ARTICLES

Endocytosis by Numb breaks Notch symmetry at cytokinesis Lydie Couturier1,2 , Nicolas Vodovar1,2,3 and François Schweisguth1,2,4 Cell-fate diversity can be generated by the unequal segregation of the Notch regulator Numb at mitosis in both vertebrates and invertebrates. Whereas the mechanisms underlying unequal inheritance of Numb are understood, how Numb antagonizes Notch has remained unsolved. Live imaging of Notch in sensory organ precursor cells revealed that nuclear Notch is detected at cytokinesis in the daughter cell that does not inherit Numb. Numb and Sanpodo act together to regulate Notch trafficking and establish directional Notch signalling at cytokinesis. We propose that unequal segregation of Numb results in increased endocytosis in one daughter cell, hence asymmetry of Notch at the cytokinetic furrow, directional signalling and binary fate choice. A conserved mechanism to generate daughter cells with distinct fates is through the asymmetric inheritance of regulatory proteins leading to differential gene regulation in daughter cells. The protein Numb was the first cell-fate determinant shown to be unequally partitioned at mitosis and to regulate cell-fate decisions in Drosophila sensory organ lineages1 . Sensory organ precursor cells (SOPs) divide asymmetrically along the Drosophila body axis to produce an anterior pIIb cell and a posterior pIIa cell2 . The pIIa/pIIb binary fate choice is regulated by Notch, a transmembrane receptor with an extracellular domain binding its ligand Delta (Dl) and an intracellular domain acting as a membrane-tethered transcriptional coactivator3 . Notch is activated in pIIa and inhibited in pIIb (Fig. 1a). Inhibition of Notch in pIIb requires Numb1 . Numb localizes at the anterior cortex of dividing SOPs and is specifically inherited by the anterior SOP daughter. Whereas the molecular mechanisms underlying asymmetric localization of Numb are largely understood4 , it is not clear how Numb inhibits Notch in pIIb. As Notch inhibition by Numb is evolutionarily conserved5–12 and may underlie adult-onset diseases in humans8,13,14 , understanding the mode of action of Numb is important. Genetic and molecular studies have indicated that Numb inhibits Notch through an endocytic process. Numb directly interacts with the α-adaptin subunit of the adaptor protein complex 2 (AP-2; ref. 15), and α-adaptin mutations disrupting this interaction phenocopied numb7 . These data indicated that Numb might regulate the endocytosis of Notch in pIIb, hence preventing Notch from interacting with its ligand. However, no Numb-dependent change in Notch localization has ever been observed in vivo, raising doubts about this model. Alternatively, Numb might indirectly inhibit Notch by promoting the endocytosis of a positive regulator

of Notch. In support of this view, the four-pass transmembrane protein Sanpodo (Spdo) is required for Notch signalling in the context of Numb-mediated binary fate decisions, directly interacts with Numb and is endocytosed by Numb16–21 . Nevertheless, how Spdo regulates Notch is not known21 and whether Numb acts by inhibiting Spdo is controversial19 . Thus, how Numb inhibits Notch is still a matter of debate. RESULTS Notch transiently localizes at the apical pIIa/pIIb interface To study the in vivo dynamics of Notch localization and signalling in Drosophila, we generated a functional green fluorescent protein (GFP)-tagged Notch. First, four Notch versions with an intracellular yellow fluorescent protein (YFP) tag were designed on the basis of primary-sequence divergence within the Drosophila genus and reintroduced into the fly genome as bacterial artificial chromosome (BAC) transgenes (Fig. 1b). Two transgenes, NiYFP4 and NiYFP5, were found to fully rescue a Notch null mutation (N 55e11 ), whereas one, NiYFP3, only partially rescued viability and another, NiYFP2, was dominant negative. We then generated a GFP-tagged version, NiGFP (Notch intracellular fusion with GFP), similar to NiYFP5. The BAC-encoded NiGFP transgene fully rescued the N 55e11 mutation (Fig. 1b). N 55e11 NiGFP flies were fully viable and fertile with no detectable phenotype, indicating that NiGFP is fully functional. Furthermore, one copy of NiGFP generated a dominant Confluens phenotype (as did duplications of wild-type Notch22 ) and N 55e11 females with one copy of NiGFP showed the haploinsufficient Notch wing phenotype, indicating that NiGFP is expressed at physiological levels.

1

Institut Pasteur, Developmental Biology Department, F-75015 Paris, France. 2 CNRS, URA2578, F-75015 Paris, France. 3 Present address: Institut Pasteur, Virology Department, F-75015 Paris, France. 4 Correspondence should be addressed to F.S. (e-mail: [email protected]) Received 26 April 2011; accepted 12 December 2011; published online 22 January 2012; DOI: 10.1038/ncb2419

NATURE CELL BIOLOGY VOLUME 14 | NUMBER 2 | FEBRUARY 2012 © 2012 Macmillan Publishers Limited. All rights reserved.

131

ARTICLES NECD

21 8 22 5 5 23 9 88

b

Numb

pIIb

NICD 18 49

a

Rescue

Notch off

Wild type +

No Numb

NiYFP2 NiYFP3 NiYFP4 NiYFP5

– + – + +

NiGFP

+

SOP pIIa

Notch on

c SOP

SOP

d pIIb

‘pIIb’

SOP

neur

t=0

Epidermal

e

t = 20

f

g Epidermal

pIIb

pIIb

pIIa

pIIb

pIIa

t = 60

t = 60

h

t = 15

pIIa

t = 40

i pIIb

pIIb

pIIa

pIIb

pIIa

pIIa

t = 10

t = 20

Figure 1 Notch localization at the pIIa/pIIb interface. (a) Asymmetric SOP division. (b) Functional rescue by tagged Notch. The positions of the YFP (yellow) and GFP (green) tags are indicated below a domain representation of Notch (white, epidermal growth factor repeats; brown, Lin/Notch repeats; black, transmembrane; blue, RAM domain; red, ankyrin repeats). (c) Live imaging of SOPs (H2B–RFP, red) showing that SOPs had low NiGFP (green) levels at the cortex (first two panels; surface views) and in the nucleus (last two panels) relative to epidermal cells. (d) Live analysis of NiGFP in neur mutant cells (loss of nuclear GFP, green). n = 3 pupae. High-magnification views (surface views in second and fourth panels and corresponding basal views in third and fifth panels) of the region boxed in the first panel are shown at t = 0 (second and third panels; chromatin decondensation in nuclei indicated by red dots) and t = 20 (fourth and fifth panels). NiGFP specifically accumulated at the apical interface at t = 20 (white arrow in fourth panel). neur mutant epidermal cells exhibit high NiGFP levels whereas neur mutant

SOPs (H2B–RFP, red) have very low NiGFP levels (neur+ cells are indicated by white dots in second and fourth panels). (e,f) NiGFP in pIIa/pIIb (e) and epidermal (f) cell pairs at t = 60 min after mitosis. (g) NiGFP (green, surface view) transiently accumulated at the apical pIIa/pIIb interface (apical surface of pIIa and pIIb indicated by green arrows). NiGFP was detected at t = 15 but not at t = 40. The pIIa and pIIb nuclei (H2B–RFP, red; located below the surface) are also shown (red arrows). (h) SOPs (light green) have low Notch relative to epidermal cells (green). Notch (pink) may localize at the apical pIIa/pIIb interface by diffusion along the cortex (arrows above the cell surface). Notch might also be targeted to the cytokinetic furrow through vesicular trafficking (arrows inside the cell). (i) Intracellular dots containing NiGFP were detected in pIIb, but not pIIa, at t = 20. A PH–RFP membrane marker (red) was used to outline the pIIa and pIIb cells. In this and all other figures, anterior is to the left (pIIb is anterior, pIIa is posterior); time (t ) is in minutes (min); scale bars are 5 µm.

N 55e11 NiGFP flies, that is, devoid of endogenous Notch but expressing NiGFP, were used throughout this study. We first analysed the distribution of NiGFP in the pupal notum, a single-layered neuroepithelium composed of SOPs and epidermal cells.

NiGFP localized in a similar way to endogenous Notch at the apical cortex and within endocytic vesicles (Fig. 1c). SOPs exhibited a lower level of NiGFP than epidermal cells (Fig. 1c), as reported earlier for Notch23 . Interestingly, a weak nuclear NiGFP signal was detected in

132

NATURE CELL BIOLOGY VOLUME 14 | NUMBER 2 | FEBRUARY 2012 © 2012 Macmillan Publishers Limited. All rights reserved.

ARTICLES a

pIIa pIIb

Fixed, anti-GFP

b

c

pIIb

pIIa

pIIa

pIIb

t = 60

Fixed, anti-NICD

e

GFP (a.u.)

1.6

pIIa

1.4 1.2

pIIb

1.0 0.8

0

5

10

15

20

25

1.4

GFP (ratio)

d

30

Live, GFP

1.2 pIIa/pIIb

1.0 0.8 0.6

0

5

10

Time (min)

15 20 Time (min)

25

30

Figure 2 Live analysis of Notch signalling. (a) NiGFP (green, anti-GFP) was detected in the nucleus of pIIa but not of pIIb (blue, 4,6-diamidino-2-phenylindole (DAPI); red, Spdo). (b) Nuclear Notch (green, anti-NICD) was detected in the nucleus of pIIa but not of pIIb (blue, DAPI; red, senseless, sens) in wild-type pupae. (c) Snapshot of a time-lapse sequence showing low nuclear NiGFP (green) levels in pIIa (red, H2B–RFP) at t = 60. Images are representative examples from more than 4 experiments

(a–c). (d) Nuclear accumulation of NiGFP increased over time after mitosis and was detected in pIIa within 10 min of furrow ingression (monitored using PH–RFP). No significant nuclear NiGFP was detected in pIIb. Each dot represents the mean value of n ≥ 12 nuclei. (e) Ratiometric analysis of nuclear NiGFP in pIIa and pIIb. Nuclear NiGFP signals in pIIa and pIIb were plotted as ratio values over time after chromatin decondensation (monitored using H2B–RFP). Each dot represents the mean value of n ≥ 3 pIIa/pIIb pair.

epidermal cells but not SOPs (Fig. 1c, fourth panel), consistent with NiGFP being processed by γ-secretase3 in epidermal cells (see below). To determine whether the cortical NiGFP signal detected in SOPs originated from SOPs or from neighbouring epidermal cells, clones with clusters of SOPs were studied. Clones of neuralized (neur) mutant cells were generated in N 55e11 NiGFP pupae (see Supplementary Table S1 for detailed genotypes). We observed that neur mutant SOPs had very low levels of NiGFP when compared with epidermal cells (Fig. 1d; see below for similar results with Dl), clearly indicating that SOPs have greatly reduced levels of Notch. Consistent with this, low NiGFP levels were detected in dividing SOPs at the pIIa/pIIb interface (Fig. 1e; compare with epidermal cells in Fig. 1f). NiGFP was best detected at the pIIa/pIIb interface during cytokinesis, defined here as the time period extending from furrow ingression to 10 min afterwards (Fig. 1g and Supplementary Movie S1; chromosome decondensation, monitored using histone H2B–red fluorescent protein (RFP), precisely correlated with the ingression of the cytokinetic furrow and was taken as t = 0; Supplementary Movie S2). During cytokinesis, NiGFP predominantly localized at the apical interface, that is within 1 µm below the cell surface. By contrast, NiGFP was barely detectable along the basal cytokinetic furrow, that is 3–5 µm below the surface. In neur mutant clones, NiGFP specifically accumulated at the apical interface between daughter cells (Fig. 1d, fourth panel; Supplementary Fig. S1). This localization of Notch at cytokinesis may involve directed transport

along the apical cortex and/or vesicular trafficking towards the newly formed plasma membrane (Fig. 1h). This localization of Notch at the pIIa/pIIb interface may be important to restrict signalling to sister cells. Live monitoring of nuclear Notch indicates that directional signalling is established at cytokinesis Activation of Notch by Dl leads to the processing of Notch by the γ-secretase and the nuclear translocation of the Notch intracellular domain (NICD; ref. 3). In Drosophila, NICD has only been indirectly detected in the nucleus using a nuclear activity assay24,25 . Here, we were able to detect both NiGFP and endogenous NICD in the nucleus of pIIa in fixed and live specimens (Fig. 2a–c). Nuclear NiGFP was also seen in epidermal cells but not in SOPs (Fig. 1c, fourth panel). Thus, nuclear NiGFP correlated with receptor activation. We therefore used nuclear NiGFP and time-lapse imaging to follow the dynamics of Notch activity in living flies. A significant increase in nuclear NiGFP was detected in pIIa, but not pIIb, as early as 10 min after cytokinesis (Fig. 2d). Consistent with this, the ratio of nuclear NiGFP in pIIa versus pIIb gradually increased from telophase onwards (Fig. 2e). These data indicated that NICD accumulates in the nucleus of pIIa but not pIIb at cytokinesis. We conclude that Notch is activated soon after mitosis and that directionality in Notch signalling is established at cytokinesis. These results also indicate that it is now possible to monitor the dynamics of Notch signalling in vivo.

NATURE CELL BIOLOGY VOLUME 14 | NUMBER 2 | FEBRUARY 2012 © 2012 Macmillan Publishers Limited. All rights reserved.

133

ARTICLES a

Wild type

z=0

b

numbRNAi

c

spdoRNAi

d

numbRNAi spdoRNAi

e

numb-/-

f

spdo-/-

t=5

z=0

z = –1

z = –4.5

Figure 3 Regulation of Notch by Numb and Spdo at cytokinesis. (a–f) Live analysis of NiGFP (green) in wild-type (a), numb RNAi (b), spdo RNAi (c) and spdo RNAi numb RNAi (d) pupae and in negatively marked clones of numb 15 (e) and spdo G 104 (f) mutant cells at t = 5 min after chromatin decondensation (red, H2B–RFP). Cross-section as well as surface (z = 0 µm), subapical (z = −1 µm) and basal (z = −4.5 µm)

views are shown. Silencing and loss of numb led to a transient localization of NiGFP at the basal cytokinetic furrow (arrows in b,e). Silencing and loss of spdo led to a transient apical accumulation of NiGFP at the pIIa/pIIb interface (arrows in c,f). Apical accumulation of NiGFP was also observed after double numb spdo silencing (arrow in d).

Numb is required to prevent Notch localization along the pIIa/pIIb interface at cytokinesis We next studied the role of Numb in establishing directionality at cytokinesis. RNA interference (RNAi)-mediated silencing of numb resulted in a pIIb-to-pIIa transformation with both SOP daughter cells accumulating similar levels of nuclear NiGFP (Supplementary Fig. S2), confirming that Numb inhibits Notch and indicating that silencing of numb was effective. Analysis of NiGFP distribution at cytokinesis, that is 5 min after chromatin decondensation, revealed that the subcellular distribution of NiGFP was altered on numb silencing. Whereas NiGFP was detected at the level of the apical pIIa/pIIb interface in wild-type controls (Fig. 3a, 100%, n = 24, and Supplementary Movie S1), NiGFP was detected at the pIIa/pIIb interface both apically and along the basal cytokinetic furrow (3–5 µm below the surface) in 80% of the SOPs silenced for numb (Fig. 3b; n = 35). NiGFP was detected along the basal furrow as soon as the furrow completed its ingression (Fig. 4a). This localization of NiGFP at the basal cytokinetic interface was transient (Supplementary Movie S3). NiGFP was also detected along the basal cytokinetic furrow in numb mutant clones at cytokinesis (Fig. 3e, 75%, n = 16). By contrast, silencing of numb had no effect on the distribution of E-Cadherin–GFP (not shown). Thus, we conclude that Numb is required to prevent the transient accumulation of Notch along the basal pIIa/pIIb interface at cytokinesis. This is in vivo evidence that Numb regulates the localization of Notch during asymmetric cell division.

interface was examined at cytokinesis (timing in fixed samples was estimated on the basis of condensed chromatin and distance separating the pIIa/pIIb nuclei; Supplementary Movie S1). Whereas accumulation of endogenous Notch was not detected in wild-type controls (Fig. 4b, 0%, n = 13), Notch-containing dots were observed 3–5 µm below the apical surface along the pIIa/pIIb interface in shits pupae at cytokinesis (Fig. 4c, 70%, n = 17), suggesting that Notch is normally endocytosed from the basal pIIa/pIIb interface. To more precisely determine when Notch is endocytosed relative to cytokinesis, the dynamics of NiGFP localization was studied in living shits NiGFP pupae. Endocytosis was conditionally blocked by shifting the temperature to 32 ◦ C at telophase (as determined using H2B–RFP). In wild-type control pupae, NiGFP localization remained unchanged at restrictive temperature (Fig. 4d). By contrast, marked changes were observed in shits pupae. First, NiGFP concentrated within 5 min into distinct foci at the apical cortex (Fig. 4e, third panel, and Supplementary Fig. S3). These foci might correspond to cortical sites of endocytosis where NiGFP accumulates on inhibition of vesicle release from the plasma membrane. Second, apical vesicles containing NiGFP rapidly disappeared (Supplementary Fig. S3), indicating that this pool of endocytic vesicles was not replenished on dynamin inactivation. Third, dots of NiGFP accumulated along the pIIa/pIIb interface 3–5 µm below the apical surface within 5 min at 32 ◦ C (Fig. 4e, fourth panel, 100%, n = 5, and Supplementary Fig. S3), indicating that inhibition of dynamin led to the rapid accumulation of Notch along the basal pIIa/pIIb interface. We conclude that Notch is transiently present at the cell surface along the basal pIIa/pIIb interface at cytokinesis but that it is rapidly removed by endocytosis. Thus, rapid endocytosis of Notch may explain why NiGFP was barely detected along the basal cytokinetic furrow in wild-type pupae (Fig. 2a).

Dynamin-dependent endocytosis of Notch at cytokinesis As Numb acts as an AP-2 adaptor in pIIb (refs 7,16–20), we hypothesized that Notch is removed from the pIIb plasma membrane by an endocytic process. To test this hypothesis, endocytosis was conditionally blocked using shits , a thermosensitive allele of the gene shibire, which encodes the fly dynamin. Staged pupae were heat-pulsed for 20 min at restrictive temperature (32 ◦ C) and then rapidly dissected, fixed and stained for Notch. Notch distribution along the pIIa/pIIb

134

Numb-dependent endocytosis of Notch in pIIb To test whether the endocytosis of Notch at cytokinesis is pIIb specific, endocytosis was monitored in living tissues using an antibody uptake

NATURE CELL BIOLOGY VOLUME 14 | NUMBER 2 | FEBRUARY 2012 © 2012 Macmillan Publishers Limited. All rights reserved.

ARTICLES a

numbRNAi

t=0

t=2

b

t=6

t = 10

shits

32 °C

c

Wild type

32 °C

Wild type

t = –1

22 °C

t = –1

22 °C

d

t = 15

32 °C

t = 15

32 °C

e

shits

f

g

Wild type

h

numbRNAi

Figure 4 Notch is endocytosed from the pIIa/pIIb interface. (a) Time-lapse analysis of NiGFP (green; green channel shown in bottom panels) in numb RNAi pupae (n = 4) showing that the onset of NiGFP localization along the basal cytokinetic furrow was at ingression (red, PH–RFP). (b,c) Endogenous Notch (green, anti-NICD/NECD) accumulated into dots at the pIIa/pIIb interface (n ≥ 12), 3–5 µm below the surface at cytokinesis (based on condensed chromatin and distance separating the pIIa/pIIb nuclei; red, sens) on inhibition of dynamin in shi ts pupae at 32 ◦ C (arrow in c, right panel). No accumulation was detected in wild-type pupae (b, right panel). (d,e) Time-lapse analysis of NiGFP (green) in wild-type (d; n = 3) and shi ts (e; n = 5) pupae. Surface (first and third panels) and basal (-4 µm; second and fourth panels) views are shown before temperature shift (end of

spdoRNAi

telophase at t = −1, as monitored using H2B–RFP, red; temperature shift was at t = 0) and at t = 15 min. NiGFP was specifically detected at the basal pIIa/pIIb interface on conditional inhibition of dynamin (arrow in e, fourth panel). (f–h) Endocytosis of endogenous Notch was monitored using an antibody uptake assay (green, i-Notch). Lower panels show cross-section views. At cytokinesis (distant nuclei with condensed chromatin; red, sens), more i-Notch dots were observed in pIIb and no staining was seen at the pIIa/pIIb interface in wild-type pupae (vertical arrows in f). By contrast, fewer i-Notch dots were detected in numb RNAi (g) and spdo RNAi pupae (h). Anti-NECD antibodies were detected at the basal pIIa/pIIb interface on numb silencing (arrows in g) and at the apical interface on spdo silencing (h; horizontal arrows in cross-section views).

NATURE CELL BIOLOGY VOLUME 14 | NUMBER 2 | FEBRUARY 2012 © 2012 Macmillan Publishers Limited. All rights reserved.

135

ARTICLES a

b

c Epidermal SOP

SOP SOP Wild type

spdoRNAi

Delta

d

Epidermal

SOP Delta spdo

e

ap > spdo

f

Figure 5 Endocytosis of Notch by Spdo. (a) Wild-type SOPs (arrow) had low NiGFP relative to epidermal cells. (b) Silencing of spdo abolished this difference. (c,d) Live analysis of NiGFP (green) in Dl (c) and Dl spdo double-mutant cells (d). Mutant cells were marked by the loss of nlsGFP (green) and SOPs by H2B–RFP (red). Wild-type epidermal cells (white dots in first panels) and SOPs (pink dots) are indicated in c,d. Dl mutant SOPs had very low NiGFP levels relative to mutant epidermal cells (c; n = 4 pupae). In contrast, Dl spdo double-mutant SOPs and epidermal cells had similar levels of NiGFP at the apical cortex (d, first two panels; n = 3 pupae). Of note, higher nuclear NiGFP was observed in mutant cells located along the clone border (thin outline) relative to mutant cells located away from the

border (thick outline). This difference is consistent with the non-autonomy of Dl and suggests that this nuclear NiGFP signal is Dl dependent. (e) Spdo (green) expression reduced Notch apical accumulation (red, NECD) but had no effect on Bazooka (blue, fourth panel) in ap-GAL4 UAS-spdo wing discs (first panel, low-magnification view showing the ectopic expression of Spdo in dorsal cells; other panels, high-magnification views at the dorsoventral boundary). (f) Low Notch levels at the apical cortex (red, NECD, first panel) correlated with Notch accumulation into Sara-positive endosomes (green, Sara, last two panels; z = −1.5 µm in last three panels) in dorsal cell ap-GAL4 UAS-spdo discs. Images are representative examples from ≥ 5 experiments (e,f).

assay. Briefly, dissected pupae were incubated with an antibody recognizing an extracellular epitope of Notch, then rapidly washed and fixed. Internalized antibodies, corresponding to internalized Notch (i-Notch), were detected using secondary antibodies. Analysis of newly born pIIa/pIIb cells, identified based on chromatin condensation and the distance separating the pIIa/pIIb nuclei, showed that i-Notch was mostly detected in pIIb at cytokinesis (Fig. 4f; see also Fig. 6a–c below): pIIb had 1.9 ± 1.3 dots containing i-Notch (n = 15) whereas pIIa had 0.5 ± 0.7 dots (P < 0.001, t -test). As i-Notch was not detected at anaphase (not shown), no significant endocytosis of Notch seemed to take place during mitosis. We therefore conclude that the rate of Notch internalization and/or sorting towards detectable endosomes is higher in pIIb than in pIIa at cytokinesis. Consistent with this, intracellular dots

of NiGFP were detected in pIIb, but not pIIa, 15–45 min after division (Fig. 1i). We next tested whether this endocytosis of Notch in pIIb was dependent on numb. First, similar i-Notch levels were detected in the anterior ‘pIIa-like’ (0.7 ± 0.8 dots, n = 21) and posterior pIIa (0.9 ± 1.0 dots) on numb silencing. Second, anti-Notch extracellular domain (NECD) antibodies were detected along the basal cytokinetic furrow (Fig. 4g, arrow; 76%, n = 21). This observation was consistent with NiGFP accumulating along the basal interface at cytokinesis on silencing of numb (Fig. 3b,e). It further suggests that Notch remained at the cell surface and was less efficiently internalized in the anterior ‘pIIa-like’. Together, our data indicate that Numb regulates the internalization and/or endosomal sorting of Notch in pIIb such that i-Notch became more rapidly detectable in pIIb endosomes. We

136

NATURE CELL BIOLOGY VOLUME 14 | NUMBER 2 | FEBRUARY 2012 © 2012 Macmillan Publishers Limited. All rights reserved.

ARTICLES a

b

i-Spdo

SOP

pIIb

i-Notch

c

pIIa

i-Spdo

i-Notch

d

pIIb

pIIa

e

Wild type

pIIb

i-Notch

i-Spdo

f

Wild type

pIIa

PLA spdo

g

PLA

Spdo

Dlg

numb

h

epi SOP

pIIb

pIIa

pIIb

'pIIb'

'pIIa'

pIIa

Figure 6 Endocytosis of Spdo–Notch complexes. (a–c) Co-endocytosis of Notch with Spdo–CherryL2 was monitored using a double-antibody uptake assay (green, i-Spdo; red, i-Notch; blue, sens). i-Spdo and i-Notch co-localized in SOPs before mitosis (a) and during cytokinesis (b; timing based on distant nuclei with condensed chromatin) as well as after cytokinesis (c). At cytokinesis, i-Notch (arrows in b, right panel) co-localized with i-Spdo in pIIb. At later stages, co-internalization was observed in both pIIa and pIIb (c). For a–c, images are representative examples from 2 experiments (SOPs, n = 14; early pIIa/pIIb cells, n = 16). (d) Spdo–NICD epitope pairs (red, PLA dots) were observed in pIIb and pIIa and at the pIIa/pIIb interface (green, Spdo; white, DAPI; blue, Dlg), indicating that Spdo physically interacts with Notch. Most PLA dots co-localized with Spdo (second panel), at the pIIa/pIIb interface (arrowheads in third panel), in endosomes (pIIb, horizontal arrows; pIIa, vertical arrow pointing down, third

panel) or at the plasma membrane (pIIa, vertical arrows pointing up, third panel). (e–h) Models. Before division (e), Notch (pink) is endocytosed in a Spdo-regulated manner in SOPs (blue arrow). At cytokinesis (f), cortical Notch is transported along the cortex to localize at the apical pIIa/pIIb interface (black arrows at the cell surface) whereas i-Notch is targeted to the cytokinetic furrow (black arrows in pIIa/pIIb). Numb regulates the endocytosis of Notch in pIIb (red arrow), thereby creating an asymmetry in the Notch distribution along the basal pIIa/pIIb interface. In the absence of spdo (g), Notch accumulates at the apical cortex before mitosis and at the apical pIIa/pIIb interface at mitosis. Lower levels of endocytosed Notch might correlate with lower levels of Notch along the basal pIIa/pIIb interface. In the absence of numb (h), Notch endocytosis is not upregulated in pIIb. This results in symmetric distribution of Notch along the basal pIIa/pIIb interface.

therefore propose that Numb promotes the removal of Notch from the pIIb side of the pIIa/pIIb interface.

spdo had no effect on E-Cadherin–GFP levels or distribution (data not shown). We therefore conclude that spdo acts before mitosis to downregulate Notch levels in SOPs. We next tested whether accumulation of NiGFP along the basal furrow on loss of numb (Fig. 3b,e) required the activity of spdo. The combined silencing of spdo and numb (Supplementary Fig. S2) resulted in the accumulation of NiGFP at the apical pIIa/pIIb interface (Fig. 3d) as in spdoRNAi pupae (Fig. 3c). NiGFP was not detected basally as seen on numb silencing (Fig. 3b). Thus, the localization of NiGFP at the basal interface seen on loss of numb required the activity of spdo. We suggest that Spdo acts before Numb in the regulation of Notch, possibly by increasing the endosomal pool of Notch before SOP division so that a significant amount of Notch is delivered to the newly formed membrane at cytokinesis.

Sanpodo is required to downregulate Notch levels before mitosis This proposed role of Numb differs from the one proposed earlier whereby Numb inhibits Notch indirectly by regulating the endocytosis of Spdo16–21 . In light of our data, we wondered whether Spdo acts together with Numb to regulate the trafficking of Notch. The silencing of spdo led to pIIa-to-pIIb transformations with low levels of nuclear NiGFP in SOP progeny cells (Supplementary Fig. S2), confirming that spdo is required for Notch signalling20 . We next examined the distribution of Notch in spdoRNAi SOPs. Whereas wild-type SOPs had low levels of NiGFP relative to epidermal cells, SOPs and epidermal cells exhibited similar levels of NiGFP on silencing of spdo (Fig. 5a,b), indicating that Spdo regulates Notch levels in SOPs. Using Dl mutant clones to create clusters of SOPs, we showed that Dl spdo double-mutant SOPs had high NiGFP levels, similar to those seen in epidermal cells (Fig. 5d) whereas Dl mutant SOPs had low levels (Fig. 5c). This indicated that spdo is required for the downregulation of NiGFP in SOPs. Consistent with spdo mutant SOPs entering mitosis with high NiGFP levels, NiGFP accumulated at the apical ‘pIIb-like’/pIIb interface in spdoRNAi SOPs (100%; n = 13; Fig. 3c and Supplementary Movie S4) as well as in spdo mutant cells (Fig. 3f). In contrast, the silencing of

Sanpodo regulates the endocytosis of Notch As Spdo interacts with Notch and co-localizes with Notch in endocytic vesicles18–20 , Spdo may downregulate Notch by promoting its endocytosis. Consistent with this, ectopic expression of Spdo in wing imaginal discs decreased cortical Notch levels and led to the accumulation of Notch into Rab5- and Sara-positive endosomes (Fig. 5e,f; data not shown). By contrast, ectopic Spdo had no effect on the distribution of Bazooka (Fig. 5e), Dl, E-Cadherin and

NATURE CELL BIOLOGY VOLUME 14 | NUMBER 2 | FEBRUARY 2012 © 2012 Macmillan Publishers Limited. All rights reserved.

137

ARTICLES Discs-large (not shown). We conclude that Spdo can promote the endocytosis of Notch. As reported earlier18,21 , ectopic Spdo also decreased Notch activity (Supplementary Fig. S4). We therefore suggest that endocytosis of Notch by ectopic Spdo decreased receptor levels at the cell surface and thereby inhibited Notch activity. As ectopic Spdo inhibited Notch in numb mutant cells (Supplementary Fig. S4), the effect of Spdo on Notch endocytosis may not strictly depend on Numb. Consistent with this, deletion of the Numb-binding motif of Spdo19 did not prevent Spdo from reducing cortical Notch levels (Supplementary Fig. S5). However, overexpression of Numb enhanced the inhibition of Notch by ectopic Spdo (Supplementary Fig. S6), indicating that Spdo can cooperate with Numb to inhibit Notch. Together, our data indicate that Spdo can regulate the endocytosis of Notch both with and without Numb. We next tested the role of Spdo in Notch endocytosis using the anti-NECD antibody uptake assay. At cytokinesis, i-Notch levels were only slightly reduced in spdoRNAi pupae (anterior pIIb, 0.7 ± 0.8; posterior ‘pIIb-like’, 0.3 ± 0.4, n = 19) relative to control pupae (pIIb, 1.9 ± 1.3; pIIa, 0.5 ± 0.7, n = 15). This suggests that Notch can be internalized in an spdo-independent manner in sensory cells, as it is in most tissues that do not express Spdo. Also, the silencing of spdo led to the accumulation of anti-NECD antibodies at the apical pIIa/pIIb interface (Fig. 4h; 68%, n = 19), suggesting that Notch remained at the cell surface. This observation was consistent with Notch accumulating at the apical pIIa/pIIb interface on loss of spdo activity (Fig. 3c,f). Together, these data suggest that Spdo regulates the endocytosis of Notch and that Spdo-independent mechanisms also contribute to the endocytosis of Notch in SOPs. As Spdo interacts with Notch20 and co-localizes with Notch in Rab5 endosomes in pIIb (refs 18,19), we wondered whether Notch is endocytosed, at least in part, as Spdo–Notch complexes. To test this hypothesis, we carried out a double-antibody uptake assay with mouse anti-NECD to reveal i-Notch and rabbit anti-RFP to detect internalized Spdo–CherryL2 (i-Spdo; Spdo–CherryL2 is a version of Spdo fused to Cherry in its second extracellular loop26 ). i-Notch and i-Spdo co-localized in SOPs, pIIa and pIIb (Fig. 6a–c). At cytokinesis, i-Spdo and i-Notch co-localized in pIIb (Fig. 6b). We conclude that Notch and Spdo are co-internalized. We next tested whether Spdo and Notch physically interact using a polymerase linear amplification (PLA) method that detects pairs of epitopes that are in close proximity (AA19 (F. Roegiers), UAS–Cherry–SpdoL2 (R. Le Borgne), neur– H2B–RFP (ref. 29), DE-Cad–GFP (ref. 30), sGMCA (ref. 31) (moesinABD–GFP, BDSC).The detailed genotypes used in this study are listed in Supplementary Table S1.

Transgenes. Notch genomic DNA was cloned from BAC RP98-1A14 (http://bacpac.chori.org) into the attB-P[acman]-ApR vector using recombineeringmediated gap repair32 . The duplication spans 44,868 nucleotides from −6,233 nucleotides 50 to the transcription start site to 1,245 nucleotides downstream of the 30 UTR. DNA fragments corresponding to the 50 and 30 homology arms were PCR amplified with oligonucleotides DpN-UF (50 cccggccgCTGATATGGCTGAAATTCGAG-30 )/DpN-UR (50 -CAAGAGggatccTACATTTCTGTTCGATACGC-30 ) and DpN-DF (50 -GAAATGTAggatccCACTTGTGTTTGGTATTGAG-30 )/DpN-DR (50 -cgaattcAGAACTGTCACAATCAATGG-30 ) and cloned into the attB-P[acman]-ApR vector using NotI and EcoRI. The resulting clone was linearized using BamHI and dephosphorylated using EXOSAP before being used for recombineering with BAC RP98-1A14 in Escherichia coli SW102 as described at http://recombineering.ncifcrf.gov/Protocol.asp. The resulting BAC was used as a template to introduce YFP and GFP, flanked by GVG linkers, by recombineering as described in ref. 32. Constructs were verified by restriction digests and sequencing of the recombined regions before phiC31-mediated transgenesis using the strain y w M{vas-int.Dm}ZH − 2A;M{3xP3-RFP.attP}ZH-51D. Transgenesis was carried out by BestGene and correct integration at the attP site was verified by PCR. The neur–PH–mRFP1 transgene was obtained by combining a SOP-specific enhancer of the neur gene29 , the PH domain from PLCδ1 (ref. 33) and mRFP1. Transgenic flies were obtained by P-element transformation. Immunostainings. Dissection of staged pupae and antibody staining were carried out using standard procedures. shits pupae were grown at 18 ◦ C–22 ◦ C and heatpulsed for 20 min at 31 ◦ C in a water bath. The following antibodies were used: mouse anti-Notch (NECD, C458.2H, 1:100, DSH; NICD, C17 9C6, 1:100), mouse anti-Cut (2B10 ascite, 1:500, DSHB), mouse anti-Delta (C594.9B, 1:1,000, DSHB), rabbit anti-GFP (Molecular Probe, 1:1,000), rabbit anti-Bazooka (1:1,000; gift from A. Wodarz), rat anti-E-Cad2 (1:500; DSHB), mouse anti-Arm (N2 7A1 ascite, 1:100), rabbit anti-Dlg (1:2,000; gift from P. Bryant), guinea pig anti-Spdo (1:1,000; gift from J. Knoblich), rat and rabbit anti-Spdo (1:1,000, gift from J. Skeath), guinea pig anti-sens (1:3,000, gift from H. Bellen), anti-Sara (1:500, gift from M. Gonzalez-Gaitan), anti-Su(H) (ref. 34) and rabbit anti-HRP (1:500, Cappel). All secondary antibodies were Cy2-, Cy3- and Cy5-coupled antibodies from Jackson’s Laboratories. Endocytosis assay was carried out as described earlier35 using mouse anti-NECD (1:10, concentrate from DSHB) and rabbit anti-DsRed (1:100, Clonetech). A PLA method was used to detect in situ pairs of either Notch (mouse NICD, 1:1,000) or Arm epitopes (mouse anti-Arm, N2 7A1 supernatant, 1:20; DSHB) with Spdo (rabbit polyclonal, 1:1,500) epitopes using a Duolink I kit (http://www.olink.com/). In this method, two different epitopes that are closely located (AA (green in d-f’; the NPAF motif of Spdo that mediates interaction with Numb is deleted in GFP-SpdoNP>AA 2) in wing imaginal cells (ectopic expression of Spdo was directed in dorsal cells using ap-GAL4) resulted in reduced levels of Notch at the apical cortex (NECD, red). Note that GFP-SpdoNP>AA localized baso-laterally whereas GFP-Spdo localized more apically. Panels

b, b’, e and e’ show high magnification views of a and d, respectively. Cross section views are shown in panels c, c’, f and f’. We conclude that the SpdoNumb interaction is not strictly required for the down-regulation of Notch by Spdo. Genotypes: a-c’ : w1118 ; ap-GAL4 / UAS-GFP-Spdo d-f’: w1118 ; ap-GAL4 / + ; UAS-GFP-SpdoNP>AA / +

WWW.NATURE.COM/NATURECELLBIOLOGY

5 © 2012 Macmillan Publishers Limited. All rights reserved.

S U P P L E M E N TA R Y I N F O R M AT I O N

Figure S6 Numb enhanced the down-regulation of Notch by Spdo. Analysis of the density of SOPs (marked by Senseless) in 16 hr APF pupae was used to investigate the effect of Numb and Spdo on the activity of Notch. Mild overexpression of Numb (b) and Spdo (c) had a weak inhibitory effect on Notch (compare with the negative control in a), as revealed by a mild increase in SOP number. Concomitant overexpression of Numb and Spdo resulted in a significant increase in SOP density (d),

indicating that Numb enhanced the inhibitory effect of Spdo on the activity of Notch. Genotypes: a: w1118 ; ; eq-GAL4 / + b: w1118 ; ; eq-GAL4 / P[UAS-Numb] c: w1118 ; ; eq-GAL4 / P[UAS-Spdo]G1 d: w1118 ; ; eq-GAL4 / P[UAS-Numb] P[UAS-Spdo]G1

6

WWW.NATURE.COM/NATURECELLBIOLOGY © 2012 Macmillan Publishers Limited. All rights reserved.

S U P P L E M E N TA R Y I N F O R M AT I O N

Supplemental Information - Movies

Movie 1: live imaging of Notch in a wild-type SOP Spinning disk confocal sections corresponding to surface (z=0), subapical (z=-1mm) and basal (z=-5mm) views of NiGFP (green/white) are shown together with H2B-RFP (magenta; maximal projection of several confocal sections). Genotype: N55e11 w1118 P[neur-Histone2B-RFP]619 / Y ; M[3xP3-RFP.attP.w+.NiGFP]51D / + Movie 2: live imaging of cytokinesis in a wild-type SOP Actin (sGMCA, white) and Histone2B (H2B-RFP, red) were imaged using spinning disk confocal microscopy. Confocal sections corresponding to surface (z=0), subapical (z=-3mm) and basal (z=-5mm) views are shown. A maximal projection of H2B-RFP channel is shown. Cytokinetic furrow initially forms basally. Furrow ingression is concomitant to chromatin decondensation. Note that apical actin accumulates in the newly born SOP daughter cells. This accumulation has been described as the Actin Rich Structure (ARS)3. Genotype: w1118 P[pNeur-H2B-RFP]619 / Y; P[pSqh-MoeABD-GFP] (sGMCA ) /+ Movie 3: live imaging of Notch in a numbRNAi SOP Spinning disk confocal sections corresponding to surface (z=0) and basal (z=-5mm) views of NiGFP (green/white) are shown together with H2B-RFP (magenta; maximal projection of several confocal sections). A transient basal accumulation of NiGFP is indicated by an arrow. Genotype: N55e11 w1118 P[neur-Histone2B-RFP]619; M[3xP3-RFP.attP.w+.NiGFP]51D / + ; pnr-GAL4 P[UAS-dsRNA numb]-3779R-3 / + Movie 4: live imaging of Notch in a spdoRNAi SOP Spinning disk confocal sections corresponding to surface (z=0) and subapical (z=-1mm) views of NiGFP (green/white) are shown together with H2B-RFP (magenta; maximal projection of several confocal sections). A transient apical accumulation of NiGFP is indicated by an arrow. Genotype: N55e11 w1118 P[neur-Histone2B-RFP]619; M[3xP3-RFP.attP.w+.NiGFP]51D / P[UAS-dsRNA spdo]-104092 ; pnr-GAL4 / +

WWW.NATURE.COM/NATURECELLBIOLOGY

7 © 2012 Macmillan Publishers Limited. All rights reserved.

GENOTYPES

PANELS

w1118

Fig. 2b,b’; Fig. 4b,b’; Fig. 4f; Fig. 6d-d’’

shits1

Fig. 4c,c’

N55e11 w1118 / Y ; M[3xP3-RFP.attP.w+.NiGFP]51D / + 55e11

N

1118

w

P[neur-Histone2B-RFP]619 / Y ; M[3xP3-

RFP.attP.w+.NiGFP]51D / +

Fig. 2a,a’ Fig. 1c-c’’’; Fig. 1e-g’; Fig. 2c,c’; Fig. 2e; Fig. 3a; Fig. 4d-d’’’; Fig. 5a

N55e11 w1118 P[neur-PH-mRFP1]-965 / Y ; P[neur-PH-

Fig. 2d

mRFP1]-967 M[3xP3-RFP.attP.w+.NiGFP]51D / + d-d’’’’: N55e11 w1118 P[Ubx-flp] P[neur-Histone2B-

Fig. 1d-d’’’’

RFP]619 / Y ; M[3xP3-RFP.attP.w+.NiGFP]51D / + ; FRT82B P[ubi-nlsGFP] / FRT82B neurIF65 N55e11 w1118 P[neur-PH-mRFP1]-965 / Y ; P[neur-PH-

Fig. 1i,i'

mRFP1]-967 M[3xP3-RFP.attP.w+.NiGFP]51D / + N55e11 w1118 P[neur-Histone2B-RFP]619 / Y ; M[3xP3-

Fig. 3b; Fig. 4g

RFP.attP.w+.NiGFP]51D / + ; pnr-GAL4 P[UAS-dsRNA numb]-3779R-3 / + N55e11 w1118 P[neur-Histone2B-RFP]619 / Y ; M[3xP3-

Fig. 3c; Fig. 4h; Fig. 5b

RFP.attP.w+.NiGFP]51D / P[UAS-dsRNA spdo]104092 ; pnr-GAL4 / + N55e11 w1118 P[neur-Histone2B-RFP]619 / Y ; M[3xP3-

Fig. 3d

RFP.attP.w+.NiGFP]51D / P[UAS-dsRNA spdo]104092 ; pnr-GAL4 P[UAS-dsRNA numb]-3779R-3 / + N55e11 w1118 P[Ubx-flp] P[neur-Histone2B-RFP]619 / Y ;

Fig. 3e

numb15 FRT40A M[3xP3-RFP.attP.w+.NiGFP]51D / P[ubi-nlsGFP] FRT40A N55e11 w1118 P[Ubx-flp] P[neur-Histone2B-RFP]619 / Y ;

Fig. 3f

M[3xP3-RFP.attP.w+.NiGFP]51D / + ; FRT82B P[ubinlsGFP] / FRT82B spdoG104

© 2012 Macmillan Publishers Limited. All rights reserved.

N55e11 w1118 P[neur-PH-mRFP1]-965 / Y ; M[3xP3-

Fig. 4a-a’’’

RFP.attP.w+.NiGFP]51D / + ; pnr-GAL4 P[UAS-dsRNA numb]-3779R-3 / + shits1 / Y ; M[3xP3-RFP.attP.w+.NiGFP]51D / + ; P[neur-

Fig. 4e-e’’’

Histone2B-RFP] / + N55e11 w1118 P[Ubx-flp] P[neur-Histone2B-RFP]619 / Y ;

Fig. 5c,c’

M[3xP3-RFP.attP.w+.NiGFP]51D / + ; FRT82B P[ubinlsGFP] / FRT82B DlRev10

N55e11 w1118 P[Ubx-flp] P[neur-Histone2B-RFP]619 / Y ;

Fig. 5d-d’’’

M[3xP3-RFP.attP.w+.NiGFP]51D / + ; FRT82B P[ubinlsGFP] / FRT82B DlRev10 spdoG104 w1118 ; ap-GAL4 / + P[UAS-Spdo]G1 / +

Fig. 5e-f’’’

w1118 ; neurP-GAL4 / P[UAS-SpdoL2-Cherry]

Fig. 6a-c’’

© 2012 Macmillan Publishers Limited. All rights reserved.