Characterization of TNF-induced cell death in Drosophila reveals caspase- and JNK-dependent necrosis and its role in tumor suppression

Abstract

Tumor-necrosis factor (TNF) and its superfamily members are pleiotropic cytokines. Activation of TNF can lead to distinct cellular outcomes including inflammation, cell survival, and different forms of cell death, such as apoptosis and necrosis in a context-dependent manner. However, our understanding of what determines the versatile functions of TNF is far from complete. Here, we examined the molecular mechanisms that distinguish the forms of cell death induced by Eiger (Egr), the sole homolog of TNF in Drosophila. We show that expression of Egr in the developing Drosophila eye simultaneously induces apoptosis and apoptosis-independent developmental defects indicated by cellular disorganization, both of which rely on the c-Jun N-terminal kinase (JNK) signaling activity. Intriguingly, when effector caspases DrICE and Dcp-1 are defective or inhibited, expression of Egr triggers necrosis which is characterized by loss of cell membrane integrity, translucent cytoplasm, and aggregation of cellular organelles. Moreover, such Egr-induced necrosis depends on the catalytic activity of the initiator caspase Dronc and the input from JNK signaling but is independent of their roles in apoptosis. Further mosaic analysis with mutants of scribble (scrib), an evolutionarily conserved tumor suppressor gene regulating cell polarity, suggests that Egr/JNK-mediated apoptosis and necrosis establish a two-layered defense system to inhibit the oncogenic growth of scrib mutant cells. Together, we have identified caspase- and JNK-dependent mechanisms underlying Egr-induced apoptosis versus necrosis and their fail-safe roles in tumor suppression in an intact organism in vivo.

Fig. 1. The cDcp1 antibody recognizes cleaved DrICE and Dcp-1 in Drosophila.

Late 3rd instar larval eye disks labeled with either the TUNEL assay (a, c, e, g) or the cDcp1 antibodies (b, d, f, h). Expression of hid under the control of GMR (GMR-hid) induces two apoptotic waives indicated by either TUNEL (a, arrows) or cDcp1 (b, arrows) staining. GMR-hid-induced two apoptotic waves are recognized by both TUNEL (c) and cDcp1 (d) in dcp-1 null mutants. GMR-hid-induced apoptosis is blocked in drICE null mutants as indicated by lack of TUNEL labeling (e). However, cDcp1 recognizes a low level of signals in the whole GMR domain in the same genetic background (f). No GMR-hid-induced signals were detected by either TUNEL (g) or cDcp1 (h) in drICE and dcp-1 double null mutants

Fig. 2. GMR > egr induces apoptosis via hid in the Drosophila eye.

a–d Adult eye images. Compared with wild type (a), expression of egr under the control of GMR (GMR-GAL4 UAS-egr, GMR > egr) induces a strong eye ablation phenotype (b). This phenotype is completely suppressed by an RNAi knockdown of egr (egr^RNAi, c) or expression of puckered (puc, d), a negative regulator of JNK. Late 3rd instar larval eye disks labeled with cDcp1 (e–h, j) or cDcp1 and GFP (i–i′′). Compared with wild type (e), GMR > egr induces massive apoptosis indicated by cDcp1 labeling (f). This apoptosis is suppressed in dronc null mutants (g) or by expression of a GMR-p35 transgene (h). In GMR > egr disks with hid mutant clones marked by lack of GFP (i), apoptosis is blocked in the clones (highlighted by yellow dotted lines in i′ and i′′ which are enlarged images of the outlined area in i). In contrast, rpr mutants (rpr^87/XR38a, combination of a deletion and a null mutant of rpr) do not suppress GMR > egr-induced apoptosis (j). k–n Adult eye images. hid mutant clones (k), expression of a RING domain-deleted, therefore stabilized, form of Diap1 (GMR-BIR, l) or dronc null mutants (pharate adults were dissected out of the pupal cases, m) strongly suppress GMR > egr-induced eye ablation phenotype. In contrast, rpr mutants (rpr^87/XR38a) do not suppress the small eyes induced by GMR > egr (n)

Fig. 3. GMR > egr induces nonapoptotic, but JNK-dependent, cellular disorganization in the Drosophila eye.

a, b Adult eye images. Expression of a dominant negative form of bsk (bsk^DN, bsk = Drosophila JNK) (a) or a hemizygous mutant of Tak1 (Tak1^-/Y, b), an upstream kinase of JNK, almost completely suppresses GMR > egr-induced eye ablation phenotype. (c) A diagram showing that JNK signaling induced by Egr can lead to both apoptosis and nonapoptotic defects in the developing eye. APF22h pupal eye disks labeled with a cellular membrane maker Dlg (green in d–k and d′–k′) and a neuronal marker ELAV (red in d–k and d′′–k′′). In wild-type disks (d–d′′), ommatidia (each is composed of eight photoreceptor neurons), as indicated by ELAV, and interommatidial cells, as indicated by Dlg, are well-patterned. In contrast, defective cellular organization was observed in GMR > egr disks (e–e′′). Examples of these defects such as ommatidial fusion (arrowheads in e′′, g′′, i′′) and increased interommatidial spacing (arrows in e′′, g′′, i′′) are highlighted. dronc null mutants (f–f′′, dronc^−/−) or expression of a stabilized form of Diap1 (h–h′′, GMR-BIR) neither alter the ommatidial patterning in wild-type eye disks (compare f–f′′ and h–h′′ to d–d′′) nor suppress the irregular ommatidial organization in GMR > egr disks (compare g–g′′ and i–i′′ to e–e′′). In contrast, expression of a dominant-negative form of JNK (bsk^DN, j–j′′) or a hemizygous mutant of Tak1 (k–k′′) strongly suppresses the cellular disorganization induced by GMR > egr

Fig. 4. Nonapoptotic cell death is induced in GMR > egr when effector caspases DrICE and Dcp-1 are inhibited.

APF22h pupal disks labeled with cDcp1 (green in a, b, c and gray in a′, b′, c′) and ELAV (red in a, b, c), a neuronal marker. At this stage, no apoptotic cells were detected in wild-type disks (cDcp1, a, a′). Ommatidia are also well-patterned (ELAV, a). In contrast, strong apoptosis was detected at APF22h in GMR > egr eye disks (cDcp1, b and b′). Cellular disorganization indicated by increased interommatidial spacing (arrows, b) and ommatidial fusion (arrowheads, b) was observed. Although GMR > egr-induced apoptosis is almost completely blocked by expression of P35 (cDcp1, c and c′), the irregular ommatidial organization (c, arrows and arrowheads) is not suppressed. d–f Adult eye images. Expression of P35 (GMR-p35, d), RNAi knockdown of drICE and dcp-1 (e), or drICE null mutants (f) do not or only slightly suppress GMR > egr-induced eye ablation phenotype (compare 4d, e, f to 2b). APF22h pupal eye disks labeled with Propidium Iodide (PI, green in g, h, i and gray in g′, h′, i′) and Hoechst (red in g, h, i and gray in g′′, h′′, i′′). In GMR > egr disks, PI detects a background level of signals (arrows, g–g′′) which often do not co-localize with the Hoechst labeling, a nucleus marker. In contrast, expression of P35 (GMR > egr/GMR-p35) results in a strong increase of PI-positive nuclei, majority of which are also Hoechst-positive (arrowheads, h–h′′). Suppression of these PI signals in dronc heterozygous mutants can be reversed by expression of a wild-type dronc transgene (arrowheads, i–i′′). Asterisks indicate irregular cellular spacing caused by expression of Egr in the corresponding eye disks. j Quantification of PI-positive cell numbers in APF22h pupal eye disks of various genetic backgrounds as indicated. One-way ANOVA with Bonferroni multiple comparison test was used to compute p-values. Asterisks indicate statistically significant changes (**** p< 0.0001). A background low level of PI-labeling was observed in both wild type and GMR > egr disks. This low PI-labeling in GMR > egr is not increased in dronc mutants or by expression of a stabilized form of Diap-1 (GMR-BIR). The PI-labeling is also low in GMR-p35 disks. In contrast, strong PI-labeling was observed in GMR > egr/GMR-p35 disks. This PI-labeling is largely suppressed in dronc heterozygous mutants (GMR > egr/GMR-p35, dronc^+/−). In this background, further expression of a wild-type form of Dronc (GMR > egr-dronc^wt/GMR-p35, dronc^+/−), but not a catalytic site-mutated form of Dronc (GMR > egr-dronc^C318A/GMR-p35, dronc^+/−), is sufficient to restore the PI signals

Fig. 5. Nonapoptotic cell death in GMR > egr/GMR-p35 shows morphological features of necrosis.

TEM images of wt (a), GMR > egr/ + (b, b′) and GMR > egr/GMR-p35 (c–c′′) disks at APF22h. b′ is an enlarged image for the outlined area in b. c′, c′′ are enlarged images for the outlined areas in c. Compared with the wild-type eye disk cell which has a large nucleus (yellow arrow, a), apoptotic features such as high-electron-density chromatin condensation (yellow arrows, b) and apoptotic bodies (dark aggregates in b and b′) are frequently observed in GMR > egr disks (b). Expression of P35 in GMR > egr (c), however, induces necrotic cell features such as translucent cytoplasm, mal-shaped (arrow, c′′) or unidentifiable nuclei (c′), and aggregation of endoplasmic reticulum and other cellular organelles (arrowheads, c′ and c′′). Asterisk indicates a phagolysosome

Fig. 6. GMR > egr induces Dronc-dependent necrosis when DrICE and Dcp-1 are inhibited.

a, a′ APF22h pupal eye disks labeled with cDcp1 (green in d and gray in d′) and ELAV (red in d). Loss of one copy of dronc does not or only slightly suppress GMR > egr-induced apoptosis. b–e Adult eye images. Although loss of one copy of dronc only slightly inhibits GMR > egr-induced small eye phenotype (compare 6b to 2b), it strongly suppresses the eye ablation phenotype induced by GMR > egr/GMR-p35 (compare 6c to 4d). This suppression is neutralized by expression of a wild-type form of Dronc (d), but not a catalytic site-mutated form of Dronc (e). f, g Adult eye images. Expression of Dronc and P35 in dronc heterozygous mutants does not reduce the eye size although it causes an eye pigmentation defect (f). Expression of a catalytic site-mutated form of Dronc does not result in any eye defects (g). h Quantification of the average adult eye size (mean ± SD) of various genetic backgrounds as indicated. One-way ANOVA with Bonferroni multiple comparison test was used to compute p-values. Asterisks indicate statistically significant changes (*P < 0.05 or ****P < 0.0001). Suppression of GMR > egr by expression of P35 is not statistically significant (n.s.). Heterozygous dronc mutants only weakly suppress GMR > egr-induced small eyes (GMR > egr/+, dronc^+/−). But they strongly suppress GMR > egr/GMR-p35-induced eye ablation phenotype (GMR > egr/GMR-p35, dronc^+/−). In this background, further expression of a wild-type form of Dronc (GMR > egr-dronc^wt/GMR-p35, dronc^+/−), but not a catalytic site-mutated form of Dronc (GMR > egr-dronc^C318A/GMR-p35, dronc^+/−), is sufficient to restore the eye ablation phenotype

Fig. 7. JNK signaling contributes to Eiger- and Dronc-induced necrosis when apoptosis is blocked.

a–h Adult eye images. Heterozygous mutants of bsk¹ (a), hep¹ (b), MKK4^G680 (c) and Tak1² (d) can only weakly or moderately inhibit GMR > egr-induced eye ablation phenotype (compared with Fig. 2b). In contrast, GMR > egr/GMR-p35-induced small eyes are strongly suppressed by heterozygous mutants of bsk¹(e), MKK4^G680 (g) and Tak1² (h), but not hep¹(f) (compared with Fig. 4d). i–l Late 3rd instar larval eye disks labeled with Propidium Iodide (PI). Compared with expression of a wild-type form of Dronc (GMR > dronc^wt/+, i), co-expression of Dronc and P35 (GMR > dronc/GMR-p35) induces PI-positive necrosis (arrowhead, j). In contrast, PI-labeling is not observed when a catalytic site-mutated form of Dronc is expressed instead (GMR > dronc^C318A/GMR-p35, k). Loss of one copy of bsk (bsk^+/−) strongly suppresses necrosis induced in GMR > dronc^wt/GMR-p35 (l). m Quantification of PI-positive cell numbers in late 3rd instar larval eye disks of various genetic backgrounds as indicated. One-way ANOVA with Bonferroni multiple comparison test was used to compute p-values. Asterisks indicate statistically significant changes (**** p< 0.0001). Loss of one copy of Tak1 (Tak1^+/−) or a Tak1 null mutant (Tak1^−/−) strongly suppresses necrosis induced in GMR > dronc^wt/GMR-p35. n A diagram showing comparable molecular mechanisms of regulated necrosis in Drosophila and mammals. The Drosophila TNF (Egr), similar to its mammalian counterparts, has multiple context-dependent functions including induction of necrosis when apoptosis is blocked. In mammals, necrosis can occur when inhibition of caspase-8 on RIPK1 and RIPK3 is removed. JNK contributes to both apoptosis and necrosis. While in Drosophila, effector caspases DrICE and Dcp-1 inhibit Egr-induced necrosis. Once this inhibition is removed, the initiator caspase Dronc can activate necrosis with an additional input(s) from JNK signaling. Key factors that mediate this necrosis downstream of caspases and JNK are currently unknown (indicated by the question mark). Moreover, energy metabolism regulators have been implicated in regulation of Egr/TNF-induced signaling responses although their exact roles remain to be determined (see “Discussion”)

Fig. 8. Egr/JNK-mediated necrosis restricts oncogenic growth of scrib mutant cells.

a–c′ Late 3rd instar larval eye disks with GFP-positive scrib mutant (scrib^−/−, a), scrib^−/−-p35 (b) or scrib^−/−-p35-bsk^DN (c) clones in an otherwise wild-type background. These disks are labeled with Propidium Iodide (PI, red in a–c and gray in a′–c′). Compared with scrib mutant clones (a, a′), PI-positive necrosis is induced in scrib^−/−-p35 clones (arrows, b, b′). This necrosis is suppressed by inhibition of JNK via expression of bsk^DN (c, c′). d, e Adult eye images. scrib mutant clones only cause a mild rough-eye phenotype (d). In contrast, necrotic patches (arrow, e) are observed in over 90% of scrib^−/−-p35 mosaic eyes (n = 44). f scrib^−/−-p35-bsk^DN mosaic animals are pupal lethal. g Quantification of the clone/disk size ratio in late 3rd instar larval eye disks with various genetic backgrounds as indicated. The ratio is a comparison of the total clone size in each eye disk to the full disk size. One-way ANOVA with Bonferroni multiple comparison test was used to compute p-values. Asterisks indicate statistically significant changes (****P < 0.0001). Compared with wild-type clones which occupy an average of 40% of the whole eye disk, scrib mutant (scrib^−/−) clones are much smaller with an average of 8% coverage on the disk. Expression of P35 in scrib^−/− clones (scrib^−/−-p35) moderately increases their sizes leading to an average disk coverage of 28%. Further expression of bsk^DN in these clones results in their massive overgrowth and increases the clone/disk ratio to an average of 78%