Human AKTIP interacts with ESCRT proteins and functions at the midbody in cytokinesis

To complete mitosis, the intercellular bridge that links daughter cells needs to be cleaved. This abscission step is carried out by the sequential recruitment of ESCRT proteins at the midbody. We report here that a new factor, named AKTIP, works in association with ESCRTs. We find that AKTIP binds to the ESCRT I subunit VPS28, and show by high resolution microscopy that AKTIP forms a ring in the dark zone of the intercellular bridge. This ring is positioned in between the circular structures formed by ESCRTs type III. Functionally, we observe that the reduction of AKTIP impinges on the recruitment of the ESCRT III member IST1 at the midbody and causes abscission defects. Taken together, these data indicate that AKTIP is a new factor that contributes to the formation of the ESCRT complex at the midbody and is implicated in the performance of the ESCRT machinery during cytokinetic abscission.


42
To complete cytokinesis, cells need to cleave the intercellular bridge, a membrane 43 structure enriched in microtubules linking the two daughter cells. This cleavage step, 44 named abscission, is operated by the endosomal sorting complex required for transport 45 (ESCRT) (1, 2). The core of the ESCRT machinery is divided into four subfamilies, ESCRT 46 type I, II and III and VPS4. ESCRT subunits are sequentially positioned at the midbody of 47 7 In order to assess the size of the AKTIP structure we measured its diameter in mid, 145 late and cut midbodies ( Fig. 2A). The average internal diameter of AKTIP ring is of 146 1.05±0.03μm, and the external diameter of 1.89±0.077μm (Fig. 2B). When these data are 147 compared to available measurements for ESCRTs and ESCRT associated factors, we 148 notice that the AKTIP ring is similar to that formed by ESCRT I TSG101 and by ESCRT II 149 VPS36, and slightly larger than that calculated for members of the ESCRT III complex 150 (Fig. 2C). 151 Altogether these results provide evidence that AKTIP forms a supra-molecular ring 152 structure associated with microtubules. The fully formed supra-molecular structure 153 localizes in the dark zone in mid-phase midbodies and is progressively disaggregated in 154 late to cut stage midbodies. 155

156
The rings formed by components of the ESCRT III complex IST1, CHMP4B and 157

CHMP2A flank the central ring formed by AKTIP at the midbody 158
We then asked whether the AKTIP localization was temporally and spatially linked 159 to subunits of the ESCRT machinery. We focused on the elements of the ESCRT III 160 complex IST1, CHMP4B and CHMP2A, for which the spatial and temporal localization at 161 the midbody had been previously defined (7,10,20). In early forming midbodies, a low 162 AKTIP signal follows the tubulin profile ( Fig. 3A-B, top panel), and neither AKTIP or IST1 163 are as yet organized in a supra-molecular structure. In the mid-stage, both AKTIP and 164 IST1 assemble into ring shaped supra-molecular structures. Two IST1 rings flank the 165 central, single, larger and thicker AKTIP ring at the midbody ( Fig. 3A-B, second panel and 166 supplementary video S3). In late-to-cut stages, IST1 spirals become apparent on the 167 asymmetric tubulin bridge, and AKTIP progressively loses its circular organization (Fig. 168 3A-B, bottom two panels and supplementary video S3). AKTIP is absent from the 169 secondary ingression, suggesting it is an early component of the abscission machinery. AKTIP shows that the AKTIP ring is sandwiched in between the two rings composed by 173 these ESCRT III subunits in mid-stage ( Fig. 3C-D, top panels). In late stages, the ESCRT 174 III subunits CHMP4B and CHMP2A spiral towards the constriction site and AKTIP starts to 175 lose its structural regularity ( Fig. 3C-F). 176 These data show that AKTIP is in the form of a large, circular supra-molecular 177 structure in proximity with the ESCRT III subunits IST1, CHMP2A and CHMP4B, when 178 these get organized into double rings in the central portion of mid-stage midbodies. 179 180

Reduction of AKTIP impairs the recruitment of the ESCRT III member IST1 at the 181 midbody 182
Taken together our data indicate that AKTIP forms a circular supra-molecular 183 structure in the dark zone, at the center of the intercellular bridge that links the two 184 daughter cells. This AKTIP structure is flanked by ESCRT III subunits in the mid-stage 185 midbody. To provide an understanding on the role played by this AKTIP structure in the 186 context of the ESCRT machinery, we investigated whether AKTIP was required for ESCRT 187 complex assembly. The expression of AKTIP was reduced by RNA interference (Fig. S2) 188 and the presence of ESCRT III subunits at the midbody monitored, focusing on the 189 recruitment of CHMP4B, CHMP2A and IST1. We observe that in cells with reduced AKTIP 190 expression (Fig. S2A), the signal of CHMP4B and CHMP2A at the midbody is only 191 modestly affected (Fig. 4 A, B and D). In contrast to this, the reduction in AKTIP 192 expression impinges significantly on the localization signal of IST1 at the midbody  Together these data indicate that, not only AKTIP has a spatial and temporal 195 connection with the ESCRT machinery, but it is also functionally implicated in the correct 9 recruitment of components of the ESCRT machinery at the midbody.  Altogether these data suggest that AKTIP impacts on the IST1 recruitment at the 219 midbody and on the formation of the ESCRT complex, and contributes to the completion of 220 cytokinesis. 221

AKTIP interacts with the member of the ESCRT I complex VPS28 223
Taken together our data demonstrate a physical contiguity of AKTIP with members 224 of the ESCRT machinery at the midbody, an impact of AKTIP on the assembly of the 225 ESCRT machinery, and a role for AKTIP in the completion of cytokinesis. To further 226 understand the possible role of AKTIP as an ESCRT member we searched the 3D 227 structure protein data bank for AKTIP homologues. This analysis identifies TSG101 as an 228 AKTIP homologue with high probability (E-value: 6.3E-7). Significantly, TSG101 is part of 229 the ESCRT I complex, together with VPS37, VPS28 and MVB12 or UBAP1 (1, 2, 23) (24). 230 TSG101 interacts with the ESCRT I member VPS28, which, in turn, bridges the ESCRT I 231 complex to the ESCRT II, and therefore to the ESCRT III complexes. Computer-modeling 232 of AKTIP shows that it superimposes with the UEV domain structure of TSG101 with a root 233 mean square deviation of 1.9 Å (Fig. 6A). Given the structural homology between AKTIP 234 and the ESCRT I TSG101, we asked whether AKTIP was biochemically associated with 235 the ESCRT I complex as is TSG101. To investigate this possibility we carried out a yeast 236 two hybrid screen. Yeast cells were transformed with a plasmid encoding AKTIP fused to 237 the Gal4 DNA-binding domain, in combination with ESCRT I, II, III subunits or associated 238 factors fused to the VP16 activation domain. By measuring LacZ activity in co-239 transformants we find that AKTIP significantly and selectively interacts with the ESCRT I 240 VPS28 (Fig. 6B). To confirm this interaction in mammalian cells, VPS28 was cloned as a 241 GST-fusion, and AKTIP as HA-or MYC-tagged fusion. 293T cells were co-transfected 242 with control GST or GST-VPS28 and either MYC-AKTIP or HA-AKTIP. Pull down assays 243 followed by Western blotting show that AKTIP interacts with VPS28 ( Fig. 6C Taken together, these results indicate that AKTIP interacts with VPS28, which is 247 bound also by TSG101. We then asked whether AKTIP and TSG101 were both present in 248 the dark zone or whether the midbody could contain either AKTIP or TSG101. To this end, 249 we immunostained HeLa cells with both anti-AKTIP and anti-TSG101 antibodies. This 250 analysis shows that 99% (n=100) of the midbodies are positive for both AKTIP and 251 TSG101 (Fig. 6E). 252 Altogether these data suggest that AKTIP works in association with the ESCRT I 253 complex. In addition, the results point to the hypothesis that AKTIP could contribute to the 254 assembly of the ESCRT machinery at the midbody via its interaction with the ESCRT I 255 VPS28, which can act as a bridge to the ESCRT II, which in turn would impinge on ESCRT 256 III assembly and functional abscission (Fig. 6F). 257 258 259

260
AKTIP belongs to a subfamily of ubiquitin-conjugating E2 enzyme variants which 261 cannot function directly in the ubiquitination pathway since they lack the cysteine residue 262 required for ubiquitin binding (18). The mechanism of action of members of this family is 263 not yet understood. In the case of AKTIP, we made two observations that suggest that its 264 activity could be associated with the ESCRT complex. A first indication comes from the 265 work by Xu and co-authors, who observed that AKTIP (named FTS) functions in vesicle 266 trafficking (25), a process in which the ESCRT machinery plays a role. The second 267 indication comes from a bioinformatics search in the 3D structure data bank, which 268 indicated that AKTIP has homology with a member of the ESCRT machinery, the ESCRT I 269 subunit TSG101. Here we report that AKTIP is indeed associated with the ESCRT 270 machinery and contributes to its function during abscission.

12
The first experimental evidence linking AKTIP to the ESCRT complex is based on 272 its supra-molecular organization. We show that AKTIP forms a structure around 273 microtubules at the center of the intercellular bridge where the ESCRT complex is 274 recruited and acts to finalize abscission (5, 7). The supra-molecular structure of AKTIP has 275 the shape of a ring, which in mid phase midbody has an average outer diameter of 276 1.89μm. This circular organization of AKTIP (showed in figures 1 to 3 and in the related 277 videos) is reminiscent of that of ESCRTs and ESCRT associated factors. These factors 278 form a series of disks and rings in the dark zone, the central area of the midbody, serving 279 as a platform for the successive assembly and structural evolution of the ESCRT 280 machinery. The diameter of the AKTIP ring is similar to that of the ESCRT I member 281 TSG101, which also forms circular structures at the center of the dark zone. 282 The temporal dynamics of AKTIP at the midbody indicates that the AKTIP supra-283 molecular assembly happens in early abscission. The assembly of the AKTIP ring is 284 preceded by a phase where AKTIP and tubulin have similar localizations. In the mid-stage 285 of abscission, when the ESCRT III elements have formed full circular structures, the 286 AKTIP ring is found at the center and in proximity with the ESCRT III subunits IST1, 287 CHMP2A and CHMP4B. In the late stages of abscission, when the spiral formed by the 288 ESCRT III factors become evident, AKTIP supra-molecular organization showes a loss of 289 regularity, suggesting that the AKTIP circular structure is needed in early abscission, prior 290 to the final constriction and severing stages. 291 A further piece of experimental evidence linking AKTIP to the ESCRT machinery is 292 the observation that AKTIP interacts with the ESCRT I VPS28 subunit. Since also the 293 ESCRT I TSG101 also binds to VPS28 (23, 26, 27), we asked whether the presence of 294 AKTIP and TSG101 at the midbody was mutually exclusive. However, both AKTIP and 295 TSG101 are detected simultaneously at the center of the intercellular bridge, which 296 suggests that the regions involved in the interactions between AKTIP and TSG101 with 297 13 VPS28 are different. This interpretation is consistent with the information obtained by 298 bioinformatic modeling. In fact, while VPS28 binds to the conserved C-terminal region of 299 TSG101 (24), AKTIP differs from TSG101 in that region (see figure 6A), and VPS28 300 binding sites that are used by TSG101 are predicted to be buried in AKTIP by two C-301 terminal helices. 302 Since VPS28 bridges the ESCRT I to the ESCRT II complex (27), the interaction of 303 AKTIP with VPS28 suggests a sequential pathway in which AKTIP via VPS28 is 304 connected to ESCRT II and further then to ESCRT III (see figure 6F). This in turn points to 305 a role of AKTIP in the assembly of the ESCRT III complex. The implication of AKTIP in this 306 process was demonstrated by analyzing the localization of ESCRT III members at the 307 midbody upon depletion of AKTIP. AKTIP impacts the recruitment of ESCRT III complex 308 member IST1, but not on that of ESCRT III CHMP2A and CHMP4B. IST1 is an atypical 309 ESCRT III member that is needed both at the nuclear envelope and in cytokinesis. The 310 fact that an AKTIP reduction impinges more significantly on the ESCRT III IST1 311 recruitment at the midbody, as compared to its impact on ESCRT IIII CHMP4 and 312 CHMP2B, suggests a further element of distinction of IST1 as compared to the other two 313

ESCRT III elements. 314
Consistently with the observation that AKTIP impacts on the recruitment of ESCRT 315 III members at the midbody, the reduction of AKTIP expression affects cell division. Cells 316 with lowered AKTIP have significantly longer abscission times and are more frequently 317 binucleated as compared to controls. 318 In summary, we present evidence that the ubiquitin-conjugating E2 enzyme variant 319 family member AKTIP associates with the ESCRT machinery. AKTIP interacts with the 320 ESCRT I VPS28 and localizes at the midbody, where it forms a ring in proximity to ESCRT 321 III subunits. AKTIP affects the recruitment of the pivotal ESCRT factor IST1 and impinges 322 on cell division (see Fig. 6F). Further work will be required to define the rules for the 14 assembly of AKTIP and other ESCRT factors at the midbody during cell division. It will be 324 also interesting to investigate whether the AKTIP pool localizing at the nuclear envelope 325 plays a role in combination with the ESCRT machinery at the nuclear membrane. In fact, 326 beyond its role in abscission, recent work on IST1 has shown that it is recruited to at the 327 nuclear envelope to seal the membrane of the daughter nuclei by CHMP7 (6). It will be 328 interesting to study whether AKTIP subunits localized at the nuclear envelope are 329 associated with the ESCRT machinery operating at this site and if this activity is related to 330 IST1. In this respect, it is also tempting to speculate that the phenotype of telomere and 331 chromatin damage that we observed in AKTIP depleted cells could be due to defects in 332 the nuclear envelope sealing processes, which would be consistent with the observed 333 lentivirus (LV) mediated interference, viruses were produced as previously described (29). 348 The LV-shAKTIP (shAKTIP) and LV-scramble (ctr) vectors were described previously (12). 349 The multiplicity of infection (moi) used was 5pg p24/cell. Transduction was performed in 350 complete medium supplemented with 8µg/ml polybrene (Sigma). After viral addition, cells 351 were centrifuged for 30min at 1800rpm at RT, incubated for 3hrs at 37°C and then 352 transferred to fresh complete medium. Seventy-two hrs post-infection, cells transduced 353 with LVs were subjected to selection in complete medium supplemented with 2µg/ml 354 puromycin (Sigma) and kept under these conditions for further analyses.  and analyzed with the 2 -ΔΔCq method as previously described (31). For Western blotting, 365 72hrs post-transfection with siRNAs, protein extracts were obtained as previously 366 described (12) and quantified by Bradford assay. 100µg protein extracts were loaded onto 367 pre-cast 4-12% gradient acrylamide gels (Novex, Life Technology). After electro-blotting 368 filters were incubated with anti-AKTIP (HPA041794 Sigma) and anti-actin-HRP conjugated 369 (sc-1615, Santa Cruz Biotechnology) antibodies. Filters were then incubated with anti 370 rabbit HRP-conjugated secondary antibody (sc-2357, Santa Cruz Biotechnology). 371 Detection was performed using the enhanced chemiluminescence system (Clarity ECL, 372 Biorad). 373 374

Microscopy 375
HeLa cells were seeded onto glass coverslips in 6-well plates and fixed with 3.7% 376 formaldehyde in PBS for 10min. Cells were then permeabilized with 0.25% Triton X-100 in 377 PBS for 5min and treated with PBS 1% BSA for 30min, then stained with primary 378 antibodies in PBS 1% BSA for 1hr at RT. In the case of AKTIP-TSG101 co-379 immunofluorescence, HeLa cells were seeded on slides, fixed as previously described (5), 380 permeabilized in PBS-0.1% Triton X-100 for 2hrs and successively blocked as described 381 above. The following primary antibodies were used: anti-AKTIP (WH0064400M2 clone 382  Detection of exogenous AKTIP-FLAG by immunofluorescence using anti-FLAG antibody. (C-E) Immunofluorescence with anti-AKTIP (WH0064400M2 clone 2A11) (C) and qPCR (D) showing that AKTIP reduction causes a drop to 47% AKTIP positively staining midbodies as opposed to 91.7% of control cells (E). Results shown are the mean value of two replicates  SEM ***p < 0.001; Student's t-test; 60 midbodies per condition were analyzed. Scale bar 5μm.