Caspase-mediated Cleavage of Hematopoietic Progenitor Kinase 1 (HPK1) Converts an Activator of NFκB into an Inhibitor of NFκB

Hematopoietic progenitor kinase 1 (HPK1), a mammalian Ste20-related protein kinase, is a potent stimulator of the stress-activated protein kinases (SAPKs/JNKs). Here we report activation of NFκB transcription factors by HPK1 that was independent of SAPK/JNK activation. Overexpression of a dominant-negative SEK1 significantly inhibited SAPK/JNK activation, whereas NFκB stimulation by HPK1 remained unaffected. Furthermore, activation of NFκB required the presence of full-length, kinase-active HPK1, whereas the isolated kinase domain of HPK1 was sufficient for activation of SAPK/JNK. We also demonstrate that overexpression of a dominant-negative IKKβ blocks HPK1-mediated NFκB activation suggesting that HPK1 acts upstream of the IκB kinase complex. In apoptotic myeloid progenitor cells HPK1 was cleaved at a DDVD motif resulting in the release of the kinase domain and a C-terminal part. Although expression of the isolated HPK1 kinase domain led to SAPK/JNK activation, the C-terminal part inhibited NFκB activation. This dominant-negative effect was not only restricted to HPK1-mediated but also to NIK- and tumor necrosis factor α-mediated NFκB activation, suggesting an impairment of the IκB kinase complex. Thus HPK1 activates both the SAPK/JNK and NFκB pathway in hematopoietic cells but is converted into an inhibitor of NFκB activation in apoptotic cells.

Constant turnover and the capacity to adapt efficiently to a changing environment are hallmarks of the hematopoietic system. Inflammatory cytokines like tumor necrosis factor ␣ (TNF␣) 1 and interleukin 1 trigger intracellular pathways resulting in the activation of the stress-activated protein kinases SAPKs/JNKs and p38s as well as of NFB family transcription factors (reviewed in Refs. 1 and 2). SAPKs are executing enzymes acting at the basal level of a hierarchical three-tiered kinase cascade (3), which upon activation enter the nucleus and phosphorylate nuclear transcription factors (4,5).
In mammals, two families of serine/threonine kinases have been identified that contain a catalytic domain with extensive homology to Sterile 20 (Ste20) kinase of the yeast Saccharomyces cerevisiae. Kinases prototypically represented by p21-activated kinase (PAK) are characterized by a C-terminal kinase domain and an N-terminal p21-binding domain, flanked by proline-rich sequences that serve as SH3 domain-binding sites (6). PAKs are activated by the GTP-bound forms of the small GTPases Rac/Cdc42 and have been implicated in regulation of cytoskeletal dynamics, cell cycle, and oxidant generation in neutrophils (7). The second family comprises kinases related to germinal center kinase (GCK), which are defined by a N-terminally located kinase domain. Based on homologies within their Cterminal domains GCKs can be grouped into six subfamilies. Four kinases, GCK, GCKR/KHS, GLK, and HPK1, which will be referred to as subfamily I, share a C-terminally located regulatory domain, called citron homology domain (CNH) (8,9). Of this group GCKR/KHS (10,11) and GLK (12) are ubiquitously expressed, whereas GCK (13) and HPK1 (14,15) display tissue specificity, with HPK1 being exclusively expressed in the hematopoietic cells of the adult. Upon overexpression subfamily I kinases are rendered active and activate potently and selectively the SAPK/JNK pathway via MAP3Ks. Kinase activity of endogenous GCK, GCKR/KHS, and GLK is stimulated in response to TNF␣ (11,12,16). In addition GCK and GCKR/ KHS have been shown to bind to the TNF receptor-associated factor 2 (TRAF2) (17,18). In contrast to GCK, GCKR/KHS, and GLK are responsive to UV light. HPK1, which already displays significant kinase activity when immunoprecipitated from nonstimulated tissues or cell lines, has been reported to be activated in response to erythropoietin receptor engagement (19) as well as T and B cell immunoreceptor cross-linking (20,21).
Four proline-rich stretches located between the kinase domain and CNH domain of HPK1 contain a PXXP motif, the minimal sequence requirement for SH3 domain ligands. Three of these have been shown to interact with small adaptor proteins including Grb2 (22), Nck (22), HS1 (19), and Crk (23,24), providing a possible link to activated transmembrane receptors. SH3 domain-mediated coupling to possible downstream MAP3Ks like mixed lineage kinase 3 (MLK3) has also been described (14). Despite their structural similarity, the SH3 domain ligand motifs are poorly conserved between subfamily I kinases. Therefore, subfamily I kinases appear to be subject to different regulatory mechanisms and most likely serve distinct physiological functions.
NFB/Rel proteins are dimeric, sequence-specific transcription factors that control many important biological processes, including development, immune responses, cell growth, and apoptosis. NFB family transcription factors are rendered inactive within the cytoplasm by interaction with IB inhibitory proteins. In response to extracellular signals, a high molecular weight IB kinase (IKK) complex is activated resulting in IB phosphorylation followed by ubiquitinylation and degradation. De-repressed NFB proteins translocate to the nucleus, where they bind and transactivate B sites within the promoter region of NFB-regulated genes (25).
Tissue-specific signaling molecules such as HPK1 are likely to provide specific inputs in ubiquitous transduction pathways and may function as signaling integrators or branch points. The increasing number of kinases that activate both SAPKs and NFB family transcription factors prompted us to investigate a possible function of HPK1 in NFB signaling. Here we report a robust stimulation of NFB activity by HPK1 in hematopoietic cells. In apoptotic cells HPK1 was cleaved by a caspase 3-like activity, resulting in the generation of a dominant-negative C-terminal fragment that inhibited NFB stimulation.

EXPERIMENTAL PROCEDURES
Generation of NFB Reporter and Expression Plasmids-The pGL8xNFB-fos reporter plasmid contains 8 repeats of the mouse major histocompatibility complex class I h2dk gene B site fused to the mouse c-fos minimal promoter driving a luciferase reporter gene. For normalization we generated pfos-LacZ that contains the identical mouse c-fos minimal promoter lacking the B-binding sites. pSP64T-HPK1(D383E):HA and pSP64T-HPK1(D383N):HA were generated by site-directed mutagenesis using the U.S.E. Mutagenesis Kit (Amersham Pharmacia Biotech) according to the manufacturer's protocol. pMT2-based HPK1 expression plasmids were described previously (14). For detection by immunoblotting T7-tagged versions of the GC family kinases GCK, GCKR, and GLK were generated in pCAT7. For pCAT7: FL:GLK an NcoI/Bsp120I fragment from pCR3.1-FL:GLK was inserted into the SmaI site of pCAT7-neo. For pCAT7:GCKR an EcoRI/SmaI fragment from pFLAG-GCKR was inserted into pCAT7-neo, and pCAT7:GCK was created by blunt insertion of an EcoRV/XbaI fragment from pRC/CMV-GCK into the EcoRI site of pCAT7-neo.
Transient Transfections, NFB Reporter Assays, and Retroviral Infection of FDC-P1 Cells-For reporter assays 2 ϫ 10 5 COS1 cells were transfected by the Ca 2ϩ -phosphate coprecipitation method with 2 g of expression plasmid, 1 g of pGL8x NFB-fos, and 0.2 g of pfos-LacZ.
After 48 h NFB activity was determined using the Tropix, Inc. Dual-Light TM Chemiluminescent Reporter Assay System following the manufacturer's protocol. All experiments were at least repeated three times. Bars depict the average of duplicate transfections. Values presented within a diagram are derived from a single transfection experiment.
Jurkat cells were transiently transfected by electroporation of 20 g of NFB luciferase reporter plasmid, 2 g of ␤-galactosidase reporter construct, and 0 -10 g of wild-type or kinase-inactive HPK1 as indicated in 400 l of serum-free RPMI 1640 medium.
FDC-P1 cells were infected using transient supernatants of the ecotropic packaging cell line GPϩE 86 transfected with 10 g of retroviral vector. After 48 h successfully infected cells were selected by a 12-day selection period at 1 mg/ml G418 (Life Technologies, Inc.).
In Vitro Kinase Assays, SAPK Activation, and Immunoblotting-HPK1 and p54-SAPK␤ kinase assays were performed as described previously (14). For Western blotting the polyclonal rabbit anti-HPK1 sera 3, 5, 6, and 7 were used. Sera 5 and 6 have been described previously (14), and serum 3 was raised against the peptide KSGYQPPRLKEKSRWSSSC located in sub-domain X of the HPK1 kinase domain and serum 7 against the Cterminal peptide TRPTDDPTAPSNLYIQE. The HA tag was detected using the mouse monoclonal antibody 12CA5. A phosphospecific antibody against serine 32 of IB␣ was obtained from Cell Signaling Technology, the IB␣ (antibody 2/PC142) antibody from Oncogene Research Products. The T7 tag was detected by the monoclonal antibody T7⅐Tag (Novagen).
Generation of Apoptotic HL60 Cell Extracts and in Vitro Cleavage of HPK1-Cytoplasmic extracts from apoptotic HL60 cells were prepared as described (29,30). Lysates were centrifuged for 50 min at 100,000 ϫ g, and the resulting clear supernatant was stored in aliquots at Ϫ80°C until further usage. The presence of a caspase 3-like activity in the cytoplasmic extract was verified by cleavage of Ac-DEVD-7-amino-4methylcoumarin substrate peptide.
In vitro translated HPK1:HA, HPK1(D383E):HA, and HPK1(D383N):HA were generated as described (14). 2 l of [ 35 S]methionine-labeled reaction product were mixed with 5 l of apoptotic HL60 cell extract and incubated at 37°C for the indicated times. Caspase inhibitors (Peptide Institute, Osaka, Japan) were added in Me 2 SO at a final concentration of 0.1 M (Ac-DEVD-CHO) and 1.0 M (Ac-YVAD-CHO).

HPK1 Activates the NFB Pathway in Hematopoietic Cell
Lines-We tested the capacity of HPK1 to activate NFB in Jurkat T cells, which endogenously express HPK1. Cotransfection of increasing amounts of HPK1 expression plasmid and a NFB-dependent reporter resulted in HPK1-dependent NFB activation ( Fig. 1A), whereas expression of the kinase-deficient mutant HPK1(K46E) failed to activate NFB. These results suggest that HPK1 is an activator of NFB in Jurkat T cells.
To investigate HPK1-dependent NFB activation in a myeloid progenitor cell line, FDC-P1 cells (26) were used, which also endogenously express HPK1. We demonstrated HPK1-dependent NFB activation taking advantage of two retrovirally transduced FDC-P1 clones FDC-P1/C9 and FDC-P1/D4 that stably express an exogenous, HA-tagged variant of HPK1. FDC-P1/D4 cells harbor approximately double the amount of HPK1 kinase activity present in FDC-P1/C9 cells (Fig. 1B). FDC-P1/C9 cells contain about double the amount of HPK1 kinase activity of FDC-P1 wild-type cells. Comparing nuclear extracts derived from FDC-P1 wild-type cells and the cell clones C9 and D4, we observed an increase in NFB bandshift activity that paralleled HPK1 kinase expression levels. These results provide an independent demonstration of HPK1-mediated NFB transcription factor activation.
Activation of NFB Depends on Full-length Kinase-active HPK1-To analyze the HPK1-driven NFB activation in more detail and to circumvent the presence of endogenous HPK1, we used COS1 cells for further analysis. For subsequent transfection experiments we generated a double reporter gene system, in which NFB-dependent transcription of a luciferase reporter Equal amounts of nuclear extracts prepared from 4 ϫ 10 6 cells were incubated with a 32 P-labeled DNA fragment comprising an NFB-binding site. Nucleoprotein complexes were separated by native PAGE, visualized by autoradiography, and quantified using image analysis software. Middle panel, expression of endogenous and HA-tagged HPK1 in FDC-P1/WT, FDC-P1/C9, and FDC-P1/D4 cells was assessed by anti-HPK1 Western blotting using rabbit serum 7. Lower panel, endogenous and HA-tagged HPK1 was immunopurified from FDC-P1/WT, FDC-P1/C9, and FDC-P1/D4 cells using anti-HPK1 antisera 5/6 and tested for the ability to autophosphorylate in vitro. Electrophoretic mobilities of endogenous and HA-tagged HPK1 are indicated by arrows. C, dual reporter plasmid system used to assess activation of NFB in transiently transfected COS1 cells. Transcription of a luciferase reporter gene driven by 8 NFB-binding sites, fused to a c-fos minimal promoter, was normalized on the basis of ␤-galactosidase expression from an identical c-fos promoter lacking NFB-binding sites. D, upper panel, COS1 cells were transiently cotransfected with plasmids encoding HPK1, the kinase-deficient version HPK1(K46E), or the C-terminal deletion mutant HPK1-Ko in the presence of the double reporter system depicted in C. After 36 h control cells were stimulated with 2 nM recombinant hTNF␣ and incubated for additional 12 h followed by cell lysis. After 48 h NFB-driven luciferase activity was determined and normalized against ␤-galactosidase activity using a chemiluminescence assay system. Relative activation of NFB-driven luciferase activity normalized for transfection efficacy is shown. Depicted are averages of a representative experiment, in which all transfections were performed in duplicate. Middle panel, in parallel identical amounts of the same plasmids were assayed for p54-SAPK␤ activation. After 48 h cells were lysed, and p54-SAPK␤ was immunopurified. SAPK/JNK activation was determined by in vitro phosphorylation of a bacterially expressed c-Jun N-terminal fragment fused to GST. Phosphoproteins were separated on SDS-PAGE and visualized by autoradiography. Lower two panels, expression of the different proteins was visualized by Western blotting using the polyclonal anti-HPK1 rabbit serum 3 (directed against kinase domain subdomain XI) or 7 (directed against the HPK1 C terminus). E, upper panel, increasing amounts of expression plasmids for the indicated GCK-related kinases or empty vector DNA were cotransfected into COS1 cells. The total amount of transfected DNA was kept constant. Lower panel, expression of the different proteins was visualized by Western blotting using the polyclonal anti-HPK1 rabbit serum 7 or anti-T7 tag antibody. F, identical amounts of expression plasmids for the GCK-related kinases were cotransfected in COS1 cells and assayed for their ability to undergo autophosphorylation (upper panel) or activate an HA-tagged p54-SAPK␤ (middle panel). p54-SAPK␤ expression levels were visualized by anti-HA Western blotting (lower panel).
gene was normalized against basal transcription of a galactosidase reporter gene (Fig. 1C). To delineate the requirements for NFB activation, we transiently expressed wild-type HPK1, the kinase-deficient variant HPK1(K46E), or the isolated kinase domain HPK1-Ko in COS1 cells. NFB activity levels were compared with the activity level observed after treatment with the inflammatory cytokine TNF␣, a well established inducer of NFB (Fig. 1D). HPK1 caused a robust activation of NFB, comparable to TNF␣ stimulation, whereas HPK1(K46E) was not able to activate NFB. Interestingly, HPK1-Ko failed to stimulate NFB, although it still activated the SAPK/JNK pathway.
These results confirm our observation in Jurkat T cells dem-onstrating again that HPK1 kinase activity is essential for NFB activation. Furthermore, NFB activation required the presence of full-length HPK1, whereas the C-terminal regions of HPK1 are dispensable for SAPK/JNK activation. The Ste20-related Kinases of the GC Family Show Variable Capacities to Activate NFB-We next addressed the question whether the potential to activate NFB is shared by other members of subfamily I GCK-related kinases, besides HPK1. After transient expression of increasing amounts of HPK1, GLK, GCKR/KHS, or GCK in COS1 cells, we found a dose-dependent and profound activation of NFB by HPK1 closely followed by GCKR/KHS (Fig. 1E). Moderate NFB activation by GCK was only seen at the highest expression level, whereas FIG. 2. HPK1-mediated NFB activation is independent of its SAPK/JNK activation and involves the IKK complex. A, COS1 cells were transiently cotransfected with expression plasmids for HPK1, MEKK1, and MLK3, either alone or in combination. 48 h after transfection NFB activation was determined. Lower panel, in parallel, identical amounts of the indicated plasmids were cotransfected with a p54-SAPK␤ expression plasmid, and p54-SAPK␤ kinase activity was assayed. B, expression plasmids for MEKK1, HPK1, and SEK1(AL) were transfected into COS1 cells alone or in combination, and their capacity to activate NFB was determined. Lower panel, the same plasmids were tested for their capacity to activate p54-SAPK␤. Relative increase of p54-SAPK␤ activity (numerical values) was determined using a phosphorimaging system. C, the indicated expression plasmids were transiently transfected into COS1 cells, and after 48 h NFB activation was determined. Inset, expression of HPK1 was visualized by Western blotting using the polyclonal anti-HPK1 rabbit serum 7.

FIG. 3. NFB activation is dependent on the proline-rich SH3-binding sites in HPK1.
A, upper panel, COS1 cells were transiently cotransfected with expression plasmids for HPK1 or HA-tagged MLK3⌬ alone or in combinations. 48 h after transfection NFB activation was determined. Lower two panels, expression of the proteins was visualized by Western blotting using the polyclonal anti-HPK1 rabbit serum 7 or anti-HA antibody. B, the indicated expression plasmids were transiently transfected into COS1 cells, and after 48 h NFB activation was determined. Lower panel, expression of the HA-tagged forms of HPK1 and Grb2 was visualized by anti-HA Western blotting. C, HPK1 or HA-tagged deletion mutants lacking the indicated SH3 domain-binding motifs and the NFB reporter system were cotransfected into COS1 cells. After 48 h NFB activation was determined. Expression of HPK1 proteins was demonstrated by anti-HPK1 Western blotting using rabbit serum 7. Localization of the proline-rich sites (P1, P2, and P4) and the amino acid sequence DDVD on the HPK1 protein is depicted (inset). D, HPK1 polyproline stretch deletion mutants are not impaired in their ability to activate p54-SAPK␤. Top panel, the HPK1 mutants assayed in C were transiently coexpressed in COS1 cells with p54-SAPK␤, immunopurified, and tested for their ability to autophosphorylate in vitro. Middle panel, equal expression was demonstrated by anti-HA Western blotting. Bottom panel, p54-SAPK␤ activation was determined as described in Fig. 1D. NFB activation by GLK was largely blunted.
At expression levels that resulted in comparable levels of autophosphorylation activity (Fig. 1F, upper panel), the capacity of the GCK-related kinases to activate NFB correlated with their ability to activate the SAPK/JNK p54␤ (Fig. 1F,  middle panel).
These experiments clearly demonstrate that within the GCK subfamily I of Ste20 kinases HPK1 and the most closely related GCKR/KHS are potent activators of NFB.
HPK1-mediated NFB Activation Is Independent of HPK1mediated SAPK Activation-We wondered whether NFB activation by HPK1 was secondary to HPK1-mediated SAPK/ JNK activation or caused by a component of the SAPK/JNK pathway. Three MAP3Ks, MEKK1 (15), MLK3 (14), and TAK1 (33), have been implicated as potential downstream elements of HPK1 in SAPK/JNK activation. In our assay system HPK1 and the established NFB activator MEKK1 (34, 35) displayed comparable potency in activating NFB, whereas MLK3 failed to activate NFB ( Fig. 2A, upper panel). By using identical conditions HPK1, MEKK1, and MLK3 all activate the SAPK/JNK p54␤ to a comparable extent ( Fig. 2A, lower panel). Furthermore, we detected no synergism between HPK1 and MEKK1 in NFB nor in SAPK/JNK p54␤ activation. Surprisingly, we found that coexpression of MLK3 potently inhibits HPK1-mediated NFB activation, whereas it had no influence on SAPK/ JNK p54␤ activation, suggesting that NFB activation by HPK1 does not involve MLK3.
To demonstrate formally that HPK1-mediated NFB and SAPK/JNK activation utilize distinct effector pathways, we took advantage of a mutant form of SEK1/MKK4, SEK1(S220A,T224L), which we will refer to as SEK1(AL). Mutations of the critical activation loop residues Ser-220 and Thr-224 render SEK1 refractory to upstream activating kinases and turn it into a potent dominant-negative inhibitor of SAPK/JNK activation at the MAP2K level. SEK1(AL) potently inhibits SAPK/JNK activation in a dominant-negative fashion at the MAP2K level (36). Whereas SEK1(AL) caused no change in HPK1 or MEKK1-induced NFB activation (Fig. 2B, upper  panel), SAPK/JNK activation was profoundly inhibited (Fig.  2B, lower panel), demonstrating that HPK1-mediated NFB activation is independent of HPK1-mediated SAPK/JNK activation.
Kinase-deficient IB Kinase ␤ (IKK␤) Abrogates HPK1-mediated Activation of NFB-To define further the level of HPK1 action and to test whether IKK functions downstream of HPK1, we coexpressed HPK1 and dominant-negative forms of NIK and IKK␤, NIK(KK429,430AA), and IKK␤(K44A). The dominant-negative variants of NIK and IKK␤ both inhibited TNF␣stimulated as well as HPK1-mediated NFB activation (Fig.  2C). Kinase-active NIK displayed an NFB activation potential comparable to that of HPK1, whereas no synergy between both kinases was detectable. Blockage of HPK1 signaling to NFB by overexpression of dominant-negative NIK or IKK␤ protein suggested that HPK1 might act upstream of the IKK complex.
HPK1 kinase activity and phosphorylation status were not responsive to TNF␣, neither did overexpression of HPK1 augment TNF␣-mediated NFB activation (not shown). Taken together these findings argue against a direct involvement of HPK1 in TNF␣ signaling.

HPK1-mediated NFB Activation Is Blocked by Overexpression of SH3 Domain-containing Molecules-
The surprising finding that MLK3 potently inhibited HPK1-mediated NFB activation ( Fig. 2A) led us to speculate that the SH3 domaindriven interaction between MLK3 and HPK1 (14) could result in sequestration of HPK1 from an NFB-activating complex. To test this hypothesis we coexpressed MLK3⌬, a truncation mutant of MLK3 consisting of the N terminus, which includes the SH3 domain and 21 amino acids of the adjacent kinase domain (14) with HPK1 (Fig. 3A). According to our hypothesis we detected a potent suppression of HPK1-driven NFB activation. Therefore we reasoned that other HPK1-binding SH3 domain-bearing molecules should also interfere with NFB activation. The small adaptor Grb2, which consists of a central SH2 domain flanked by SH3 domains, has been shown to associate with HPK1 (22). Coexpression of HPK1 and Grb2 resulted in a strong inhibition of HPK1-mediated NFB activation, whereas MEKK1-mediated NFB activation was not affected (Fig. 3B). These results lend further support to our notion that SH3 domain interactions are likely critically involved in HPK1-mediated NFB activation.
NFB Activation Is Dependent on the Proline-rich SH3-binding Sites in HPK1-Nonspecific inhibition of HPK1-mediated NFB activation could be excluded by cotransfection of an SH3 domain-containing molecule that does not bind to HPK1. Because of the possibility of residual interactions between overexpressed SH3 domains and HPK1, we decided to generate HPK1 mutants, in which the three proline-rich SH3 domainbinding sites (P1, P2, and P4) were deleted either singularly or in combination. When we tested the HPK1 proline deletions for their ability to activate NFB, none of them displayed an activity comparable to wild-type HPK1 (Fig. 3C). The mutations did not impair protein stability, as they did not decrease HPK1-associated kinase activity or the capacity of HPK1 to activate SAPK/JNK (Fig. 3D). These results demonstrate a strong dependence of NFB activation on the proline-rich SH3binding sites in HPK1, whereas those sites were dispensable FIG. 5. The isolated C-terminal portion of HPK1 inhibits NFB activation. A, top panel, HPK1, the isolated kinase domain (HPK1-Ko) or the C-terminal portion (HPK-⌬N) were transiently expressed alone or in combination and tested for their ability to activate NFB in COS1 cells. 2nd panel from top, the HPK1 proteins assayed in A were transiently coexpressed in COS1 cells with p54-SAPK␤, immunopurified, and tested for their ability to phosphorylate a c-Jun N-terminal fragment fused to GST in vitro. Lower two panels, expression of the different proteins was visualized by anti-HPK1 Western blotting using rabbit serum 3 or 7. B, increasing amounts of HPK1-⌬N cDNA were expressed either alone or in combination with NIK and the NFB reporter system in COS1 cells. Luciferase activity was measured as described above. Inset, expression of HPK-⌬N was visualized by anti-HPK1 Western blotting using rabbit serum 7. C, FDC-P1 myeloid progenitor cell pools infected with either the empty retroviral vector MSCV (FDC-P1/vector only) or a virus transducing HPK1-⌬N (FDC-P1/HPK1-⌬N) were stimulated with 2 nM recombinant hTNF␣ and incubated for 2, 5, or 10 min followed by cell lysis. Endogenous phospho-IB␣ was visualized by Western blotting using a phospho-IB␣ (Ser-32)-specific antibody (upper panel). Total levels of IB␣ and HPK1 proteins were visualized by anti-IB␣ and by anti-HPK1 (rabbit serum 7) Western blotting.
for SAPK/JNK activation. Furthermore, we found a C-terminally HA-tagged version of HPK1 to be less efficient in activating NFB as compared with the native protein (Fig. 3C), indicating a critical role of the HPK1 C terminus in NFB activation.
Taken together our findings demonstrate that SH3 domainmediated interactions are a prerequisite for NFB activation and that HPK1-mediated SAPK/JNK and NFB activation differ significantly in their molecular requirements.
HPK1 Is Proteolytically Degraded in FDC-P1 Cells Rendered Apoptotic by IL-3 Withdrawal-NFB target genes have been implicated in a plethora of pro-and anti-apoptotic processes, and a number of NFB regulators have been shown to be subject to caspase cleavage in apoptotic cells. Growth and survival of the hematopoietic progenitor cell line FDC-P1 is strictly IL-3-dependent (37). We tested HPK1 stability in apoptotic FDC-P1/D4 hematopoietic progenitor cells after induction of apoptosis by IL-3 withdrawal for 18 h. At this time point DNA fragmentation, a hallmark of the apoptotic cell death program, was maximal (not shown). By using Western blot analysis, we detected reduced HPK1 levels after IL-3 deprivation and observed the appearance of HPK1 cleavage products (Fig. 4A). These results were reproduced under identical conditions using the IL-3-dependent myeloid progenitor cell line 32D-Cl3 indicating that HPK1 is proteolytically degraded in hematopoietic cells rendered apoptotic after growth factor withdrawal.

The Hinge Region between the HPK1 Kinase and the Citron Homology Domain Contains a Caspase Recognition Motif-
Caspases, the effector proteases during apoptosis (38), display overlapping substrate specificity, with the four N-terminal amino acids preceding their cleavage site being most important for substrate recognition and turnover. An Asp residue at the position N-terminally flanking the cleavage site is indispensable for all caspases. Caspase 3/7 cleavage sites are defined by a DXD motif, where X denotes a wide variety of amino acids and a hydrophobic amino acid (39,40). Inspection of the primary sequence of the HPK1 hinge region between kinase and CNH domain, which also contains the proline-rich motifs, revealed several potential caspase recognition sites with a DDVD motif immediately preceding the P2 proline-rich stretch being most prominent (Fig. 4B).

HPK1 Is Efficiently Cleaved by Apoptotic Cell Extracts in Vitro-
To address the question, if the HPK1 DDVD motif is recognized by caspases, we first employed an in vitro test system, in which [ 35 S]methionine-labeled HPK1 generated by in vitro translation was exposed to cytoplasmic extracts of apoptotic HL60 cells. Such extracts contain abundant activated caspases. During a 60-min incubation HPK1 was efficiently cleaved into two fragments, the size of which correlated well with the calculated fragment sizes of 43 and 48 kDa. Heat pretreatment of the apoptotic extracts completely abolished proteolytic degradation (Fig. 4C, top panel). HPK1 Cleavage Is Blocked by the Caspase 3/7 Inhibitor Ac-DEVD-CHO and Depends on a DDVD Motif-Caspases 3 and 7 can be blocked by incubation with the inhibitory peptide Ac-DEVD-CHO, whereas the Ac-YVAD-CHO inhibitor blocks caspases 1 and 4 preferentially. When HPK1 was incubated with apoptotic HL60 extracts in the presence of Ac-DEVD-CHO, proteolytic cleavage was efficiently blocked, whereas the Ac-YVAD-CHO inhibitor showed no effect (Fig. 4C, middle  panel). Addition of the vehicle Me 2 SO alone had no inhibiting effect on HPK1 cleavage (Fig. 4C, top panel).
Two point mutants HPK1(D383N) and HPK1(D383E), in which the DDVD motif was either changed to DDVN or DDVE, were found to be completely cleavage-resistant (Fig. 4C, bottom   panel). These data identify the DDVD motif in the hinge region as a relevant target site for HPK1 cleavage by apoptotic extracts in vitro.
Our data indicated that in apoptotic myeloid progenitor cells proteolytic cleavage of HPK1 occurs after growth factor withdrawal.
The Isolated HPK1 C Terminus Suppresses NFB Activation by HPK1-To address possible implications of HPK1 proteolytic cleavage during apoptosis, we tested the signaling capacity of the isolated HPK1 kinase domain (HPK1-Ko) or the HPK1 C-terminal fragment (HPK1-⌬N). HPK1-⌬N comprises the proline-rich motifs P2 and P4 as well as the CNH domain. Whereas expression of full-length HPK1 caused robust stimulation of NFB-mediated transcription, none of the two fragments caused a detectable activity (Fig. 5A, upper panel). In agreement with our previous results the intact kinase domain HPK1-Ko was necessary and sufficient for SAPK/JNK activation (14), whereas HPK1-⌬N failed to activate SAPK/JNK (Fig.  5A, 2nd panel from top). We did not detect an augmented kinase activity of HPK1-Ko as a result of the removal of the C-terminal part (not shown).
When we coexpressed HPK1 fragments in combination with full-length HPK1, HPK1-⌬N inhibited HPK1-induced NFB stimulation, whereas HPK1-Ko failed to exert an effect. Under identical conditions activation of SAPK/JNK by full-length HPK1 was not altered.
These data indicate that the C-terminal part of HPK1 was capable of inhibiting HPK1-driven NFB activation, although it did not alter HPK1-mediated SAPK/JNK activation.
HPK1-⌬N Inhibits NFB Activation by NIK-To investigate whether the C-terminal part of HPK1 was able to inhibit NFB activation by stimuli other than HPK1 itself, we tested a potential effect on NIK that displayed an NFB activation capacity comparable to HPK1 (Fig. 2C). Coexpression of increasing amounts of NIK and HPK1-⌬N resulted in a dramatic dose-dependent decrease in NFB activation (Fig. 5B). This observation clearly showed that a C-terminal fragment of HPK1, like HPK1-⌬N, is a potent inhibitor of NFB activation.
HPK1-⌬N Reduces Phosphorylation of IB␣ in FDC-P1 Cells-To assess the capacity of HPK1 cleavage products to act as inhibitors of NFB activation at physiological levels, we derived retrovirally infected FDC-P1 cells that stably expressed the C-terminal fragment HPK1-⌬N. Selected clones were pooled to avoid effects due to clonal variation. These clones expressed endogenous full-length HPK1 and HPK1-⌬N at a similar ratio to that observed in FDC-P1/D4 cells rendered apoptotic by growth factor withdrawal (see Fig. 4A and Fig. 5C, bottom panel). To test for a possible impact of HPK1-⌬N on NFB activation, we compared the appearance of IB␣ phosphorylated on serine 32 in FDC-P1/HPK1⌬N cells and FDC-P1 cells infected with the parental retroviral vector pMSCV (FDC-P1/vector only) after application of TNF␣.
In FDC-P1/vector only cells IB␣ phosphorylation sharply peaked at 5 min after stimulation and was barely detectable after 10 min (Fig. 5C, top panel), at the same time a decrease in total IB levels became apparent (Fig. 5C, middle panel). In FDC-P1/HPK1-⌬N cells the accumulation of phospho-IB␣ was found to be significantly reduced indicating a suppression of NFB activation (Fig. 5C, top panel). Therefore, it appears that cleavage of HPK1 in apoptotic hematopoietic cells may be utilized as an effective tool to block NFB activation. DISCUSSION Hematopoietic progenitor kinase (HPK1), a GC kinase-related mammalian Ste20 homologue, has been implicated as an upstream regulator of SAPK activity. We show here that HPK1 also potently activates NFB transcription factors in hemato-poietic and COS1 cells. These findings corroborate a previous report showing stimulation of IKK␣/␤-mediated IB␣ phosphorylation after forced expression of HPK1 (41). Among the GCK subfamily I kinases, we also observed NFB activation by the closest HPK1 homologue GCKR/KHS, whereas the more distantly related GCK caused only moderate NFB activation, and GLK failed to elicit any activity in COS1 cells. NFB activation by GCK-related kinases may be highly cell type-dependent as GCK and GCKR/KHS failed to affect NFB in HEK293 cells (11), whereas in melanoma cells GCK activated NFB moderately (42).
The molecular requirements for NFB activation differed substantially from those for activation of SAPK, which appear to necessitate the HPK1 kinase domain mainly (14). Our data suggest the existence of distinct HPK1-containing complexes responsible for SAPK/JNK and NFB activation. The scaffolding protein JIP1 has been described to coordinate a SAPK/ JNK-activating complex that contains SAPK/JNK, MKK7/ SEK2, MLK3, and HPK1 and likely facilitates the interaction of the pathway compounds (43). NFB activation has been shown to be dependent on the activity of a high molecular mass complex of 700 -900 kDa containing the IB-kinases (IKKs) ␣ and ␤ (44). Whether HPK1 is physically associated with this complex remains to be established. Kinase-deficient mutants of NFB-inducing kinase (NIK) and IKK␤ efficiently inhibited HPK1-mediated NFB activation, suggesting that HPK1 acts upstream of the IKK complex. Suppression of NIK-induced NFB activation by an N-terminal HPK1 deletion mutant implicated HPK1 downstream of NIK. The apparent paradox of HPK1 being able to act upstream and downstream of NIK might result from the action of overexpressed mutant forms of both proteins on the same NFB-inducing complex. Furthermore, SH3 domain interactions are likely to contribute crucially to the formation of the HPK1-containing NFB-activating complex, as coexpression of SH3 domain-containing molecules strongly interfered with this process, and NFB activation was dependent on the presence of intact polyproline sites in HPK1.
MLK3, which has been shown to interact with HPK1 via its SH3 domain (14), was recently reported to phosphorylate directly IKK␣ and IKK␤ (45) raising the interesting possibility that MLK3 might mediate NFB stimulation by HPK1. Although we and others (46) failed to detect NFB activation by MLK3, this could be a consequence of the different cell types used in the assay systems (HeLa versus COS1) or different amounts of DNA transfected (45).
In the adult, HPK1, which is exclusively hematopoietic, displays the most restricted expression pattern of the GCK subfamily I kinases. In contrast to GCK, GCKR/KHS and GLK, which appear to be elements of TNF signaling pathways, HPK1 is not activated by TNF␣ in vivo. GCK and GCKR/KHS were reported to bind TRAF2 (11,18) and therefore are likely to act at a receptor proximal position in TNF signaling. However, GCK binding is dispensable for TRAF2-mediated p38, SAPK/ JNK, and NFB activation (47). The TRAF2-GCK complex was recently described to protect melanoma cells against UV-induced apoptosis (42). Increasing expression of TRAF2 and GCK during melanoma progression was positively linked to JNK and NFB activity.
In T cells HPK1 has been shown to be constitutively associated with the adaptor protein Grb2 and its homologue Grap (20), whereas association with the Grb2 family member Gads was only observed after T cell receptor ligation (21). Both B and T cell receptor engagement cause stimulation of HPK1 kinase which is dependent on Src and Syk/ZAP-70 tyrosine kinases and the adaptor proteins LAT, SLP76/BLNK (20). Taken to-gether these studies suggest that HPK1, which is dependent on inducible tyrosine phosphorylation of immunospecific adaptors like LAT and SLP76/BLNK, mediates immunoreceptor signals in a receptor proximal position resulting in the stimulation of SAPK/JNK and NFB effector pathways. The subfamily I GC kinase HPK1 therefore appears to have adopted a specific function in hematopoietic cells coupling cell type-specific receptor systems to ubiquitous effector pathways.
The precise role of HPK1 during apoptosis of murine hematopoietic cells is not well understood. We found HPK1 to be cleaved by a caspase 3-like activity in vitro, and we detected corresponding proteolytic products in apoptotic cells in vivo. The resulting C-terminal fragment inhibited NFB activation by HPK1 and NIK in a dominant-negative manner and strongly reduced phosphorylation of IB␣ in FDC-P1 cells. Although HPK1 is not likely to be involved in the induction of apoptotic processes in hematopoietic cells, its conversion from an activator of NFB to an inhibitor of NFB may contribute significantly to the efficient execution of the apoptotic program. While our manuscript was in preparation, caspase-mediated cleavage of human HPK1 was demonstrated in Jurkat T cells following Fas ligation (48).
Caspase cleavage has been reported for several mammalian Ste20 kinases. Proteolysis of SPAK/PASK has been implicated in the regulation of subcellular localization (49), whereas caspase 3 cleavage of MST/Krs (50,51), SLK (52) and PAK2 (53,54) results in the generation of an activated apoptosisinducing kinase domain fragment. Interestingly, we did not detect an enhanced kinase activity of the isolated HPK1 kinase domain, suggesting that the HPK1 regulatory C terminus fulfills no autorepressive function as has been described for PAK2.
Mammalian GCK-related kinases are a rapidly growing family of cytoplasmic serine/threonine kinases. Here we have presented data that show activation of NFB by HPK1 and proteolytic cleavage of HPK1 during apoptosis in growth factordeprived myeloid progenitor cells. Proteolytic cleavage converts HPK1 into an inhibitor of NFB. These findings significantly enhance our knowledge on HPK1 activity and open novel avenues to test the biological roles of HPK1 which is likely to fulfill yet undefined functions that rely on NFB activation.