TDP2 negatively regulates axon regeneration by inducing SUMOylation of an Ets transcription factor

Abstract

In Caenorhabditis elegans, the JNK MAP kinase (MAPK) pathway is important for axon regeneration. The JNK pathway is activated by a signaling cascade consisting of the growth factor SVH‐1 and its receptor tyrosine kinase SVH‐2. Expression of the svh‐2 gene is induced by axonal injury in a process involving the transcription factors ETS‐4 and CEBP‐1. Here, we find that svh‐14/mxl‐1, a gene encoding a Max‐like transcription factor, is required for activation of svh‐2 expression in response to axonal injury. We show that MXL‐1 binds to and inhibits the function of TDPT‐1, a C. elegans homolog of mammalian tyrosyl‐DNA phosphodiesterase 2 [TDP2; also called Ets1‐associated protein II (EAPII)]. Deletion of tdpt‐1 suppresses the mxl‐1 defect, but not the ets‐4 defect, in axon regeneration. TDPT‐1 induces SUMOylation of ETS‐4, which inhibits ETS‐4 transcriptional activity, and MXL‐1 counteracts this effect. Thus, TDPT‐1 interacts with two different transcription factors in axon regeneration.

Introduction

A fundamental and conserved property of neurons is their ability to regenerate their axons following injury. Axon regeneration following injury is governed by an interaction between the local extracellular environment and the neuron's intrinsic growth capacity. Upon axon severance, regeneration signals are transported in a retrograde manner from the sites of damage to the nucleus, where several transcription factors are upregulated that increase the synthesis of proteins participating in neurite outgrowth 1. Since manipulation of these processes can improve the chances for successful axon regeneration, these signaling processes are potential targets for regeneration therapies. However, our understanding of these intrinsic signaling pathways is incomplete at present.

The nematode Caenorhabditis elegans is a useful model for investigating the mechanisms of axon regeneration following injury 2. Genetic studies in C. elegans have identified a large number of pathways specifically involved in adult axon regeneration 3. The C. elegans JNK MAP kinase (MAPK) pathway, consisting of MLK‐1(MAPKKK)–MEK‐1(MAPKK)–KGB‐1 (MAPK), plays a critical role in the initiation of axon regeneration 4. The JNK pathway is inactivated at the KGB‐1 activation step by VHP‐1, a member of the MAPK phosphatase family 5. The vhp‐1 null mutation causes hyper‐activation of the JNK pathway, resulting in the larval arrest of animals 5. We recently identified new components functioning in JNK‐mediated signaling by employing a genome‐wide RNAi screen for suppressors of vhp‐1 lethality (svh genes) 6. The svh‐1 gene encodes a growth factor‐like protein homologous to mammalian HGF, and the svh‐2 gene encodes a homolog of mammalian Met, a receptor for HGF 6. SVH‐2 is a receptor tyrosine kinase (RTK) that activates the JNK pathway via tyrosine phosphorylation of the MAPKKK MLK‐1. The svh‐2 gene is normally not expressed in motor neurons, but is induced in damaged motor neurons following axotomy 6. This axotomy‐induced expression of svh‐2 is regulated by ETS‐4 and CEBP‐1, an Ets‐like transcription factor and a C. elegans homolog of the CCAAT/enhancer‐binding protein (C/EBP), respectively 7. Therefore, the ets‐4 mutation is defective in axon injury‐induced svh‐2 expression and axon regeneration. When an axon is damaged, intracellular cAMP levels increase and cAMP‐dependent protein kinase (PKA) is activated. This in turn phosphorylates ETS‐4, causing the formation of a complex between ETS‐4 and CEBP‐1, which then activates svh‐2 transcription (Fig 1A) 7.

Figure 1. Identification of MXL‐1 and its involvement in axon regeneration.

1. Regulation of svh‐2 expression in response to axon injury. Axon injury initiates cAMP signaling and the Ca²⁺–p38 MAPK pathway, which together function to induce the formation of an ETS‐4–CEBP‐1 transcription factor complex. This complex binds to the svh‐2 promoter to induce svh‐2 expression. MXL‐1 forms a complex with MDL‐1.
2. Structure of MXL‐1. Schematic diagrams of MXL‐1 and its mammalian counterpart Max are shown. The domains (B‐HLH‐LZ) shown are the basic region (B; blue), helix‐loop‐helix (HLH; yellow), and leucine zipper (LZ; red). The bold line underneath indicates the extent of the deleted region in the tm1530 deletion mutant.
3. Representative D‐type motor neurons in wild‐type and mxl‐1 mutant animals 24 h after laser surgery. In wild‐type animals, a severed axon has regenerated a growth cone (arrow). In mutants, proximal ends of axons failed to regenerate (arrowhead). Scale bar = 10 μm.
4. Percentages of axons that initiated regeneration 24 h after laser surgery. The numbers (n) of animals examined are shown. Error bars indicate 95% confidence intervals (CI).

Data information: In (D), statistical significance was determined by Fisher's exact test; *P < 0.05; ***P < 0.001; NS, not significant.

In the present study, we investigated the role of the svh‐14/mxl‐1 gene in axon regeneration. The mxl‐1 gene encodes a Max‐like transcription factor, and we have discovered that it also acts as an activator of svh‐2 expression in response to axon injury and, accordingly, that the mxl‐1 mutation is defective in axon regeneration. Furthermore, we isolated TDPT‐1, a homolog of mammalian TDP2 (also called EAPII), as an MXL‐1‐binding protein. We show that TDPT‐1 inhibits the transcriptional activity of ETS‐4 by inducing its SUMOylation. Thus, TDPT‐1 acts as a negative regulator of axon injury‐induced svh‐2 expression. These results suggest that TDPT‐1 interacts with two different transcription factors and modulates their transcriptional functions in axon regeneration.

Results

SVH‐14/MXL‐1 is required for axon regeneration

We have previously performed a genome‐wide RNAi screen for suppressors of vhp‐1 lethality and isolated 92 svh RNAi clones that regulate the JNK signaling pathway 6, 8. In order to find new transcription factors involved in axon regeneration, we searched among the svh RNAi clones for genes encoding transcription factors. In this study, we investigated the svh‐14 gene, which encodes MXL‐1 9, a C. elegans homolog of the mammalian Max transcription factor (Fig 1A and B). Max belongs to a member of the basic region–helix‐loop‐helix–leucine zipper (B‐HLH‐LZ) protein family 10. Caenorhabditis elegans MXL‐1 exhibits similarity to human Max within its B‐HLH domain (Fig 1B).

To examine the effect of mxl‐1 on axon regeneration, we assayed the regrowth of laser‐severed axons in γ‐aminobutyric acid (GABA)‐releasing D‐type motor neurons (Fig 1C) 4. In wild‐type animals at the young adult stage, the regeneration of axons severed by laser was initiated within 24 h (Fig 1C and D, and Appendix Table S1). However, in mxl‐1(tm1530) deletion mutants (Fig 1B), the frequency of axon regeneration was reduced (Fig 1C and D, and Appendix Table S1). The mxl‐1(tm1530) mutation has no obvious effect on nerve development (Appendix Fig S1); therefore, the effect of mxl‐1 is injury‐specific. To test whether MXL‐1 can act in a cell‐autonomous manner, we expressed mxl‐1 from the unc‐25 or mec‐7 promoter in mxl‐1(tm1530) mutants. The mxl‐1 defect in axon regeneration was rescued by expression of mxl‐1 in D‐type motor neurons by the unc‐25 promoter but not by expression in sensory neurons by the mec‐7 promoter (Fig 1D and Appendix Table S1). Furthermore, expression of mxl‐1 in D neurons of mxl‐1(tm1530) mutants rescued regeneration to wild‐type levels (Fig 1D and Appendix Table S1). These results demonstrate that MXL‐1 acts to promote regeneration in a cell‐autonomous manner in the damaged neuron.

MXL‐1 is involved in axotomy‐induced svh‐2 expression in injured neurons

Since MXL‐1 is a transcription factor 9 and expression of svh‐2 is induced by axonal injury 6, we examined the genetic interaction between mxl‐1 and svh‐2. When we generated mxl‐1(tm1530); svh‐2(tm737) double mutants, we found that the regeneration defect of the double mutant was no greater than that of the mxl‐1 single mutant (Fig 2A and Appendix Table S1), suggesting that MXL‐1 and SVH‐2 act in the same pathway. Furthermore, we found that constitutive expression of svh‐2 from the Punc‐25::svh‐2 transgene was able to rescue the axon regeneration defect of mxl‐1(tm1530) mutants (Fig 2A and Appendix Table S1). This indicates that MXL‐1 acts upstream of SVH‐2 in the axon regeneration pathway. These results support the possibility that MXL‐1 is involved in the induction of svh‐2 expression in response to axon injury.

Figure 2. MXL‐1 is required for the transcriptional induction of the svh‐2 gene in response to axon injury.

1. Percentages of axons that initiated regeneration 24 h after laser surgery. The numbers (n) of animals examined are shown. Error bars indicate 95% CI.
2. Induction of Psvh‐2::nls::venus expression in D‐type motor neurons by laser surgery. Expression of fluorescent proteins in D‐type motor neurons of wild‐type, mxl‐1, tdpt‐1, and mxl‐1; tdpt‐1 animals 3 h after laser surgery is shown. Yellow arrowheads and white arrows indicate axon and cell bodies, respectively, of D‐type neurons. White arrows indicate the sites of laser surgery. D neurons are visualized by CFP under control of the unc‐25 promoter. Cell bodies of D‐type neurons are magnified and shown within the red boxes. Most of the intestinal fluorescence in these photos is from endogenous and variable background autofluorescence. A schematic representation of D‐type motor neurons is shown on the bottom. Scale bars = 10 μm.

Data information: In (A), statistical significance was determined by Fisher's exact test; ***P < 0.001; NS, not significant.

Next, we examined whether MXL‐1 regulates axon injury‐induced expression of svh‐2 using a reporter construct consisting of the svh‐2 promoter, a nuclear localization signal (NLS), and the fluorescent protein VENUS (Psvh‐2::nls::venus) 6. In wild‐type animals, axon injury induced expression of Psvh‐2::nls::venus in D‐type motor neurons (Figs 2B and EV1), as reported previously 6, 7; however, in mxl‐1(tm1530) mutants, no induction was observed (Figs 2B and EV1). These results suggest that MXL‐1 is required for axotomy‐induced svh‐2 expression.

Figure EV1. Induction of Psvh‐2::nls::venus expression in D‐type motor neurons by laser surgery.

• Percentages of animals expressing Psvh‐2::nls::venus in D‐type motor neurons with (+) or without (−) laser surgery are shown. The numbers (n) of animals examined are shown. Error bars indicate 95% CI.

Data information: Statistical significance was determined by Welch's t‐test; **P < 0.01; ***P < 0.001; NS, not significant.

MXL‐1 promotes axon regeneration by inhibiting TDPT‐1 function

In mammalian systems, the transcriptional regulatory function of Max depends on its heterodimerization partner, Myc or Mad 11. Caenorhabditis elegans does not have an obvious Myc ortholog but does contain a Mad‐like protein MDL‐1. Indeed, it has been shown that MDL‐1 physically associates with the MXL‐1 protein (Fig 1A) 9, 12 and that the MDL‐1–MXL‐1 complex acts as a heterodimeric transcription factor in the regulation of aging and proteostasis 13. Therefore, we investigated whether MDL‐1 might also be involved in axon regeneration. However, we found this was not the case: In contrast to the mxl‐1 mutation, the mdl‐1(tm311) null mutation had no effect on axon regeneration (Fig 3A and Appendix Table S1).

Figure 3. TDPT‐1 negatively regulates axon regeneration.

1. Percentages of axons that initiated regeneration 24 h after laser surgery. The numbers (n) of animals examined are shown. Error bars indicate 95% CI.
2. Interaction of MXL‐1 with TDPT‐1 by yeast two‐hybrid assays. The reporter yeast strain L40u was co‐transformed with expression vectors encoding LexA DBD‐MXL‐1 and GAL4 AD‐TDPT‐1 as indicated. Yeasts carrying the indicated plasmids were grown on a selective plate lacking histidine and containing 10 mM 5‐aminotriazole for 4 days.
3. Structure of TDPT‐1. Schematic diagrams of TDPT‐1 and human TDP2 are shown. The domains shown are the UBA (red) and tyrosyl‐DNA phosphodiesterase (TDP; green) domains. The bold line underneath indicates the extent of the deleted region (8 bp) in the km68 mutation. An asterisk indicates a premature stop codon in tdpt‐1 caused by the km68 deletion.
4. Percentages of axons that initiated regeneration 24 h after laser surgery. The numbers (n) of animals examined are shown. Error bars indicate 95% CI.

Data information: In (A and D), statistical significance was determined by Fisher's exact test; *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant.

To understand how MXL‐1 regulates svh‐2 expression in response to axon injury, we searched for additional MXL‐1‐binding proteins using a yeast two‐hybrid screen. From this screen, we isolated three genes, tdpt‐1, ubc‐9, and mdl‐1 (Figs 3B and EV2). The tdpt‐1 gene encodes a homolog of mammalian TDP2 (tyrosyl‐DNA phosphodiesterase 2; Fig 3C) 14. To ask whether TDPT‐1 is involved in axon regeneration, we used the CRISPR/Cas9 system to generate the mutant tdpt‐1(km68), which encodes a frameshift and generates a premature stop of the tdpt‐1 ORF (Fig 3C). To our surprise, we found that axon regeneration was significantly improved in tdpt‐1(km68) mutants compared to wild‐type animals (Fig 3A and Appendix Table S1). This result implies that in contrast to MXL‐1, TDPT‐1 negatively regulates axon regeneration. To confirm this possibility, we overexpressed TDPT‐1 in D‐type motor neurons of wild‐type animals using the unc‐25 promoter, and indeed found that axon regeneration was decreased (Fig 3D and Appendix Table S1). These results further suggest that TDPT‐1 inhibits axon regeneration cell‐autonomously.

Figure EV2. Effect of TDPT‐1 on the interaction between MXL‐1 and MDL‐1.

The reporter strain L40u was co‐transformed with expression vectors encoding LexA DBD‐MXL‐1, GAL4 AD‐MDL‐1, and TDPT‐1, as indicated. Yeasts carrying the indicated plasmids were grown on a selective plate lacking histidine (−His) or containing histidine (+His) and containing 10 mM 5‐aminotriazole for 4 days.

Since TDPT‐1 interacts with MXL‐1 but has the opposite effect on axonal regeneration, we next examined the genetic interaction between mxl‐1 and tdpt‐1. We found that the tdpt‐1(km68) mutation strongly suppressed the regeneration defect in mxl‐1(tm1530) mutants (Fig 3A and Appendix Table S1). Furthermore, expression of Psvh‐2::nls::venus in D‐type neurons was induced by laser surgery in mxl‐1(tm1530); tdpt‐1(km68) double mutants (Figs 2B and EV1). However, expression of Psvh‐2::nls::venus in D neurons of tdpt‐1(km68) single mutants was still inducible but not constitutive (Figs 2B and EV1). These results suggest that TDPT‐1 acts downstream of MXL‐1 to inhibit axon regeneration through repression of svh‐2 expression in injured neurons. This raises the possibility that MXL‐1 negatively regulates TDPT‐1 activity, thereby maintaining the regenerative capacity of injured neurons.

Since MXL‐1 associates with both MDL‐1 and TDPT‐1 (Figs 3B and EV2), we asked whether these interactions compete with each other. We found that co‐expression of TDPT‐1 decreased the interaction between MXL‐1 and MDL‐1 (Fig EV2). These results suggest that TDPT‐1 and MDL‐1 compete for binding to MXL‐1. Consistent with this, mdl‐1 could inhibit axon regeneration when overexpressed from the unc‐25 promoter in wild‐type animals, but not when overexpressed in tdpt‐1(km68) mutants (Fig 3D and Appendix Table S1). Furthermore, in contrast to tdpt‐1, the mdl‐1(tm311) mutation did not suppress the mxl‐1(tm1530) defect in axon regeneration (Fig 3A and Appendix Table S1). These results can be explained by assuming that overexpression of MDL‐1 and its interaction with MXL‐1 induces the dissociation of TDPT‐1 from MXL‐1, resulting in an increase free TDPT‐1, which in turn inhibits axon regeneration. Thus, it is likely that overexpression of MDL‐1 inhibits axon regeneration independent of its transcriptional activity.

To examine whether MDL‐1 is physiologically relevant for axon regeneration, we investigated the expression pattern of mdl‐1 using a reporter consisting of Pmdl‐1::nls::gfp. Consistent with the cell‐autonomous function of MDL‐1 in the inhibition of axon regeneration, we found that the mdl‐1 gene is expressed in D‐type motor neurons, and its expression was not affected by axon injury (Fig EV3A and B). We next examined the effect of axon injury on MDL‐1 protein levels in D neurons using the Punc‐25::mdl‐1::gfp gene. We observed that MDL‐1::GFP was localized in D neuron cell bodies (Fig EV3C). Following laser ablation of the axons, MDL‐1::GFP expression decreased in D neurons (Fig EV3C and D). Taken together, these results suggest that axon injury induces degradation of MDL‐1 resulting in the formation of free MXL‐1, which associates with TDPT‐1 and inhibits its negative effect on axon regeneration.

Figure EV3. Effects of axon injury on expression of mdl‐1 .

1. Expression of the Pmdl‐1::nls::gfp gene in ventral nerve neurons. Fluorescent images of animals carrying Pmdl‐1::nls::gfp and Punc‐47::mcherry are shown. D‐type motor neurons are visualized by mCherry under control of the unc‐47 promoter. Blue arrowheads indicate D‐type motor neurons. Anterior is the left. Scale bar = 10 μm.
2. Expression of Pmdl‐1::nls::gfp after axon injury. Fluorescent images of severed axons in animals carrying Pmdl‐1::nls::gfp and Punc‐47::mcherry are shown. Images were taken at 0 or 5 h after laser surgery. Yellow and blue arrowheads indicate cell bodies of injured and uninjured D neurons, respectively. Red arrowheads indicate the sites of laser surgery.
3. Expression of MDL‐1::GFP in D‐type motor neurons after axon injury. Fluorescent images of animals carrying Punc‐25::ndl‐1::gfp and Punc‐47::mcherry are shown. Images were taken at 0 or 6 h after laser surgery. Yellow and red arrowheads indicate cell bodies of injured D neurons and the sites of laser surgery, respectively.
4. Quantification of the relative fluorescent levels of MDL‐1::GFP in D neurons with (6 h) or without (0 h) laser surgery. The data are presented as a boxplot with median (thick line within the box) and interquartile range (edge of box). White circles indicate outliers.

Data information: In (D), statistical significance was determined by Wilcoxon rank sum test; *P < 0.05.

TDPT‐1 induces SUMOylation of ETS‐4

How does TDPT‐1 negatively regulate axon regeneration? We have previously demonstrated that the ETS‐4–CEBP‐1 complex activates svh‐2 expression in response to axon injury (Fig 4A) 7. Mammalian TDP2 was also independently identified as EAPII (Ets1‐associated protein II) because of its association with the Ets1 transcription factor and was shown to modulate the transcriptional activity of Ets1 15. Since C. elegans ETS‐4 belongs to the Ets transcription factor family 16, we examined the relationship between TDPT‐1 and ETS‐4. In contrast to MXL‐1, we found that the tdpt‐1(km68) mutation was unable to suppress the regeneration defect caused by the ets‐4(ok165) mutation (Fig 3A and Appendix Table S1), suggesting that ETS‐4 functions downstream of TDPT‐1 in the axon regeneration pathway. These findings are consistent with a model in which TDPT‐1 modulates ETS‐4 activity, similar to the mammalian EAPII/TDP2, and thereby inhibits axon regeneration.

Figure 4. TDPT‐1 induces SUMOylation of ETS‐4.

1. Relationship among MXL‐1, TDPT‐1, ETS‐4, CEBP‐1, and PKA in the regulation of axon injury‐induced svh‐2 expression.
2. Interaction between ETS‐4 and TDPT‐1. COS‐7 cells were transfected with plasmids encoding T7‐TDPT‐1, HA‐ETS‐4 (WT), HA‐ETS‐4(S73E) (SE), and HA‐ETS‐4(S73A) (SA), as indicated. Whole‐cell extracts (WCE) and immunoprecipitated complexes obtained with anti‐T7 antibody (IP: T7) were analyzed by immunoblotting (IB).
3. In vitro SUMOylation of ETS‐4. COS‐7 cells were transfected with plasmids encoding HA‐ETS‐4 (WT) and HA‐ETS‐4(K32A; K83A) (2KA), as indicated. Cell lysates were immunoprecipitated with anti‐HA antibody. The immunoprecipitates were subjected to in vitro SUMOylation assays. Reaction mixtures were analyzed by immunoblotting (IB) with anti‐SUMO‐2/3 and anti‐HA antibodies. Arrows mark the SUMOylated‐ETS‐4 bands.
4. Schematic representation of the two putative consensus SUMOylation sites and domain structure in ETS‐4. Domains are shown as follows: a pointed domain (PNT; yellow) and an Ets DNA‐binding domain (ETS; blue).
5. SUMOylation of ETS‐4 in mammalian cells. HEK293 cells were transfected with plasmids encoding HA‐ETS‐4 (WT), HA‐ETS‐4(K32A; K83A) (2KA), T7‐TDPT‐1, His‐SUMO‐1, His‐SUMO‐2, and MXL‐1‐Myc, as indicated. SUMO‐conjugated proteins were isolated by cobalt affinity chromatography. ETS‐4 was detected by immunoblotting with anti‐HA antibody. Whole‐cell extracts (WCE) were analyzed by immunoblotting (IB).

Next, we examined whether TDPT‐1 can physically interact with ETS‐4 by co‐immunoprecipitation in mammalian COS‐7 cells. T7‐tagged TDPT‐1 was co‐expressed with HA‐tagged ETS‐4 in COS‐7 cells, immunoprecipitated with anti‐T7 antibody, and HA‐ETS‐4 probed by immunoblotting with anti‐HA antibody. We found that ETS‐4 indeed co‐immunoprecipitated with TDPT‐1 (Fig 4B). Since ETS‐4 is phosphorylated by PKA at Ser‐73 7, we investigated whether PKA phosphorylation of ETS‐4 Ser‐73 would affect its interaction with TDPT‐1. We found that the phosphorylation‐negative mutant ETS‐4(S73A) and a phosphorylation‐mimicked form of ETS‐4, ETS‐4(S73E), associated with TDPT‐1 to a similar extent as wild‐type ETS‐4 (Fig 4B). These results suggest that ETS‐4 interacts with TDPT‐1 and that the phosphorylation state of ETS‐4 has no effect on this interaction.

Mammalian TDP2 interacts with SUMO‐2, a ubiquitin‐like modifier; UBC9, a SUMO E2 conjugating enzyme; and ZNF451, a SUMO E3/E4 ligase 17, 18. Therefore, we examined whether TDPT‐1 affects SUMOylation of ETS‐4. We first tested whether ETS‐4 is SUMOylated in vitro. We expressed HA‐ETS‐4 from mammalian COS‐7 cells and then incubated the immunopurified HA‐ETS‐4 with human SUMO‐2, E1, UBC9, and ATP. We found that HA‐ETS‐4 was efficiently SUMOylated in vitro (Fig 4C). To identify the SUMOylated residue, we analyzed the ETS‐4 protein sequence for potential SUMOylation sites. SUMOylation occurs on a lysine residue within the consensus motif φ‐Lys‐Xxx‐Glu (φ; large hydrophobic residue) 19. We found that ETS‐4 contains two consensus SUMOylation motifs: Val‐Lys(32)‐Ser‐Glu and Val‐Lys(83)‐Asn‐Glu (Fig 4D). Consistent with this, a mutant version of ETS‐4 in which both lysine residues were replaced with alanines, ETS‐4(K32A; K83A), showed no SUMOylation (Fig 4C).

We next examined whether TDPT‐1 enhances SUMOylation of ETS‐4. When T7‐TDPT‐1 was added in SUMOylation assay by its co‐expression with HA‐ETS‐4 in COS‐7 cells, we observed little enhancement of ETS‐4 SUMOylation in vitro by TDPT‐1 (Fig EV4). We therefore investigated the effect of TDPT‐1 on ETS‐4 SUMOylation in vivo of mammalian cells. To test whether ETS‐4 is SUMOylated in mammalian cells, we co‐expressed HA‐ETS‐4 and His‐tagged SUMO‐1 or SUMO‐2 with or without T7‐TDPT‐1 in HEK293 cells. SUMO‐modified proteins were purified from cell lysates by cobalt affinity chromatography, and SUMO‐ETS‐4 conjugates were analyzed by immunoblotting with anti‐HA antibody. When ETS‐4, SUMO‐2, and TDPT‐1 were co‐expressed, SUMO‐modified ETS‐4 proteins were detected (Fig 4E). No bands were observed in the absence of TDPT‐1 co‐expression. In addition, when His‐SUMO‐1 was co‐expressed, we did not detect SUMOylation of ETS‐4 (Fig 4E). When HA‐ETS‐4(K32A; K83A) was co‐expressed with His‐SUMO‐2 and T7‐TDPT‐1, no SUMOylation was observed (Fig 4E). These results indicate that TDPT‐1 expression induces SUMOylation of ETS‐4 at Lys‐32 and Lys‐83 in mammalian cells. Furthermore, we found that co‐expression of MXL‐1‐Myc inhibited TDPT‐1‐induced SUMOylation of HA‐ETS‐4 (Fig 4E). The fact that enhancement of ETS‐4 SUMOylation by TDPT‐1 was detected in mammalian cells but not in vitro suggests that TDPT‐1 expression induces SUMOylation of ETS‐4 by inhibiting its de‐SUMOylation.

Figure EV4. Effect of TDPT‐1 on in vitro SUMOylation of ETS‐4.

COS‐7 cells were transfected with plasmids encoding HA‐ETS‐4 and T7‐TDPT‐1, and cell lysates were immunoprecipitated with anti‐HA antibody. The immunoprecipitates were subjected to in vitro SUMOylation assays. Reaction mixtures were analyzed by immunoblotting (IB) with anti‐SUMO‐2/3, anti‐HA, and anti‐T7 antibodies. Arrows mark the SUMOylated‐ETS‐4 bands.

Next we examined the interplay between the phosphorylation and SUMOylation of ETS‐4. We co‐expressed T7‐TDPT‐1 and His‐SUMO‐2 with HA‐ETS‐4(S73A) or HA‐ETS‐4(S73E) in HEK293 cells. We found that both ETS‐4(S73A) and ETS‐4(S73E) were SUMOylated, although SUMOylation of ETS‐4(S73E) was slightly decreased (Fig 5A). Thus, the phosphorylation state of ETS‐4 S73 has little effect on SUMOylation. This is also consistent with the observation that phosphorylation of ETS‐4 S73 had no effect on the interaction between ETS‐4 and TDPT‐1 (Fig 4B). Next, we tested the effect of ETS‐4 SUMOylation on Ser‐73 phosphorylation. HA‐ETS‐4 was expressed in COS‐7 cells, and we performed in vitro kinase assays with active PKA and immunopurified HA‐ETS‐4. We confirmed that PKA could phosphorylate HA‐ETS‐4 but not HA‐ETS‐4(S73A) (Fig 5B). Thus, PKA phosphorylates the Ser‐73 residue of ETS‐4 in vitro. However, when we took immunopurified HA‐ETS‐4, subjected this to in vitro SUMOylation and then incubated it with active PKA, the non‐SUMOylated form of HA‐ETS‐4 was phosphorylated but the SUMOylated form was not (Fig 5B). These results suggest that SUMOylation of ETS‐4 inhibits its PKA‐mediated Ser‐73 phosphorylation.

Figure 5. The interplay between the phosphorylation and SUMOylation of ETS‐4.

1. Effect of phosphorylation of ETS‐4 on SUMOylation of ETS‐4 in mammalian cells. HEK293 cells were transfected with plasmids encoding HA‐ETS‐4 (WT), HA‐ETS‐4(K32A; K83A) (2KA), HA‐ETS‐4(S73A) (SA), HA‐ETS‐4(S73E) (SE), T7‐TDPT‐1, and His‐SUMO‐2, as indicated. SUMO‐conjugated proteins were isolated by cobalt affinity chromatography. ETS‐4 was detected by immunoblotting with anti‐HA antibody. Whole‐cell extracts (WCE) were analyzed by immunoblotting (IB).
2. Effect of SUMOylation of ETS‐4 on PKA phosphorylation of ETS‐4 in vitro. COS‐7 cells were transfected with plasmids encoding HA‐ETS‐4 (WT), HA‐ETS‐4(S73A) (SA), and HA‐ETS‐4(K32A; K83A) (2KA), as indicated. Cell lysates were immunoprecipitated with anti‐HA antibody. The immunoprecipitates were subjected to in vitro PKA kinase assays or in vitro SUMOylation and then in vitro PKA kinase assays, using recombinant GST‐fused active PKA, as indicated. Reaction mixtures were analyzed by immunoblotting (IB) with anti‐phospho‐PKA substrate, anti‐SUMO‐2/3, and anti‐HA antibodies. Arrows mark the SUMOylated‐ETS‐4 bands. Asterisk indicates the non‐SUMOylated‐ and phosphorylated‐ETS‐4 bands.

The effect of SUMOylation on ETS‐4 function in axon regeneration

If SUMOylation is important for the inhibition of ETS‐4 function, then mutations defective in SUMOylation might affect axon regeneration. UBC‐9 is an E2 SUMO conjugating enzyme that catalyzes the formation of an isopeptide bond between the C‐terminal glycine of SUMO and lysine 20. Caenorhabditis elegans contains the single gene smo‐1 encoding SUMO. We found that similar to the tdpt‐1 mutation, ubc‐9(tm2610) and smo‐1(ok359) mutations caused a significant improvement in axon regeneration compared to animals carrying the wild‐type genes (Fig 6A and Appendix Table S1). Overexpression of ets‐4 did not enhance axon regeneration in ubc‐9(tm2610) mutants (Fig 6A and Appendix Table S1). We next examined the genetic interaction between ubc‐9 and tdpt‐1. We found that overexpression of tdpt‐1 was unable to inhibit axon regeneration in ubc‐9(tm2610) mutants (Fig 6A and Appendix Table S1). Thus, the inhibitory effect of tdpt‐1 overexpression on axon regeneration is dependent on SUMOylation. These results suggest that TDPT‐1 functions upstream of UBC‐9‐mediated SUMOylation or inhibits de‐SUMOylation of ETS‐4. Furthermore, we found that ubc‐9(tm2610) and smo‐1(ok359) mutations suppressed the axon regeneration defect observed in mxl‐1(tm1530) mutants (Fig 6B and Appendix Table S1). In addition, the ets‐4(K32A; K83A) mutation also suppressed the mxl‐1(tm1530) defect (Fig 6C and Appendix Table S1). These results support a model in which SUMOylation of ETS‐4 is necessary for the inhibition of axon regeneration and MXL‐1 counters this by reducing TDPT‐1‐mediated SUMOylation of ETS‐4.

Figure 6. Effects of SUMOylation of ETS‐4 on axon regeneration.

• A–C
  Percentages of axons that initiated regeneration 24 h after laser surgery are shown. The numbers (n) of animals examined are shown. Error bars indicate 95% CI.

Data information: In (A–C), statistical significance was determined by Fisher's exact test; *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant.

Phosphorylation of ETS‐4 at Ser‐73 by PKA is required for complex formation with CEBP‐1, which activates transcription of the svh‐2 gene (Fig 4A) 7. If SUMOylation of ETS‐4 inhibits its phosphorylation at Ser‐73, and phosphorylation at Ser‐73 promotes axon regeneration, then the phospho‐mimetic ets‐4(S73E) mutation should suppress the mxl‐1 defect in axon regeneration. This would be similar to what was observed with the ets‐4(K32A; K83A) mutation, which is defective in ETS‐4 SUMOylation. As expected, we found that ets‐4(S73E) ameliorated the observed defect in axon regeneration in mxl‐1(tm1530) mutants (Fig 6C and Appendix Table S1). In contrast, the ets‐4(K32A; K83A; S73A) triple mutation, which is defective in both SUMOylation and phosphorylation, is unable to suppress the mxl‐1(tm1530) defect in axon regeneration (Fig 6C and Appendix Table S1). Thus, disruption of ETS‐4 SUMOylation does not bypass the requirement for phosphorylation at Ser‐73 in axon regeneration. These results suggest that ETS‐4 SUMOylation inhibits phosphorylation of ETS‐4 Ser‐73, resulting in the inhibition of its transcriptional activity. Although the ets‐4(K32A; K83A) and ets‐4(S73E) mutations were able to suppress the mxl‐1(tm1530) defect in axon regeneration (Fig 6C and Appendix Table S1), this suppression was not as strong as seen with the tdpt‐1(km68), ubc‐9(tm2610), or smo‐1(ok359) mutations (Figs 3A and 6B, and Appendix Table S1). Furthermore, expression of ets‐4(K32A; K83A) or ets‐4(S73E) in wild‐type animals did not increase the frequency of axon regeneration (Appendix Fig S2 and Table S1). These results suggest that TDPT‐1 may inhibit axon regeneration by inducing SUMOylation of another factor in addition to ETS‐4.

Discussion

In C. elegans, axon regeneration is activated by the SVH‐1 growth factor, its receptor SVH‐2, and the JNK signaling pathway 6. The svh‐2 gene is not expressed in D‐type motor neurons under normal conditions, but is induced in response to neuronal damage 6. This upregulation appears to critically involve the physical interaction of the Ets‐like factor ETS‐4 and the C/EBP‐like factor CEBP‐1, which in concert activate the svh‐2 promoter 7. Axon injury activates PKA, which phosphorylates ETS‐4, thereby promoting its association with CEBP‐1. This ETS‐4–CEBP‐1 complex binds to the promoter region of svh‐2 and activates its transcription. In this study, we identified Max‐like MXL‐1 as another transcription factor involved in axon injury‐induced expression of svh‐2. In mammals, Max regulates transcription of gene expression as part of a complex with Mad 11. Similarly, C. elegans MXL‐1 associates with the Mad‐like MDL‐1 9, 12. These two proteins need to form heterodimers in order to bind DNA, suggesting that each protein requires the other to effect its function. Indeed, the MDL‐1–MXL‐1 complex has been shown to have a role in longevity and proteostasis 13. However, in contrast to the mxl‐1 mutation, the mdl‐1 loss‐of‐function mutation has no effect on axon regeneration. Furthermore, MXL‐1 cannot form a homodimer and by itself is unable to bind to DNA 9. These results suggest that MXL‐1 may not regulate transcription of the svh‐2 gene by directly binding to the promoter region.

How does MXL‐1 regulate expression of svh‐2 in response to axon injury? This regulation seems to involve TDPT‐1, which we isolated as an MXL‐1‐binding protein. The tdpt‐1 deletion mutation is able to suppress the mxl‐1 defect but not the ets‐4 defect in axon regeneration, indicating that TDPT‐1 acts between MXL‐1 and ETS‐4 to negatively regulate axon regeneration. TDPT‐1 is homologous to mammalian TDP2, which was also independently identified as EAPII, a protein that binds to the Ets1 transcription factor 15. EAPII inhibits the transcriptional activity of Ets1. Similarly, C. elegans TDPT‐1 also associates with ETS‐4 and inhibits its ability to transactivate svh‐2 expression. These results suggest that MXL‐1 activates svh‐2 transcription in response to axon injury by interacting with TDPT‐1 and relieving its inhibition of ETS‐4 activity.

How does TDPT‐1 inhibit ETS‐4 transcriptional activity? Mammalian TDP2 is the first in a series of enzymes that removes adducts at the 5′‐phosphotyrosyl bond of DNA complexed with topoisomerases (TOPs) 14. Recently, it was shown that TDP2 is involved in the SUMOylation of TOP2 18. Consistent with this, we found that TDPT‐1 also induces SUMO modification of ETS‐4 when expressed in mammalian cells. Since TDP2 associates with SUMO‐modified TOP2 and stabilizes its SUMOylated state, it is likely that TDPT‐1 association with ETS‐4 inhibits its de‐SUMOylation. The transcriptional activity of mammalian Ets1 is also negatively regulated by SUMOylation 21. Furthermore, negative regulation by SUMOylation was also observed for the C. elegans factor LIN‐1, a member of the Ets transcription factor family 22, 23. Although SUMOylation promotes transcriptional repression in mammals and C. elegans, the molecular mechanism by which this occurs has not been well defined 24. Here, we show that SUMO modification of ETS‐4 negatively affects its transcription‐activating function. We observed that mutation of the SUMOylation sites in ETS‐4 or inhibition of the SUMOylation process by the smo‐1 and ubc‐9 mutations suppresses the mxl‐1 defect in axon regeneration. These results argue for an important role of SUMO modification in ETS‐4 transcriptional activity. This transcriptional activity requires the formation of a complex between ETS‐4 and CEBP‐1, which is promoted by phosphorylation of ETS‐4 by PKA 7. We show that SUMOylation of ETS‐4 inhibits PKA phosphorylation of ETS‐4. Furthermore, an ets‐4 mutation that mimics the phosphorylated state also suppresses the mxl‐1 defect in axon regeneration. These biochemical and genetic data raise the possibility that SUMO modification of ETS‐4 interferes with PKA‐mediated phosphorylation, thereby inhibiting formation of the ETS‐4–CEBP‐1 complex and preventing the transcriptional activation of svh‐2. Taken together, our analyses indicate that TDPT‐1 is a negative modulator of ETS‐4 activity that probably functions by inhibiting ETS‐4 interaction with its co‐activator CEBP‐1. Importantly, we found that aspects of these regulatory relationships are evolutionarily conserved: (i) Both TDP2 and TDPT‐1 inhibit the transcriptional activity of Ets family transcription factors; (ii) TDP2 and TDPT‐1 both modulate SUMOylation; and (iii) SUMOylation inactivates the activities of the Ets family transcription factors. TDP2 and TDPT‐1 may represent novel components in signal transduction that regulate eukaryotic gene expression in response to extracellular signals.

In this study, we show that MXL‐1 stimulates the transcriptional activity of ETS‐4 by inhibiting TDPT‐1‐mediated SUMOylation. Based on our findings, we propose the following molecular mechanism by which MXL‐1 regulates axonal regeneration in D‐type motor neurons after nerve injury (Fig 7). Under normal conditions, MXL‐1 forms a complex with MDL‐1, and TDPT‐1 interacts with ETS‐4 to induce its SUMOylation, resulting in the repression of ETS‐4 transcriptional activity. Axon injury leads to the degradation of MDL‐1, and free MXL‐1 can now interact with TDPT‐1. Next, ETS‐4 is released from TDPT‐1 and is de‐SUMOylated, allowing PKA to phosphorylate ETS‐4 and subsequently activate signaling. These findings thus expand our understanding of the functions of MXL‐1 and TDPT‐1 as key regulators of axon regeneration.

Figure 7.

Schematic diagram for the regulation of ETS ‐4 by TDPT ‐1 in axon regeneration.

Materials and Methods

C. elegans strains

The C. elegans strains used in this study are listed in Appendix Table S2. All strains were maintained on nematode growth medium (NGM) plates and fed with bacteria of the OP50 strain by the standard method 25.

Plasmids

Punc‐25::mxl‐1, Punc‐25::tdpt‐1, and Punc‐25::mdl‐1 were generated by inserting each cDNA isolated from a cDNA library into the pSC325 vector. Pmec‐7::mxl‐1 was generated by inserting the mxl‐1 cDNA into the pPD52.102 vector. Punc‐25::mdl‐1::gfp was constructed by inserting a GFP fragment into the Punc‐25::mdl‐1 plasmid by a standard Gibson assembly reaction using NEBuilder HiFi DNA Assembly Master Mix (NEB). Punc‐25::svh‐2, Punc‐25::ets‐4, Punc‐25::ets‐4(S73A), and Punc‐25::ets‐4(S73E) plasmids have been described previously 6, 7. Punc‐25::ets‐4(K32A; K83A) and Punc‐25::ets‐4(S73A; K32A; K83A) were made by oligonucleotide‐directed PCR using Punc‐25::ets‐4 and Punc‐25::ets‐4(S73A) as templates, respectively, and the mutations were verified by DNA sequencing. To make Pmdl‐1::nls::gfp, a 5 kbp DNA fragment just upstream of the start codon of the mdl‐1 gene was amplified from wild‐type genomic DNA by PCR and inserted into the pPD95.70 vector. Punc‐25::cfp and Psvh‐2::nls::venus plasmids have been described previously 7. The T7‐TDPT‐1, His‐SUMO‐1, and His‐SUMO‐2 plasmids were generated by inserting the tdpt‐1, SUMO‐1, and SUMO‐2 cDNAs into the pCMV‐T7 and pcDNA3.1/His B vectors, respectively. The HA‐ETS‐4, HA‐ETS‐4(S73A), and HA‐ETS‐4(S73E) plasmids have been described previously 7. The HA‐ETS‐4(K32A; K83A) was made by oligonucleotide‐directed PCR using HA‐ETS‐4 as a template, and the mutations were verified by DNA sequencing. The MXL‐1‐Myc construct was made by inserting the mxl‐1 cDNA into the pCMV‐Myc‐C vector (Clontech). The LexA DBD‐MXL‐1 plasmid was made by inserting the mxl‐1 cDNA into the pBTM116 vector. The GAL4 AD‐TDPT‐1 and GAL4 AD‐MDL‐1 plasmids were made by inserting the tdpt‐1 and mdl‐1 cDNAs into the pACTII vector, respectively 26. The pKT10‐TDPT‐1 plasmid was made by inserting the tdpt‐1 cDNA into the pKT10 vector 27.

Generation of the tdpt‐1 mutation using CRISPR–Cas9

The tdpt‐1(km68) deletion mutant was generated using the CRISPR–Cas9 system 28. The pU6::tdpt‐1_sgRNA was made by replacing the unc‐119 target sequence of pU6::unc‐119_sgRNA (Addgene) with 5′‐CGGGAGTGAGGCGTTCGCAG‐3′, corresponding to the genomic sequence within the tdpt‐1 gene. The pU6::tdpt‐1_sgRNA (50 ng/μl) was co‐injected together with the Peft‐3::cas9‐sv40_nls::tbb‐2 3′UTR (30 ng/μl) and Pmyo‐2::dsred‐monomer (25 ng/μl) plasmids into KU501 strain. An F1 animal carrying the transgene was picked, and genomic DNA from its descendants was subjected to a heteroduplex mobility assay 29 to detect the presence of short insertions or deletions in the tdpt‐1 gene. The descendants of these animals were selected to obtain the tdpt‐1 homozygous mutant. The tdpt‐1(km68) mutation is an 8‐bp deletion in the tdpt‐1 gene, causing a frameshift and premature stop codon in exon 3.

Transgenic animals

Transgenic animals were obtained by the standard C. elegans microinjection method 30. Pmyo‐2::dsred‐monomer (25 ng/μl), Punc‐25::mxl‐1 (50 ng/μl), Pmec‐7::mxl‐1 (50 ng/μl), Punc‐25::tdpt‐1 (50 ng/μl), Punc‐25::mdl‐1 (50 ng/μl), Punc‐25::ets‐4 (50 ng/μl), Punc‐25::ets‐4(K32A; K83A) (50 ng/μl), Punc‐25::ets‐4(S73E) (50 ng/μl), Punc‐25::cfp (50 ng/μl), Psvh‐2::nls::venus (75 ng/μl), Pmdl‐1::nls::gfp (25 ng/μl), and Punc‐25::mdl‐1::gfp (25 ng/μl) plasmids were used in kmEx1512 [Punc‐25::mxl‐1 + Pmyo‐2::dsred‐monomer], kmEx1513 [Pmec‐7::mxl‐1 + Pmyo‐2::dsred‐monomer], kmEx1515 [Punc‐25::tdpt‐1 + Pmyo‐2::dsred‐monomer], kmEx1517 [Punc‐25::mdl‐1 + Pmyo‐2::dsred‐monomer], kmEx1519 [Punc‐25::ets‐4 + Pmyo‐2::dsred‐monomer], kmEx1521 [Punc‐25::ets‐4(K32A; K83A) + Pmyo‐2::dsred‐monomer], kmEx1522 [Punc‐25::ets‐4(S73E) + Pmyo‐2::dsred‐monomer], kmEx1527 [Punc‐25::cfp + Psvh‐2::nls::venus + Pmyo‐2::dsred‐monomer], kmEx1531 [Pmdl‐1::nls::gfp + Pmyo‐2::dsred‐monomer], and kmEx1532 [Punc‐25::mdl‐1::gfp + Pmyo‐2::dsred‐monomer], respectively. The kmIs1523 [Punc‐25::ets‐4 + NeoR], kmIs1524 [Punc‐25::ets‐4(K32A; K83A) + NeoR], kmIs1525 [Punc‐25::ets‐4(S73E) + NeoR], and kmIs1526 [Punc‐25::ets‐4(S73A; K32A; K83A) + NeoR] integrated transgenes were made by the miniMos technique 31, using a NeoR plasmid carrying the unc‐25 promoter driving each ets‐4 cDNA and the unc‐54 3′UTR. The kmEx501 extrachromosomal and wpIs36 integrated arrays have been previously described 6, 32.

Microscopy

Standard fluorescent images of transgenic animals were observed under an ×100 objective of a Nikon ECLIPSE E800 fluorescent microscope and photographed with a Zyla CCD camera. Confocal fluorescent images were taken on Olympus FV‐500 or Zeiss LSM‐800 confocal laser‐scanning microscopes with an ×63 objective.

Axotomy

Axotomy and microscopy were performed as described previously 6. In brief, the hermaphrodite animals at the young adult stage were randomly selected and subjected to axotomy. Imaged commissures that had growth cones or small branches present on the proximal fragment were counted as “regenerated”. Proximal fragments that showed no change after 24 h were counted as “no regeneration”. A minimum of individuals with 1–3 axotomized commissures were observed for most experiments.

Quantification of VENUS expression

Expression of VENUS fluorescence was quantified using the ImageJ program (NIH). The cell bodies of severed D neurons were identified by the relative position of D neurons 33. They were outlined with closed polygons, and the fluorescent intensities within these areas were determined (I_s). As controls, the cell bodies of unsevered D neurons in the same animal were analyzed similarly (I_u). To determine the background intensity of each cell, the same polygon was placed in the area neighboring the cell body and fluorescence was measured (I_bs and I_bu, respectively). The relative signal intensity (I_r) normalized by unsevered D neuron was calculated as (I_s‐I_bs)/(I_u‐I_bu). Cells having an I_r > 5 were categorized as “expressed”.

Data acquisition and quantification of MDL‐1::GFP after axotomy

Z‐stack fluorescent images of animals expressing mCherry and MDL‐1::GFP in their D‐type motor neurons were obtained by confocal microscopy (Zeiss LSM‐800). Images were taken every 15 min for 5 h starting shortly after axotomy of some of the MDL‐1::GFP‐expressing neurons. The stability of MDL‐1::GFP after axotomy was assayed by calculating the relative intensity of MDL‐1 in the following way. First, an elliptical region of interest (ROI) roughly corresponding to the cell nucleus was determined for each cell expressing MDL‐1::GFP, and the mean fluorescent intensities of GFP and mCherry were measured for this ROI to obtain I_MDL‐1::GFP and I_mCherry, respectively. Next, the background was estimated by measuring the mean GFP (or mCherry) intensity of an adjacent region of the same size (I_{MDL‐1::GFP(BG),} I_mCherry(BG)). The ratio of background‐subtracted MDL‐1::GFP to mCherry intensity was calculated for two time points – immediately after axotomy: I_{MDL‐1::GFP/}I_mCherry (0) = (I_MDL‐1::GFP – I_GFP(BG))/(I_mCherry – I_mCherry(BG)) and 5 h after axotomy: I_{MDL‐1::GFP/}I_mCherry (5) = (I_MDL‐1::GFP – I_GFP(BG))/(I_mCherry – I_mCherry(BG)). Finally, the relative intensity of MDL‐1::GFP, RI_MDL‐1::GFP(5) at 5 h after axotomy was calculated as: RI_MDL‐1::GFP(5) = I_{MDL‐1::GFP/}I_mCherry (5)/I_{MDL‐1::GFP/}I_mCherry (0). RI_MDL‐1::GFP(5) values were calculated both for neurons with intact (n = 13) and severed axons (n = 16), and data were plotted using the R functions boxplot and stripchart in RStudio 35.

Yeast two‐hybrid screening and analysis

Yeast two‐hybrid screening using the LexA DBD‐MXL‐1 plasmid was performed as described previously 34. For yeast two‐hybrid analysis, GAL4 AD‐TDPT‐1 and pACTII plasmids were co‐transformed with either LexA DBD‐MXL‐1 or the pBTM116 vector into the Saccharomyces cerevisiae reporter strain L40u [MATa trp1 ura3 leu2 his3 LYS2::(lexAop) ₄ ‐HIS3] and allowed to grow on SC‐Leu‐Trp plates. Transformants grown on these plates were then streaked out onto SC‐Leu‐Trp‐His plates with 10 mM 5‐aminotriazole (AT) and incubated at 30°C for 4 days. The pKT10‐TDPT‐1 or pKT10 vector was further transformed into yeast carrying LexA DBD‐MXL‐1 and GAL4 AD‐MDL‐1. The transformants were selected on SC‐Leu‐Trp‐Ura plates, streaked onto SC‐Leu‐Trp‐Ura‐His (−His) or SC‐Leu‐Trp‐Ura (+His) plates with 10 mM AT, and incubated at 30°C for 4 days.

Immunoprecipitation

For immunoprecipitation, transfected COS‐7 cells were lysed in RIPA buffer [50 mM Tris–HCl, pH 7.4, 0.15 M NaCl, 0.25% deoxycholic acid, 1% NP‐40, 1 mM EDTA, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, phosphatase inhibitor cocktail 2 and 3 (Sigma), and protease inhibitor cocktail (Sigma)], followed by centrifugation at 15,000 × g for 12 min. The supernatant was added to 10 μl (bed volume) of Dynabeads Protein G (Invitrogen) with anti‐T7 antibody (PM022; MBL) and rotated for 2 h at 4°C. The beads were then washed three times with ice‐cold phosphate‐buffered saline (PBS) and subjected to immunoblotting.

Immunoblotting

After cell extracts were subjected to SDS–PAGE, proteins were transferred to a polyvinylidene difluoride membrane (Hybond‐P; GE Healthcare). The membranes were immunoblotted with anti‐HA antibody (mouse 16B12; BioLegend) or anti‐T7 antibody (mouse T7‐Tag; Merck; or rabbit PM022; MBL), and bound antibodies were visualized with horseradish peroxidase (HRP)‐conjugated antibodies against rabbit or mouse IgG using an HRP chemiluminescent substrate reagent kit (Novex ECL; Invitrogen).

In vitro SUMOylation assays

In vitro SUMOylation of ETS‐4 was carried out using the SUMOylation assay kit (Abcam) according to the manufacturer's instructions. Briefly, immunopurified HA‐ETS‐4 or HA‐ETS‐4(K32A; K83A) from transfected COS‐7 cells was suspended in reaction buffer (20 μl) containing SUMO activating enzyme (E1), UBC9 (E2), SUMO‐2, and ATP. Following 1 h of incubation at 37°C, reactions were terminated using Laemmli sample buffer and boiling. SUMOylated ETS‐4 was detected by immunoblotting with mouse anti‐SUMO‐2/3 antibody (ab139470; Abcam).

In vitro kinase assays

ETS‐4 proteins were immunopurified from transfected COS‐7 cells. SUMOylated ETS‐4 was prepared from the in vitro SUMOylation assay described above. Kinase reactions were performed in a final volume of 20 μl buffer consisting of 25 mM MOPS (pH 7.2), 12.5 mM glycerol phosphate, 25 mM MgCl₂, 2 mM EDTA, 0.25 mM DTT, 200 μM ATP, and 0.4 μg of recombinant PKA (Carna Biosciences). Samples were incubated for 20 min at 30°C and the reactions terminated by addition of Laemmli sample buffer and boiling. Phosphorylation of ETS‐4 was detected by immunoblotting with rabbit anti‐phospho‐PKA substrate antibody (100G7E; Cell Signaling).

Cobalt affinity pull‐down assay

His‐SUMO‐2‐ETS‐4 conjugates were purified by cobalt‐based immobilized metal affinity chromatography (Co‐IMAC). Transfected human HEK293 cells (mycoplasma‐free) were lysed in denaturing buffer containing 8 M urea, 20 mM imidazole, and 0.1% Triton X‐100 in phosphate‐buffered saline (PBS), followed by sonication and centrifugation at 15,000 × g for 5 min. The supernatant was added to 10 μl (bed volume) of Dynabeads His‐Tag Isolation & Pulldown beads (Invitrogen) and rotated for 1 h at 4°C. The beads were then washed three times with ice‐cold denaturing buffer and one time with ice‐cold PBS, followed by elution in 30 μl sample buffer (AE‐1430 EzApply; ATTO) supplemented with 300 mM imidazole.

Statistical analysis

Statistical analyses were carried out as described previously 6. Briefly, confidence intervals (95%) were calculated by the modified Wald method and two‐tailed P‐values were calculated using Fisher's exact test (http://www.graphpad.com/quickcalcs/contingency1/). The Welch t‐test and the Wilcoxon rank sum test (two‐tailed) were performed by using a t‐test calculator (http://www.graphpad.com/quickcalcs/ttest1/) and the R function wilcox.test(), respectively.

Author contributions

YS, HH, NH, and KM designed the experiments, analyzed the data, and wrote the manuscript. YS, HH, SIP, TS, CL, and NH performed experiments.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Review Process File

EMBO Reports (2019) 20: e47517