AGO2 localizes to cytokinetic protrusions in a p38-dependent manner and is needed for accurate cell division

Argonaute 2 (AGO2) is an indispensable component of the RNA-induced silencing complex, operating at the translational or posttranscriptional level. It is compartmentalized into structures such as GW- and P-bodies, stress granules and adherens junctions as well as the midbody. Here we show using immunofluorescence, image and bioinformatic analysis and cytogenetics that AGO2 also resides in membrane protrusions such as open- and close-ended tubes. The latter are cytokinetic bridges where AGO2 colocalizes at the midbody arms with cytoskeletal components such as α-Τubulin and Aurora B, and various kinases. AGO2, phosphorylated on serine 387, is located together with Dicer at the midbody ring in a manner dependent on p38 MAPK activity. We further show that AGO2 is stress sensitive and important to ensure the proper chromosome segregation and cytokinetic fidelity. We suggest that AGO2 is part of a regulatory mechanism triggered by cytokinetic stress to generate the appropriate micro-environment for local transcript homeostasis.

A rgonaute (AGO) proteins are at the hub of RNA-induced silencing complex (RISC), a major contributor to the finetuning of gene expression due to posttranscriptional regulation 1 . AGOs are ubiquitously expressed and present strong evolutionary conservation. The human AGO family includes four different proteins in humans: AGO1, AGO2, AGO3, and AGO4, with extremely high homology (exceeds the 80% value over the entire protein-length) among all members. Although they share the same signature domains N, MID, PAZ, and PIWI, only AGO2 has proved to exert slicer-endonuclease and microRNA (miRNA) stabilization activities in humans, till recently. The epigenetic process is directed by small RNAs 2 , the anchoring of which into concrete binding pockets allows the AGO2 guidance toward the respective targets 3 . Higher eukaryotes feature a large diversity of AGOs and small RNAs, including miRNAs, which are implicated in a plethora of cellular processes, such as cell differentiation, transposon silencing, embryonic development, and cancer 1,3,4 . In the canonical pathway, the mature miRNAs are loaded onto AGO proteins in an ATP-dependent manner 5 . The miRNA active guide strand is considered to have the lower 5' stability and interacts with AGO2 complex while the unloaded passenger strand is unwound from the guide strand, according to the degree of complementarity 6,7 . Dynamic alterations in mRNA structure, mRNA-folding, and unfolding also shape AGO2-target recognition 8 . Nevertheless, apart from the canonical, multiple non-canonical miRNA biogenesis pathways, using various combinations of the microprocessor, exportin, or miRISC loading complex proteins, have been reported 9,10 . In any case, AGO family is a prerequisite for the proper function of miRNAs.
In mammalian cells, AGOs are located and operate through canonical and non-canonical pathways both into the cytoplasm and nucleus 11 , in discrete foci [12][13][14] . However, the RISC can also reside/function in unexpected areas inside the cell, such as the GW-bodies, P-bodies and stress granules [15][16][17] , the midbody of dividing cancer cells 18 , along apical junctions 19 with a putative tumor-suppressing function 20 and in Drosophila nanotubes 21 . The GW-bodies most likely function as repositories for translationally silenced RNAs 17 , while inside the P-bodies or in the stress granules, AGO2 participates in miRNA-induced mRNA silencing 22 . In the apical zonula adherens, an alternative local regulation of selected miRNA processing was uncovered that suppresses the expression of markers critically involved in transformed cell growth 19 . Casey et al. reported the AGO2 localization in the midbody of dividing MCF10A, T47D, BT-474, and SK-BR-3 cells 18 . Furthermore, there has been an association between RISC and endosomes at synapses, through the interaction of PICK1 and AGO2, and the relocalization of the latter to endosomal compartment. This results in an elevated translation of mRNA targets locally 23 , dictating the spatio-temporal homeostasis. In conclusion, AGO2 most likely generates a niche creating the appropriate microenvironment locally for cellular and subcellular functions, maintenance, and regulation in a complex, multifaceted, and versatile manner.
In this article, we examined the AGO2 involvement in another locasome, the tubular protrusions, including tunneling nanotubes and cytokinetic bridges. Together, our results provide support for an AGO2 complex at the midzone, essential in cytokinesis and cell-division integrity.

Results
AGO2 resides in protrusional structures along with other components of the RNAi-machinery. The AGO2 protein is located into the cytoplasm as well as into the nucleus of all the examined epithelial and mesenchymal cell lines, normal (thyroid NTHY ori 3-1, primary mammary HMEC-1 and hepatic stellate LX-2) and cancer (breast cancer MDA-MB-231, melanoma A375 and liver cancer HepG2), exhibiting a punctuate pattern ( Fig. 1a and Supplementary Fig. S1a-f). Interestingly, AGO2 was also localized in protrusions, demonstrating differences in shape and density. More specifically, AGO2 resided in Actin-filopodial-like structures that are many per cell (Fig. 1a, b), and in tubular protrusions that are always formed in paired cells (Fig. 1a, c, d). The tubular protrusions demonstrate two different AGO2 distribution patterns, a low immunofluorescence AGO2 signal correlated with loosely shaped tunneling nanotubes, termed open-ended tubes (Fig. 1c), and an intense AGO2 signal in a compact and dense structure bearing a "gap" lacking AGO2, usually in the middle-line of the tube, termed close-ended tubes (Fig. 1d). To track the dynamic behavior of AGO2 in the two types of paired protrusions, time-lapse experiments were performed using cells expressing an AGO2-GFP chimeric protein.
In the AGO2 tunneling nanotubes (open-ended), AGO2 components displayed flow and motility across the tube (Supplementary Movies 1 and 2 and Fig. 1e, f). AGO2 close-ended formations exhibited high flexibility, non-homogeneous structure, with a distinctively larger AGO2 object, when compared to open-ended nanotubes; the close-ended formation breaks, eventually leading to separation of the involved cells (Supplementary Movies 3 and 4 and Fig. 1g, h). Next, we investigated the presence and distribution of Drosha, DGCR8, and Dicer proteins, in the AGO2-enriched open-and close-ended paired structures. These crucial components of the RNAi-machinery functions, although they are known to reside mainly into the nucleus and cytoplasm, were also detected in the AGO2 tubes ( Fig. 2a-f). Immunofluorescence assays demonstrated that Drosha and DGCR8 exhibited punctuate expression patterns within both tunneling open- (Fig. 2a, b) and close-ended nanotubes (Fig. 2d, e), similar to those observed with AGO2. Dicer was detected within and along the open-ended nanotubes (Fig. 2c), while in close-ended protrusions it was situated primarily in the middle-line ("gap") of the tubes (Fig. 2f). Notably, Staufen, a double-stranded RNA (dsRNA) binding protein responsible for dsRNA transport and mRNA localization, was mainly cytoplasmic but was also found, though at lower signal intensities, inside the open-ended tunneling nanotubes (Fig. 2g), while being almost undetected in close-ended cellular protrusions (Fig. 2h). Altogether, it seems that in the open-ended tunneling nanotubes, ΑGΟ2 is loaded with dsRNA species, as indicated by the presence of Staufen. This does not seem to apply to the AGO2 close-ended protrusions, at least as suggested by the absence of Staufen. Given the strong signal of AGO2 in the close-ended structures and the distinct pattern with the "gap" in the middle of the tube, their composition and functional contribution to cellular physiology needs to be further elucidated.
AGO2 close-ended protrusions constitute a different type than typical Actin-filopodial structures. To obtain more information regarding the structure of AGO2 close-ended protrusions we compared them to the well-established Actin-filopodial-like ones. A total of 168 confocal laser scanning microscopy (CLSM) images of individual cells (Supplementary Figs. S2a, b and S3a, b), from at least three independent biological experiments, were collected and processed. The number of protrusions per cell, their lengths, and the tip-to-midpoint width of both types of structures were measured. The statistical analysis of the measurements revealed significant differences that clearly distinguish between the two types of protrusions. In all, 38 Actin-filopodial protrusions on average were present per cell, whereas only one AGO2 protrusion could be maximally identified per cell. Intriguingly, although the length of Actin-filopodial protrusions ranges from 0.48 to 37.76 µm with a mean of 2.85 μm, the AGO2 close-ended protrusions exhibit less divergence (0.88-19.00 µm) and their mean length was significantly (p = 2.2e −16 ) larger (5.37 μm) ( Supplementary Fig. S2c). In an attempt to obtain a model shape of the structures, the average tipto-midpoint width ratio of all protrusions was calculated. The average ratio of AGO2-carrying structures (1.03 μm) differed significantly (p = 7.58e -07 , Supplementary Fig. S2d) from the typical triangular shape of Actin-filopodial protrusions (mean ratio 0.85 μm), showing a tendency toward a parallelogram-like formation ( Supplementary Fig. S2e). In conclusion, AGO2 close-ended protrusions differ from the typical Actin-filopodial ones as they are infrequent, parallelogram-shaped, and always formed in paired cells.
AGO2 close-ended cellular structures are cytokinetic protrusions. Based on the aforementioned findings, we investigated the possibility of the close-ended protrusions to comprise midbody structure, a narrow intercellular bridge arose during cytokinesis, the final step of cell division. To explore this scenario, we examined the presence of cytokinesis-related proteins in these structures. Citron kinase (CITK), a protein carrying a coiled-coil domain that differentially dictates its subcellular localization and leads to proper midbody stabilization during cytokinesis, was detected in the close-ended AGO2 protrusions. CITK was primarily enveloped at the area called midbody ring (Fig. 3a) that in the close-ended protrusions AGO2 signal is absent and thus depicted as "gap." In situ proximity ligation assay (iPLA) experiments of AGO2 and CITK demonstrated the accumulation of fluorescent spots to be restricted into the cytoplasm and almost Visual assessment (merged images, scatterplots, and intensity lineplots) of colocalization between paired proteins included in the yellow region of interest of the merged images. f, h AGO2 (green) with CITK and Dicer (red), respectively. e, g Aurora B in green with AGO2 and CITK in red, respectively. Lineplots indicate the signal intensity of the paired proteins along the lines overlaid on the corresponding images that appear in Supplementary Fig. S4d absent along the intercellular bridge (Fig. 3b). In contrast, Aurora B, the other critical cell-division kinase that is essential for the functional midbody architecture through CITK cross-regulation, exhibited a striking overlap with AGO2 (Fig. 3c). The AGO2-Aurora strong correlation was also confirmed with the iPLA experiments (Fig. 3d). Colocalization analysis (Fig. 3e-h and Supplementary Fig. S4d-g) and corresponding statistical tests strongly supported these findings (Supplementary Table S1). In particular, the CITK-AGO2 pair exhibits low M2 value-proportion of CITK channel occupying pixels of the AGO2 channel -(M2 = 0.336), since CITK is primarily detected at the midbody ring whereas AGO2 is absent (Fig. 3e). On the contrary, strong colocalization emerged between AGO2 and Aurora B (R = 0.718, M1 = 0.889, M2 = 0.786), as Aurora B resided almost entirely along the intercellular bridge being occupied by AGO2 (Fig. 3f). CITK and Aurora B demonstrated differential distribution, in complementary patterns, in the AGO2 close-ended cellular structures (Fig. 3g). Finally, Dicer was primarily located at the midbody ring ( Supplementary Fig. S4a-c, h, l, m), as indicated by the intensity plots (Fig. 3h). The presence of CITK and Aurora B kinases in the AGO2 close-ended cellular structures revealed that these comprise cytokinetic protrusions.
α-Tubulin alterations and cytoskeletal changes influence AGO2 localization. Given that AGO2 appears to reside in cytokinetic protrusions, the possibility of a tight configuration with α-Tubulin polymers (microtubules) was investigated. Polymerized α-Tubulin was strongly colocalized with AGO2 in the emerged blebs of cytokinesis and it was absent from the ring of midbody formation (Fig. 4a, c and Supplementary Fig. S4i). This was also demonstrated through iPLA assays with multiple foci being observed along the midbody arms, but not at the midbody ring (Fig. 4b). However, F-Actin (G-Actin polymer), the other fundamental cytoskeleton component, was presented with significantly lower degree of colocalization with AGO2 (Supplementary Table S1). Actin filaments were mainly located close to the starting point of the protrusions and were attenuated along the intercellular bridge (Fig. 4d, e and Fig. S4j, k). A representation of the distribution patterns of all the herein examined proteins (AGO2, Dicer, CITK, Aurora B, α-Tubulin, and F-Actin) across the intercellular bridge is illustrated in Fig. 4f. We further investigated the relationship of AGO2 with the two pivotal components of cytoskeleton, α-Tubulin and F-Actin, in a drug-induced depolymerization manner. To this end, Demecolcine and Cytochalasin D, specific inhibitors targeting the microtubule and Actin-filament depolymerization, respectively, were applied and their effects on cell morphology, AGO2 distribution, and formation of cytokinetic and cytoplasmic protrusions were examined. After 5 h of Demecolcine treatment, α-Tubulin was not totally depolymerized across the intercellular bridges and AGO2 exhibited decreased signal intensity (Fig. 5a, b). The midbody formation was not completely disrupted as indicated via the normal distribution of CITK (Fig. 5c). Most importantly, after 7 h of Demecolcine treatment, a depolymerization of intercellular bridges was induced as indicated by the disruption of α-Tubulin structures, with a similar fragmentation of AGO2 protrusions (Fig. 5d, f, g). The AGO2 expression signal was directed toward the cytoplasm at lamellipodia and in membrane ruffles (Fig. 5d). However, the localization profile of CITK clearly points toward an intact midbody ring formation (Fig. 5e). In conclusion, AGO2 distribution is related to the intact polymerization of α-Tubulin, indicating that α-Tubulin can constitute a capable scaffold for AGO2 transfer and function ( Fig. 5f, g). Exposure to Cytochalasin D did not cause detectable changes in AGO2 typical distribution (Fig. 6a, b, f, g).
Nevertheless, Cytochalasin D prevented Dicer accumulation in the midbody and compelled, in some cases, its distribution along the AGO2 midbody arms (Fig. 6c, d, h, i). Remarkably, the localization pattern of CITK remained unaffected, hence, indicating the specificity and component-selectivity of the druginduced molecular phenotype. (Fig. 6e). Dicer-specific foci, frequently observed inside the cytoplasm and in the milieu, proved to serve as markers of midbody remnants after abscission completion (Fig. 6c). It seems that the scattering of AGO2 and Dicer is likely dictated by different regulatory mechanisms.
AGO2 follows the α-Tubulin cellular expression patterning during cell division. As the strong correlation between AGO2 and α-Tubulin was already demonstrated, it is not surprising that AGO2 follows the α-Tubulin cytoskeleton distribution pattern during the different phases of cell cycle ( Supplementary Fig. S5). In particular, in metaphase that is typified by chromosomes being aligned during central spindle assembly, AGO2 showed a dense cytoplasmic signal with the exception of the area in-between the chromosomes. However, AGO2 and Dicer immunofluorescent topologies, following the mitotic spindle, become concentrated at the mitotic poles (Supplementary Figs. S5a and S6a). In anaphase, AGO2 was highly accumulated in the centrosomal area (Supplementary Fig. S5b), while Dicer followed the microtubulesegregation ( Supplementary Fig. S6b). In this phase, the initial cue for cytoskeleton rearrangement, the creation of microtubulebundling area called midzone, and the cleavage furrow ingression become readily recognizable. In telophase, AGO2-localization pattern demonstrated a "gap" in the middle of the midzone area, similar to the α-Tubulin configuration ( Supplementary Fig. S5c). In parallel, Dicer was mainly located at the midbody ring (Supplementary Fig. S6c). The intercellular bridge starts its formation during late telophase, with a simultaneous compression of midzone and the development of microtubule midbody. AGO2 and Dicer reside in the intercellular bridge; during the process of stabilization of cleavage furrow ingression AGO2 formed a tight arrangement inside and along the α-Tubulin bridge (Supplementary Fig. S5d), while Dicer was mainly distributed around the ring and gradiently decreased inside the arms ( Supplementary  Fig. S6d). Then, the cleavage furrow begins the narrowing until the intercellular bridge is disintegrated and the two daughter cells are emancipated (Supplementary Figs. S5e and S6e). Following the α-Tubulin constriction and abscission, AGO2 regressed toward the daughter cells following the midbody together with the arm-structure into the cell cytoplasm for autophagy-mediated elimination ( Supplementary Fig. S5f). The Dicer-labeled midbody ring was digested from one cell in the case of asymmetric division or removed to the cell milieu in case of symmetric division ( Supplementary Fig. S6f). Altogether, the AGO2 expression pattern during mitosis and its resemblance to the -polymerized-α-Tubulin respective one reinforce the notion that AGO2 sustains important roles in cell division.
AGO2 over-expression generates aneuploidy via cytokinesis errors. To further study the contribution of AGO2 to the fidelity of cell division, we examined the effect of over-expression and downregulation of AGO2 in chromosome segregation. HCT116 cells were transfected to transiently express AGO2-GFP. The AGO2 over-expression induced statistically significant micronuclei formation (Fig. 7a-d, m) and numerical chromosomal deregulation (gain or loss) ( Supplementary Fig. S7a-c and Fig. 7m). However, although there were detected lagging chromosomes in anaphase and cytokinesis (Fig. 7e-l), the percentage was similar to the control, an observation that needs further investigation. These findings lead to the conclusion of AGO2 implication in aneuploidy, reflecting post-replicative events and cytokinesis errors. By knocking down AGO2 (almost two-foldchanges) (Supplementary Figs. S7h and S8a), no statistically significant changes in numerical chromosomal deregulation and micronuclei formation were observed, but structural alterations emerged ( Supplementary Fig. S7d-g). Cytokinesis-failure events such as binuclear cell formations and midbody abnormalities such as double midbody rings ( Supplementary Fig. S7i-k) demonstrate that AGO2 influences the integrity and the succession of cytokinesis. Altogether, these results illustrate the implication of AGO2 in fine-tuning chromosomal regulatory mechanisms, safeguarding the cell-division success.
Kinases follow the AGO2/Dicer distribution in the midbody structure. The role of AGO2 in the midbody might be an  Supplementary Table S1. c, d AGO2 in green, α-Tubulin, and F-Actin (Texas red-phalloidin) in red. e α-Tubulin in green, F-Actin in red (n = 6 biologically independent samples). unrecognized contribution to cell division or related to a function of the known -typical-AGO2 pathway. In an attempt to distinguish between the two scenarios and better understand AGO2 involvement, we herein investigated the presence of poly-A + tailed transcripts along the cytokinesis bridge (Fig. 8a). Surprisingly, the hybridization (oligo dT) probe followed the AGO2 pattern. This finding together with the complementary Dicer distribution pattern (Fig. 3h) indicated a Dicer-independent AGO2 function in the arms of the midbody structures. However, as the lack of AGO2 detection from the midbody ring was striking, we further investigated the possibility that posttranslationally modified AGO2 protein molecules also resided at this site. Phosphorylation is a crucial regulation step of AGO2 activation 24 and, therefore, we explored the occurrence of different activated kinases in the cytokinetic bridge during the late steps of cell division. Colocalization analysis of a small kinome including phospho-MEK (Fig. 8b), phospho-ERK (Fig. 8c), phospho-p38 MAPK (Fig. 8d), phospho-JNK (Fig. 8e), phospho-Akt ( Supplementary Fig. S9a), and phospho-AMPK ( Supplementary Fig. S9b) showed that there are two distinct, almost complementary to each other, patterns in the cytokinetic structure. Phospho-MEK, -ERK, -p38, and -JNK, although being present along the midbody arms, were highly concentrated at the ring following Dicer's distribution, while phospho-Akt (Supplementary Fig. S9a) and phospho-AMPK ( Supplementary Fig. S9b) occupied the cytokinetic protrusions leaving a small "gap" at the midbody ring resembling the AGO2 pattern. The specificity of these distribution patterns is enhanced by the fact that other cell cycle-specific kinases such as phospho-CHK2 were absent across the cytokinetic bridge ( Supplementary Fig. S8b). Due to the midbody distribution of phospho-MEK/ERK/p38/JNK and the fact that the p38 MAPK pathway enhances the phosphorylation of AGO2 at the catalytic residue Serine 387 , we investigated the presence of phospho-AGO2 Ser 387 in the midbody. Interestingly, the phospho-AGO2 Ser 387 follows the phospho-MEK/ERK/p38/ JNK pattern, exhibiting high concentration in the midbody ring (Fig. 8f). To strengthen the p38-dependent AGO2 phosphorylation in the bridge, we treated the cells with SB203580, a selective inhibitor of p38 MAPK. Indeed, the phospho-AGO2 Ser 387 was relocated in a number of cytokinetic bridges, moving from the ring to the arms of the midbody (Fig. 8g, h). Ιn conclusion, the phospho-AGO2 Ser 387 is located in the ring of the midbody in a ΑGO2 is a stress-sensitive molecule. The above findings support the strong implication of AGO2 in successful cell division. The dividing cells are known to be under tremendous stress 25 , a finding corroborated by the existence of the abovementioned phospho-kinases locally. As the role of kinases in mechanical stress has been previously reported 26 , the presence of several activated kinases in the midbody/midzone area is likely associated with the development of locally applied stress, during late cell division. To support this, we conducted transmission electron microscopy (TEM) and scanning electron microscopy (SEM) high-resolution technologies. A proteinaceous electron-dense material was observed at the midzone of microtubule bundles with the typical distance being highly varied ( Fig. 9a-g). In SEM images, light-colored, electron-dense material was apparent (Fig. 9c, d). This material, likely containing molecular cargoes and proteins, circuited at the protrusions providing views of rough surfaces (Fig. 9e). When abscission occurs (Fig. 9f, g) the midbody is withdrawn and uptaken/fused with one daughter cell for digestion. Notably, dividing cells did not carry filopodial protrusions anti-diametrically of the intercellular bridge (Fig. 9f), indicating that during the exertion of mechanical forces -polymerized-α-Tubulin and F-Actin cytoskeleton, together with membrane depositions, are rearranged for the formation of midzone arms and midbody ring. After abscission mitotic arm protrusions become contracted (Fig. 9f) and vesicle inventories (Fig. 9h) are conspicuous, probably for membrane deposition and remodeling. Since the dividing cells are under tremendous stress and AGO2 appeared to carry a functional role in cell division, we reasoned that AGO2 is a stress-sensitive molecule. Indeed, heatshock treatment malformed or abolished AGO2 close-ended protrusions (Fig. 9i-l). The finding was strengthened by noteworthy changes in colocalization of AGO2 with the stress-related proteins, Upf1 and TIAR. In particular, under local basal stress, Upf1 and TIAR proteins were detected at the midzone area and their granules were weakly colocalized with AGO2 protein (Fig. 9i, k). After heat shock, the AGO2 foci were colocalized with Upf1 stress granules and the malformed cytokinetic bridges, wherever they developed, demonstrated a strong colocalization pattern with the Upf1 and TIAR respective ones (Fig. 9j, l). These findings supported the stress-sensitive nature of AGO2 that was accompanied by cytokinesis-failure events. We further investigated the implication of AGO2 to a specific type of stress, ER-stress, which is required for a successful cytokinesis 27 . ER-stress modulated effectors such as eIF2α translation factor and its downstream target, ATF4, are found along the intercellular bridge and into the midbody ring (Fig. 10a, e and Supplementary Fig. S4n, o). iPLA foci of AGO2-phospho-eIF2α and AGO2-ATF4, although located into the cell matrix, were also sparsely located within the bridge (Fig. 10c, d) indicating that AGO2 may trigger ER-stress effectors for ATF4 activation. The interaction foci were detected in both sides of the bridge likely implying a cytoprotective regulation of local-type transcript homeostasis of gene expression. The coexistence of the four activated phospho-kinases, MEK, ERK, p38, and JNK, together with the phospho-eIF2α and ATF4 proteins into the AGO2enriched cytokinetic structures strongly suggests the simultaneous and synergistic activation of the MAPK-signaling network and RNAi-machinery in response to stress (mechanical and/or ER), in a topology-specific manner, most likely for the restoration of local -transcript-homeostasis during cytokinesis.

Discussion
AGO2, a protein of the miRISC machinery, catalyzes the mRNA degradation or translational repression in the cytoplasm, through guidance by miRNAs that are loaded on AGO2 complexes. Nevertheless, AGO2 can also reside into the nucleus, acting in a typical RNAi manner 28 . Alternative splicing process, RNAmediated epigenetic regulation through RISC-chromatin interactions and double strand-break repair are among the emerging nuclear functions of AGO family members [28][29][30][31][32][33] . In mammalian cells AGOs are located and operate through canonical and noncanonical pathways both into the cytoplasm and nucleus 11 in discrete foci [12][13][14] in physiological and pathological conditions 34 . RISC can also function in GW-bodies, P-bodies, and stress granules [15][16][17] , along apical junctions 19 and in Drosophila nanotubes transmitting intercellular signals 21 , regulating translation of mRNA targets locally 23 and therefore dictating the spatiotemporal homeostasis. In addition to the above, we herein provided evidence of an AGO2-niche in membrane protrusions, suggesting its critical role(s) in these structures. Specifically, we demonstrated the accumulation of AGO2 in open-ended tunneling nanotubes and close-ended cytokinetic protrusions, in human cells.
Human cellular protrusions are known to be involved in intercellular trafficking of cytoplasmic or genetic material 35,36 , pathogen transmission 37,38 , mitochondria, calcium, and other cargoes transportation 35,39 , and miRNA and protein communication 40 . In dendritic spines, AGO2 constitutes an important element for derepression of dendritic mRNAs and local protein synthesis triggered by synaptic activity, thus introducing the concept of spatiotemporal participation of AGO2 in local protein regulation 41 . Our findings support a similar AGO2-dictated local regulation in other cell types such as epithelial cells. In our experimental setting the AGO2-niche correlated with loosely shaped tunneling nanotubes termed open-ended tubes. Drosha, DGCR8, and Dicer, crucial components of the miRNA machinery, as well as Staufen, a dsRNAbinding protein, were also shown to reside into these nanotubes. Live imaging demonstrated the AGO2 protein motion inside the nanotube. Combined with the presence of the abovementioned proteins, these loci indicate RNAi-machinery motility in these structures. Through this action, cells can conduct intermolecular cross-talking, exchange RNA signals, and regulate transcript instability (half-life time) or translational repression at a very specific local (micro-) environment.
In the case of AGO2 close-ended protrusions, alternative regulatory functions are likely engaged. The Dicer pattern, the detected Staufen protein, and the presence of oligo dT probe at the intercellular bridge indicated alternative, non-typical AGO2 functions, probably deployed in a miRISC-independent manner. The AGO2 close-ended tubes were different structures from the Actin-filopodial ones, as they were unique per cell, parallelogram-shaped, and always appearing in paired cells. The detection of CITK and Aurora B kinases in these cellular structures confirmed that they comprise bona fide cytokinetic protrusions (Aurora B regulates cytokinesis through microtubule interactions and stabilization of microtubule structures 42,43 ). Stably growing and rapidly shrinking microtubules actively participate in cytoskeletal remodeling, delivering dramatic cell changes 44 . AGO2 is present in a tight configuration along α-Tubulin polymers (microtubules) into the intercellular bridges of cytokinesis. Although it resided in extremely high concentrations in the midbody arms, also reported by Casey et al. in 2019 18 , it leaves an empty space, a "gap," in the area of midbody ring. α-Tubulin distribution in cells exposed to Demecolcine, a specific inhibitor of tubulin polymerization, was disorganized and AGO2 protrusions were forced to fragmentation and AGO2 is concentrated at lamellipodia. This indicates that α-Tubulin may serve as a major scaffold for AGO2 local motility. On the other hand, Dicer, highly accumulated in the midbody ring, was significantly affected by F-Actin. Architectural derangement of F-Actin induced by Cytochalasin D cell exposure prevented Dicer accumulation in the midbody and occasionally compelled it in unexpected areas along AGO2 arms. The above indicate the differential regulation of AGO2 and Dicer and reinforced the notion that AGO2 sustains an important role in cell division.
AGO involvement in chromosome division phenomena has been previously reported in a number of model organisms such as yeast, Drosophila, and mouse. TbAGO1 is known to be crucial for mitotic spindle assembly and chromosome segregation in the Trypanosome brucei parasite 45 . In yeast deletion of Dicer1 or AGO1 was found to disrupt chromosome segregation, leading to chromosome lagging and centromere-silencing abrogation 46 . In  48 . In mice AGO4 is implicated in the precise chromosome segregation and is required for the proper entry into meiosis in germ cells 49,50 . In addition, Dicer can prevent mitotic defects during meiosis in mouse oocytes 51 . In mammals AGO2 catalytic activity, but not the classical miRNA pathway, is crucial for proper chromosome segregation. Moreover, Dicer generates ASAT (α-satellite) siRNAs, which, synergistically with AGO2, control satellite RNAs 52 . These reports support the mechanistic association of AGO2 with cell-division machinery in a wide evolutionary spectrum of organisms, including humans. Our findings underpin this association, since over-expression of AGO2 results in a numerical deregulation (gain, or loss) of chromosomes. Moreover, these cells present statistically significant micronuclei formation, providing evidence for aneuploidy events. Downregulation of AGO2 induces a number of abnormalities, such as binuclear formations, blebs, double midbody rings, and structural chromosomal instabilities, thus indicating the indispensable contribution of AGO2 to the regulation, progression, and success of cytokinesis. As mitosis requires high cytoskeletal rearrangements and continuous supply of energy and biomolecules for its successful and unimpaired implementation, the presence of kinases such as phospho-MEK, -ERK, -JNK, -p38, -Akt, and -AMPK in the cytokinetic bridge was not unexpected. Previous studies have highlighted the crucial regulatory roles of the herein examined kinases in the midbody formation and cytokinesis [53][54][55][56] , with their contribution to cell division mostly being manifested through activation of critical downstream targets for regulating microtubule dynamics and F-Actin cytoskeletal integrity 57 . In our experiments phospho-Akt and -AMPK, a major energy and nutrient sensor 58 , proved to follow the AGO2-distribution pattern at the arms of the midbody structure, whereas phospho-MEK, -ERK, -p38, and -JNK showed a Dicer-like-distribution motif, being mainly concentrated in the ring and at the tips of the midbody arms. Interestingly, phospho-AGO2 Ser 387 follows the phospho-MEK/ERK/p38/JNK expression pattern, exhibiting a high concentration in the ring area. Phosphorylation of AGO2 at Ser 387 by the activated p38 MAP Kinase is induced by cellular stress 59 . In accordance, a selective inhibitor of the p38 MAP Kinase (SB203580), causes relocalization of phospho-AGO2 Ser 387 in a number of cytokinetic bridges, shifting its location from the ring to the midbody arms. The detection of both p38-dependent phospho-AGO2 Ser 387 and Dicer in the midbody ring provides evidence for the existence of an active RISC machinery, suggesting a yet unrecognized role of the complex in cell division.
The dividing cells are under tremendous stress, as corroborated by our findings regarding the existence of stress-related proteins such as phospho-Akt, -AMPK, -ERK, -MEK, -JNK, and -p38activated kinases in the midzone. In conjunction with the localization of AGO2 and phospho-AGO2, these findings underline the probable implementation of AGO2 in the regulation of spatio-temporal homeostasis in order to beneficially manage local stress. The stress-sensitive nature of AGO2 was further supported by heat-shock treatments that resulted in malformed or abolished AGO2 close-ended protrusions and by the strong colocalization pattern with the stress-related proteins Upf1 and TIAR. Moreover, the presence of phospho-eIF2α and ATF4 (ER-stress) proteins into the AGO2-enriched cytokinetic structures further suggests the simultaneous and synergistic activation of a MAPKsignaling network and the RNAi machinery in response to stress, in a topology-specific manner, most likely for the restoration of local transcript homeostasis during cytokinesis.
In conclusion, we herein demonstrate the AGO2 involvement in the tubular protrusions' locasome, including open-ended tunneling nanotubes and close-ended cytokinetic bridges. Altogether, our results provide evidence for the discovery of an AGO2-niche in the midzone carrying essential properties for successful cytokinesis and cell-division integrity, although further investigation needs to be performed to thoroughly dissect the exact roles of AGOs, and their protein-and RNA-interactors, in a cell-division setting.
In situ proximity ligation assay (iPLA). Cells were cultured on coverslips, rinsed with DUO link wash Buffer B according to the manufacturer's protocol (DUO92007, Sigma-Aldrich, USA). Then they were incubated in blocking solution for 30 min at 37°C, and the appropriate primary antibody was applied overnight at 4°C. Next, cells were exposed to secondary antibodies conjugated to oligonucleotide iPLA probes (cat DUO92001, cat DUO92005, MINUS and PLUS, Sigma-Aldrich) for 1 h at 37°C. Ligase activity performed coupled with rolling circle amplification and the products were hybridized by oligo-nucleotide probes labeled with a fluorophore. Red punctuate signals were captured using confocal scanning microscopy.
Quantitative PCR analysis. Total RNA extract from NTHY ori 3-1 cells was isolated using Trizol according to the manufacturer's protocol. Next, cDNA was synthesized using as a template ∼500 ng of total RNA and the Quantitect RT kit, Qiagen, 205311Q according to the manufacturer's recommendations. The cDNA was amplified in triplicates on a Roche Light Cycler 96 System. Relative gene expression was calculated using the 2 −ΔΔCt method. Expression of genes of interest (goi) was normalized by the reference gene Lamin. Fold change in gene expression was calculated as 2 −ΔΔCT 60 , where ΔΔC T = ΔC T(goi) − ΔC T(Lamin) . The relative change of Ago2 following the treatment with siAgo2 was calculated using the gene expression of the untreated scrambled control cells as a control.
Microscopy. Images were acquired with a Leica TCS SP5 inverted CLSM, using Leica HC PLAPO 63×1.4NA CS immersion objectives. Acquisition was performed sequentially to avoid cross-talk. Time-lapse videos were started 18 h post transfection at 30 sec or 1 min time-lapse intervals for 45 min.
For TEM, the cells were cultured in 35 mm CellStar cell culture dishes (Greiner Bio-One, GmbH, Germany), on a clarfilm, for 24-48 h, under standard conditions, fixed in 2.5% glutaraldehyde (in 0.1 M phosphate buffer), and post-fixed with 1% osmium tetroxide. Next, samples were dehydrated via a graded series of ethanol concentrations, 30, 50, 70, 90, and 100%, followed by propylene-oxide (PO) treatment. They were further processed in a mixture of Epon/Araldite resins diluted in PO, flat-embedded in fresh epoxy-resin mixture, and allowed to polymerize at 60°C for 24 h. Small epoxy pieces were peeled away from petridishes, glued on epoxy blocks, and allowed to polymerize for 24 h. Ultra-thin sections of 65-70 nm (Leica EM UC7 Ultra-microtome, Leica Microsystems, Austria) were mounted onto 200 mesh nickel grids, stained with uranyl acetate and lead citrate, examined with a Philips 420 transmission electron microscope at an acceleration voltage of 60 kV, and photographed with a MegaView G2CCD camera (Olympus SIS, Germany).
SEM was performed with a Jeol 6380LV electron microscope (JEOL Ltd., USA) that was operated at 15 kV and was equipped with an Oxford EDS system for chemical analysis. The XRD patterns were recorded by means of a Siemens D5005 (Bruker AXS, USA) using Cu K radiation. For the evaluation of the surface structural configuration, NTHY ori 3-1 cells were prepared and cleansed with ethanol. For the electron microscopy imaging and analysis, cells were attached on aluminum stubs and were analyzed in a low-vacuum mode (around 30 Pa) to minimize cell stress due to the progressive lack of humidity of the samples during the process in instrument's vacuum chamber, as well as to eliminate charging effects after the cells have dried.
Details for the materials, reagents, drug compounds, and antibodies (primary and secondary) used in the described Methods are presented in Supplementary  Table S1.
Protrusion measurements. The number and length of filopodial protrusions were measured in 102 cells from CLSM fluorescence images (three independent biological experiments) ( Supplementary Fig. S3a, b) using the FiloDetect algorithm 62 . The algorithm is implemented in Matlab and was suitably modified so that filopodia shape could also be modeled. A filopodium's midpoint width was defined as the number of pixels occupied on the minor axis of the ellipse created by Matlab's "regionprops" function. The corresponding numbers for parallel translations of the minor axis for a distance equal to 2/3 of half the major axis toward both ends of the major axis were also evaluated. Between these two points the one at greater distance from the cell's centroid (also identified using "regionprops") was defined to be the filopodium tip width. The shape model was obtained by evaluating the average tipto-midpoint width ratio among all filopodia detected in the cells measured. The same procedure was followed for the calculation of the corresponding quantities for AGO2 protrusions in 66 cells. Since this type of protrusion is unique per cell no averages were required. The AGO2 protrusion of each cell along with other protrusion-like structures was automatically identified using Matlab's "fibermetric" function. To isolate the real AGO2 protrusion among them, a region of interest was manually cropped around it and superimposed with the "fibermetric" image.
Image analysis. 3D surface reconstruction was implemented using Imaris 9.1.2 (Bitplane, South Windsor, CT, USA). Images used for colocalization analysis were acquired with Nyquist sampling; the Richardson-Lucy Total Variation algorithm 63 was selected from DeconvolutionLab2 plugin 64 for image deconvolution. Theoretical PSFs according to the Born and Wolf 3D Optical Model were constructed from the PSF Generator plugin 65 . Different channels were processed independently. Further preprocessing consisted of background subtraction (3D rolling ball algorithm 66 ) and thresholding (triangle method 67 ) all run in Fiji 68 . The Pearson correlation coefficient (R), and Manders' coefficients (M1, M2) were used to quantify colocalization. R measures the correlation between probes, while M1 and M2 are measures of co-occurrence; M1 captures the proportion of above-threshold pixels in the red channel occupying above-threshold pixels of the green channel, and vice versa for M2. Controls for specificity of primary and secondary antibodies ( Supplementary Fig. S9c, g, h), autofluorescence ( Supplementary Fig. S9d), specificity of PLA interaction foci ( Supplementary Fig. S9e), colocalization analysis ( Supplementary Fig. S9f), negative control (scramble) of siAGO2 ( Supplementary  Fig. S3a), and biological colocalization negative control ( Supplementary Fig. S3b) are provided.
Statistics and reproducibility. One-tailed Student's t-test 69 was used to evaluate the significance of Pearson (R) and Manders' values. The sample size for each test was n = 6 fluorescence images and the α-level was set to 0.01. Mann-Whitney Utests (two-sided) were used to compare the differences obtained in the lengths and ratios between Actin-filopodial and AGO2 protrusions. Although filopodial protrusions are log-normally distributed 25 , a non-parametric test was chosen since q-q plot and Shapiro test did not reveal a similar distribution for AGO2 protrusions. Actin-filopodial protrusions were measured on 102 cells (at least n = 3 biologically independent experiments, statistical sample size n = 3849 Actin-filopodial protrusions) while AGO2 protrusions on 66 cells (n = 5 biologically independent experiments, statistical sample size n = 66 AGO2 protrusions). The α-level was set at 0.01 for both tests. z-test was employed to test the proportion of non-lagging chromosomes (n1 = 176 GFP cells, n2 = 169 AGO2-GFP cells), micronuclei formation (n1 = 174 GFP cells, n2 = 168 AGO2-GFP cells), and chromosomal deregulation (n1 = 188 GFP cells, n2 = 182 AGO2-GFP cells) in GFP/AGO2-GFP assays. The same procedure was employed to test the proportion of micronuclei formation (n1 = 117 SCR cells, n2 = 117 siAGO2 cells), and chromosomal deregulation (n1 = 103 SCR cells, n2 = 114 siAGO2 cells) in SCR/siAGO2 knocked down cells. For the analyses three biologically independent experiments were conducted. The use of z-test was justified based on the expected frequencies that occurred in all our measurements 70,71 . The α-level was set at 0.01.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All raw images along with the corresponding processed ones can be found at Zenodo (https://doi.org/10.5281/zenodo.4415734). All other data are available from the corresponding author on reasonable request.

Code availability
All scripts used can be found at Zenodo (https://doi.org/10.5281/zenodo.4415734). Imaris 9.1.2, Matlab 2018a, and R version 4.0.3 were used. Further details on the procedure of measurements and corresponding statistical tests for all relevant figures can be found in the "Notes.txt" file at the Online Repository.