The key finding of this study is that Nox5 forms an oligomer, most likely a tetramer, and that oligomerization is needed for full activity. Furthermore, the oligomerization was mediated by the dehydrogenase domain. Oligomerization was demonstrated by the following approaches. First, we showed that inactive mutants of Nox5 inhibited calcium-activated ROS generation by wild-type Nox5. While this sort of dominant negative effect does not itself prove oligomerization, this is one of several possible explanations. Second, Nox5 immunoprecipitated with itself, demonstrated by immunoprecipitating epitope-tagged Nox5 with either wild-type Nox5 or a different epitope-tagged version of Nox5. Third, deletion of the DH domain eliminated oligomer formation and also eliminated the inhibitory effect on Nox5 activity (i.e., the ability of dominant negative forms of Nox5 to inhibit depended on the occurrence of oligomer formation). Fourth, the DH domain of Nox5 anchored to a heterologous transmembrane domain (from CiVSP) mediates oligomerization whereas the CiVSP transmembrane domain alone is monomeric. In addition, the CiVSP-linked Nox5 DH-domain inhibited wild-type Nox5 activity, providing convincing evidence that the DH domain is a key molecular determinant mediating Nox5 oligomerization. Fifth, radiation inactivation analysis in whole permeabilized cells revealed that the functional Nox5 size of both overexpressed and endogenous Nox5 was most consistent with a tetramer, and was much larger than the monomer molecular weight. Thus, Nox5 forms a functionally important oligomer within the membrane, and this oligomer is critically important for calcium-activated ROS generation by Nox5.
The calculation of a tetrameric size for the active form of Nox5 assumes that no additional proteins are present in a catalytically essential complex with Nox5. Indeed, Nox5 was selected for the present study because it is not known to require any additional subunits such as regulatory subunits or p22phox
. However, in this context, two additional proteins, c-Abl and calmodulin, are reported to stimulate Nox5 activity in response to H2
and elevated calcium, respectively, and to form complexes with Nox5 under certain activation conditions (42
). In the case of c-Abl, H2
stimulates a calcium flux and tyrosine phosphorylation of c-Abl, and results in its translocation to a membrane compartment containing Nox5. In addition, after cells are treated with H2
, immunoprecipitation of c-Abl results in co-immunoprecipitation of Nox5. However, both the irradiation inactivation and immunoprecipitation experiments described above were carried out without H2
stimulation of the cells, so significant complex formation with c-Abl would not have occurred under our conditions. Likewise, ionomycin was used only to assay activity, and calcium-free conditions were used for both co-immunoprecipitation experiments and radiation inactivation experiments, so calcium-induced complex formation with calmodulin is not expected to be an issue under our conditions. Thus, in addition to the phox
-like regulatory proteins, calmodulin and c-Abl can be ruled out as members of the Nox5 complex demonstrated here.
One possible explanation for dominant inhibitory effects is that the inactive Nox5 binds to signaling proteins such as calmodulin or c-Abl, described above, or protein kinase C, which decreases the concentration needed to activate Nox5. However, in the case of ionomycin activation, high calcium (around 1 mM) fully activates Nox5, and other effectors do not cause further activation (15
). In the present study, we observed inhibition of Nox5 activity by co-expression of an inactive Nox5 mutant in the ionomycin-treated HEK293 cells, as well as in cells permeabilized in the presence of a maximally activating concentration of free calcium ion. These results demonstrated clearly that the functional effect of the dominant negative forms of Nox5 is not due to binding of the mutant Nox5 to either of these activator proteins.
The molecular details by which oligomerization of Nox5 occurs are not known, but comparison with known oligomeric homologs suggests some possibilities. Our data point to the dehydrogenase domain as the primary interaction site mediating oligomerization. We are not aware of homologous flavoprotein dehydrogenases that have been proven to form oligomers in solution. However, cytochrome b5
reductase from Physarium
forms a dimer in the crystal structure (43
). It is not yet clear whether this represents an artifact of crystal packing or a true dimer structure. Nevertheless, the structure provides a reasonable starting point for proposing specific amino acid residues that function in oligomerization. While mutagenesis studies are underway to identify such key residues, we have not yet succeeded in generating an oligomer interface mutant that exists as a monomer. It is therefore not yet possible to test whether the monomer retains some activity. Nevertheless, radiation inactivation experiments imply that the functional unit as it exists in native cells and membranes is an oligomer. In addition, several EF-hand proteins, such as S100 proteins (44
), calpain (45
), and polcalins (47
), exist as dimers and this self-interaction may facilitate regulatory coupling and cooperativity, e.g.
upon calcium binding. Often, dimerization in these cases is mediated by non-canonical EF-hand regions that occur as un-paired odd-numbered motifs (45
), and such regions are absent in the Nox5 calcium-binding domain. The recently reported (48
) crystal structure of the EF-hand domain of a plant NADPH-oxidase, OsRbohB, showed a dimer structure, but it is not clear whether the dimer occurs naturally, or is a crystallization artifact. In the case of human Nox5, the present study shows that oligomers of Nox5 form in the absence of the EF-hand domain and that the dehydrogenase domain is critical for mediating oligomerization. We cannot rule out the possibility that in an assembled Nox5 oligomer, the calcium-binding domains might also associate in a functionally important manner.
Dimerization and higher order oligomerization can perform a variety of functions. First, such oligomerization may result in stabilization of the protein, extending its lifetime in the cell. The heterodimer between Nox2 and p22phox
, for example, results in stabilization of both proteins so that in the absence of either protein, its partner is degraded more rapidly (49
). In addition, oligomerization frequently has regulatory consequences, such as cooperativity seen for oxygen binding to hemoglobin, and substrate binding to many regulated enzymes. We are not aware of any published data that suggest cooperative behavior for regulators or substrates for Nox5, but this possibility cannot be ruled out, as reported experiments were not designed to reveal such kinetic behavior. Finally, inactive forms of Nox may occur naturally and may be involved in adjusting the activity of the active forms of Nox5. Five splice variants of human Nox5 ---- Nox5α, β, γ, δ and Nox5S ---- have been described. Nox5α, β, γ, and δ all include a calcium-binding domain containing all four EF-hand motifs, but with insertions or extensions of different length; other regions including PBRN, transmembrane domain, and DH domain are present in these variants. Nox5S, however, is a variant in which the calcium-binding domain is absent; this form is identical to the truncation mutant M173-F737 in which the calcium-binding domain has been deleted. Nox5S is expressed in human Barrett’s adenocarcinoma cell lines where it promotes cell proliferation (50
), and in human microvascular endothelial cells (HMEC) where it participates in thrombin signaling (12
). Nox5S appears to produce ROS under some circumstances by as-yet unknown activation mechanisms, but because it lacks the calcium-binding domain, cannot respond directly to calcium. The present study suggests that Nox5S can also function as an endogenous suppressor of calcium-activated forms of Nox5 (i.e.
, those that retain a calcium-binding domain), since Nox5(M173-F737) bound to Nox5α and inhibited its ionomycin-induced ROS generation (). HMEC cells expressed NOX5β (an active long form) as well as significant amount of Nox5S, suggesting that the activity of NOX5β in HMEC cells might be regulated by the protein level of Nox5S.
In addition, oligomerization could in theory function to concentrate and elevate local concentrations of reactive oxygen within a small area, increasing its effectiveness as a signal molecule, for example in the regulation of enzymes such as protein tyrosine phosphatases, transcription factors, and ion channels, as has been well documented previously (1
). This might be especially effective if the Nox enzyme is co-localized with its target protein in an organelle or membrane subregion. In this context, in vascular smooth muscle, Nox1 has been reported in caveolae, while Nox4 is localized to focal adhesions (53
It will be important to address in future studies whether other isoforms of Nox also form oligomers, and if so, what are the functional consequences? Inactive mutated forms of Nox1, Nox2 and Nox4 also function as dominant negative inhibitors (Kawahara, unpublished data), but in these cases they might act by competing for p22phox
and/or regulatory subunits. Because of this, it is not straightforward to distinguish this mechanism from one involving homo-oligomerization. Interestingly, cytochrome b558
(the complex formed from Nox2 plus p22phox
), when detergent extracted and purified from human neutrophil membrane, was heterogeneous in size by sedimentation equilibrium analysis, but the smallest detectable species showed a molecular mass of ~350 kDa (54
). If one assumes minimal binding by phospholipids or other bound proteins, then this size corresponds to an oligomer of n = 3–4 (the exact number depending on the contribution of glycosylation to the overall size of gp91phox
). While it is not clear in this report whether the observed oligomerization occurs naturally or was induced upon detergent extraction of the protein from its native environment, the possibility should be entertained that oligomerization may be a general property of the Nox family of enzymes, and may impact enzymatic function and ROS signaling by the several mechanisms as discussed above.