The IKK/NF-κB signalling system is proposed to be critically involved in the pathogenesis of various neurological diseases . On the one hand, it is well characterized as a central regulator of inflammatory responses by controlling the expression of multiple proinflammatory acting genes [3, 22]. On the other hand, IKK/NF-κB signalling is crucially involved in neuronal differentiation and various CNS functions [6–8]. However, due to its complex regulation in different cell types and diverse responses to different physiological and pathological conditions, the precise function of the IKK/NF-κB system in CNS physiology and pathology is only partially understood.
Former studies suggested an ambivalent role of the IKK/NF-κB system in the pathogenesis of neurological disorders [8, 43]. Due to its proinflammatory function, NF-κB activation is able to trigger neuronal dysfunction, aging and cell death, thereby increasing severity of CNS diseases [8, 11, 44, 45]. In contrast, NF-κB activation can also mediate neuroprotection [6, 8, 46]. Previously, we found that IKK2/NF-κB activation in neurons increases tissue damage in a mouse model of stroke, probably by enhancing the overall neuroinflammatory process elicited by this acute insult . Therefore, we wanted to further investigate the role of IKK2-mediated neuron-specific NF-κB activation in the induction of neuroinflammatory responses using the IKK2nCA model. We hypothesized that constitutive IKK2 activation in neurons is sufficient to induce inflammation, as it was demonstrated in several non-neural cell types as well as in astrocytes [11, 18–20, 23]. However, with the exception of microgliosis and astrogliosis observed in the DG, neuron-specific IKK2 activation did not result in a prominent inflammatory phenotype including infiltration of immune cells. Consistently, typical proinflammatory NF-κB target genes like Ccl2, Tnf, Ptgs2, Lcn2 and Cxcl10 that are highly expressed in other inflammatory conditions are either moderately or not induced in the IKK2nCA model. This argues for a specific function of IKK2/NF-κB signalling in neurons.
What could be the reason for this unexpected response? As NF-κB is activated by synaptic signalling, such kind of NF-κB activation in neurons would already create a proinflammatory environment under the physiological conditions of neurotransmission. Vice versa, inflammation-mediated NF-κB activation in neurons would lead to functional conflicts like deregulation of NF-κB-mediated neurite outgrowth and synaptic plasticity. Therefore, a functional separation of neuronal IKK/NF-κB signalling versus inflammatory IKK/NF-κB signalling in other cells could be of physiological advantage. Several studies showed important functions of NF-κB in neuronal differentiation, including neurite outgrowth, formation and remodelling of synaptic connections, axogenesis and neuronal function, e.g. hippocampal learning and memory formation [6, 26, 28, 47–50]. These studies are mainly based on experimental approaches inhibiting the IKK/NF-κB signalling system in neurons, therefore it could be anticipated that neuronal IKK/NF-κB activation might result in a phenotype that improves neuronal survival and cognitive capabilities. However, this idea appears to be in stark contrast to our findings. One plausible explanation for this discrepancy could be the duration of IKK/NF-κB signalling. In our model we induce permanent IKK/NF-κB activation over weeks, whereas in the physiological context of learning and memory rather a transient or repetitive activation is known to occur which is e.g., elicited by the neurotransmitter glutamate, known to induce NF-κB in synaptic signalling [51, 52]. As excessive glutamate signalling results in excitotoxic cell death , we can speculate that the constitutive NF-κB activation in our model is probably detrimental to the DG neurons. Interestingly, pharmacological inhibition of IKK2 was able to block NMDA-induced excitotoxic cell death in hippocampal neurons and oligodendrocytes . In line with the view that especially a transient NF-κB activation kinetic improves neuronal differentiation, Russo et al. shows that stereotactic application of a virus expressing IKK2-CA to the nucleus accumbens leads to spine formation within a short time of 3 days .
The adequate function of the adult dentate gyrus depends on both healthy mature granule cells as well as ongoing neurogenesis  and NF-κB was shown to be critically involved in different aspects of adult neurogenesis using loss-of-function approaches [28, 56]. Since we only see neurodegeneration in the DG, a hypothesis might be that constitutive IKK2 activity also interferes with neurogenesis, which then results in a depletion of neurons in the GCL. However, a roughly 50% reduction in the cell count at 9M is difficult to explain solely by blockade of ongoing adult neurogenesis in the IKK2nCA model but rather suggests active neurodegeneration. Furthermore, we could observe increased evels of Ki67-positive cells upon transgene inactivation arguing for elevated neurogenesis that may account for an active regeneration process of the DG rather than simple prevention of further neurodegeneration.
Imielski et al.  showed that the structural degeneration of the DG depends on apoptotic cell death, which was not detected in our model as measured by cleaved caspase-3 and TUNEL assay (Additional file 5). Instead, we could identify degenerating neurons in the DG but not in other brain regions by Fluoro-jade staining suggesting that IKK2-CA induces cell death but this cell death is independent of apoptosis or is due to a very slow rate of apoptosis that may escape detection. So far it remains largely open why this degeneration process is specific to the DG. However, our findings implicate that a combination of Bdnf decrease and Tnf increase (and possibly changes in other so far unkown factors) may account for the selective neurodegeneration of the dentate gyrus in the IKK2nCA model. The structural restoration of the dentate gyrus after transgene inactivation in both models implies that fine balanced levels of NF-κB are required for appropriate neuronal survival and homeostasis in this brain region. Therefore, reactivation of the IKK/NF-κB system for therapeutic measures of neuro-regeneration in the context of dementia-associated diseases as suggested by Imielski et al.  is apparently critical and surely dose-dependent.
The neurodegenerative effect of constitutive IKK2 signalling could be due to the composition of the activated NF-κB dimers. In IKK2nCA mice the canonical NF-κB pathway is active, most likely leading to the nuclear translocation of p65 containing dimers that are found to regulate apoptosis associated genes . Also, there might be an under-representation of c-Rel containing dimers, which are known to promote neuronal survival by enhancing Bcl2l1 transcription . Corresponding to this, a downregulation of pro-survival genes like Bcl2 and Bcl2l1 was detected at older age in IKK2nCA mice, although both of these are regulated by NF-κB [58, 59]. Moreover, the decreased Bdnf expression in IKK2nCA mice can be proposed as a potential mechanism that interferes with neuronal survival  because it also correlates with a decline in Bcl2 and Bcl2l1 levels . Bdnf is well known to regulate cognitive tasks, synaptic plasticity and neuronal survival by activating its receptor TrkB [39, 41, 60, 61] and its expression is compromised in brain disorders as AD, HD, Rett syndrome and schizophrenia [62, 63]. Thus, the reduced levels of Bdnf and Bdnf-regulated AMPA receptors might attribute to the impaired hippocampal learning and the atrophy of the DG observed in our model.
Nevertheless, other factors may also contribute to the impaired learning and atrophy of the dentate gyrus. There is the possibility that the microgliosis and astrogliosis observed in the DG are sufficient to cause the neurodegeneration in IKK2nCA mice [64, 65]. Together with the elevated Tnf levels, such kind of inflammatory processes may influence learning and memory as well as neuronal survival. This might contribute to the observed phenotype, although the importance of low-grade neuroinflammation for learning and memory and neurodegeneration is still controversially discussed .
What is the underlying molecular mechanism resulting in reduced Bdnf, Bcl2 and Bcl2l1 expression in IKK2nCA mice? The observed downregulation for Bdnf is rather surprising, as Bdnf is an NF-κB target gene in astrocytes  and would therefore expected to be rather upregulated in neurons, too. Although we did not investigate the mechanism behind this repression of Bdnf in IKK2nCA mice, previous studies identified a similar kind of downregulation of target genes by NF-κB, e.g. in the case of hypoxia, Tnf-dependent EAAT2 expression, or in the regulation of anti-apoptotic genes after treatment of cells with DNA-damaging agents [68–70]. Campbell et al. showed that the cytotoxic stimuli like ultraviolet light (UV-C), and daunorubicin, downregulated the expression of anti-apoptotic NF-κB target genes like Bcl2, Bcl2l1, Xiap and A20, thus providing the possibility that canonical NF-κB activation may account for induction and repression of target genes depending on the presence of coactivators, given cell type and induction mechanism . There is also the possibility that NF-κB mediated changes in epigenetic gene regulation may affect Bdnf expression [71–73]. Moreover, IKK2 has been previously described to phosphorylate Bcl-xL, a mechanism associated with reduced expression of this gene in stressed, post-mitotic neurons . A recent publication by Zhang et al.  addressed the role of IKK2/NF-κB signalling in the hypothalamus, which increases with aging and mediates suppression of hypothalamic-gonadotropin-releasing hormone (GnRH1) expression finally promoting systemic aging. They found that elevated IKK2 and NF-κB activity induces cJun/cfos and PKC levels, which are able to inhibit Gnrh1 promoter activity. This mode of NF-κB-mediated inhibition of gene expression might also account for NF-κB-mediated Bdnf repression since Bdnf expression is known to be regulated by multiple promoters.
More importantly, the IKK2-CA transgene and Bdnf expression pattern do not coincide very well in the DG of IKK2nCA mice (Figure 6D). Bdnf expression is located to the hilus region whereas IKK2-CA protein is detected in GCL neurons. In support of this, GABAergic interneurons present in the hilus are devoid of Camk2a expression thereby excluding Camk2a-driven transgene expression in these neurons . This strongly argues for a scenario that IKK2-CA mediated NF-κB activation does not directly influence Bdnf expression. Rather, a so far unknown factor/s released by IKK2-CA positive neurons suppresses Bdnf production in a paracrine manner in the vicinity of the hilus, a process that necessarily does not depend on NF-κB-mediated gene regulation.
Notably, recent work by Han et al. reported that the Camk2a-tTA transgene, also used in the present study to drive IKK2-CA expression, itself exhibits a degenerating effect on the neurons which was not recognized by the scientific community for many years. Moreover, they observed that this degeneration was permanently rescued by administration of DOX during the first 6 weeks of life . As the IKK2nCA animals were treated with DOX up to the age of 4 weeks, most likely, we also avoid the tTA-induced degeneration. Consistent with that, the Camk2a-tTA-induced neurodegeneration gets obvious already at the age of 2 months, whereas a IKK2nCA animals do not show atrophy up to the age of 3 months rather develop degeneration between 3 and 6 month age periods. Furthermore, IKK2nCA animals were bred in pure NMRI background, an outbred model, which is different from the analysed hybrid strains sensitive for tTA-induced degeneration.