Inhibition of NF-κB in astrocytes is sufficient to delay neurodegeneration induced by proteotoxicity in neurons

A hallmark of many age-related neurodegenerative diseases is the presence of protein aggregates, which precedes the onset of clinical symptoms. Genetic profiling of transcriptional changes in neurodegenerative diseases resulted in the identification of glial genes whose altered expression could potentially contribute to pathogenesis [1, 2]. Besides this, another characteristic in most, if not all neurodegenerative diseases associated with protein aggregation is the activation of astrocytes [3]. Recently, it was demonstrated that a subtype of astrocytes is neurotoxic and is present in most neurodegenerative diseases [4].

The presence of aggregates in astrocytes can result in their activation and subsequent signaling to neurons [5]; reviewed in [6]. However, it is unclear whether signals from aggregate-expressing neurons can modulate the activity of astrocytes. Furthermore, the putative consequences of signaling from astrocytes that are activated by neuronal proteotoxic stress on the functioning of neurons are not known.

Astrocytes are involved in essential CNS functions, including metabolic functions, regulating levels of neurotransmitters, maintaining the blood-brain barrier, as well as immune defense, thus contributing to neuronal homeostasis [7]. Astrocytes are activated early in neurodegenerative diseases at times that may precede the appearance of aggregates [6]. This activation only occurs in specific areas of the brain, i.e., those that are primarily affected by the respective diseases, such as for example the striatum in Huntington’s disease (HD), the pons in SCA3, and the cortex in Alzheimer’s disease (AD) [8, 9]. Alterations in function of astrocytes can contribute to pathogenesis in a mouse model for HD, and therefore astrocytes have been suggested as potential therapeutic targets [10]. In this model, the HD-associated, aggregation-prone protein was amongst others expressed in the brain, including neurons, microglia, and astrocytes. Thus, in this model, cell-autonomous activation of astrocytes (in response to proteotoxic stress) as well as cell-non-autonomous activation of astrocytes (in response to proteotoxic stress in neurons) could contribute to pathogenesis. There is ample evidence of cell-autonomous activation of astrocytes in the pathogenesis of neurodegenerative diseases (reviewed in [6, 11]). However, how signaling from aggregate-expressing neurons can influence signaling in astrocytes and how responses from astrocytes can modulate neurodegeneration is unknown.

We investigated if and how astrocytes contribute to neurodegenerative diseases in a cell-non-autonomous manner, within the context of an intact animal. Examining astrocytes in an in vivo model is key, given that their morphology and activity change when taken outside their physiological context (reviewed in [12]). For this, we conducted a dedicated RNAi screen to selectively knock down individual genes in astrocytes in a Drosophila melanogaster model of the polyQ disease SCA3. In SCA3, the ATXN3 gene contains an expanded CAG repeat (coding for glutamine) that leads to the expression of a misfolded, aggregation-prone ATXN3 polyglutamine protein. These misfolded polyQ-containing proteins accumulate intracellularly, resulting in neuronal damage and activation of astrocytes [13]. In the brains of patients as well as of animal models, proteolytic cleavage of a fragment of the gene containing the polyQ-containing stretch can mediate the toxicity of expanded polyQ [14‐16]. We expressed this biologically relevant, truncated part of the ATXN3 gene containing an expanded polyQ stretch (SCA3^polyQ78) exclusively in Drosophila eyes and simultaneously downregulated expression (by RNAi) of candidate genes exclusively in astrocytes. This setup would provide insight into the putative cell-non-autonomous contributions of astrocytes to neurodegeneration induced by proteotoxicity in neurons. In a candidate RNAi screen, we identified genes in astrocytes that cell-non-autonomously influenced degeneration. This demonstrates the relevance of cell-non-autonomous signaling in astrocytes in neurodegeneration. We further explored the contribution of NF-κB transcription factor Relish, which was identified as an enhancer in our screen, and its activity was upregulated in our SCA3 model. We investigated the role of Relish in astrocytes in SCA3, as well as in a model for Alzheimer’s disease.

Expression of a biologically relevant, truncated form of the human ATXN3 gene containing an expansion of the glutamine region (trSCA3 Q78, containing 78 glutamines, hereafter referred to as SCA3^polyQ78) in Drosophila eyes resulted in progressive degeneration of photoreceptor cells [21], accompanied by eye depigmentation as well as depigmentation accompanied by black necrotic spots (depigmented and necrotic, Fig. 1a; arrow indicates necrotic spots). No such effects were seen in control eyes or eyes expressing SCA3^polyQ27, containing a non-pathogenic length of glutamines (SCA3^polyQ27; Fig. 1a) [21]. Quantification of the fraction of the eyes with a normal appearance, displaying depigmentation or necrotic spots is shown in Fig. 1b. Only expression of SCA3^polyQ78 resulted in depigmentation or necrotic spots. We quantified the extent of degeneration by determining the fraction of the eyes containing necrotic spots. This quantification provides a quick and reproducible means of analysis. To see whether SCA3^polyQ78-induced degeneration was accompanied by decreased solubility of SCA3^polyQ78, lysates of fly heads were analyzed. Expression of SCA3^polyQ78 but not SCA3^polyQ27 resulted in the presence of high molecular weight insoluble proteins (Fig. 1c), which have previously been identified as aggregates using immunofluorescence [22].

We used another robust, sensitive way to quantify degeneration, by coexpressing membrane-targeted mCD8-GFP [23] with SCA3^polyQ27 or SCA3^polyQ78 in the eyes. Expression of SCA3^polyQ78 but not SCA3^polyQ27 resulted in a decrease of mCD8-GFP levels, indicative of degeneration (Fig. 1d, top panel). Analysis of mCD8-GFP expression levels in SCA3^polyQ78-expressing eyes in lysates of fly heads allows for quantification of degeneration (Fig. 1e). Both the fraction of necrotic eyes and GFP levels were used as readouts to determine the effect of downregulation of specific genes in astrocytes on degeneration.

To analyze a putative cell-non-autonomous role for astrocytes in SCA3^polyQ78-induced eye degeneration, we used two binary expression systems that act independently [20], UAS-Gal4 and QUAS-QF2 (Additional file 1a). Indeed, expression of fluorescent proteins in astrocytes (RFP) or the eyes (GFP) shows specific non-overlapping expression in the respective tissues (Additional file 1b). We analyzed whether the location of astrocytes (expressing RFP) was altered upon expression of SCA3^polyQ78 in the eyes. We observed the presence of astrocytes in SCA3^polyQ78-expressing eyes but not in control eyes or eyes expressing SCA3^polyQ27 (Fig. 1d, bottom panel). The response in SCA3^polyQ78-expressing flies is quite robust, as they all displayed expression of RFP in the eyes (Additional file 2). The number of astrocytes in all samples in Fig. 1d is comparable, as total RFP levels in fly heads were similar (Fig. 1e). Thus, the presence of astrocytes in SCA3^polyQ78-expressing eyes suggests a response in astrocytes, which warrants further investigation into a putative cell-non-autonomous role of astrocytes in degeneration.

Our Drosophila SCA3 model allows specific analysis of cell-non-autonomous contributions of astrocytes to the degenerative SCA3 eye phenotype. We selected genes (many of them evolutionarily conserved; Additional file 3) potentially involved in (1) communication between neurons and astrocytes, such as neuropeptides and their cognate receptors (2) immune signaling pathways, such as nuclear factor kappa B (NF-κB) and (3) potential receptors for cytokines, neuropeptides, aggregates, or damage-associated molecular patterns. Some of the genes that were selected have either been implicated in cell-autonomous activation of astrocytes [6] or are upregulated in the degenerative fly brain [24, 25]. However, the role of these genes in cell-non-autonomous activation of astrocytes in response to proteotoxic stress in neurons and subsequent contribution to neurodegeneration still remained to be determined. In a candidate RNAi screen, RNAi constructs targeting a set of 156 selected genes were expressed exclusively in astrocytes in flies expressing SCA3^polyQ78 exclusively in the eyes, and the extent of eye degeneration was analyzed by determining the fraction of necrotic spots. The screen revealed both suppressors (n = 26; RNAi against these genes enhanced degeneration) as well as enhancers (n = 20; RNAi against these genes reduces degeneration) of the SCA3 eye phenotype (Additional file 3). These results underscore the relevance of astrocytes in the progression of SCA3^polyQ78-induced eye degeneration.

One of the genes identified in the screen was NF-κB transcription factor Relish (orthologous to mammalian NF-κB1). Downregulation of Relish expression by RNAi, but also downregulation of several other genes in the Relish pathway, suppressed the SCA3 phenotype. While NF-κB has been implicated in activation of astrocytes in neurodegeneration (reviewed in [6, 26]), a cell-non-autonomous role of NF-κB activation in astrocytes through signals from neurons and subsequent effect on neurodegeneration remains to be determined. In Drosophila, there are two independent NF-κB pathways, activating transcription factors Relish or Dif/Dorsal respectively. We examined expression of anti-microbial peptides (AMPs) in our SCA3 model, transcriptional NF-κB targets which help fight infection [27]. In inflammatory responses, these peptides are considered to aid in activating and recruiting immune cells (reviewed in [28]). While both NF-κB pathways are activated upon eye-specific expression of SCA3^polyQ78, activation of Relish was stronger: AMPs specific for Relish (CecA and DptA) were upregulated more than Dif/Dorsal-specific AMPs (IM1 and IM2; Fig. 2a and shown in [29]). Activation of Relish in SCA3^polyQ78-expressing eyes occurred in astrocytes: GFP under control of the promoter of a Relish-dependent gene was expressed predominantly in the astrocytes that were present in SCA3^polyQ78-expressing eyes (Additional file 4a). In control eyes, no induction of GFP was seen.

We further focused on the cell-non-autonomous role of astrocyte-specific Relish signaling in SCA3. We first confirmed the results of our screen by using two independent Relish RNAi constructs (efficacy of knockdown, see Additional file 4b) that both decreased SCA3^polyQ78-induced degeneration (Fig. 2b). Inversely, the opposite effect was seen upon overexpression of Relish. Modulation of Relish expression in astrocytes did not affect morphology of control eyes (Additional file 4c). In flies heterozygous for Relish, degeneration was attenuated, similar to astrocyte-specific Relish RNAi. The effect of Relish in astrocytes on eye degeneration was quantified (Fig. 2c) by determining the fraction of the eyes containing necrotic spots. Importantly, the effects of Relish on degeneration are cell-non-autonomous, mediated via astrocytes: when coexpressing SCA3^polyQ78 and RNAi constructs targeting Relish in the eyes, eye degeneration was not attenuated (Additional file 4d).

When membrane-targeted mCD8-GFP was used as a readout for degeneration [23] in eyes expressing SCA3^polyQ78, the protective or enhancing effects of the astrocyte-specific down- and upregulation of Relish levels were confirmed (Fig. 2d for GFP images of different genotypes). To see whether modulating of Relish expression in astrocytes affected the localization of astrocytes, we coexpressed myr-RFP in astrocytes. We did not see an effect of modulation of Relish expression on the localization of the RFP signal (Fig. 2d for RFP images of different genotypes, Fig. 2f for western blot analysis of lysates). These data suggest that Relish in astrocytes does not have an effect on the localization of astrocytes in SCA3^polyQ78-expressing eyes (additional images are provided in Additional file 4e). We quantified the effect of modulating Relish expression in astrocytes on mCD8-GFP levels in SCA3^polyQ78-expressing eyes by western blot (Fig. 2e). GFP levels in flies heterozygous for Relish were similar to those of Relish RNAi in astrocytes, suggesting the importance of astrocyte-specific Relish signaling in SCA3^polyQ78-induced degeneration. The effect of Relish on mCD8-GFP levels in the eye is specifically related to SCA3^polyQ78-induced degeneration, as in control flies (not expressing SCA3^polyQ78) GFP levels were similar, irrespective of Relish expression (Additional file 4f). Together, these experiments demonstrate that in our SCA3 model, Relish is activated (and Dif/Dorsal to a lesser extent) and that Relish signaling in astrocytes enhances SCA3^polyQ78-induced degeneration. Thus, eye-specific expression of SCA3^polyQ78 promoted Relish activation in astrocytes, and this Relish activation enhanced degeneration in a cell-non-autonomous manner.

In the brain, Relish is activated in glia in aging flies as well as in flies that are mutant for A-T mutated (ATM), a kinase associated with ensuring genomic integrity and neurodegeneration [30, 31]. We analyzed contribution of astrocytes to overall Relish signaling in the fly head induced by eye-specific expression of SCA3^polyQ78. For this, we examined levels of Relish-dependent AMPs in fly heads expressing SCA3^polyQ78 in the eyes together with astrocytes-specific targeting of Relish RNAi or Relish overexpression constructs, or in SCA3^polyQ78-expressing flies heterozygous for Relish. Expression of Relish RNAi constructs in astrocytes in flies expressing eye-specific SCA3^polyQ78 resulted in decreased expression of Relish-specific AMPs (DptA and AttC) in the head to levels comparable to flies not expressing SCA3^polyQ78 (Fig. 3a), whereas overexpression of Relish enhanced expression. SCA3 flies heterozygous for Relish expressed levels of Relish target genes comparable to those expressing Relish RNAi constructs in the astrocytes. Dif/Dorsal-specific target genes were unaffected by \modulating Relish levels in astrocytes (Additional file 5). These data highlight the importance of Relish signaling in astrocytes regarding Relish-dependent immune signaling in the head.

The presence of aggregated or misfolded proteins is toxic to neurons [32]. Possibly, astrocytes could have an effect on levels of protein aggregates. To see whether Relish-specific signaling in astrocytes could have an effect on solubility of SCA3^polyQ78 in the eyes, we collected heads of flies expressing eye-specific SCA3^polyQ78 2 days after eclosion and analyzed the cell extract. Modulating Relish expression in astrocytes or heterozygosity for Relish did not affect total levels or solubility of SCA3^polyQ78 (Fig. 3b, quantification of three independent experiments in Fig. 3c). This suggests that the effect of Relish on SCA3^polyQ78-induced degeneration is not the result of either a clearance of SCA3^polyQ78 aggregates or of a shift of the balance of soluble-insoluble SCA3^polyQ78 towards more soluble SCA3^polyQ78.

To see whether Relish can modulate SCA3-induced eye degeneration via Relish-dependent AMPs, we expressed RNAi constructs targeting two Relish-dependent AMPs (AttA and CecA; efficacy of knockdown in Additional file 6a). Astrocyte-specific downregulation of these AMPs attenuated the degenerative SCA3 eye phenotype (Fig. 4a; quantification of degeneration in Fig. 4b), similar to decreasing Relish activity in astrocytes (Fig. 2a, b). No effect on eye morphology was seen when RNAi targeting AMPs was expressed in astrocytes (Additional file 6b).

We next analyzed mCD8-GFP levels in the eyes as a measure of degeneration in SCA3^polyQ78-expressing eyes and examined the effect of astrocyte-specific knockdown of Relish-specific AMPs. Levels of mCD8-GFP were increased upon knockdown of AMPs, demonstrating attenuation of SCA3^polyQ78-induced eye degeneration (Fig. 4c; analysis on western blot and quantification of three independent experiments in Fig. 4d). In control eyes, no effects of downregulation of Relish-specific AMPs expression in astrocytes on mCD8-GFP levels in the eyes were found (Additional file 6c). We also examined whether astrocyte-specific knockdown of Relish-specific AMPs could influence the presence of astrocytes in SCA3^polyQ78-expressing eyes, but did not see an effect (Fig. 4c), and the numbers of astrocytes were also not affected, as analyzed by levels of myr-RFP in astrocytes (Fig. 4e).

When we examined the effect of astrocyte-specific downregulation of AMPs on SCA3^polyQ78 levels or solubility, we did not observe an effect (Fig. 4f; quantification of three independent experiments), similar to modulating Relish levels in astrocytes. These data show that the effect of Relish on degeneration in our SCA3 model is, to a significant extent, mediated via the generation of AMPs, and that the effect of AMPs does not occur via alterations in SCA3^polyQ78 solubility.

We next tested whether the observed effects of the Relish pathway in astrocytes on (short term) degeneration also translates into a (long term) survival benefit in flies expressing SCA3^polyQ78 in neurons. Astrocyte-specific Relish inhibition may also extend lifespan in neurons expressing SCA3^polyQ78, given our observations of Relish in astrocytes on the SCA3 eye phenotype. We expressed SCA3^polyQ78 pan-neuronally (not in astrocytes) and examined effects of modulation of Relish expression specifically in astrocytes (specificity of expression shown in Additional file 7a) using the following two independent binary expression systems (Additional file 7b): UAS-Gal4 and QUAS-QF2. To avoid potential detrimental effects of SCA3^polyQ78 during development, we expressed SCA3^polyQ78 in adult flies only, using the inducible “Q system,” in which the neuronally expressed transcription factor (QF2) that induces SCA3^polyQ78 expression is coexpressed with QF-suppressor QS [20] (Additional file 7b). This way, expression of SCA3^polyQ78 is suppressed during development (Fig. 5a). Feeding quinic acid suppresses QS, thus alleviating inhibition of QF2 resulting in the neuron-specific expression of SCA3^polyQ78, of which the extent of aggregation increased over time (Fig. 5a). The lifespan of flies expressing SCA3^polyQ78 in neurons was significantly shortened (Fig. 5b). Quinic acid by itself did not affect lifespan (Additional file 7c), excluding non-specific effects. Relish-dependent gene expression was increased in flies that expressed SCA3^polyQ78 in neurons (Fig. 5c), similar to expression of SCA3^polyQ78 in the eyes (Fig. 2a).

We next wished to examine whether modulating Relish levels in astrocytes in adult flies could have an effect on lifespan of flies that expressed SCA3^polyQ78 pan-neuronally. For this, levels of Relish were modulated in astrocytes of adult flies to exclude effects on development. Hereto, we used the transcription factor Gal4 to drive the expression of the Relish constructs. At a lower temperature (18 °C), activity of Gal4 is low and Relish RNAi constructs do not decrease Relish expression (Additional file 7d). 4 days after eclosion, we induced SCA3^polyQ78 expression in neurons by adding quinic acid to the food. Flies were shifted to 25 °C to induce expression of the Relish constructs in astrocytes. Importantly, reducing Relish expression in astrocytes extended the lifespan of SCA3 flies, whereas overexpression shortened the lifespan (Fig. 5d). In control flies, downregulating Relish expression in astrocytes had no effect on lifespan (Additional file 7e), indicating that the effects of Relish RNAi are linked to SCA3^polyQ78-mediated degeneration of neurons. However, overexpression of Relish alone in astrocytes significantly shortened lifespan (Additional file 7e), precluding analysis on Relish overexpression on SCA3^polyQ78-mediated shortening of lifespan.

In parallel, SCA3-related effects on motor function were analyzed, measured as climbing ability. The SCA3^polyQ78-induced decrease in climbing ability was partially alleviated by knockdown of Relish in astrocytes (Additional file 7f). As was observed with modulating Relish expression in astrocytes in SCA3^polyQ78-expressing eyes, no effects of Relish in astrocytes on total brain SCA3^polyQ78 aggregation load or levels could be detected (Additional file 7g).

The effects of Relish modulation on SCA3-related degeneration were not related to alterations of the levels of SCA3^polyQ78 aggregates (Fig. 3b and Additional file 7g). Therefore, we hypothesized that modulation of the Relish pathway in astrocytes could be more generic and relevant to a neurodegenerative disease associated with the presence of extraneuronal aggregates, such as in Alzheimer’s disease. For this, we turned to an established Drosophila model of Alzheimer’s disease [18], in which the Amyloid beta (Abeta42; Aβ42) peptides are fused to a secretion signal peptide (of the necrotic gene), resulting in the secretion of Aβ42 peptides. When expressed in neurons (Additional file 7h), these Aβ42 peptides are secreted and form extracellular aggregates in flies, reminiscent of the plaques in human AD [33]. In this AD fly model, flies displayed progressive neurodegeneration, with mobility and memory deficits, and reduced lifespan [18, 34]. In line with our hypothesis, the lifespan of flies expressing Aβ42 in neurons was shortened, but extended when Relish expression was suppressed in astrocytes, using two independent RNAi lines targeting Relish (Fig. 5e). This, together with the data on SCA3, demonstrates that modulating Relish signaling in astrocytes has beneficial effects in neurodegeneration, irrespective of the aggregation process that initiates it.

Our data show that astrocytes can modulate the rate of neurodegenerative disease progression in a cell-non-autonomous manner. This is the first time to show that neuron-specific expression of a pathological, degeneration-inducing protein results in activation of astrocytes, which can subsequently influence symptoms of degeneration.

Astrocytes become reactive during the course of a number of age-related neurodegenerative diseases associated with aggregates [6], and several reports have suggested that during the course of disease astrocytes gain neurotoxic properties ([4] reviewed in [6]). However, potential mechanisms for how astrocytes affect progression of disease have yet been lacking, in part due to the fact that in most studies disease-causing genes are expressed in both neuronal and non-neuronal cells in the brain. Expression of aggregation-prone proteins in astrocytes can cell-non-autonomously influence neuronal viability [5], reviewed in [6]. However, putative cell-non-autonomous contributions of astrocytes to neurodegeneration in response to neurons expressing aggregation-prone proteins are unclear. Here, we investigate how signaling in astrocytes can influence neurons that express aggregation-prone, disease-causing proteins, thus investigating cell-non-autonomous contributions of astrocytes to neurodegeneration. We identified astrocytes in Drosophila eyes expressing SCA3^polyQ78 but not in control eyes or eyes expressing SCA3^polyQ27 (Fig. 1d), suggesting a role for astrocytes. We further investigated putative cell-non-autonomous contributions of genes in astrocytes in a candidate RNAi screen, where the effect of RNAi-mediated downregulation of individual genes in astrocytes on SCA3^polyQ78-induced eye degeneration was determined. This resulted in the identification of astrocyte-specific enhancers and suppressors of the SCA3-related neurodegeneration (Additional file 3). These results demonstrate, for the first time, the importance of signaling in astrocytes induced by specific expression of aggregation-prone, neurodegeneration-associated proteins. Astrocytes in Drosophila share structural and functional similarities with mammalian astrocytes: amongst others, they provide trophic support to neurons and are involved in neurotransmitter recycling (reviewed in [35]). Therefore, genes that we identified in our screen may be relevant for mammalian astrocytes as well.

Conserved NF-κB transcription factor Relish was identified as an enhancer in our SCA3 model, and its activity was upregulated in the head upon eye-specific SCA3^polyQ78 expression. We therefore focused on Relish signaling for more in-depth analyses. Relish was specifically activated in astrocytes present in eyes expressing SCA3^polyQ78, but not in astrocytes elsewhere in the head (Additional file 4a). Downregulation of expression in astrocytes of both Relish, but also Relish-dependent AMPs alleviated SCA3^polyQ78-induced eye degeneration (Figs. 2 and 4). In contrast, we did not see an effect of downregulation of Relish expression on degeneration in eyes coexpressing SCA3^polyQ78 and Relish RNAi constructs (Additional file 4d). Thus, eye-specific expression of aggregation-prone SCA3^polyQ78 results in cell-non-autonomous activation of NF-κB in astrocytes. We next investigated possible cell-non-autonomous contributions of astrocytes to neurodegeneration.

Downregulation of Relish expression in astrocytes not only delayed neurodegeneration (suggested by enhanced mobility, Additional file 7f) and extended lifespan of flies expressing SCA3^polyQ78 in neurons (Fig. 5d), but also extended lifespan in flies expressing neuronal Aβ42 peptides (Fig. 5e). We did not see an effect of downregulating Relish expression in astrocytes on lifespan (Additional file 7e), underscoring the efficacy of using this pathway as a target to inhibit the detrimental effects of SCA3^polyQ78. However, overexpression of Relish in astrocytes was sufficient to shorten lifespan, demonstrating toxic effects of Relish. This is in agreement with earlier studies demonstrating that expression of Relish or Relish-specific AMPs was sufficient to induce neurodegeneration [30, 36]. Of particular interest is that in aging flies Relish activity and levels of AMPs were increased in both neurons and glia, and constitutes an important determinant of lifespan control and the onset of neurodegeneration [31]. In addition to this, our data demonstrate the importance of cell-non-autonomous Relish signaling in astrocytes in neurodegeneration associated with aggregates, which is generally aging-associated. The presence of aggregation-prone, neurodegeneration-associated proteins in neurons resulted in Relish signaling in astrocytes, which subsequently influenced neuronal functioning. Together, these findings suggest the importance of NF-κB activity in astrocytes in age-associated neurodegenerative diseases.

Our data thus provide direct evidence for earlier suggestions that astrocytes are not only activated in response to neurodegeneration (as, e.g., demonstrated by elevation of GFAP, glial fibrillary acidic protein, a marker for astrocyte activation) [5], but indeed can modulate disease progression. The relevance of the cell-non-autonomous Relish activation in astrocytes is underscored by the observation that downregulation of Relish in astrocytes can influence disease progression. A toxic gain of function in astrocytes was also demonstrated in a mice expressing human aggregation-prone TAR DNA-binding protein 43 (TDP-43) in neurons [37], resulting in alterations in protein expression in astrocytes. Depletion of an upregulated, astrocyte-specific factor reduced neurotoxicity, underscoring the importance of astrocytes to neurotoxicity. Our findings that genes in the NF-κB pathway are key mediators of such a response extends the observation that astrocyte-specific, transient expression of IKK (IκB kinase 2), resulted in neurodegeneration in mice [38].

While Lattke et al. showed sufficiency of astrocyte-specific NF-κB signaling in inducing neurodegeneration, our data now demonstrate that signals from neurons expressing aggregation-prone proteins suffice to induce cell-non-autonomous NF-κB activation in astrocytes. The importance of NF-κB activation in astrocytes was suggested by date presented in Fig. 3a and Additional file 5a. The presence of astrocytes in SCA3^polyQ78-expressing eyes (Fig. 1d, Additional file 4) suggests recruitment of astrocytes, as shown for Alzheimer’s disease (reviewed in [39]). Downregulation of NF-κB in astrocytes did not appear to alter the presence of astrocytes in SCA3^polyQ78-expressing eyes (Fig. 2d, Additional file 4e), suggesting that additional signaling in astrocytes promotes the recruitment of astrocytes to degenerating cells. Together, these data demonstrate the existence and importance of intercellular signaling between neurons and astrocytes as an important determinant for the speed of neurodegeneration.

The nature of the neuronal signals that trigger NF-κB activation in astrocytes remains to be elucidated. Possibly, damage-associated molecular patterns (DAMPs), including SCA3^polyQ78 aggregates, released from neurons can result in activation of NF-κB in astrocytes. It is unlikely that the Relish canonical pathway, headed by PGRP-LC (reviewed in [27]), accounts for Relish activation, since modulation of PGRP-LC did not significantly affect the SCA3 eye phenotype (Additional file 1). Previous reports have demonstrated Relish activation independent of PGRP-LC, either by neuropeptides or nitric oxide [40, 41], which may account for Relish activation in our model.

Whereas we show that astrocytes can directly modulate neuronal health, this appeared not to be associated with alteration in the burden of protein aggregates in the brain. However, decreasing the levels of misfolded or aggregated proteins has been suggested as a therapeutic strategy [42]. Furthermore, reduction of oxidative damage can decrease degeneration [43]. Thus, a combination of therapeutic strategies targeting aggregates, astrocyte-mediated neurotoxicity as well as oxidative damage may have optimal effects in alleviating aggregates-associated neurodegeneration.

Targeting NF-κB signaling in astrocytes to alleviate neurodegeneration may be more broadly applicable: the lifespan of flies expressing Aβ peptides, present in patients with AD, was extended upon astrocyte-specific inhibition of NF-κB (Fig. 5e). Thus, despite intracellular localization of aggregates in SCA3 and extracellular localization of Aβ peptides in AD, astrocyte-specific NF-κB inhibition extended lifespan in fly models for both diseases. This suggests that targeting NF-κB in astrocytes may be beneficial in other aggregates-associated neurodegenerative diseases.

In conclusion, our data demonstrate that astrocytes respond to cellular stress or damage in neurons induced by neuronal expression of disease-associated, aggregation-prone proteins. These astrocytic responses are important modulators of neurodegeneration. We demonstrate that activation of transcription factor NF-κB Relish occurs in astrocytes in response to proteotoxic stress in neurons, and that in this situation astrocyte-specific inhibition of Relish enhances vitality and extends lifespan. While proteotoxic stress in astrocytes can contribute to neurodegeneration, these data highlight the importance of astrocytic responses to cellular stress or damage in neurons, and provide putative therapeutic potential.