Mad1 facilitates α5 integrin trafficking from the Golgi to promote abscission during cytokinesis

Abstract

Mad1 is an essential component of the mitotic spindle assembly checkpoint. During interphase, Mad1 regulates the trafficking of α5 integrin from the Golgi to the plasma membrane. Here, we show that depletion of Mad1 or α5 integrin induces cytokinesis failure. Though the cytokinetic furrow ingresses with normal timing, it ultimately regresses, resulting in cytokinesis failure. We identify an ~300 amino acid internal fragment of Mad1 that is necessary and sufficient for the Golgi localization of Mad1. This fragment, which we term Mad1-Golgi, can rescue α5 integrin secretion and cytokinesis in Mad1-depleted cells. Expression of exogenous α5 integrin is sufficient to overcome the cytokinesis defect caused by Mad1 depletion. The contribution of both Mad1 and α5 integrin to cytokinesis is observed specifically under adherent growth conditions, and a pool of both proteins localizes to the midbody in adherent cells. These results demonstrate a previously uncharacterized role for Mad1 in cytokinesis by regulating α5 integrin secretion from the Golgi apparatus.

Mad1 is an essential component of the mitotic spindle assembly checkpoint with emerging roles in interphase. Here, the authors show that Mad1 plays an unexpected role in cytokinesis via its interphase function in facilitating α5 integrin secretion.

Introduction

Mitosis (nuclear division) and cytokinesis (cell division) are tightly controlled, evolutionarily conserved processes that ensure the timely and accurate production of two genetically identical daughter cells after DNA replication^1–4. Mitosis is regulated by a major cell cycle checkpoint known as the mitotic checkpoint (or spindle assembly checkpoint), which delays mitotic progression until every pair of replicated sister chromatids is properly attached to the microtubules of the mitotic spindle^5,6. Mitotic Arrest Deficient 1 (Mad1) plays an essential role in the mitotic checkpoint throughout evolution^6–11. Depletion of Mad1 reduces the length of time the mitotic checkpoint can maintain cells in mitosis in response to unattached chromatids, and increases chromosome missegregation^12,13.

Aside from its canonical role in mitosis, there is growing evidence that Mad1 performs key functions in interphase^13–17. Mad1 localizes to the nucleus and nuclear pores^14,17 where it facilitates the assembly of inhibitors of the anaphase promoting complex/cyclosome^13,18,19 and nuclear organization¹⁵. In addition to its nuclear localization, we and others demonstrated that Mad1 localizes to the Golgi apparatus during interphase^20,21. Mad1 depletion affects the cell surface localization of α5 integrin but does not delay accumulation of EGFR or VSVG at the plasma membrane, indicating Mad1 depletion does not induce a global defect in protein secretion²⁰. α5 integrin forms a heterodimer with β1 integrin to mediate interactions with the extracellular matrix (ECM) component fibronectin^22–25. As expected for cells with decreased cell surface expression of α5 integrin, Mad1 depleted cells are deficient in attaching and migrating on fibronectin. Conversely, overexpressing Mad1 resulted in increased migration on fibronectin, suggesting that Mad1 directly regulates the post-Golgi transport of α5 integrin to the plasma membrane²⁰. The full functional significance of this has remained unclear.

Integrins are heterodimeric proteins that control cellular adhesion to the ECM^22,26,27. Upon binding to the ECM, integrins can modulate intracellular and extracellular signaling pathways, resulting in migration and spatial reorganization of the ECM^28–30. Despite numerous insights on how integrins are recycled from the plasma membrane and internalized^31,32, understanding of mechanisms that control the transport of newly synthesized integrins remains a critical gap³³.

The role of adhesion to the ECM for cell differentiation and migration is well-defined^34–38. Adhesion has also been implicated in cytokinesis, at least in certain cell types^39–43. Initial evidence of a role for adhesion in cytokinesis came from Dictyostelium cells deficient in myosin heavy chain (mhcA), which were able to divide only under adherent growth conditions^41,42. This led to the proposal of multiple pathways of cell-cycle-coupled cytokinesis, including one that depends on actomyosin ring constriction and another that relies on adhesion^43,44. Adhesion also contributes to cytokinesis in certain vertebrate cell types, which exhibit binucleation in response to anchorage independent growth, including Xenopus blastomeres, mouse fibroblasts and melanoma cells, monkey kidney epithelial cells, and human retinal pigment epithelial cells^39,40,45. Integrin signaling has been implicated in the adhesion-mediated pathway of cytokinesis^46,47. Dictyostelium cells lacking the paxillin homolog (PAXB), which links integrins to the actin cytoskeleton via vinculin at focal adhesions⁴⁸, show defects in cytokinesis under adherent conditions⁴⁹. In mammalian cells, addition of fibronectin was sufficient to rescue binucleation in rat follicular cells cultured in serum-free media⁵⁰. Additionally, chondrocyte-specific loss of β1 integrin, which forms heterodimers with at least 10 α integrins, resulted in cytokinesis failure in growth plates in a mouse model⁵¹. These findings suggest a plausible link between adhesion and cytokinesis, yet no specific integrin complex has been directly implicated.

Though Mad1 plays a crucial role in mitosis, our understanding of the functional significance of its effect on α5 integrin biosynthetic transport from the Golgi has remained incomplete. Mad1 interacts with α5 integrin, but it has been unknown whether either of these proteins participates in cytokinesis. Here we show that depleting Mad1 or α5 integrin induces cytokinesis failure in multiple cell types, which is not due to chromosome missegregation. Interestingly, the cytokinesis defect caused by Mad1 knockdown can be rescued with a Mad1 fragment that is sufficient for Golgi localization of Mad1 and the directional release of α5 integrin from the Golgi, as well as by α5 integrin itself. Consistent with this, newly trafficked α5 integrin localizes to the midbody region in a Mad1-dependent manner. These data support the conclusion that Mad1 functions in an adhesion-mediated pathway of cytokinesis through facilitating the transport of newly synthesized α5 integrin from the Golgi apparatus.

Results

Mad1 depletion induces cytokinesis failure

While culturing clonal lines of HeLa cells expressing ~50% of the Mad1 protein found in wild type cells due to stable expression of 3 distinct shRNAs²⁰, we observed an unexpected ~3-fold increase in binucleation, a hallmark of cytokinesis failure (Fig. 1a–c). Modest, stable knockdown (KD) of Mad1 in T47D cells, which reduced the expression of Mad1 at the Golgi apparatus, also increased binucleation (Supplementary Fig. 1). To rule out the possibility that binucleation was induced due to the aneuploid nature of HeLa and T47D cell lines, we depleted Mad1 with siRNA in two near-diploid non-transformed epithelial cell lines, RPE1 retinal pigment epithelial cells and MCF10A non-transformed breast cells. ~2-fold depletion of Mad1 substantially increased binucleation in both RPE1 (Fig. 1d, e) and MCF10A (Fig. 1f, g) cells. Multinucleate cells with 3 or 4 nuclei were rarely observed (≤2%), and their incidence did not increase after Mad1 depletion.

Fig. 1. Mad1 functions in cytokinesis.

a Representative images of mononucleated wildtype or binucleated Mad1 knockdown (KD) HeLa cells. b–g Mad1 depletion increases binucleation in multiple cell types. Immunoblots with quantitation showing ~50% depletion of Mad1 protein in HeLa cells with stable Mad1 knockdown (b) or RPE1 (d) and MCF10A cells (f) transfected with siRNA against Mad1 for 4 days. Immunoblots are representative of three independent replicates. Color indicates biological replicate. NT, nontargeting. n = 3. Quantitation of binucleation in (c) HeLa, (e) RPE1 and (g) MCF10A cells. n = 100 cells from each of three biological replicates. h-i Representative still images from timelapse microscopy of RPE1 cells treated with non-targeting siRNA control (h; see also Supplementary Movie 1) or siMad1 (i; see also Supplementary Movie 2). Solid arrows show the presence of an intercellular bridge connecting the two daughter cells, whereas dotted arrows show the disappearance of the intercellular bridge. Asterisks show separating daughter cells in h and single daughter cell after failed cytokinesis in (i). Numbers indicate time in minutes from anaphase onset. Quantitation of (j) time between anaphase onset and completion of furrow ingression, (k) cytokinesis failure in cells with (n = 50) or without mitotic errors (n = 20) (see also Supplementary Fig. 2a, b) or (l) time from midbody appearance to abscission in cells that successfully divide in timelapse movies. In j, the violin plots show data from all three biological replicates. Horizontal lines indicate median. In l, small dots represent individual cells while large dots represent the mean for each biological replicate. n = 20 cells from each of three biological replicates (colored dots). *=p < 0.05, **=p < 0.01, unpaired t-test (j, k, l) or Tukey’s post hoc test with no adjustment (c, e, g). Data represent mean ± SE. All statistical tests are two-sided.

To confirm that binucleation due to Mad1 depletion occurred due to cytokinesis failure, we performed timelapse microscopy in RPE1 cells stably expressing fluorescent H2B-iRFP, tubulin-mNeonGreen, and HaloTag-CAAX, to label chromosomes, microtubules, and the plasma membrane, respectively. Cells were transfected with siRNA against Mad1 or a non-targeting control prior to timelapse imaging for 72 hours (Fig. 1h, i). Both control and Mad1-KD cells completed furrow ingression in a timely manner (Fig. 1j and Supplementary Movies 1 and 2), suggesting that the initiation of cytokinesis was not impaired by Mad1 depletion. However, Mad1-KD cells were significantly more likely to fail abscission, leading to regression of the cleavage furrow and binucleation (Fig. 1i, k and Supplementary Movie 2). To gain insight into potential steps leading to cytokinesis failure in Mad1-KD cells, we measured the timing from the appearance of the midbody to abscission in control cells and Mad1-KD cells that successfully completed cytokinesis. Interestingly, even Mad1-KD cells that succeed in cell division have an approximately two-fold increase in the time between the appearance of the midbody and abscission (Fig. 1l). Taken together, these data reveal that an approximate 2-fold depletion of Mad1 delays abscission and increases the incidence of cytokinesis failure.

Mad1 depletion can cause chromosome missegregation^10,13,52 due to weakened mitotic checkpoint surveillance. Lagging chromosomes and DNA bridges trapped in the cytokinetic furrow can delay or prevent cytokinesis through activation of the Aurora B-mediated abscission checkpoint^53–55. Therefore, we tested whether binucleation was an indirect consequence of mitotic errors. In our timelapse analysis, Mad1 depletion resulted in a 3-fold increase in lagging chromosomes (Supplementary Fig. 2a, b). However, none of the cells with lagging chromosomes failed cytokinesis (Fig. 1k and Supplementary Movie 3). Additionally, Mad1-KD cells that exhibited cytokinesis failure did not show evidence of chromosome missegregation (Fig. 1k), suggesting that cytokinesis failure did not occur as a consequence of chromosome missegregation. To further test this, we induced chromosome missegregation with an inhibitor of the Bub1 mitotic checkpoint kinase (BAY 1816032)⁵⁶, which recruits the mitotic checkpoint proteins Mad1, Mad2, BubR1, and Bub3 to unattached kinetochores during checkpoint activation^57,58. BAY 1816032 significantly increased metaphase and anaphase errors consistent with chromosome missegregation (Supplementary Fig. 2c–e). However, Bub1 inhibition did not induce cytokinesis failure (Supplementary Fig. 2f, g), demonstrating that mitotic errors are not sufficient to cause binucleation. We conclude that Mad1 participates in cytokinesis independent of its mitotic role in regulating chromosome segregation.

An internal fragment of Mad1 is sufficient for Golgi localization

Since the function of Mad1 in cytokinesis is independent of its role in mitosis, we questioned whether Mad1 participates via its role in the secretion of α5 integrin. We noted that cytokinesis failure in RPE1 or MCF10A cells depleted with siMad1 #2 was somewhat higher than cells transfected with siMad1 #1 (Fig. 1e, g), though Mad1 protein levels at the population level appeared slightly higher after treatment with siMad1 #2 (Fig. 1d, f). Interestingly, though both siRNA sequences depleted Mad1 at both the nucleus and the Golgi, Mad1 levels at the Golgi were significantly lower after transfection of siMad1 #2 (Supplementary Fig. 3), suggesting that the Golgi-localized pool of Mad1 may contribute to its role in cytokinesis.

To determine which regions of Mad1 are necessary for its localization to the Golgi, we created Mad1 truncations fused to a Flag tag (Supplementary Fig. 4a) and tested their localization (Supplementary Fig. 4b, c). A central portion of Mad1 spanning amino acids (aa) 180–485 showed robust Golgi localization (Supplementary Fig. 4b, c). A C-terminal fragment of aa 481–718 also demonstrated Golgi localization, though to a lesser extent than aa 180–485 (Supplementary Fig. 4b,c). Dissection of aa 481–718 revealed that aa 481–596 localized to the Golgi, though not as efficiently as aa 180–485 (Supplementary Fig. 4d, e). Removing aa 180–485 from Mad1 remarkedly reduced Golgi localization, though this version of Mad1 was nuclear in 35% of cells (Supplementary Fig. 4f). Mutation of the nuclear localization signal (NLS)⁵⁹ in the fragment lacking aa 180-485 produced a cytoplasmic version of Mad1 with poor Golgi recruitment (Supplementary Fig. 4f, g). As an additional validation that the Mad1 fragments were localizing to the Golgi apparatus, we treated cells with the microtubule poison vinblastine, which disperses Golgi stacks. Mad1 fragments were also dispersed after vinblastine treatment (Supplementary Fig. 4h, i). We conclude that Mad1 aa 180–485 (hereafter called Mad1-Golgi) is necessary and sufficient for recruiting Mad1 to the Golgi. We refer to Mad1 with a mutant NLS that lacks aa 180-485 as Mad1-ΔGolgi.

Mad1-Golgi supports α5 integrin trafficking from the Golgi and cytokinesis

To test the role of Mad1 and Mad1-Golgi in biosynthetic transport of α5 integrin, we designed an inducible protein trafficking system to follow α5 integrin as it is secreted from the endoplasmic reticulum (ER) through the Golgi to the plasma membrane. α integrins possess a conserved GFFKR amino acid motif in their transmembrane domain, which is required for insertion into the plasma membrane⁶⁰. They also contain a short cytoplasmic tail to mediate binding to cytoplasmic interactors^61,62, including Mad1. Since the cytoplasmic tail of integrins is sufficient for binding to cytoplasmic interactors⁶³, we fused the transmembrane domain and cytoplasmic tail of α5 integrin to four aggregation-prone DmrD (Dimerize reverse dimerization) repeats^64,65, HaloTag, and GFP (hereafter called 4xDmrD-Halo-GFP-α5^TMCT; Fig. 2a). Upon inducible expression with doxycycline, 4xDmrD-Halo-GFP-α5^TMCT misfolds and aggregates in the ER. The addition of a dimer-dimer solubilizer (DDS) allows 4xDmrD-Halo-GFP-α5^TMCT to fold correctly and undergo export from the ER (Fig. 2b). The release of 4xDmrD-Halo-GFP-α5^TMCT from the ER occurs rapidly after addition of DDS, as it begins localizing to the Golgi within 5 min. To test whether Mad1 depletion affects Golgi morphology, we measured Golgi area and volume in HeLa (Fig. 2c–e) and RPE1 (Fig. 2f, g) cells, but observed no difference. Trafficking of newly synthesized 4xDmrD-Halo-GFP-α5^TMCT from the ER to the Golgi was unaffected by Mad1 depletion, however, reduction of Mad1 substantially delayed trafficking of 4xDmrD-Halo-GFP-α5^TMCT from the Golgi in both fixed and timelapse analysis (Fig. 2h, i, Supplementary Fig. 5, Supplementary Movies 4 and 5). The use of a cell impermeable HaloTag dye, JF635i, confirmed that Mad1 depletion delayed the arrival of 4xDmrD Halo-GFP-α5^TMCT at the plasma membrane (Fig. 2j–l).

Fig. 2. An internal fragment of Mad1 promotes α5 integrin trafficking.

a Schematic of α5 integrin trafficking construct. b Trafficking sequence of α5 integrin construct in (a). c Immunoblot (top) and quantitation (bottom) showing Mad1 protein level in HeLa parental wild type and stable Mad1-KD cells. n = 3. Dot colors represent biological replicate. Quantitation of Golgi area (d, f) or volume (e, g) in HeLa (d, e) and RPE1 (f, g) cells transfected with siNT or siMad1 siRNA for 72 h. n = 33 cells from each of three biological replicates. h–i Quantitation of time required for trafficking 4xDmrD-Halo-GFP-α5^TMCT to and from the Golgi in (h) live HeLa cells (n = 15 cells from each of three biological replicates) expressing fluorescent Mannosidase II (Golgi marker; see also Supplementary Movie 4 and Supplementary Movie 5) or (i) fixed cells using GM130 Golgi antibody. n = 12 cells from each of three biological replicates. Representative images (j–k) and quantitation (l) showing reduced cell surface expression of 4xDmrD-Halo-GFP-α5^TMCT in Mad1-KD cells, detected using cell impermeable HaloTag dye. n = 12 cells from each of three biological replicates. m Schematic showing Mad1 domains fused to mScarlet used for reconstitution experiments in (o–q). KRAR is the wildtype NLS, whereas, KLAL is a mutated form. NES, nuclear export signal. CTD, C-terminal domain. M2ID, Mad2 interacting domain. n Representative images from WT HeLa cells showing colocalization of endogenous Mad1 (top, red arrow) and Flag-tagged Mad1-Golgi (middle, red arrow) but not Mad1-ΔGolgi (bottom) with the GM130 Golgi marker. Quantitation of time required for trafficking of 4xDmrD-Halo-GFP-α5^TMCT from the Golgi in HeLa cells transfected with (o) Mad1FL-mScarlet, (p) Mad1-Golgi-mScarlet, or (q) Mad1-ΔGolgi-mScarlet. Full length (FL) Mad1 and Mad1-Golgi rescue α5 integrin trafficking while Mad1-ΔGolgi does not. o–q n = 12 cells from each of three biological replicates. r Quantitation of binucleation in RPE cells + siRNA (48 h) + Mad1 transgenes (24 h) showing that full length Mad1 and Mad1-Golgi rescue binucleation while Mad1-ΔGolgi does not. n = 100 cells from each of three biological replicates. ns=not significant. *=p < 0.05, **=p < 0.01, unpaired t test (d–g) or Tukey’s post hoc test without adjustments (r), logistic regression (h, i, l, o–q). Data represent mean ± SE. All statistical tests are two-sided.

To ascertain which portions of Mad1 could complement the effect of Mad1 depletion on 4xDmrD-Halo-GFP-α5^TMCT trafficking from the Golgi to the plasma membrane, we generated parental wild type or stable Mad1-KD cell lines expressing shRNA-resistant full length Mad1, Mad1-Golgi, and Mad1-ΔGolgi fused to mScarlet (Fig. 2m, n). We found that expression of either full length Mad1 or Mad1-Golgi was sufficient to rescue the trafficking delay for 4xDmrD-Halo-GFP-α5^TMCT in Mad1-depleted cells (Fig. 2o, p and Supplementary Fig. 5a–c). In contrast, expression of Mad1-ΔGolgi was not, and these cells still showed a significant delay in trafficking 4xDmrD-Halo-GFP-α5^TMCT from the Golgi (Fig. 2q and Supplementary Fig. 5d). We conclude that Mad1-Golgi, but not Mad1-ΔGolgi, supports biosynthetic trafficking of the truncated α5 integrin reporter.

To determine the role of specific regions of Mad1 in cytokinesis, we transfected Mad1-depleted RPE1 cells with Flag-tagged full-length Mad1, Mad1-Golgi, and Mad1-ΔGolgi. Consistent with the α5 integrin trafficking data, full length Mad1 and Mad1-Golgi were able to rescue binucleation, while Mad1-ΔGolgi was not (Fig. 2r). These data demonstrate that complementing Mad1-KD cells with Mad1-Golgi rescues both 4xDmrD-Halo-GFP-α5^TMCT trafficking and cytokinesis failure. We conclude that the function of Mad1 at the Golgi contributes to its role in cytokinesis.

α5 integrin depletion induces cytokinesis failure

Since the portion of Mad1 that supports biosynthetic trafficking of 4xDmrD-Halo-GFP-α5^TMCT is sufficient to rescue binucleation in Mad1-KD cells, we next tested whether α5 integrin is necessary for cytokinesis. Indeed, depletion of α5 integrin in RPE1 and MCF10A cells substantially increased binucleation (Fig. 3a–c). Since α5 integrin, unlike Mad1, is not an essential gene, we generated α5 integrin knock out (KO) cells by using CRISPR/Cas9 to edit exon 3 of ITGA5 (Supplementary Fig. 6a, b). Exon 3 harbors a β-propeller domain, which is responsible for binding to β1 integrin and is therefore necessary for α5 integrin function²⁵. An antibody that recognizes a conformational state of α5 bound to β1 integrin, SNAKA51²⁴, as well as an antibody recognizing a C terminal epitope showed loss of α5 in the KO clones (Fig. 3d, e and Supplementary Fig. 6c). α5 KO RPE1 and MCF10A cells were particularly prone to cytokinesis failure when plated on fibronectin (Fig. 3d, e and Supplementary Fig. 6d, e), suggesting that α5 integrin KO cells adapt and rely on other integrins or adhesion molecules for cytokinesis when not cultured on fibronectin.

Fig. 3. α5 integrin functions in cytokinesis.

a Representative images showing mononucleated (wild type) and binucleated (α5 integrin knockdown (KD)) cells. Depletion (b, c) or knockout (d, e) of α5 integrin increases binucleation in multiple cell types. Immunoblots with quantitation showing α5 integrin protein levels after α5 integrin depletion (b, c; 4 days after transfection) or knockout (d, e). n = 3 biological replicates. Quantitation of binucleation in the indicated cell lines (b–e, bottom). α5 integrin KO cells were plated on 5 µg/mL fibronectin-coated plates for 24 hours prior to quantitation of binucleation. n = 100 cells from each of three biological replicates. Colored dots represent biological replicates. Representative images from timelapse microscopy of RPE1 cells treated with siNT control (f; see also Supplementary Movie 6) or siα5 integrin (g; see also Supplementary Movie 7). Solid arrows show the presence of the intercellular bridge connecting the two daughter cells. Dotted arrows show disappearance of the intercellular bridge. Asterisks show separating daughter cells in (f) and single daughter cell after failed cytokinesis in (g). Merged images show magnification of cleavage furrow and intercellular bridge. Numbers indicate time after anaphase onset in minutes. Quantitation of (h) time from anaphase onset to furrow ingression, (i) percentage of cells exhibiting cytokinesis failure, and (j) time from midbody appearance to abscission. n = 20 cells from each of three biological replicates, indicated with different colors. Small dots, individual cells. Large dots, mean for each biological replicate. In (h), violin plots show data from all three biological replicates. Quantitation of binucleation in (k) RPE1 or (l) MCF10A cells after depletion of α5 integrin for 48 h ± rescue with α5 cDNA for 24 h. n = 100 cells from each of three biological replicates. m Quantitation of binucleation in RPE1 α5 integrin KO clones on 5 µg/mL fibronectin-coated plates and transfected with indicated plasmids for 24 h. ns=not significant. *=p < 0.05, **=p < 0.01, unpaired t-test (h–j) or Tukey’s post hoc test without adjustment (b–e, k–m). Data represent mean ± SE. All statistical tests are two-sided.

To confirm that binucleation observed after loss or depletion of α5 integrin was due to cytokinesis failure, we performed timelapse microscopy of RPE1 cells stably expressing fluorescent H2B-iRFP, tubulin-mNeonGreen, and HaloTag-CAAX to label chromosomes, microtubules, and the plasma membrane, respectively. We transfected fluorescently labelled RPE1 cell lines with siRNA against α5 integrin or non-targeting control and performed timelapse microscopy for 72 h, imaging every 5 min (Fig. 3f, g and Supplementary Movies 6 and 7). Both control and α5 integrin-KD cells completed furrow ingression in a timely manner (Fig. 3h), suggesting that the onset of cytokinesis was not impaired by loss of α5 integrin. However, after what initially appeared to be successful cytokinesis, abscission ultimately failed in a proportion of α5 integrin-KD cells, and the cytokinetic furrow regressed, resulting in binucleation (Fig. 3g, i and Supplementary Movie 7). As seen for Mad1 depletion, even α5 integrin-KD cells that successfully completed cytokinesis exhibited a significant delay in the time between midbody formation and abscission (Fig. 3j). This was not due to mitotic errors, as these were not increased due to α5 integrin knockdown (Supplementary Fig. 6f). To confirm cytokinesis failure was due to reduced expression of α5 integrin and not an off-target effect, we reconstituted α5 integrin depleted cells and α5 integrin KO cells with GFP-Flag or α5 integrin-Flag. Expression of α5 integrin-Flag but not GFP-Flag rescued cytokinesis in RPE1 (Fig. 3k) and MCF10A (Fig. 3l) cells in which endogenous α5 integrin was depleted. Similarly, α5 integrin-Flag reduced binucleation to control levels in RPE1 α5 integrin KO clones, while GFP-Flag did not (Fig. 3m). We conclude that α5 integrin participates in cytokinesis in adherent cells.

α5 integrin localizes to the midbody region in a Mad1-dependent manner

To test if a pool of α5 integrin localizes to the intercellular bridge, we used cells expressing 4xDmrD-Halo-GFP-α5^TMCT and tracked newly synthesized α5 integrin 1 h after release from the ER. Indeed, a portion of 4xDmrD-Halo-GFP-α5^TMCT trafficked to the midbody region in the intercellular bridge (Fig. 4a). Interestingly, depletion of Mad1 significantly reduced the level of newly synthesized 4xDmrD-Halo-GFP-α5^TMCT on either side of the midbody (Fig. 4a–c).

Fig. 4. Mad1 directs α5 integrin trafficking to the intercellular bridge.

a–c HeLa cells stably expressing 4xDmrD-Halo-GFP-α5^TMCT were treated with doxycycline overnight to express the inducible secretion construct, which aggregates in the ER. DDS was added for 1 h to permit folding and trafficking from the ER. a Representative images showing 4xDmrD-Halo-GFP-α5^TMCT (GFP-α5^TMCT) localizes to the midbody region in control cells (top). Mad1 depletion reduces GFP-α5^TMCT localization to the midbody region (bottom). b Quantitation of nuclear Mad1 signal showing ~50% depletion 24 h after transfection with siMad1. n = 15 cells in each of three biological replicates. c Quantitation of the midbody intensity of GFP-α5^TMCT after background subtraction (see “Methods”). n = 12 cells from each of three biological replicates. Representative images from HeLa cells showing reduction of endogenous α5 integrin, identified with Abcam AB150361 (d) or SNAKA51 (e) antibodies, to the midbody region after siMad1 depletion for 72 h. Quantification of Mad1 nuclear intensity (f, i) or α5 integrin intensity at the midbody with Abcam (g, j) or SNAKA51 (h, k) antibodies in HeLa (f–h) or RPE1 (i–k) cells transfected with siRNA for 72 h. n = 12 cells in each of three biological replicates. l–o Validation of Abcam and SNAKA51 α5 integrin antibodies. Representative images showing midbody localization of (l) Abcam or (n) SNAKA51 α5 integrin antibodies in RPE1 WT cells, which is lost in α5 integrin KO cells. Quantitation of (m) Abcam or (o) SNAKA51 α5 integrin midbody localization. n = 12 cells in each of 3 biological replicates. p–r Mad1-Golgi-3xFlag but not GFP-3xFlag localizes to the midbody. p Representative images of HeLa cells transfected with GFP-3xFlag or Mad1-Golgi-3xFlag. q–r Percentage of cells showing midbody localization of Flag (q), or (r) fluorescence intensity of Flag at the midbody. n = 40 (q) or n = 12 (r) telophase cells per condition in each of three biological replicates. For plots in b, c, f–k, m, o, q, r, small dots represent individual cells while large dots represent the mean for each biological replicate. Colors indicate biological replicate. All insets show magnification of the midbody region. *=p < 0.05, **=p < 0.01, unpaired t-test. Data represent mean ± SE. All statistical tests are two-sided.

To test if endogenous α5 integrin also localizes to the midbody in a Mad1-dependent manner, we used α5 integrin antibodies that recognize the C terminal transmembrane domain and cytoplasmic tail (Fig. 4d and Supplementary Fig. 6a) or a conformational state of α5 bound to β1 integrin, SNAKA51²⁴ (Fig. 4e). Consistent with the results using 4xDmrD-Halo-GFP-α5^TMCT, both antibodies showed robust localization to the midbody (Fig. 4d–k). Mad1 depletion reduced localization of endogenous α5 integrin in the midbody in both HeLa (Fig. 4d–h) and RPE1 (Fig. 4i–k) cells. Midbody localization was lost in RPE1 α5 KO cells, confirming the specificity of the α5 integrin antibodies (Fig. 4l–o).

Integrins contribute to the formation of focal adhesions^29,30,48, which support cell adhesion and migration. The localization of α5 integrin at the midbody suggested that focal adhesions could form at or near the midbody and potentially contribute to cytokinesis. Using vinculin as a marker for focal adhesions, we observed that ~50% of cells had focal adhesions at or near the midbody in wild type RPE1 cells, which was reduced by ~half in α5 KO cells (Supplementary Fig. 6g, h). Vinculin intensity was reduced by ~40% in α5 KO cells in telophase (Supplementary Fig. 6i, j), suggesting that α5 integrin contributes to the formation of focal adhesion sites at the midbody region.

Mad1 antibodies produced in-house⁶⁶ showed a portion of Mad1 also localized to either side of the midbody, consistent with the localization pattern for 4xDmrD-Halo-GFP-α5^TMCT (Fig. 4a) and endogenous α5 integrin (Fig. 4d, e, l, m). To confirm this observation, we obtained three additional Mad1 antibodies from commercial vendors and validated them based on proper localization of Mad1 at kinetochores during prometaphase in HeLa (Supplementary Fig. 7a) and RPE1 (Supplementary Fig. 7b) cells. These antibodies also showed robust localization of a pool of Mad1 to the midbody in HeLa (Supplementary Fig. 7c) and RPE1 (Supplementary Fig. 7d) cells. As expected, depleting Mad1 with siRNA reduced Mad1 midbody localization and fluorescence intensity (Supplementary Fig. 7e,f). Since Mad1-Golgi was sufficient to rescue 4xDmrD-Halo-GFP-α5^TMCT trafficking (Fig. 2p) as well as cytokinesis failure (Fig. 2r), we tested if it also localizes to the midbody. Mad1-Golgi-3xFlag showed robust localization at the midbody, while GFP-3xFlag did not (Fig. 4p–r). These data are consistent with previous studies that independently recovered Mad1^67,68 and α5 integrin^68,69 in the midbody proteome using affinity mass spectrometry analysis (Supplementary Table 1). These data provide support for a previously unidentified role of Mad1 in regulating the transport of α5 integrin to the midbody region.

Mad1 and α5 contribute to the adhesion-mediated pathway of cytokinesis

Two major pathways contribute to the physical separation of cells during cytokinesis following mitosis, actomyosin ring constriction and traction via adhesion molecules (Fig. 5a), though the reliance on traction varies by cell line^{43–45,49,70,71}. Since α5 integrin participates in adhesion, we reasoned that Mad1 and α5 integrin contribute to the traction-mediated arm of cytokinesis, specifically through α5β1-mediated interactions with fibronectin. In this case, depletion of Mad1 should not impair cytokinesis on collagen, which binds to α1β1, α2β1, α10β1, and α11β1 integrins^38,51. To test this, we depleted Mad1 with siRNA for 48 h and plated cells on glass coverslips coated with 50 µg/mL collagen for 24 h (Supplementary Fig. 8a–d). Indeed, Mad1-KD cells successfully divided on collagen, whereas, control cells treated with blebbistatin, a non-muscle myosin II inhibitor that blocks actomyosin ring constriction during cytokinesis, exhibited binucleation (Supplementary Fig. 8e–g), supporting our hypothesis that Mad1 mediates cytokinesis via α5β1-mediated binding to fibronectin.

Fig. 5. Mad1 and α5 integrin function in the traction pathway in cytokinesis.

a Schematic depicting two cytokinesis pathways. Red bar: actomyosin ring. Red arrows: direction of cell movement. b Normal (mononucleate) and binucleate cell. Binucleation in (c) HeLa or (d) RPE1 cells transfected with siRNA for 48 h. Binucleation in RPE1 (e) or MCF10A (f) cells transfected with siRNA for 48 h, trypsinized, and plated on low attachment or standard plates for 24 h. g Normal (mononucleate) and binucleate B cells. Mad1 (h) or α5 integrin (i) fluorescence intensity in siRNA (Cy-3)-positive B cells. n = 22 cells in each of three biological replicates. j Binucleation in B cells treated with siRNA or myosin II inhibitor blebbistatin for 72 h. Binucleation in (k) α5 integrin-depleted or (l) α5 integrin KO RPE1 cells transfected with full length (FL) Mad1-Flag or GFP-Flag. n = 100 cells per biological replicate. Binucleation in (m) RPE1 or (n) MCF10A cells transfected with siRNA for 48 h or (o) HeLa cells with stable Mad1 knockdown (KD) scored for binucleation 24 h after α5-Flag or GFP-Flag transfection. p–r Midbody localization of α5 integrin in siMad1 cells is rescued by overexpressing α5 integrin-Flag. p Representative images from HeLa cells. Insets show magnification of the midbody region. Quantification of (q) Mad1 nuclear intensity or (r) midbody α5 integrin intensity in HeLa cells treated with siRNA for 72 h ± transfection with α5 integrin-Flag for 24 h. n = 34 (q) or 17 (r) cells in each of three biological replicates. s Model of Mad1 and α5 integrin function in cytokinesis. Mad1 promotes the transport of α5 integrin from the Golgi to the plasma membrane and midbody, generating traction forces. After Mad1 depletion, α5 integrin accumulates at the Golgi, decreasing traction and increasing cytokinesis failure. c–f, m–o n = 100 cells in each of three biological replicates. h–i, q, r, Small dots, individual cells. Large dots, mean of biological replicate, indicated by color. ns=not significant. *=p < 0.05, **=p < 0.01, unpaired t-test (h, i), Tukey’s post hoc test without adjustments (c–e, j–o, q, r). Data represent mean ± SE. All statistical tests are two-sided.

We next tested whether Mad1 and α5 integrin contribute to the same pathway of cytokinesis. We found that dual depletion of Mad1 and α5 integrin in HeLa (Fig. 5b, c) or RPE1 (Fig. 5d) cells was not additive but yielded essentially the same rate of binucleation as depleting either protein alone, suggesting that Mad1 and α5 integrin act in the same pathway.

Next, we tested if Mad1 and α5 integrin contribute to cytokinesis via the traction or actomyosin pathway by depleting Mad1 or α5 integrin and plating cells on standard plates to which they could adhere or on low attachment plates. If Mad1 and α5 integrin contribute to the traction mediated pathway, we expected that their depletion would compromise cytokinesis in adherent cells, but not those grown under conditions of low attachment. Though RPE1 and MCF10A cells showed an increase in binucleation when grown on low attachment plates, this was not exacerbated by depletion of Mad1 or α5 integrin (Fig. 5e, f). Indeed, depletion of Mad1 or α5 integrin only increased cytokinesis failure in cells grown under normally adherent conditions (Fig. 5e, f, blue). To confirm that Mad1 and α5 integrin function via the traction-mediated pathway of cytokinesis, we tested whether depletion of Mad1 or α5 integrin increased cytokinesis failure in A++ 11 F B lymphocytes grown in suspension. We transfected cells with Cy3-labelled siRNA against Mad1 or α5 integrin and scored cells positive for Cy3. While treatment with blebbistatin increased cytokinesis failure, ~2-fold depletion of Mad1 or α5 integrin did not (Fig. 5g–j). Thus, non-adherent B cells depend on actomyosin ring constriction but not Mad1 or α5 integrin for cytokinesis. These results support the conclusion that Mad1 and α5 integrin function in the traction-mediated arm of the cytokinesis pathway.

We next performed rescue experiments to test whether the role of Mad1 in cytokinesis is due to its function in promoting the biosynthetic trafficking of α5 integrin. First, we asked whether overexpression of Mad1 was sufficient to rescue binucleation caused by depletion or loss of α5 integrin. We reasoned that Mad1 overexpression could rescue cytokinesis failure in cells with reduced expression of α5 integrin, but not in cells completely lacking α5 integrin. Indeed, transfection of full length Mad1-Flag but not GFP-Flag rescued binucleation to control levels in RPE1 cells depleted of α5 integrin (Fig. 5k). However, neither Mad1-Flag nor GFP-Flag was capable of rescuing cytokinesis in α5 integrin-KO RPE1 cells (Fig. 5l).

We next tested whether the cytokinesis defect caused by depleting Mad1 could be rescued simply by overexpressing α5 integrin. GFP-Flag or α5 integrin-Flag were transfected into RPE1, MCF10A, or HeLa cells in which Mad1 was depleted transiently (Fig. 5m, n) or stably (Fig. 5o). In all cases, α5 integrin-Flag rescued binucleation to control levels while GFP-Flag did not (Fig. 5m–o). These data suggest that in Mad1-depleted cells, α5 integrin overexpression is sufficient to bypass the reliance on Mad1 and restore midbody localization of α5-integrin. To test this, we quantified α5 integrin accumulation at the midbody in HeLa cells depleted of Mad1 with siRNA (Fig. 5p–q). Depletion of Mad1 reduced endogenous α5 integrin at the midbody (Fig. 5p; see also Fig. 4d–k), but overexpression of α5 integrin-Flag in Mad1-depleted cells rescued α5 integrin midbody intensity to control levels (Fig. 5p–r). We conclude that Mad1 functions in the traction-mediated pathway of cytokinesis by promoting the secretion of α5 integrin from the Golgi apparatus (Fig. 5s).

Discussion

Mad1 plays a well-characterized role in chromosome segregation during mitosis but is expressed throughout the cell cycle. We previously identified an interphase role of Mad1 in supporting the trafficking of α5 integrin from the Golgi apparatus to the plasma membrane²⁰, though the ramifications of this function remained incompletely understood. Here, we demonstrate that the role of Mad1 in trafficking α5 integrin from the Golgi provides an important contribution to cytokinesis in adherent cells. Mad1 depletion induces cytokinesis failure in cells that do not display chromosome missegregation, suggesting the role of Mad1 in cytokinesis is independent of its canonical role in the mitotic checkpoint. Consistent with this, α5 integrin depletion induces cytokinesis failure but not chromosome missegregation, while inhibition of the Bub1 kinase produces chromosome missegregation without cytokinesis failure. Overexpression of α5 integrin is sufficient to rescue the defect in cytokinesis caused by depletion of Mad1. Hence, we have identified previously uncharacterized, interdependent roles for Mad1 and α5 integrin in cytokinesis.

Adhesion has been implicated in cytokinesis in certain cell types, and integrins participate in adhesion, but the evidence directly linking integrins to cytokinesis has thus far been limited. Existing evidence has relied on conditional deletion or mutation of β1 integrin, which heterodimerizes with at least 10 α integrins^51,72,73, depletion of Rab21, a small GTPase involved in membrane trafficking, including recycling α/β1 integrin heterodimers^74,75, or downstream signaling pathways^46,47. Though these reports have suggested a role for integrins in cytokinesis, the specific integrin heterodimers and their precise role in cytokinesis have remained elusive. We show here that α5 integrin, which forms a heterodimer with β1 integrin, is necessary for efficient completion of the final stages of cytokinesis in adherent cells. Though the cytokinetic furrow initiates in a timely manner and in the proper cellular location, abscission is delayed when levels of α5 integrin are depleted, and ultimately fails in a subset of cells, resulting in binucleation. This represents the first example of a specific integrin complex directly shown to contribute to cytokinesis.

Non-muscle myosin II contributes to the generation of contractile forces in actomyosin ring constriction. However, experiments in wildtype and myosin null Dictyostelium cells showed similar traction forces, suggesting that actin polymerization could be involved⁷⁶. Consistent with this, treatment with latrunculin B, an inhibitor of actin polymerization, resulted in loss of traction force⁷⁶. This suggests that traction forces generated at the midbody region rely on actin polymerization rather than myosin, coupled with adhesion. In adherent human cell lines on substrate that contains fibronectin, this is mediated by α5β1 integrin, whose secretion is facilitated by Mad1.

In addition to the pool of α5 integrin at the midbody region, we have also observed a portion of Mad1 at this location. Though Mad1 localization to the midbody has not been widely reported, we observed this localization with four distinct antibodies, and our data agree with recent publications that recovered Mad1, α5 integrin, or β1 integrin in the midbody proteome in human cells (Supplementary Table 1)^67–69,77. α5 integrin could arrive at the midbody region directly from the Golgi, from other regions of the plasma membrane, or both, and our data cannot distinguish between these possibilities. However, a portion of Mad1 and the Golgi-targeting region in Mad1 (Mad1-Golgi) are recruited to the midbody region but do not localize generally to the plasma membrane. This raises the question of whether Mad1 is co-transported with α5 integrin from the Golgi to the midbody.

Traction at the cell body aids in successful cytokinesis, but it is unknown whether traction at the midbody generates forces that also contribute. Here we observed α5 integrin and focal adhesions at the midbody region in human cells. Consistent with this, traction forces were previously inferred to occur at the midbody region of 3T3 fibroblasts based on the formation of wrinkles on an elastic substrate⁷⁸. The inferred midbody traction forces were parallel to the cytokinetic furrow prior to and during furrow ingression and became perpendicular to it after completion of furrow ingression, prior to the daughter cells migrating apart⁷⁸. However, this assay could not precisely localize force generation, and more recent experiments in Dictyostelium did not identify traction force at the midbody^76,79. While integrins mediate traction forces, focal adhesions, and cell-substrate adhesion in many eukaryotic cells, Dictyostelium lacks canonical α and β integrins. Instead, they rely on SibA, a transmembrane protein with features similar to β integrin chains, and its homolog SibC for substrate adhesion^80,81. Though SibA and SibC play a key role in cell-substrate adhesion, this differs from focal adhesions in animals due to the lack of integrins and the tyrosine kinases FAK and SRC in Dictyostelium⁸². Hence, though traction force was not observed at or near the midbody region in Dictyostelium, it is unclear if the specifics of force generation during cytokinesis are conserved throughout evolution. Consistent with our findings in human cells, focal adhesions and traction forces were observed near the midbody region in zebrafish epicardium⁸³. It is tempting to speculate that adhesion mediated by α5β1 integrin could contribute to cytokinesis via traction at the midbody region. Based on the apparent difference between midbody traction forces in Dictyostelium versus vertebrate cells, direct measurements of traction forces during cytokinesis in vertebrate cells are warranted.

Though traction forces that result in pulling forces on the intercellular bridge clearly contribute to successful cytokinesis in a subset of adherent cells, there is also evidence that release of tension serves as the trigger for abscission⁸⁴. In HeLa Kyoto cells, faster daughter cell separation, which strongly correlated with tension on the cytokinetic bridge, extended the time required for abscission. Daughter cells typically stopped separating ~10 min prior to abscission. Artificially releasing tension on the bridge by using laser ablation to sever one side substantially shortened the time required for the cell to sever the microtubules on the other side of the midbody to complete abscission⁸⁴. Attachment of α5 integrin to the substrate at the midbody region may provide a brake on daughter cell separation, ultimately reducing the separation speed of the daughter cells to zero. This would relieve tension on the intercellular bridge and trigger abscission.

Integrins play a major role in cancer relevant pathways such as cellular proliferation, migration, and metastasis²⁸. α5 integrin is a marker of metastasis and poor prognosis^85–88 that promotes multiple stages of metastasis, including migration, invasion, expression, and activation of matrix metalloproteinases, and activation of FAK^{29,87,89–97}. Extensive study of endocytic recycling of integrins from the plasma membrane enabled clinical trials testing integrin antagonists in cancer patients^96,98,99. Unfortunately, however, effective modulation of integrin activity in patients is not yet achievable. Interfering with the transport of newly synthesized α5 integrin offers an opportunity to develop novel approaches to inhibit integrin signaling, though the mechanisms that direct integrins from the Golgi to the plasma membrane have remained largely unknown. Here we identified an ~300 amino acid region in Mad1 (aa 180–485) that supports α5 integrin trafficking. While general inhibition of secretory protein trafficking is highly toxic, Mad1 plays a much more cargo-specific role in trafficking, and interference with this function may reduce cancer cell migration and metastasis.

Our observation that siRNA sequences that target different exons of Mad1 have differential effects on the pool of Mad1 at the Golgi suggests the expression of alternatively spliced variants of Mad1 that differ in their propensity for Golgi recruitment. Consistent with this, publicly available databases report numerous Mad1 mRNA transcripts due to exon skipping or inclusion of alternate exons. Indeed, an alternatively spliced transcript of Mad1 that lacks exon 4 was identified in hepatoma cell lines and human hepatocellular carcinomas⁵⁹. Assessing the prevalence and functional impact of alternatively spliced variants of Mad1 will be an important future direction.

Here, we have uncovered a functional role for Mad1 in cytokinesis through facilitating the transport of newly synthesized α5 integrin. Both Mad1 and α5 integrin participate in the traction-mediated pathway of cytokinesis. Additional questions raised by these findings include identifying the binding partners that recruit Mad1 to the Golgi, testing whether Mad1 accompanies α5 integrin en route to the plasma membrane, and defining methods to exploit this pathway to prevent the contribution of α5 integrin to metastasis.

Methods

Cell culture

HeLa cells were grown in DMEM supplemented with 10% FBS, 2 mM L-glutamine, and 50 µg/mL penicillin/streptomycin. MCF10A cells were grown in DMEM/F12 with 5% horse serum, 50 µg/mL penicillin/streptomycin, 20 ng/mL EGF, 0.5 mg/mL hydrocortisone, 100 ng/mL cholera toxin, and 10 μg/mL insulin. T47D cells were grown in RPMI-1640 supplemented with 10% FBS, 2 mM L-glutamine, and 50 µg/mL penicillin/streptomycin. A++ 11 F B lymphocytes were cultured in RPMI-1640 supplemented with 10% FBS, 2 mM L-glutamine, and 50 µg/mL penicillin/streptomycin. B cells were electroporated with Cy3-labelled siRNA at 1200 volts and 10 mA for 5 s. All cell lines were cultured in 5% CO₂ at 37 °C. HeLa cells stably expressing the 4xDmrD-Halo-GFP-α5^TMCT trafficking construct were stably transduced with retroviruses expressing full length Mad1, Mad1-Golgi, or Mad1-ΔGolgi fused to mScarlet. GFP and mScarlet double-positive cells were FACS sorted into 96 well plates to obtain monoclonal cell lines. T47D Mad1-KD cells were achieved using retroviruses expressing Mad1 shRNA (5’- TGAGATCTTTGAACAACTT-3′ for Mad1-shRNA#40, or 5′-AGCGATTGTGAAGAACATG-3′ for Mad1-shRNA#717)¹⁶. Cell lines were verified by STR profiling.

Plasmids, cloning and virus production

The full list of plasmids used in this study is provided in Supplementary Table 2. Gibson assembly was used to clone all plasmids except for pTalen-Tet-ON-4xDmrD-Halo-GFP-α5^TMCT, which was synthesized directly (Vector Builder). Briefly, four DmrD repeats were fused to Halo-tag, GFP, the transmembrane domain of ITGA5, and the cytoplasmic tail of ITGA5, downstream of the TRE3G promoter. 4xDmrD-Halo-GFP-α5^TMCT was then inserted into the AAVS1 safe harbor site in HeLa cells using separate left and right TALEN arm plasmids. Mad1 truncation plasmids fused to 3x-Flag were previously used in ref. ¹⁶. The additional Mad1 truncations, Mad1-NLSmutant-Δ180-485, Mad1-FL-NLSmut, Mad1-480-496, and Mad1-500-596 were cloned into a pcDNA5 vector. Viral particles were generated in HEK-293 Gag-Pol expressing cell line. Briefly, 2X HEPES buffered saline (pH 7.05) was complexed with premixed 2 M CaCl₂ and equimolar concentrations of the retroviral and VSVG plasmids. Transfection complex was added to the cells for 10 hours. Viral particles generated were collected 72 h post-transfection and filtered with a 0.45 µm sterile filter before freezing at −80 °C.

Generation of ITGA5 KO strains

RPE1 and MCF10A ITGA5 KO strains were generated using guide RNAs targeting exon 3 of ITGA5. Single guide RNA sequences (5’-TTGCTGGCCCCTCACCAAGATGG-3’ or 5’-CAAGTCCTTGCAGTGGTTCGGGG-3’) were cloned into px330-Cas9 (Addgene #42230) backbone¹⁰⁰. ITGA5 sgRNA plasmids were transfected into cell lines together with a GFP vector in a 9:1 ratio. 48 hours after transfection, cells with high GFP expression were FACS sorted into 24 well plates. After polyclonal cells reached confluency, they were re-sorted into 96 well plates for monoclonal selection. Wells with single colonies were expanded into 24 well plates and screened by immunoblotting for ITGA5 KO using ITGA5 (AB150361, Abcam) antibody. Genomic DNA was extracted from KO clones using QuickExtract DNA Extraction Solution (LGC Biosearch Technologies) to perform PCR amplification. Primers used were as follows: Forward- 5’- TTCCACAAGGCTCTCGGC-3’, Reverse- 5’- GAGGGGCCTTCAGATTCAGTCCAATAAG-3’. The PCR band was excised from an agarose gel, purified, and sent for Amplicon sequencing using NGS.

Microscopy

All immunofluorescence images were acquired on a Nikon Ti-2 inverted microscope using a CoolSNAP HQ2 camera controlled by Nikon Elements imaging software. Immunofluorescence images were deconvolved using AQI 3D deconvolution tool in Nikon Elements software.

For live-cell imaging, cells were seeded in a 24 well glass bottom plate (Cellvis, CA) overnight at 37 °C and 5% CO₂. Thirty minutes prior to imaging, cells were equilibrated in a Tokai Hit stage top incubation system set to 37 °C and 5% CO₂.

For cytokinesis movies, JFX 549 HaloTag dye was diluted to 50 nM in cell culture media and incubated with RPE1 cells expressing HaloTag-CAAX, tubulin-mNeonGreen, and H2B-iRFP for 30 minutes. Media containing HaloTag dye was replaced with fresh media prior to imaging. Cells were imaged every 5 min for 72 h on a Nikon Ti-2 inverted microscope equipped with a CoolSNAP HQ2 camera at 475-nm, 550-nm, and 621-nm excitation using a multiband quad filter cube (409/493/573/652-Di02). Images were acquired using a Plan Apo 20x objective lens (NA = 0.80 NA) with 2 µm z steps over a range of 12 µm.

For 4xDmrD-Halo-GFP-α5^TMCT live trafficking experiments, Mannosidase II-mApple and 4xDmrD-Halo-GFP-α5^TMCT labelled cells were treated with D/D solubilizer (DDS, Takara Bio) and immediately imaged on a Nikon Ti-E inverted microscope equipped with an ORCA Flash4 camera at 555-nm excitation (605/55 emission filter) and at 470-nm excitation (525/50 emission filter) every 1 min for 45 min. Images were acquired using a Plan Apo 60x oil objective lens (NA = 1.4) at 0.5 µm z steps over 5 µm. Images were compiled into montages using Nikon Elements software.

Immunofluorescence and western blotting

For standard immunofluorescence (IF), cells were seeded at 50% confluency on glass coverslips in a 12 well plate for 24 hours. For quantifying cytokinesis failure, cells were treated with siRNA for the indicated amount of time. 24 h before the endpoint, cells were trypsinized and plated on coverslips at 20% confluency. Coverslips were rinsed with microtubule stabilizing buffer (MTSB: 100 mM K-Pipes, pH 6.9, 30% (wt/vol) glycerol, 1 mM EGTA, and 1 mM MgSO₄) and replaced with 4% formaldehyde prepared in MTSB for 10 min at room temperature. Cells were extracted using 0.5% Triton X-100 in MTSB for 1 min at 37 °C, washed twice with 25 °C PBS, and then blocked with Triton block (0.2 M glycine, 2.5% FBS, and 0.1% Triton X-100 in 1× PBS) for 1 h at room temperature on an orbital shaker. For Mad1 antibodies GTX105079 and GTX109519, coverslips were fixed with methanol at −20 °C for 10 min, rinsed twice with 25 °C PBS, and blocked with Triton block. Primary antibodies were diluted in Triton block and incubated at room temperature for 1 h. Coverslips were washed three times with 0.1% Triton-X100 in PBS, five minutes each. Alexa Fluor (Invitrogen) secondary antibodies were diluted at 1:200 in Triton Block and incubated on coverslips for 45 min. Coverslips were then washed three times with 0.1% Triton-X100 in PBS. Coverslips were counterstained with DAPI for 5 min, washed, and mounted onto glass slides using Vectashield (Vector Laboratories) or Prolong Diamond (Invitrogen) mounting media.

A++ 11 F B lymphocytes cells were processed for immunofluorescence as previously described¹⁰¹. Briefly, 72 h after electroporation with Cy-3-labelled siRNAs, cells were centrifuged at 250 × g for 5 min and resuspended in 25 °C PBS after which cells were slowly pipetted onto a glass coverslip in a 12 well plate and allowed to sediment at room temperature for 30 min. PBS was carefully aspirated after inclining the plate. Coverslips were fixed with 1 mL 4% PFA diluted in PBS for 10 min at room temperature and post extracted with 0.5% Triton X-100 for 1 min at 37 °C. Primary and secondary antibody staining was then performed similar to standard immunofluorescence, as described above.

For quantifying cell membrane fluorescence intensity, cells expressing 4xDmrD-Halo-GFP-α5^TMCT and mApple-Mannosidase II were fixed with 4% PFA diluted in PBS for 10 min at room temperature. To preserve plasma membrane integrity, coverslips were not extracted with any detergent. Coverslips were blocked with 0.5% BSA diluted in 1X PBS for 1 h at room temperature. HaloTag dye JFX635i (gift from Lavis lab) was diluted in 1X PBS and incubated on coverslips for 45 min. All washes were performed with 1X PBS.

For immunoblotting, cells were seeded in 6 well plates at 80% confluency and allowed to grow for 24 h unless otherwise indicated. Cells were rinsed with PBS, trypsinized, transferred to ice-chilled 1.5 mL tubes, and spun down at 7000 × g for 5 min. Cell pellet was washed once with ice-cold PBS and flash frozen. Frozen pellet was resuspended in 150 µL RIPA lysis buffer (50 mM Tris/Cl pH 7.4, 150 mM NaCl, 1 mM EDTA, 0.1% SDS, 1% Triton X-100, 0.5 % deoxycholate) supplemented with 1 mM PMSF and cOmplete protease inhibitor cocktail (Roche) and incubated on ice for 30 min, vortexing every 10 min. Cell lysate was centrifuged at 17,000 × g for 10 min at 4 °C, and the protein concentration quantified using DC protein assay kit (BIO-RAD Laboratories). Proteins were separated on SDS-Page gels, transferred to nitrocellulose membrane, blocked with 5% milk in TBS-tween, and probed with primary and secondary antibodies. Primary antibodies were diluted in 2% BSA + 0.02% sodium azide in 1X PBS and incubated overnight at 4 °C. Secondary antibodies (IRDye, LI-COR) were diluted 1:10,000 in 5% milk in TBS-tween and incubated for 45 min at room temperature. Membranes were imaged on the LI-COR Odyssey imager and analyzed using LI-COR Image Studio.

Antibodies

Primary antibodies used in this study: rabbit anti-Mad1 (affinity purified antibody produced against amino acids 333–617⁶⁶, used at 1:3000 for immunoblotting or 1:1000 for IF); rabbit anti-Mad1 (HPA003635, Atlas Antibodies, 1:400 for IF), rabbit anti-Mad1 (GTX105079, GeneTex, 1:200 for IF), rabbit anti-Mad1 (GTX109519, GeneTex, 1:500 for IF), mouse anti-Mad1 (Gift from Andrea Musacchio, 1:10 for IF), mouse anti-Flag (F1804, Sigma-Aldrich, 1:800 for IF), rabbit anti-Flag (F7425, Sigma, 1:800 for IF), mouse anti-GM130 (610822, BD Biosciences, 1:450 for IF), mouse anti-vinculin (V450S, Sigma, 1:200), rabbit anti-ITGA5 (AB150361, ABCAM, 1:500 for immunoblotting or 1:150 for IF), mouse anti-ITGA5 (SNAKA51, Novus Bio, 1:125 for IF), mouse anti-tubulin (12G10, Developmental Studies Hybridoma Bank, 1:2000 for immunoblotting or 1:500 for IF), rat anti-alpha tubulin (MCA77G, clone YL1/2, BioRad, 1:700), Phalloidin (ActinGreen 488 ReadyProbes, Thermo Fisher, used according to manufacturer’s protocol), mouse anti-Plk1 (sc17783, Santa Cruz, 1:400 for IF).

4xDmrD-Halo-GFP-α5^TMCT trafficking assay

HeLa cells stably expressing 4xDmrD-Halo-GFP-α5^TMCT at the AAVS1 safe harbor site were treated with doxycycline overnight to induce expression and aggregation in the ER. Addition of D/D solubilizer (DDS, Takara Bio) initiates trafficking by allowing the construct to properly fold and be secreted to the Golgi and then the plasma membrane. D/D solubilizer was added at a concentration of 1 µM to induce trafficking. Cells were fixed at indicated time points. Images were quantified in Fiji using the Golgi marker, GM130, to localize the Golgi and recording the Raw integrated density signal (RawIntDen) values for the cargo in the Golgi at indicated timepoints.

Cell surface quantification was done in Fiji. Briefly, frames in focus were summed, and the same LUTS set for all images. Background intensity was subtracted using a rolling ball radius of 100 pixels. An ROI of the same width was drawn around the plasma membrane using JFX635i, a HaloTag cell-impermeable dye to label 4xDmrD-Halo-GFP-α5^TMCT on the cell surface. Cell surface 4xDmrD-Halo-GFP-α5^TMCT was quantified by computing the ratio of cell surface HaloTag signal to the internal GFP signal.

For live-cell trafficking, WT and stable Mad1-KD²⁰ HeLa cells expressing both 4xDmrD-Halo-GFP-α5^TMCT and mApple-Mannosidase II were used.

Midbody quantification

Midbody quantification of 4xDmrD-Halo-GFP-α5^TMCT, endogenous Mad1, and α5 integrin was performed in Fiji (version 1.54p) using a local background correction, as described previously¹⁰². Briefly, raw (not deconvolved) images were loaded into Fiji, and in-frame Z-planes were summed using the Sum of slices feature. The same LUTs were used for all conditions. A small bounding box with a defined area was drawn around the midbody followed by a larger box 1.5x the size of the smaller box. Midbody signal intensity was determined by subtracting the integrated background intensity from the Raw integrated density (RawIntDen) signal in the smaller box.

Quantification of Golgi area and volume

Quantification of Golgi area and volume was performed in Nikon Elements imaging software. Golgi area was defined by GM130 staining using the automated area feature in Nikon Elements. For Golgi volume, in-frame Z sections were defined based on GM130 staining, and subsequently multiplied by its length and width, calculated using Nikon Elements.

Statistical analysis

All experiments were performed using three independent biological replicates quantified in a blinded manner during imaging and analysis. For comparison between two groups, an unpaired two-tailed Students t-test was performed. For multiple comparison data, Tukey’s post hoc test was used without any adjustments. P values < 0.05 were considered statistically significant. All statistical analysis was performed in GraphPad Prism software (version 10.1.1).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

D.K.S., A.A., and B.A.W. conceived and designed experiments. D.K.S. performed experiments, collected and analyzed data. G.G cloned plasmids and generated cell lines. D.K.S. and B.A.W. wrote the manuscript. All authors reviewed the manuscript.

Peer review

Peer review information

Nature Communications thanks Shigehiko Yumura and the other anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

Source data are provided with this paper. The authors declare that all the data supporting the findings of this study are available within the paper and the supplementary materials. The source data for Figs. 1–5 Supplementary Figs. 1–4 and 6–8 are provided in the Source Data file. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-70928-9.