We identified a novel form of HGK in a human glioblastoma cell line cDNA library. Compared with other forms of HGK isolated from macrophage and brain, and with the mouse ortholog NIK, this form showed multiple different splice variations. We analyzed the overall genomic structure of the HGK gene and found it to include 33 exons, with nine regions (modules) of alternative splicing. EST data suggest that all of these alternative modules are present in expressed proteins, resulting in a complex family of HGK isoforms. While many genes are alternatively spliced, the HGK gene has an unusually large amount of alternative splicing. At least five different isoforms were identified from each of four tumor cell line sources, indicating that multiple isoforms are present in the same cell. Alternative splicing modulated the length of the coiled-coil domain, the composition of the CNH domain, and an array of PxxP motifs.
Given the variations in these protein-protein interaction domains, different splice forms are likely to be differentially localized and/or regulated. For example, the HGK clone isolated from glioblastoma cells has the PXXP-containing M3 module, while an alternative PXXP-containing module, M4, replaces M3 in the clones isolated from macrophage (51
) and mouse adipocytes (43
). The consensus sequences of the PXXP motifs in M3 and M4 differ so that HGK isoforms containing one or the other may interact with different SH3-containing proteins. Two PXXP motifs in M4 from NIK were shown to bind to the adapter protein Nck (43
). The binding partners for the PXXP motifs in M3 are unknown and under investigation. The interactions mediated by these PXXP motifs may not be important for the function of HGK in tumor cells, as M3 and M4 are absent from the majority (88% and 98%, respectively) of the isoforms identified in tumor cell lines.
With a probe that detects all HGK splice forms, we found this kinase to be broadly overexpressed at the RNA level in a large number of tumor cell lines relative to most normal tissues. HGK does not map to a known tumor amplicon, yet the overabundance of HGK RNA in tumor cell lines versus normal tissue provides correlative evidence for a role for this kinase in tumorigenesis. An active role for HGK kinase activity in cell transformation is suggested by the fact that ectopic expression of inactive forms blocked oncogenically relevant processes in tissue culture cell models. It is important to note that both the NIH 3T3 and RIE-1 cells used in these studies express detectable levels of HGK, suggesting that the inactive mutant acts in a dominant negative manner (data not shown).
Expression of the inactive dominant negative mutant blocked focus formation and anchorage-independent growth, increased cell adhesion and spreading rates, and blocked invasion and morphogenesis in response to HGF. In contrast, overexpression of wild-type or active alleles slightly increased focus formation, had no effect on anchorage-independent growth, decreased cell adhesion and spreading, and potentiated the cells' invasive and morphogenic responses to HGF. Together these results suggest that HGK catalytic activity is required for anchorage-independent growth, causes the reduction of cell adhesion, and increases invasive properties.
The ability of dominant negative HGK to block the growth of RIE-1 cells in soft agar suggests that this kinase may function in pathways that regulate the anchorage dependence of normal cell growth, implicating HGK in the so-called outside-in signaling from integrin receptors (24
). Focal adhesion kinase (FAK) has been shown to function in such pathways (21
). There was no measurable difference in autophosphorylation of FAK or phosphorylation of the FAK substrate p130CAS
during cell attachment in the stable cell lines expressing the various HGK mutants (data not shown), indicating that HGK may function either downstream or independently of FAK.
The effects of HGK mutants on integrin function and on integrin receptor expression indicate that this kinase does appear to induce changes in cell adhesion or inside-out signaling. Integrin mediated inside-out and outside-in signaling are often described as independent pathways, but there is evidence that effects on one impinge on the other (reviewed in references 4
). We observed that cells with higher HGK kinase activity were less adherent to fibronectin. This effect may be explained, at least in part, by the reduced surface expression of integrin receptor α5 subunit in cells expressing active HGK. While a comprehensive analysis of all integrin subunits levels could not be done with available reagents for rat cells, we did note that HGK also had similar effects on α2 subunit levels but not on α1 subunit levels. Interestingly, NIK−/−
mice show a phenotype similar to those of fibronectin−/−
and α5 integrin receptor−/−
mice, suggesting that NIK may also be involved in the α5 integrin receptor regulation during embryogenesis (50
As the changes in α5 subunit cell surface expression were smaller than might be expected from the phenotypic changes observed, HGK may also affect integrin signaling by an additional mechanism. The recent report showing physical association, as well as colocalization in attaching cells of β1-integrin receptor with NIK (mouse HGK) strongly suggests that HGK plays a direct role in integrin receptor signaling in addition to affecting α-integrin receptor levels, as we have shown (37
). Taken together, these observations point to both direct and indirect involvement of HGK in the regulation of integrin function and signaling.
This role for HGK in negative regulation of cell adhesion may be conserved in other HGK family members, since TNIK also appears to interfere with cell spreading in a kinase activity-dependent manner in transient-transfection assays (22
). A reduction in cell spreading rates was also observed previously with other STE20 kinases, including STE20-like kinase (SLK) and the p21-activated kinases αPAK and PAK4 (32
). In these three reports, the effects on cell spreading were attributed to effects on the cell cytoskeleton. Whether or not these other STE20 kinases modulated integrin-ligand binding in addition to effects on the cytoskeleton was not investigated. STE20 kinases show significant homology within the kinase domain and may share some cellular substrates, yet the lack of homology in the regulatory domains of these kinases suggest that they are not localized and/or regulated by the same mechanism. For example, the PAK kinases are directly regulated by rho family G-proteins, while HGK is not regulated by Rac or CDC42 and furthermore has been implicated to function upstream of Rac in the activation of JNK (45
Increased HGK kinase activity through overexpression caused a striking increase in the rate of cell invasion and morphogenesis, while expression of a kinase-inactive mutant blocked invasion and morphogenesis. Ectopic expression of HGK was not sufficient to induce invasion on its own and instead potentiated the invasive response of cells to HGF. A similar potentiation role in HGF-induced invasion was shown for the cytoplasmic domain of α6β4 integrin receptors (47
). Perhaps HGK also participates in the signaling complex with the Met receptor and the α6β4 integrin receptor. In examination of signaling downstream of the Met receptor, we observed a decrease in STAT3 phosphorylation on both Y705 and S727 with dominant negative HGK mutant expression. Both of these phosphorylation events have been shown to increase the activity of STAT3 as a transcription factor (reviewed in reference 8
STAT3 does not appear to be a direct substrate of HGK in that STAT3 protein isolated from cells was not phosphorylated when active, recombinant glutathione S
-transferase-HGK kinase domain was added in the presence of ATP (data not shown). However, HGK protein was found to copurify with STAT3 in immunoprecipitations of STAT3 from cells, suggesting that they are part of the same protein complex. HGK could modulate STAT3 activation by affecting its localization in the cell or by regulating the kinases and phosphatases that directly regulate STAT3. As STAT3 plays a role in HGF-induced invasion, blocking STAT3 activation may be one mechanism by which the inactive HGK mutants blocked RIE-1 cell invasion and morphogenesis (6
). STAT3 has also been implicated in tumorigenesis in a number of different cell systems (7
) and thus could also be important for the dominant negative effects of inactive HGK on focus formation and anchorage-independent growth. A more detailed investigation of the function of HGK in STAT3 activation will be the subject of a future study.
In conclusion, we present an analysis of the HGK gene with its complex family of isoforms. We present evidence that this human homolog of the msn
kinases from Drosophila melanogaster
and Caenorhabditis elegans
functions in morphogenic pathways not only during embryogenesis (50
) but also during tumorigenesis and/or invasion. Based on our observations and those of others, we suggest that this kinase may function in the integration of extracellular signals from soluble growth factors and from the extracellular matrix.