SUMOylation of Warts kinase promotes neural stem cell reactivation – Nature Communications

Smt3, the single SUMO protein in Drosophila, is required for NSC reactivation and brain development

To identify regulators of NSC lineage development, we carried out a genetic screen using a collection of 504 CRISPR-Cas9-mediated gene deletion mutants on chromosome 2L53. From this screen, we isolated four mutants, M2L-2483, M2L-2484, M2L-2401 and M2L-2402, that displayed defects in NSC reactivation (Supplementary Fig 1a, b). At 24 h ALH, the majority of NSCs in control larval brain were reactivated and incorporated with 5-ethynyl-2’-deoxyuridine (EdU), with only 9.6% of NSCs remaining quiescent and negative for EdU (Supplementary Fig 1a, b). In contrast, the percentage of quiescent NSCs that were EdU-negative was significantly higher at 28.0% and 28.8% in M2L-2483 and M2L-2484 heterozygous mutants, respectively, and at 47.8% and 42.6% in M2L-2401 and M2L-2402 homozygous mutants, respectively (Supplementary Fig 1a, b). The M2L-2483 and M2L-2484 mutants contained mutations in smt3/SUMO, which encodes the solo SUMO protein in Drosophila, whereas the M2L-2401 and M2L-2402 mutants contained mutations in lwr/Ubc9, which encodes the SUMO E2 conjugating enzyme Ubc9 in Drosophila. M2L-2483, M2L-2484, M2L-2401 and M2L-2402 are hereinafter termed smt3M2L-2483, smt3M2L-2484, lwrM2L-2401 and lwrM2L-2402, respectively. The mutation in smt3M2L-2483 is a result of a silent mutation on Q26 and the addition of 2 extra base pairs (bps), which results in a missense mutation from the 27th amino acid onwards (Fig. 1b, Supplementary Table 1a). The mutation in smt3M2L-2484 is a result of the addition of two extra nucleotide bases, leading to a missense mutation from 27th amino acid and early termination of translation at 34th amino acid (Fig. 1b, Supplementary Table 1a). Furthermore, mutations in lwrM2L-2401 and lwrM2L-2402 are caused by a 7- and 1-bp deletions, resulting in missense mutations from the 40th or 39th amino acids, respectively, and early termination of translation at 59th and 61th amino acids, respectively (Fig. 1b, Supplementary Table 1b).

Since smt3M2L-2483 and smt3M2L-2484 mutants were homozygous lethal at 24 h ALH, we examined a known smt3 loss-of-function allele, smt304493, which was generated by a P-element insertion at 20 bp upstream of the transcription start site and survived to early second instar stage54. At 24 h ALH, 64.2% of NSCs in smt304493 homozygous mutants failed to incorporate EdU as compared to 10.4% in control larval brains (Fig. 1c, d). Furthermore, the percentage of EdU-negative, quiescent NSCs in a trans-heterozygous mutant between smt304493 and smt3M2L-2483 and a hemizygous mutant between smt304493 and a smt3-deficient line (smt3Df24652) dramatically increased to 83.6% and 88%, respectively (Fig. 1c, d). We further quantified the percentage of quiescent NSCs with primary cellular protrusions, the hallmark of quiescent NSCs21,55. Using Miranda (Mira) as a marker for the cellular extensions, we found that there was a significant percentage of Mira-positive NSCs that still extended their cellular protrusions in smt304493 mutant brains (12.6%) as compared to 4.5% in control larval brains (Fig. 1e, f). Moreover, the number of mitotic NSCs that are positive for phospho-Histone H3 (PH3) was significantly reduced from 22.2% in control larval brains to 7.1% in smt304493 mutant brains (Fig. 1e, g). Smt3 fluorescence intensity was significantly reduced to 0.52-fold in smt304493 mutant NSCs as compared to control NSCs (Supplementary Fig 1c, d), which validated smt304493 as a loss-of-function allele. Additionally, at 24 h ALH, knockdown of smt3 in NSCs by two independent RNAi lines resulted in prominent reactivation defects in NSCs: the percentage of EdU-negative quiescent NSCs significantly increased from 8.8% in control larval brains to 53.2% and 58.5% in smt3RNAi lines respectively (Supplementary Fig 1e, f). Both smt3RNAi lines still displayed strong NSC reactivation defects at 48 h and 72 h ALH at 29 °C (Supplementary Fig 3a, b), suggesting a block in NSC reactivation. Upon smt3 knockdown in NSCs, only 34.7% and 17.9% of the NSCs were Smt3/SUMO-positive in these two RNAi stocks, compared with 100% in control NSCs, confirming the efficiency of smt3 knockdown (Supplementary Fig 1g, h). While smt304493 mutant had a decrease in the total NSC number (68 NSCs/brain lobe vs 87.1 in control), the total number of NSCs in both smt3RNAi lines appears to be normal (82.3 and 79.3 vs 83.3 in control; Supplementary Table 2).

Remarkably, the volume of smt304493 mutant brain lobes was dramatically reduced to 0.84 × 106 µm3, compared to control larval brain volume of 1.66 × 106 µm3 (Fig. 1h, i), mimicking microcephaly phenotypes. There was a significant decrease in the number of Dpn-positive neuroblasts at 4 h after egg laying in smt304493 homozygous embryos (Supplementary Fig 1i, j), suggesting that this embryonic phenotype may also contribute to the reduction of brain lobe size in smt3 mutants. These data indicate that Smt3/SUMO is required for NSC reactivation and brain development.

Overexpression of smt3/SUMO leads to premature NSC reactivation

We generated Venus-tagged smt3 transgenic flies and verified them with anti-SUMO and anti-Venus antibodies (Supplementary Fig 1l, m). At 9 h ALH, most of the wild-type NSCs were still in a quiescent state, with only about 28.4% of NSCs incorporating EdU (Fig. 1l, m). Remarkably, overexpression of Venus-tagged smt3 by NSC-specific driver grainy head-Gal4 (grh-Gal4) significantly increased the percentage of proliferative NSCs marked by EdU incorporation: from 28.4% in control to 59.4% and 59.6% in two independent smt3 overexpression strains (Fig. 1l, m). Moreover, the percentage of quiescent NSCs with primary cellular protrusions significantly decreased from 12.4% in control to 5.6% and 5.7% in smt3 overexpression strains (Fig. 1n). In addition, the percentage of PH3-positive mitotic NSCs significantly increased from 7.2% in control to 14.3% and 13.5% in smt3 overexpression strains (Fig. 1o), suggesting that smt3/SUMO overexpression triggers premature NSC reactivation. Furthermore, overexpression of Venus-tagged smt3 driven by tubulin-Gal4 (tub-Gal4) fully rescued NSC reactivation defects (Fig. 1c, d), microcephaly-like phenotypes (Fig. 1h, i) and Smt3/SUMO intensity (Supplementary Fig 1n, o) in smt304493 allele, suggesting that Venus-tagged Smt3 is functional. Overexpression of smt3 driven by grh-Gal4 partially rescued the microcephaly-like phenotypes (Fig. 1h, i) in smt304493 allele, suggesting that NSC reactivation phenotype contributes to the microcephaly phenotype in smt3 mutant. grh-Gal4 drives the expression of UAS transgenes in only a subset but not all NSCs in both embryonic ventral nerve cord (Supplementary Fig 1k) and larval brains18, which likely accounts for the incomplete rescue of the microcephaly-like phenotypes using this driver. Indeed, knocking down of smt3 using another NSC-specific driver insc-Gal4, which is known to drive expression in all NSCs, but not grh-Gal4, dramatically decreased the brain size at 72 h ALH, causing microcephaly-like phenotype (Fig. 1j, k), consistent with smt3 mutant phenotype. Upon overexpression of smt3 or lwr at 9 h ALH, there was a slight increase in NSC number, compared with the UAS control line (Supplementary Table 2). Taken together, our results indicate that Smt3/SUMO is both necessary and sufficient for promoting NSC reactivation and indispensable for brain development.

SUMO conjugating enzyme Lwr/Ubc9 promotes NSC reactivation

To further understand the role of the SUMO conjugating enzyme Ubc9/Lwr in Drosophila, we analyzed two other lwr loss of function alleles, lwr05486 and lwr13, both of which were generated from the imprecise excision of a P-element inserted in the regulatory zone56,57. Interestingly, at 24 h ALH, 60.5% of NSCs in lwr05486 and 55.2% of NSCs in lwr13 mutants failed to incorporate EdU as compared to 9% in control (Fig. 2a, b), suggesting severe NSC reactivation defects in these mutants. In addition, the percentage of quiescent NSCs retaining the primary cellular protrusion increased significantly from 4.5% in control to 12.9% and 12.1% in lwr05486 and lwr13 mutant brains, respectively (Fig. 2c). Moreover, the number of PH3-positive mitotic NSCs was greatly reduced from 23.4% in control to 15% in lwr05486 and 15.3% in lwr13 mutant brains (Fig. 2d). lwr13 mutant brains showed a mild decrease in NSC number (79.2 vs 87.1 in control; Supplementary Table 2). Additionally, knockdown of lwr in NSCs by two independent RNAi lines also resulted in NSC reactivation defects at 24 h ALH, wherein, the percentage of EdU-negative NSCs significantly increased from 4.9% in control to 36.8% and 30.8% in the two lwrRNAi lines (Supplementary Fig 2a, b). lwr13 still exhibited NSC reactivation defects at 48 h, 72 h and 96 h ALH at 25 °C (Supplementary Fig 3c, d), suggesting a failure of NSC reactivation.

Fig. 2: SUMO E2 and E1 are required for NSC reactivation.
figure 2

a, k, o At 24 h ALH, larval brain lobes from various genotypes were analyzed for EdU incorporation. NSCs were marked by Dpn and Mira. White arrows point to EdU- qNSCs. b Quantification of EdU- NSCs per BL for (a). Control (yw), 9 ± 3.3, n = 11; lwr05486, 60.5 ± 4.2, P = 8.1E-18, n = 10; lwr13, 55.2 ± 8.5, P = 1.9E-13, n = 11. c Quantification for percentage of NSCs retaining cellular protrusion per BL for (a). Control, 4.5 ± 1.1, n = 10; lwr05486, 12.9 ± 2.7, P = 4.2E-08, n = 10; lwr13, 12.1 ± 1.6, P = 9.6E-08, n = 10. d Quantification of PH3+ NSCs per BL for (a). Control, 23.4 ± 3, n = 10; lwr05486, 15 ± 2.2, P = 1.1E-06, n = 10; lwr13, 15.3 ± 3.8, P = 5.1E-05, n = 10. e Maximum intensity z-projection of larval brains from control (yw), lwr13, lwr13+tub>lwr and lwr13+grh>lwr were stained with DAPI. f Quantification of brain volume for (e). Control, 1.66 ± 0.23, n = 10; lwr13, 1.06 ± 0.19, P = 6.9E-06, n = 10; lwr13+tub>lwr, 1.65 ± 0.4, P = 0.98, P = 0.0005, n = 10 BL; lwr13+grh>lwr, 1.3 ± 0.12, P = 0.0004, P = 0.0038, n = 10. g At 9 h ALH, larval brain lobes in control (β-galRNAi) and UAS-lwr line driven by grh-Gal4 were analyzed for EdU incorporation. White arrows point to EdU+ NSCs. h Quantification of EdU+ NSCs per BL for (g). Control, 30.3 ± 6, n = 10; UAS-lwr, 49.2 ± 8, P = 2.6E-05, n = 10. i Quantification of NSCs retaining cellular protrusion per BL for (g). Control, 12.4 ± 1, n = 10; UAS-lwr, 6.3 ± 1.1, P = 1.4E-10, n = 10. (j Quantification of PH3+ NSCs per BL for (g). Control, 7.2 ± 1.7, n = 10; UAS-lwr, 12.1 ± 1.5, P = 2.9E-06, n = 10. l Quantification of EdU- NSCs per BL for (k). Control (yw), 9.4 ± 2.5, n = 10 BL; aos1c06048, 32.8 ± 8, P = 6.2E-08, n = 10. m Quantification of NSCs retaining cellular protrusion per BL for (k). Control, 4.4 ± 1.1, n = 10; aos1c06048, 11.1 ± 2.4, P = 2.7E-07, n = 10. n Quantification of PH3+ NSCs per BL for (k). Control, 20.8 ± 2.3, n = 10; aos1c06048, 14.4 ± 2, P = 3.5E-06, n = 10. p Quantification of EdU- NSCs per BL for (o). Control (β-galRNAi), 6.1 ± 3, n = 11; uba2RNAi-1, 32.5 ± 7.1, P = 2.1E-10, n = 11; uba2RNAi-2, 29.7 ± 3.8, P = 5.7E-13, n = 11. q Quantification of NSCs retaining cellular protrusion per BL for (o). Control, 4.2 ± 1.3, n = 10; uba2RNAi-1, 9.2 ± 1.9, P = 2.1E-06, n = 10; uba2RNAi-2, 8.8 ± 2, P = 8.4E-06, n = 10. r Quantification of PH3+ NSCs per BL for (o). Control, 24.7 ± 5.4, n = 10; uba2RNAi-1, 16.9 ± 1.8, P = 0.0004, n = 10; uba2RNAi-2, 17.3 ± 1.9, P = 0.0007, n = 10. Data are presented as mean ± SD. **** for P ≤ 0.0001, *** for P ≤ 0.001, ns indicates p > 0.05. Scale bars: 10 μm.

Consistent with NSC reactivation defects in lwr mutants, the volume of the brain lobes significantly reduced from 1.66 × 106 µm3 in control to 1.06 × 106 µm3 in lwr13 mutant at 24 h ALH (Fig. 2e, f). The microcephaly-like phenotype of lwr13 mutants was more severe at 96 h ALH (34.98 × 107 µm3 vs 3.88 × 107 µm3 in control, Supplementary Fig 3e, f). These data suggest that lwr is important for the reactivation of quiescent NSCs and brain development.

Overexpression of SUMO E2 Lwr/Ubc9 drives premature exit of NSC quiescence

Overexpression of lwr in NSCs at 9 h ALH resulted in premature reactivation of NSCs, evident by a significant increase in the percentage of EdU-positive NSCs from 30.3% in control to 49.2% in lwr overexpression strain at 9 h ALH (Fig. 2g, h). Moreover, the percentage of NSCs with primary cellular protrusion dramatically decreased from 12.4% in control brains to 6.3% in lwr overexpression brains (Fig. 2i). In addition, the percentage of PH3-positive mitotic NSCs significantly increased from 7.2% in control brains to 12.1% in lwr overexpression brains (Fig. 2j). Interestingly, overexpression of a dominant-negative lwr (lwrDN) stain58,59 at 24 h ALH resulted in reactivation defects with increased number of EdU-negative quiescent NSCs (Supplementary Fig 2c, d). Furthermore, the overexpression of wild-type lwr driven by grh-Gal4 or tub-Gal4 in lwr13 homozygous mutant at 24 h ALH significantly rescued NSC reactivation defects (Supplementary Fig 2e, f), while overexpression of lwrDN led to more severe NSC reactivation defects (Supplementary Fig 2e, f). Besides, overexpression of lwr driven by tub-Gal4 and grh-Gal4 significantly rescued the microcephaly-like phenotype caused by lwr13 mutant (Fig. 2e, f). However, overexpression of smt3 or lwr failed to promote NSC reactivation in sucrose-only food (Supplementary Fig 2g, h), suggesting that Smt3 and Ubc9 promote NSC cell cycle reentry relying on the presence of nutrition. There was no significant difference between control and smt3 or lwr overexpression at 24 h ALH (Supplementary Fig 3g, h), suggesting their overexpression effect was not sustained at a later stage. Taken together, we propose that the SUMO conjugating enzyme Ubc9/Lwr, like the SUMO protein, is a key regulator of NSC reactivation and brain development.

The heterodimeric SUMO activating enzyme Aos1/Uba2 is required for NSC reactivation

Next, we analyzed the functions of the heterodimeric SUMO activating enzyme Aos1/Uba2 during NSC reactivation. A known aos1 loss of function allele, aos1c06048, which was generated by a PiggyBac transposon insertion60 in the 5′ UTR of the aos1 gene, 41 bp upstream from the ATG61, showed obvious NSC reactivation defects (Fig. 2k–n). At 24 h ALH, the percentage of EdU-negative quiescent NSCs significantly increased from 9.4% in control brains to 32.8% in aos1c06048 mutant brains (Fig. 2k, l). Furthermore, the percentage of quiescent NSCs retaining primary cellular protrusions increased from 4.4% in control brains to 11.1% in aos1c06048 mutant brains (Fig. 2m), and the percentage of PH3-positive mitotic NSCs reduced from 20.8% in control brains to 14.4% in aos1c06048 mutant brains (Fig. 2n). Consistently, knockdown of aos1 by RNAi at 24 h ALH resulted in a significant increase in the number of quiescent NSCs lacking EdU incorporation (Supplementary Fig 2i, j). aos1c06048 mutant brains showed a mild decrease in NSC number (80.2 vs 87.1 in control; Supplementary Table 2).

Uba2, the other component of the SUMO E1 heterodimeric enzyme alongside Aos1, is also required for NSC reactivation. Knockdown of uba2 in NSCs by two independent RNAi lines at 24 h ALH led to a notable increase in the percentage of EdU-negative quiescent NSCs: 6.1% in control to 32.5% and 29.7% in the two knockdown strains (Fig. 2o, p). In addition, the percentage of quiescent NSCs with primary cellular protrusions increased significantly from 4.2% in control to 9.2% and 8.8% in the two uba2RNAi lines (Fig. 2q). Alternately, the number of PH3-positive mitotic NSCs reduced from 24.7% in control to 16.9% and 17.3% in uba2RNAi lines (Fig. 2r). These findings point to an essential role for the heterodimeric SUMO activating enzyme Aos1/Uba2 in NSC reactivation.

Smt3/SUMO levels increase with NSC reactivation

To assess the expression pattern of smt3 in larval brains, we analyzed a published dataset of single cell RNA-sequencing obtained from late first instar larvae at 16 h ALH62. smt3 expression level was low in quiescent NSCs, but increased notably in active NSCs (Fig. 3a). Dpn, an NSC marker that has a higher expression level in active than quiescent NSCs27, was included as a positive control, while Rpl32, a housekeeping gene that has similar expression levels in these two cell populations (Fig. 3a). Consistent with changes in mRNA levels, at 6 h ALH, the relative SUMO intensity increased from 1.25 in quiescent NSCs (qNSCs) to 2.57 in active NSCs (aNSCs) (Fig. 3b, c). While at 12 h ALH, the relative SUMO intensity in quiescent NSCs significantly increased to 1.46 (compared with 1.25 at 6 h ALH), the relative SUMO intensity in active NSCs remained relatively constant at 2.45 (Fig. 3b, c). These results suggest that the Smt3/SUMO mRNA and protein levels in quiescent NSCs are increased when quiescent NSCs re-enter the cell cycle but remain constant following the reactivation.

Fig. 3: Akt stabilizes SUMO protein level during NSC reactivation.
figure 3

a Analysis of smt3 and dpn mRNA expression in qNSCs and aNSCs from the dataset of single-cell RNA seq62. RpL32 works as a house keeping gene. b Larval brains expressing mCD8-GFP driven by grh-GAL4. White arrows point to aNSCs. Yellow arrows point to qNSCs. c Quantification of SUMO intensity (normalized to GFP) at 6 h ALH for (b) in aNSCs (big Dpn+ cells) (2.57 ± 1.79, P = 4.4E-06, n = 10) and qNSCs (Dpn+ cells with protrusion) (1.25 ± 0.42, n = 10), at 12 h ALH in aNSCs (2.45 ± 1.66, P = 0.0003, P = 0.75, n = 10) and qNSCs (1.46 ± 0.57, P = 0.041, n = 10). d Fly brains from yw control and akt3 at 24 h ALH were dissected to extract proteins for western blot. e Quantification of relative SUMO protein levels in (d), n = 3. Control, 1; akt3, 0.65 ± 0.08, P = 0.002. At 24 h ALH (f) and 9 h ALH (i), larval brain lobes were labeled with SUMO, Dpn and Mira. The yellow circles labeled NSCs, lower panels are enlarged views of cells in white squares in upper panels, yellow dotted circles labeled the nucleus of NSCs. g Quantification of SUMO intensity (normalized to Dpn) in qNSCs and aNSCs in (f). QNSCs in control (yw), 1 ± 0.21, n = 10; qNSCs in akt3, 0.82 ± 0.22, P = 0.0007, n = 10; aNSCs in control, 1.96 ± 0.38, P = 1.9E-20, n = 10; aNSCs in akt3, 1.49 ± 0.17, P = 5.2E-38, P = 7.4E-15, n = 10. h Quantification of smt3 mRNA fold enrichment in qPCR assay. At 24 h ALH, larva brains from yw control and akt3 were dissected for qPCR, P = 0.07, n = 3. i Larval brain lobes from control (β-galRNAi), UAS-InRAD and UAS-Myr-Akt lines driven by grh-Gal4. j Quantification graph of SUMO intensity (normalized to Dpn) in qNSCs and aNSCs in (i). qNSCs in control, 1 ± 0.26, n = 11; qNSCs in UAS-Myr-Akt, 1.32 ± 0.35, P = 2.1E-25, P = 7.8E-11, n = 10; qNSCs in UAS-InRAD, 1.14 ± 0.33, P = 2.5E-08, P = 0.003, n = 10; aNSCs in control, 1.59 ± 0.56, P = 3.6E-17, n = 11; aNSCs in UAS-Myr-Akt, 2.39 ± 0.72, P = 4.1E-16, n = 10; aNSCs in UAS-InRAD, 2.05 ± 0.63, P = 2.5E-08, n = 10. k Quantification of the % of NSCs with higher SUMO intensity per BL in (i). Control, 26 ± 11.8, n = 11; UAS-Myr-Akt, 33.3 ± 4.6, P = 0.001, n = 10; UAS-InRAD, 30.2 ± 3.7, P = 0.027, n = 10. Yellow dotted circles labeled cell nucleus. Data are presented as mean ± SD. **** for P ≤ 0.0001, *** for P ≤ 0.001, ** for P ≤ 0.01, for 0.05≤p ≤ 0.01, ns indicates p > 0.05. Scale bars: 10 μm.

Akt promotes the increase of Smt3/SUMO protein during NSC reactivation

To test whether InR/PI3K/Akt pathway regulates SUMO protein during NSC reactivation, we examined the SUMO protein levels in akt3 mutant63 larva brains. SUMO protein level at 24 h ALH significantly decreased to 0.65-fold in akt3 mutant compare to control (Fig. 3d, e). Consistent with this observation, the SUMO fluorescence intensity in both quiescent NSCs and active NSCs in akt3 is significantly lower than that in control (0.82-fold vs 1-fold in quiescent NSCs, 1.49-fold vs 1.96-fold in active NSCs, Fig. 3f, g). Quantitative real-time PCR revealed no significant change in smt3 mRNA levels in akt3 mutant larval brains compared to control at 24 h ALH (Fig. 3h), suggesting that Akt promotes Smt3 protein levels, but not transcript levels. On the contrary, overexpression of an active form of akt (Myr-Akt) in NSCs led to significant increase of SUMO intensity in both quiescent NSCs and active NSCs compared to control (Fig. 3i, j; 1.32-fold vs 1-fold in quiescent NSCs and 2.39-fold vs 1.59-fold in active NSCs). Moreover, overexpression of an active form of InR (InRAD), the Akt upstream activator driven by grh-Gal4 led to significant increase of SUMO intensity in both quiescent NSCs and active NSCs compared to control (1.14-fold vs 1-fold in quiescent NSCs and 2.05-fold vs 1.59-fold in active NSCs, Fig. 3i, j). Besides, the percentage of NSCs with higher SUMO intensity per brain lobe in these brains are significantly higher (Fig. 3i, k). While Arf1 overexpression promotes NSC reactivation via the regulation of microtubule growth32, it did not affect SUMO intensity (Supplementary Fig. 4a, b). These results suggest that InR/Akt specifically promotes the increase of SUMO protein during NSC reactivation to promote quiescent NSCs reentering cell cycle.

SUMO pathway promotes Wts SUMOylation at Lys766

Utilizing the Joined Advanced SUMOylation Site and Sim Analyser (JASSA)64, Wts-Lys766 (Wts-K766) and Wts-Lys561 (Wts-K561) were predicted to be SUMOylated with a high probability. Wts-K766 resides in the kinase domain (Supplementary Table 3a; Fig. 4a) within the consensus sequence [Ψ]-[K]-[x]-[E/D]-[x]-[E/D]6 (ψ stands for a hydrophobic amino acid, Supplementary Table 3b), which belongs to NDSM (negatively charged amino acid-dependent SUMOylation site)65. Wts-K561 is within the consensus sequence [E/D]-[x]-[K]-[Ψ] (Supplementary Table 3b), an inverted SUMOylation motif66. Since Wts-K766, but not Wts-K561, is conserved in human LATS1 and LATS2 (Supplementary Table 3c), we focused on Wts-K766 in our subsequent study. To investigate whether Wts can be SUMOylated, we conducted SUMOylation assays in S2 cells. Cells were transiently transfected with Flag-Wts (~150 kD), either alone or in combination with Myc-tagged wild-type (WT) Smt3. Following precipitation with a Flag antibody, the resulting protein complexes exhibited an anticipated >150 kD bands corresponding to SUMOylated Flag-Wts (Fig. 4b). Notably, smt3 overexpression resulted in a remarkable 2.59-fold increase in SUMOylated Flag-Wts levels (Fig. 4b, c), suggesting that Smt3 promotes Wts SUMOylation. It is noteworthy that the conjugation of Smt3 with target proteins relies on its two carboxy-terminal glycine residues, G87 and G88 (Supplementary Table 1a in yellow color). Substituting these two amino acids with alanine (G87A, G88A) is sufficient to disrupt this conjugation67,68. Consistently, when Flag-Wts and Myc-Smt3G87A, G88A (denoted as Myc-Smt3AA) were co-expressed in S2 cells, we observed a significantly reduced enhancement of Flag-Wts SUMOylation from 2.59-fold to 1.28-fold (Fig. 4b, c). These results highlight the specificity of Wts SUMOylation by Smt3.

Fig. 4: Wts is SUMOylated by Smt3 at Lys766.
figure 4

a The illustration of Wts protein. PPxY motif is required for Yki interaction. The SUMOylation site K766 is within the kinase domain of Wts protein. HM is hydrophobic motif, T1077 is the Thr phosphorylated by Hpo, and this site is within the hydrophobic motif of Wts protein. SUMOylation assays in S2 cells. IP was conducted with anti-Flag antibody (b, f, h) or anti-Wts antibody (d), and western blots were performed by using anti-SUMO and anti-Flag to detect SUMOylated Wts and overall levels of Flag-Wts, respectively. Tubulin serves as a loading control. b Flag-Wts and Myc-Smt3WT or Myc-Smt3AA were co-transfected into S2 cells. d S2 cells were treated with double stranded RNA against gfp, smt3 or lwr for 72 h to induce knockdown. f Flag-Wts, Myc-Smt3 and HA-Ubc9 were co-transfected into S2 cells. 6 h before harvested, S2 cells were treated with DMSO or ML-792. h Flag-WtsWT or Flag-WtsKR was co-transfected with Myc-Smt3 into S2 cells. c, e, g, i Quantification of Wts SUMOylation for (b, d, f, h) respectively, n = 3. c Control (Flag-Wts), 1; Flag-Wts + Myc-Smt3WT, 2.59 ± 0.18, P = 0.0001; Flag-Wts + Myc-Smt3AA, 1.28 ± 0.35, P = 0.24, P = 0.005. e Control (gfpRNAi), 1; smt3RNAi, 0.64 ± 0.12, P = 0.007; lwrRNAi, 0.68 ± 0.19, P = 0.004. g Control (Flag-Wts, DMSO), 1; Flag-Wts + Myc-Smt3 + HA-Ubc9 (DMSO), 2.71 ± 0.62, P = 0.009; Flag-Wts + Myc-Smt3 + HA-Ubc9 (ML-792), 0.93 ± 0.05, P = 0.07, P = 0.008. i Control (Flag-WtsWT), 1; Flag-WtsWT + Myc-Smt3, 1.51 ± 0.05, P = 7.4E-05; Flag-WtsKR, 1.05 ± 0.29, P = 0.79; Flag-WtsKR + Myc-Smt3, 0.98 ± 0.22, P = 0.87, P = 0.015. j At 24 h ALH, larval brain lobes from control (β-galRNAi) and UAS-smt3AA lines driven by grh-Gal4 were analyzed for EdU incorporation. NSCs were marked by Dpn and Mira. White arrows point to EdU- qNSCs. k Quantification of EdU- NSCs per BL for (j). Control, 8.9 ± 3, n = 10; UAS-smt3AA-1, 29.8 ± 5.8, P = 7.6E-09, n = 10; UAS-smt3AA-2, 35 ± 3.6, P = 7.8E-13, n = 10. l Quantification of NSCs that retaining cellular protrusion per BL for (j). Control, 3.2 ± 1.2, n = 10; UAS-smt3AA-1, 8.4 ± 1.8, P = 4.2E-07, n = 10; UAS-smt3AA-2, 10.9 ± 1.7, P = 6.2E-10, n = 10. m Quantification of PH3+ NSCs per BL for (j). Control, 24.4 ± 4.3, n = 10; UAS-smt3AA-1, 13.9 ± 3, P = 5.5E-06, n = 10; UAS-smt3AA-2, 12.7 ± 2.3, P = 5.5E-07, n = 10. Data are presented as mean ± SD. **** for P ≤ 0.0001, ** for P ≤ 0.01, indicates 0.05≤p ≤ 0.01, ns indicates p > 0.05. Scale bars: 10 μm.

Overexpression of wts in NSCs resulted in obvious increase of Wts intensity in NSCs (1.7-fold compared to 1-fold in control, Supplementary Fig 5a, b), while Wts intensity decreased dramatically to 0.69-fold of control in wtse26-1 mutant NSCs (Supplementary Fig 5c, d). In addition, in S2 cells upon wts knockdown, Wts protein level decreased to 0.36-fold in Western blot (Supplementary Fig 5e, f) and 0.25-fold in immunostaining (Supplementary Fig 5g, h), compared with control groups, respectively, further validating the specificity of anti-Wts antibodies. Moreover, clonal experiments that allowed us to directly compare endogenous Wts signal intensity in wild-type cells with that of neighboring wts-depleted or wts-overexpressing cells all strongly supported the specificity of anti-Wts antibodies in larval brains (Supplementary Fig 5i–n).

Subsequently, we examined the SUMOylation of endogenous Wts in S2 cells following the depletion of SUMO/smt3 or SUMO E2 lwr with double-stranded RNA (dsRNA) treatment. In input, there was a significant reduction in overall SUMOylation in smt3– or lwr– knockdown groups compared to the control (gfpRNAi) (Fig. 4d). Within the IP samples, the SUMOylation of endogenous Wts decreased significantly from 1-fold in control to 0.64-fold and 0.68-fold, in the smt3– and lwr– knockdown groups respectively (Fig. 4d, e), indicating that both SUMO and SUMO E2 are essential for Wts SUMOylation. The high efficiency of smt3 and lwr knockdown was confirmed by anti-SUMO blotting of the input samples (Fig. 4d) and anti-HA (Ubc9 tagged with HA) immunostaining (Supplementary Fig 6a), respectively.

ML-792 is an effective inhibitor of SUMO activating enzyme (SAE, SUMO E1) that impedes the process of SUMOylation69. In the DMSO-treated control group within the IP samples, the overexpression of both Smt3 and Ubc9 substantially augmented Flag-Wts SUMOylation, with a 2.71-fold increase (Fig. 4f, g). However, when subjected to ML-792 inhibitor treatment, Flag-Wts SUMOylation was significantly reduced from 2.71-fold to 0.93-fold (Fig. 4f, g). This reduction is even more pronounced than the baseline level of SUMOylation observed without Smt3 and Ubc9 overexpression, which stands at 1-fold (Fig. 4f, g). Moreover, ML-792 inhibitor treatment also suppressed the overall SUMOylation in the input group (Fig. 4f). Furthermore, the SUMOylation of endogenous Wts in S2 cells is similarly disrupted by ML-792 inhibitor (Supplementary Fig 6b). Therefore, SUMO E1 enzyme is indispensable for facilitating Wts SUMOylation.

To determine whether Wts-K766 serves as the primary site for SUMOylation on Wts, the K766 residue in Wts was replaced with an arginine (R) residue in the SUMOylation assay. In S2 cells, Myc-Smt3 overexpression led to a 1.5-fold increase in the SUMOylation of Flag-Wts, while the SUMOylation of Flag-WtsK776R (WtsKR) remained unchanged (Fig. 4h, i). This observation strongly suggests that K766 represents a major site of SUMOylation on Wts.

SUMO conjugation is required for NSC reactivation

We have generated transgenic flies expressing Venus-tagged conjugation-deficient SUMO (smt3AA), verified its expression (Supplementary Fig 6c, d), and examined its effect on NSC reactivation. Notably, Venus-Smt3 localized to the nucleus (Supplementary Fig 1l, 6e), while Venus-Smt3AA localized to both cytoplasm and nucleus of the NSCs (Supplementary Fig 6c, e), consistent with a previous report37. Interestingly, at 24 h ALH, overexpression of smt3AA led to NSC reactivation defects, with a significant increase in the percentage of NSCs lacking EdU incorporation (Fig. 4j, k, 8.9% in the control; 29.8% and 35% in the smt3AA lines). Additionally, the percentage of NSCs retaining their cellular protrusion significantly increased from 3.2% in control to 8.4% and 10.9% in smt3AA lines (Fig. 4l). Furthermore, there was a significant decrease in the percentage of PH3-positive NSCs in smt3AA mutant brains (13.9% and 12.7%) compared to control (24.4%; Fig. 4m). These findings suggest that SUMO relies on its conjugation activity to regulate NSC reactivation.

SUMOylation of Wts at Lys766 attenuates its phosphorylation and kinase activity in vitro

The phosphorylation level of Wts-Thr1077 (p-Wts) by Hpo kinase is the principal readout of the Hippo pathway activity70. We examined p-Wts levels in S2 cells, wherein we expressed Flag-Wts or Flag-WtsKR, both with and without co-expression of Myc-Hpo. As expected, Hpo overexpression significantly increased Wts phosphorylation on Thr1077 from 0.57-fold to 1-fold (Fig. 5a, b, first 2 lanes). Even without hpo overexpression, the phosphorylation level of WtsKR by Hpo well exceeded that of wild-type Wts (Fig. 5a, b, lane 3 vs lane 1, 0.96-fold vs 0.57-fold), and with hpo co-overexpression, WtsKR phosphorylation level remained notably higher than that of wild-type Wts (Fig. 5a, b, lane 4 vs lane 2, 2.25-fold vs 1-fold). These observations suggest that the SUMOylation deficient Wts has higher phosphorylation level than wild-type Wts.

Fig. 5: SUMOylation of Wts at Lys766 suppresses its phosphorylation and attenuates its kinase activity.
figure 5

a, c, e, g, j S2 cells transfected with various constructs or with indicated treatment were collected for IP followed by Western blotting. b, d, f, h, i, km Quantification of the indicated protein levels or fold changes in (a, c, e, g, j) respectively, from three-four independent repeats. b, d Relative p-Wts levels (normalize to Flag-Wts or Flag-WtsKD) in (a, c). b Control (Flag-Wts), 0.57 ± 0.17, P = 0.012; Flag-Wts + Myc-Hpo, 1; Flag-WtsKR, 0.96 ± 0.13, P = 0.036; Flag-WtsKR + Myc-Hpo, 2.25 ± 0.3, P = 0.002, P = 0.30. d Control (Flag-Wts + gfpRNAi), 1; Flag-Wts + hpoRNAi, 0.92 ± 0.05, P = 0.07, P = 0.054; Flag-WtsKR + gfpRNAi 2.25 ± 0.45, P = 0.008; Flag-WtsKR + hpoRNAi, 1.12 ± 0.11, P = 0.15, P = 0.013. e S2 cells were treated with cycloheximide (CHX) prior to IP. f Relative level of p-Wts for (e) (p-Wts normalized to Flag-Wts(KR)). WtsWT 0 h: 1, 2 h: 0.95 ± 0.07; 4 h: 1.38 ± 0.26; 6 h: 1.17 ± 0.29; WtsKR 0 h: 1.82 ± 0.22, 2 h: 1.79 ± 0.27; 4 h: 1.76 ± 0.45; 6 h: 1.55 ± 0.3. g Yki and p-Yki are detected in S2 cells transfected with various constructs. h Relative p-Yki levels in (g). Control (HA-Yki), 0.7 ± 0.01, P = 6.6E-09; HA-Yki + Flag-Wts, 1; HA-Yki + Flag-WtsKR 2.07 ± 0.22, P = 5.9E-05. i Quantification of relative p-Yki levels, which was first normalized to HA-Yki and then further normalized to Flag-Wts from four repeats in (g). HA-Yki + Flag-Wts, 1; HA-Yki + Flag-WtsKR 1.83 ± 0.28, P = 0.0009. j S2 cells transfected with various constructs were treated with DMSO or ML-792. Wts, p-Wts, Yki, and p-Yki were detected following IP. km Quantifications for (j). k Wts SUMOylation fold change. DMSO, 1; ML-792, 0.58 ± 0.03, P = 2.2E-05. l Relative p-Wts level. DMSO, 1; ML-792, 2.35 ± 0.36, P = 0.003. m p-Yki level. DMSO, 1; ML-792, 2.42 ± 0.66, P = 0.02. Data are presented as mean ± SD. ns indicates p ≥ 0.05; indicates 0.05≤ p ≤ 0.01, indicates p ≤ 0.01, indicates p ≤ 0.0001. Scale bars: 10 μm.

hpo knockdown in S2 cells significantly reduced the phosphorylation levels of WtsKR from 2.25-fold to 1.12-fold (Fig. 5c, d, lane 3 vs lane 4; Supplementary Fig 7a), suggesting that either Hpo kinase exhibits a preference for phosphorylating SUMOylation-deficient Wts or phosphorylation of non-SUMOylated Wts is stabilized.

To distinguish between these two possibilities, we conducted cycloheximide (CHX) chase assays in S2 cells to block protein biosynthesis and detect Wts or WtsKR protein levels in a time-course experiment. Notably, the phosphorylation level of wild-type Wts remained relatively stable (from 1-fold to 1.17-fold), albeit much lower than WtsKR (Fig. 5e, f), whereas the phosphorylated WtsKR decreased slightly from 1.82-fold to 1.55-fold within 6 h in the IP group (Fig. 5e, f). This finding suggests that phosphorylation of non-SUMOylated Wts did not result in its stabilization. Therefore, we favor the model that Hpo kinase preferentially phosphorylates SUMOylation-deficient Wts than the wild-type Wts.

Upon its phosphorylation by Hpo kinase, Wts further phosphorylates Yki primarily at Ser168 (p-Yki), leading to its cytoplasmic retention and inactivation23,70,71. In S2 cells, co-expression of Flag-Wts with HA-Yki resulted in a slight increase in Yki phosphorylation from 0.7-fold to 1-fold (Fig. 5g, h). Notably, co-expression of Flag-WtsKR further increased Yki phosphorylation to 2.07-fold (1.83-fold after further normalized to Flag-Wts protein levels, Fig. 5g, i). Moreover, in input group, Yki band exhibited a slight upward shift when co-expressed with wild-type wts, and a more pronounced upward shift (implying Yki phosphorylation) when co-expressed with wtsKR (Fig. 5g). This observation confirms that the SUMOylation-deficient mutant WtsKR functions as an active form of Wts, exhibiting a higher kinase activity towards Yki than wild-type Wts.

Treatment of a SUMOylation inhibitor, ML-792, significantly reduced Wts SUMOylation to 0.58-fold (Fig. 5j, k), compared to untreated cells. Meanwhile, the phosphorylation levels of Wts and Yki increased 2.35- and 2.42-folds, respectively, as compared to control cells (Fig. 5j, l, m). Our data suggests that Wts SUMOylation at K766 diminishes its kinase activity, thereby reducing p-Yki levels to allow NSC reactivation.

SUMOylation-deficient mutant of Wts (WtsKR) represents a hyperactive form of Wts in vivo

We generated transgenic flies expressing Venus-tagged WtsKR and verified them by immunostaining (Supplementary Fig 7b, c). Importantly, wild-type Wts and WtsKR strains had similar Wts protein levels in NSCs (Supplementary Fig 7b, c). Similar to previous reports that overexpression of wild-type wts resulted in NSC reactivation defects20,22,26, we observed that 23.2% of Wts-overexpressing NSCs failed to incorporate EdU, compared with 6.7% in control (Fig. 6a, b). Remarkably, two independent wtsKR-overexpressing NSC lines showed a more severe defect, with 39.7% and 34.4% of these NSCs remaining without EdU incorporation (Fig. 6a, b). Additionally, there were more NSCs exhibiting primary cellular protrusions, with percentage increased from 3.3% in control to 6.7% in wild-type wts overexpression line, to 11.9% and 10.2% in wtsKR overexpression lines (Fig. 6c). Conversely, the percentage of PH3-positive NSCs decreased to 12% and 14.9% in wtsKR overexpressing lines, compared to 24.4% in control, 17.5% in wild-type wts-overexpressing line, (Fig. 6d).

Fig. 6: Wts SUMOylation deficient mutant functions as an active form of Wts kinase to suppress NSC reactivation.
figure 6

a At 24 h ALH, larval brain lobes from control (β-galRNAi), UAS-wts and UAS-wtsKR driven by grh-Gal4 were analyzed for EdU incorporation. NSCs were marked by Dpn and Mira. White arrows point to EdU qNSCs. b Quantification of EdU NSCs per BL for (a). Control, 6.7 ± 2.6, n = 10; UAS-wts, 23.2 ± 9.1, P = 0.0012, n = 10; UAS-wtsKR−1, 39.7 ± 8.9, P = 1.4E-09, P = 0.0007, n = 10; UAS-wtsKR-2, 34.4 ± 6.7, P = 4.5E−10, P = 0.0059, n = 10. c Quantification of NSCs retaining cellular protrusion per BL for (a). Control, 3.3 ± 1.2, n = 10; UAS-wts, 6.7 ± 2.3, P = 0.0005, n = 10; UAS-wtsKR−1, 11.9 ± 2.6, P = 1.2E-08, P = 0.00014, n = 10; UAS-wtsKR−2, 10.2 ± 1.9, P = 7.9E-09, P = 0.0014, n = 10. d Quantification of PH3+ NSCs per BL for (a). Control, 24.4 ± 4.3, n = 10; UAS-wts, 17.5 ± 1.9, P = 0.007, n = 10; UAS-wtsKR-1, 12 ± 3.4, P = 0.0002, n = 10; UAS-wtsKR−2, 14.9 ± 2.1, P = 6.8E-06, n = 10. e Proteins were extracted from larval brains of control (β-galRNAi), UAS-wts and UAS-wtsKR driven by actin-Gal4 for Western blot. f, g Quantification of four independent repeats in (e). f Control 1; UAS-wts, 1.25 ± 0.05, P = 3.4E-05; UAS-wtsKR 2.33 ± 0.45, P = 0.0009, P = 0.003. g Control, 1; UAS-wts, 1.48 ± 0.22, P = 0.005; UAS-wtsKR 3.04 ± 0.68, P = 0.001, P = 0.005. Data are presented as mean ± SD. indicates 0.05≤ p ≤ 0.01, indicates p ≤ 0.01, indicates p ≤ 0.001, indicates p ≤ 0.0001. Scale bars: 10 μm.

Consistent with our data in S2 cells, wtsKR overexpression in larval brains had a similar effect on the phosphorylation levels of endogenous Wts and Yki (Fig. 6e–g). At 24 h ALH, larval brains overexpressing wild-type wts under actin-gal4 (act-gal4), exhibited a slight increase in Wts phosphorylation levels (1.25-fold) and Yki (1.48-fold) phosphorylation levels compared to control brains (1-fold, Fig. 6e–g). Remarkably, larval brains overexpressing wtsKR displayed a significantly higher increase in Wts phosphorylation levels (2.33-fold) and Yki (3.04-fold) phosphorylation levels (Fig. 6e–g). These in vivo findings confirmed that SUMOylation-deficient mutant of Wts (wtsKR) is a hyperactive form of Wts.

SUMOylation of Wts promotes Yki nuclear localization in NSCs

At 24 h ALH, Yki predominantly localized within the nucleus of NSCs in control larval brains. Interestingly, the percentage of nuclear Yki (nYki) significantly reduced to 13.5% in smt3 and 31.8% in lwr mutant NSCs (Fig. 7a, b) as compared to 90.8% in control NSCs. Furthermore, the intensity of nYki in NSCs dramatically decreased from 1-fold in control NSCs to 0.5-fold and 0.59-fold in smt3- and lwr- mutant NSCs, respectively (Fig. 7a, c). These observations suggest that the Hippo pathway is activated upon depletion of SUMOylation pathway components. Moreover, overexpression of the SUMO-deficient mutant smt3AA in NSCs at 24 h ALH also significantly inhibited nuclear localization of Yki (69% and 64.9% in smt3AA lines compared to 88.2% in control; Fig. 7d, e), with a significant reduction in nYki intensity in NSCs (Control, 1-fold; smt3AA lines, 0.61-fold and 0.61-fold; Fig. 7d, f). At 24 h ALH, overexpression of wild-type wts in NSCs led to a minor reduction in the percentage of NSCs with nYki: 87.1% in control vs 60.5% in wts-overexpressing NSCs, and in nYki intensity: 1-fold in control vs 0.79-fold in wts-overexpressing NSCs (Fig. 7g–i). However, overexpression of wtsKR in NSCs resulted in a more dramatic decrease in the percentage of nYki NSCs (20.5%) and nYki intensity (0.49-fold) (Fig. 7g–i). These data suggest that the SUMO pathway-dependent downregulation of the Hippo pathway promotes the nuclear localization of Yki, and consequently, NSC reactivation.

Fig. 7: Wts SUMOylation promotes Yki nuclear localization and Yki target genes expression in NSCs.
figure 7

a, d, g, j, l, n At 24 h ALH, larval brain lobes were labeled with indicated proteins. NSCs were marked by Dpn and Mira or Msps. a, d, g White arrows point to NSCs without nYki localization. The Yki+ Dpn cells pointed by yellow arrows are optic lobe cells. b, e, h Quantification of NSCs with nYki per BL in (a, d, g) respectively. b Control (yw), 90.8 ± 2.3, n = 11; smt304493, 13.5 ± 4.6, P = 2.8E−21, n = 12; lwr13, 31.8 ± 7.8, P = 5.0E-13, n = 10. e Control (β-galRNAi), 88.2 ± 3.1, n = 10; UAS-smt3AA-1, 69 ± 9.5, P = 9.7E-06, n = 10; UAS-smt3AA-2, 65 ± 11.1, P = 5.0E-06, n = 10. h Control (β-galRNAi), 87.1 ± 4.2, n = 10; UAS-wts, 60.5 ± 7.3, P = 1.6E-08, n = 10; UAS-wtsKR, 20.5 ± 8.5, P = 2.0E-14, P = 1.5E-09, n = 10. c, f, i Quantification of nYki intensity (normalized to Dpn) in NSCs in (a, d, g) respectively. (c) Control, 1 ± 0.18, n = 11; smt304493, 0.5 ± 0.09, P = 4.0E-82, n = 12; lwr13, 0.59 ± 0.12, P = 1.9E-56, n = 10. f Control (β-galRNAi), 1 ± 0.35, n = 10; UAS-smt3AA-1, 0.61 ± 0.14, P = 2.4E-27, n = 10; UAS-smt3AA-2, 0.61 ± 0.15, P = 7.3E-23, n = 10. i Control (β-galRNAi), 1 ± 0.35, n = 10; UAS-wts, 0.79 ± 0.18, P = 4.6E-09, n = 10; UAS-wtsKR, 0.49 ± 0.12, P = 4.3E-42, P = 1.8E-42, n = 10. k, m Quantification of CycE intensity (normalized to Dpn) in NSCs in (j, l) respectively. k Control (yw), 1 ± 0.47, n = 10; smt304493, 0.33 ± 0.13, P = 4.1E-37, n = 15; lwr13, 0.36 ± 0.11, P = 4.3E-29, n = 13. m control (β-galRNAi), 1 ± 0.14, n = 10; UAS-wts, 0.86 ± 0.22, P = 8.7E-05, n = 10; UAS-wtsKR, 0.69 ± 0.18, P = 3.2E-17, P = 2.1E-05, n = 10. n Larval brain lobes from control (β-galRNAi), UAS-wts and UAS-wtsKR lines driven by grh-Gal4; Ban-lacZ were labeled with Dpn, Msps, and β-gal (Ban-lacZ). o Quantification of Ban-lacZ intensity (normalized to Dpn) in NSCs in (n). Control, 1 ± 0.28, n = 10; UAS-wts, 0.74 ± 0.26, P = 1.6E-13, n = 13; UAS-wtsKR, 0.54 ± 0.2, P = 2.3E-13, P = 4.0E-41, n = 14. The yellow circles labeled the NSCs, the lower panels are enlarged views of the cells in the white squares in the upper panels, the yellow dotted circles labeled the nucleus of the NSCs. Data are presented as mean ± SD. ** for P ≤ 0.01, **** for P ≤ 0.0001. Scale bars: 10 μm.

SUMOylation of Wts promotes the expression of Yki target genes in NSCs

We re-analyzed gene expression data from the single-cell RNA sequencing dataset62 and found the expression patterns for diap1 and cycE, potential targets of Yki, to be similar to that of smt3 and lwr—exhibiting significantly higher expression levels in the active NSCs compared to quiescent NSCs (Supplementary Fig 8a). Moreover, loss of smt3 or lwr function led to a substantial decrease in CycE protein intensity in NSCs from 1-fold to 0.33-fold in smt3 mutant and 0.36-fold in lwr mutant NSCs (Fig. 7j, k). Furthermore, smt3 knockdown in NSCs also resulted in a noticeable reduction in CycE protein levels (Supplementary Fig 8b, c). The intensity of CycE protein and Ban-lacZ also decreased in these NSCs, with a more pronounced reduction in wtsKR-overexpressing NSCs (Fig. 7l, m and Fig. 7n, o). These data further support that the SUMO pathway downregulates the Hippo pathway to allow Yki to function for NSC reactivation.

The SUMO pathway downregulates Wts protein levels in vitro

We test whether SUMOylation of Wts is responsible for the downregulation of Wts protein levels. Indeed, the western blots showed that SUMO and Ubc9 downregulated Wts protein levels in a dosage-dependent manner in S2 cells (Fig. 8a–d). Moreover, in CHX chase assays, Smt3- and Ubc9-overexpression accelerates, the rate of Wts protein degradation, compared to control conditions without Smt3 and Ubc9 overexpression (Fig. 8e, f). By contrast, Wts degradation is abrogated by smt3AA overexpression (Fig. 8g, h), suggesting that Wts degradation requires the SUMO conjugation activity of Smt3.

Fig. 8: The SUMO pathway downregulates Wts protein level in vitro.
figure 8

a, c Western blotting in S2 cells expressing indicated proteins. e, g S2 Cells expressing indicated proteins were treated with cycloheximide (CHX) for indicated intervals and collected for Western blotting with indicated antibodies. b, d, f, h Quantification of Wts protein levels for (a, c, e, g) respectively (normalized to Tubulin). Data here is the mean of 3 independent experiments. b Flag-Wts, 1; Flag-Wts+HA-Smt3, 0.68 ± 0.06, P = 0.0007; Flag-Wts+2 HA-Smt3, 0.54 ± 0.07, P = 0.0003. d Flag-Wts, 1; Flag-Wts+HA-Ubc9, 0.72 ± 0.08, P = 0.004; Flag-Wts+2 HA-Ubc9, 0.55 ± 0.06, P = 0.0002. Flag-Wts, 0 h, 1; 2 h, 0.77 ± 0.02; 4 h, 0.65 ± 0.07; 6 h, 0.49 ± 0.12; Flag-Wts+Myc-Smt3, 0 h, 1; 2 h, 0.7 ± 0.1; 4 h, 0.49 ± 0.1; 6 h, 0.29 ± 0.04; (f) Flag-Wts, 0 h, 1; 2 h, 0.72 ± 0.07; 4 h, 0.61 ± 0.09; 6 h, 0.52 ± 0.12; Flag-Wts+Myc-Smt3AA, 0 h, 1; 2 h, 0.74 ± 0.1; 4 h, 0.55 ± 0.06; 6 h, 0.53 ± 0.18. Data are presented as mean ± SD. *** for 0.001 ≤ P ≤ 0.0001, ** for 0.01 ≤ P ≤ 0.001.

The SUMO pathway downregulates Wts protein levels in vivo

Next, we investigated whether the SUMOylation pathway regulates Wts protein levels in larval brains. Notably, we observed that Wts intensity increased significantly in smt304493 and lwr13 mutants in both quiescent NSCs (Fig. 9a, b; 1.34-fold and 1.23-fold vs 1-fold in control) and active NSCs (1.58-fold and 1.5-fold vs 1.22-fold in control). Conversely, Wts protein intensity decreased significantly from 1-fold in control to 0.84-fold and 0.75-fold in two smt3 overexpression lines (Fig. 9c, d) and to 0.81-fold in lwr overexpression line (Fig. 9g, h). Intriguingly, Wts intensity in NSCs is significantly increased when conjugation-deficient SUMO (smt3AA) or dominant-negative lwr (lwrDN) was overexpressed (1.52-fold and 1.6-fold in smt3AA overexpression lines, Fig. 9e, f; and 1.25-fold in lwrDN overexpression line, Fig. 9g, h) as compared to control. Quantitative real-time PCR results showed that wts mRNA levels had no significant difference in smt3 or lwr mutant larval brains, compare to the control (Fig. 9i), suggesting a regulation independent of gene expression.

Fig. 9: The SUMO pathway downregulates Wts protein level in vivo.
figure 9

At 24 h ALH (a, e, g) or 16 h ALH (c), larval brain lobes from indicated genotypes were labeled with Wts, Dpn and Mira, Venus tag in (c) and (e) also showed expression in NSCs. The yellow dotted circles labeled the Dpn+ and Mira+ NSCs; white arrows pointed to Venus positive NSCs. The lower panels are enlarged views of the cells in yellow dotted circles in upper panels. b, d, f, h Quantification of Wts intensity in NSCs normalized to Dpn in various genotypes in (a, c, e, g) respectively. b, qNSCs in control (yw), 1 ± 0.1, n = 12; qNSCs in smt304493, 1.34 ± 0.26, P = 4.2E-11, n = 11; qNSCs in lwr13, 1.23 ± 0.25, P = 1.3E-06, n = 10; aNSCs in control, 1.22 ± 0.14, P = 1.6E-14, n = 12; aNSCs in smt304493, 1.58 ± 0.17, P = 1.0E-30, P = 1.3E-07, n = 11; aNSCs in lwr13, 1.5 ± 0.19, P = 3.5E-20, P = 5.4E-09, n = 10. d Control (β-galRNAi), 1 ± 0.14, n = 13; UAS-smt3-1, 0.84 ± 0.16, P = 3.6E-18, n = 11; UAS-smt3-2, 0.75 ± 0.1, P = 1.2E-45, n = 11. f Control (β-galRNAi), 1 ± 0.12, n = 12; UAS-smt3AA-1, 1.52 ± 0.28, P = 1.2E-51, n = 10; UAS-smt3AA-2, 1.6 ± 0.34, P = 5.1E-49, n = 10. h Control (β-galRNAi), 1 ± 0.2, n = 10; UAS-lwr, 0.81 ± 0.18, P = 1.4E-14, n = 10; UAS-lwrDN, 1.25 ± 0.2, P = 5.0E-21, n = 10. i Quantification of wts mRNA fold enrichment in qPCR assay. At 24 h ALH, larval brains from control (yw), smt304493 and lwr13 were dissected for qPCR, n = 3. Control, 1; smt304493, 0.99 ± 0.07, P = 0.82; lwr13, 0.98 ± 0.02, P = 0.15. The yellow circles labeled the NSCs, the lower panels are enlarged views of the cells in the white squares in the upper panels. Data are presented as mean ± SD. **** for P ≤ 0.0001. Scale bars: 10 μm.

We analyzed wtsVenus, in which a Venus-tagged wts was expressed under its endogenous promoter72. The localization pattern of wtsVenus by anti-GFP (Venus) staining was similar to that of anti-Wts antibody staining in the larval brain but with brighter signal (Supplementary Fig 9a). Consistent with the results obtained by anti-Wts antibody staining, loss of function of smt3 led to a significant increase in WtsVenus intensity (Supplementary Fig 9b–d). Moreover, overexpression of wild type lwr resulted in a decrease in WtsVenus intensity (0.89-fold), while overexpression of dominant negative lwrDN resulted in an increase in WtsVenus intensity (1.2-fold) (Supplementary Fig 9e, f). These findings suggest that Smt3 and Ubc9 play a pivotal role in promoting Wts SUMOylation, subsequently leading to Wts protein degradation during NSC reactivation.

At 96 h ALH, lwr13 optic lobes were much smaller compared to control brains, with the total number of Dpn+ cells in lwr13 optic lobes being significantly lower than that in control (41.9 vs 223.2 in control, Supplementary Fig 9g, h). In addition, 46.8% of Dpn+ cells in lwr13 optic lobes were EdU-negative, while all Dpn+ cells in the control optic lobes were EdU-positive (Supplementary Fig 9g, i). In addition, Wts intensity in Dpn+ cells in lwr13 optic lobes was significantly higher than that in control optic lobes (1.41-fold, Supplementary Fig 9j, k). Thus, SUMOylation may also regulate Wts in other cell types.

Wts predominantly localizes in the cytoplasm but is also present in the nucleus in NSCs (Supplementary Fig 9a, b). Given that Smt3 localizes to the nucleus (Supplementary Figs 1l, 6e), it is likely that Wts is SUMOylated in the nucleus. This is consistent with the findings that most SUMO substrates are localized in the nucleus73 and that CRL4-mediated inhibition of the hippo pathway kinases Lats1/2 occurs in the nucleus74.

SUMO pathway promotes NSC reactivation by inhibiting the Hippo pathway

Simultaneous loss of smt3 and wts by RNAi significantly suppressed the EdU-negative phenotype seen in smt3-depleted brains alone (Fig. 10a, b; smt3RNAi = 54.3%. smt3RNAi; wtsRNAi = 28.6%), while wtsRNAi brains showed no reactivation defects (Fig. 10a, b; 9.2%). Similarly, at 24 h ALH, the number of EdU-negative NSCs in brains depleted of both lwr and wts dramatically reduced to 37.8% compared with 65.7% in lwr13 mutant brains (Fig. 10c, d). Wts is known to phosphorylate and render Yki inactive, maintaining NSCs in a quiescent state20,22. In line with this, the overexpression of a constitutively active form of Yki (YkiS168A) also markedly alleviated NSC reactivation deficits caused by smt3 or lwr depletion: the percentage of EdU-negative quiescent NSCs decreased from 54.3% to 31% (Fig. 10a, b) and 65.7% to 38.1%, in smt3- and lwr-depleted brains, (Fig. 10c, d), respectively. These data strongly suggest that the SUMO pathway promotes NSC reactivation by inhibiting Wts activity and function (Fig. 10e).

Fig. 10: SUMO pathway promotes NSC reactivation by inhibiting the Hippo pathway.
figure 10

a, c At 24 h ALH, larval brain lobes in indicated genotypes were analyzed for EdU incorporation. NSCs were marked by Dpn and Mira. White arrows point to EdU negative quiescent NSCs. b, d Quantification graph of EdU- NSCs per BL for genotypes in (a, c) respectively. b Control (β-galRNAi; β-galRNAi), 11.9 ± 3.9, n = 13; smt3RNAi; β-galRNAi, 54.3 ± 8.7, P = 1.8E-13, n = 11; β-galRNAi; wtsRNAi, 9.2 ± 2.5, P = 0.07, n = 10; smt3RNAi; wtsRNAi, 28.6 ± 4.5, P = 4.1E-09, P = 9.3E-08, n = 10; β-galRNAi; ykiS168A, 16.1 ± 4.6, P = 0.025, n = 10; smt3RNAi; ykiS168A, 31 ± 5.3, P = 1.8E-09, P = 6.5E-07, n = 10. d Control (β-galRNAi; β-galRNAi),8.6 ± 4.1, n = 13; lwr13; β-galRNAi, 65.7 ± 10, P = 1.6E-15, n = 12; β-galRNAi; wtsRNAi, 8.8 ± 3, P = 0.89, n = 10; lwr13; wtsRNAi, 37.8 ± 7, P = 3.0E-11, P = 3.9E-07, n = 10; β-galRNAi; ykiS168A, 14.4 ± 4.7, P = 0.003, n = 11; lwr13; ykiS168A, 38.1 ± 5.7, P = 1.5E-06, P = 2.2E-07, n = 10. Data are presented as mean ± SD. **** for P ≤ 0.0001, ** for P ≤ 0.01 and ns for P > 0.05. Scale bars: 10 μm. e A working model illustrating the mechanism by which SUMO pathway promotes NSC reactivation. Wts kinase can be phosphorylated and activated by Hpo kinase. Phosphorylated Wts promotes Yki phosphorylation and cytoplasmic retention. Akt promotes the increase of SUMO protein. SUMO E1 Aos1/Uba2 and E2 Ubc9 promote Wts SUMOylation, resulting in Wts phosphorylation suppression and Wts protein degradation. As a result, Yki phosphorylation is decreased and more non-phosphorylated Yki enters the nucleus and binds to the transcription factor Sd, activating the target genes, such as cycE and bantam expression and, in turn, promoting quiescent NSC reactivation. P: Phosphate; S: SUMO.