 pixels); 3) 422 mCherry-Dram1 ROI that overlaps with a GFP-2xFYVE ROI; and 4) mCherry-Dram1 ROI that is 423 contained within a GFP-2xFYVE ROI. The Fiji/ImageJ plugin created to automate this analysis, called 424 'FYVE DRAM Analysis', is openly available for download via the Leiden University update site 425 (http://sites.imagej.net/Willemsejj/). 426 To analyse colocalisation between mCherry-Dram1 and LysoTracker Deep Red fluorescent signals, a 427 maximum intensity Z-projection was generated for a single layer of epithelial cells imaged in the 428 zebrafish tailfin tissue. The Gaussian Blur (sigma = 1) function was applied to decrease noise. After this, 429 the Li Threshold method, followed by the Analyze Particles function ('Show Mask'; size cut off of 10 430 pixels) was used to create a binary mask that excludes zero-zero pixels from the colocalisation analysis. 431 Finally, we used the Coloc 2 Fiji/ImageJ plugin (available via https://imagej.net/Coloc_2) to determine 432 the Pearson correlation coefficient between the two fluorescent signals. 433

Statistical analysis and data representation 434
Statistical analyses were performed using GraphPad Prism software (Version 5.01; GraphPad). All 435 experimental data (mean ± SEM) was analyzed using unpaired, two-tailed Mann-Whitney U tests for 436 comparisons between two groups and Kruskal-Wallis one-way analysis of variance with Dunn's 437 multiple comparison methods as a posthoc test for comparisons between more than two groups. (ns, 438 no significant difference; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001). The data sets from 439 each group are shown in a scatter plot (left) and a boxplot (right). In the scatter plots each dot 440 represents a data point, with the mean indicated by a horizontal line. Boxplots include 50% of the data 441 points, with a vertical line indicating the 95% confidence interval and a horizontal line indicating the 442 median. The only exception to this is Figure  isoforms that regulate autophagy.