Selective autophagy is a catabolic process with which cellular material is specifically targeted for degradation by lysosomes. The function of selective autophagic degradation of self-components in the regulation of innate immunity is still unclear. Here we show that Drosophila Kenny, the homolog of mammalian IKKγ, is a selective autophagy receptor that mediates the degradation of the IκB kinase complex. Selective autophagic degradation of the IκB kinase complex prevents constitutive activation of the immune deficiency pathway in response to commensal microbiota. We show that autophagy-deficient flies have a systemic innate immune response that promotes a hyperplasia phenotype in the midgut. Remarkably, human IKKγ does not interact with mammalian Atg8-family proteins. Using a mathematical model, we suggest mechanisms by which pathogen selection might have driven the loss of LIR motif functionality during evolution. Our results suggest that there may have been an autophagy-related switch during the evolution of the IKKγ proteins in metazoans.
Kenny is selectively degraded by autophagy
Given the observed interaction between Kenny and Atg8a, we examined whether Kenny is degraded by autophagy. Western blot analysis of whole body fly lysates showed a robust accumulation of Kenny in Atg8a and Atg7 mutants compared to wild-type flies (Fig. 3a–c). Accumulation of Kenny protein was found to be a post-translational event since mRNA levels for Kenny were found to be similar between wild-type and autophagy mutant flies (Supplementary Fig. 2). Immunofluorescence analysis of endogenous Kenny in various tissues revealed that Kenny accumulates and forms cytoplasmic aggregates that co-localize with Ref(2)P and ubiquitinated proteins in autophagy mutant fly tissues, such as adult brain, fat body and midgut, compared to wild-type flies (Fig. 3d–l, Supplementary Fig. 3). This observation was further reinforced by mosaic analysis which clearly showed an accumulation of endogenous Kenny in autophagy-depleted (Atg5-RNAi and Atg8a-RNAi) and autophagy mutant (Atg1 and Atg13) clonal cells but not in their wild-type neighbors (Fig. 4). Furthermore, expression of mCherry-eYFP-KennyWT and mCherry-eYFP-KennyF7A/L10A in HeLa cells showed that Kenny was degraded in the lysosomes in a manner dependent both on an intact LIR motif and co-expression of Atg8a (Fig. 5).
The ubiquitin-proteasome system (UPS) constitutes another major intracellular proteolytic system in eukaryotes27. To explore the contribution of the UPS in the degradation of Kenny, adult flies were fed with food supplemented with the proteasome inhibitor Bortezomib or vehicle only. We found that there was a modest (compared to the accumulation observed in autophagy mutants) accumulation of Kenny upon proteasomal inhibition, suggesting that Kenny is predominantly degraded by autophagy (Supplementary Fig. 4). Taken together these results show that Kenny is preferentially degraded by autophagy in a LIR-dependent and Atg8a-dependent manner in basal conditions.
Autophagy mutant flies accumulate phosphorylated Kenny
We observed that Kenny protein resolved into two bands in lysates from Atg8a and Atg7 mutant flies while a single band was observed in wild-type flies (Fig. 3b). The presence of a doublet of bands can be associated with post-translational modifications, such as phosphorylation or ubiquitination, which have been shown to be involved in mammalian NEMO regulation28,29. We investigated the possibility of phosphorylation as a cause of appearance of Kenny doublet bands by treating lysates from autophagy-deficient flies with either calf intestinal phosphatase (CIP) or lamba phosphatase (λPP). In each case, we noticed that the doublet of bands converted into a single band corresponding to the lower molecular weight band (Supplementary Fig. 5a). To test whether ubiquitinated Kenny accumulates in autophagy-deficient flies, GFP-Kenny was expressed in wild-type or Atg8a mutant flies and consequently immunoprecipitated from denaturated fly lysates. No ubiquitination pattern could be detected using a pan-ubiquitin chain antibody (Supplementary Fig. 5b). These results show that Kenny is post-translationally modified by phosphorylation and is accumulated in its phosphorylated form in autophagy mutants.
Kenny is required for the autophagosomal degradation of ird5
Kenny is a component of the IκB kinase (IKK) complex which consists of IKKβ/ird5 and IKKγ/Kenny, and is crucial for nuclear translocation of transcription factor Relish and induction of the expression of antimicrobial peptide (AMP) genes, including Diptericin (Dpt)23,30,31,32,33. Using UAS-mCherry-ird5 transgenic flies, we observed that mCherry-ird5 is diffused in the cytoplasm of fat body cells in fed conditions (Fig. 6a). However, mCherry-ird5 displayed a dotted localization after starvation, and co-localized with Atg8a and Cathepsin-L (Fig. 6b–d). Using a UAS-GFP-mCherry-ird5 transgenic line to distinguish acidic and non-acidic compartmentalization of ird5, we observed an accumulation of mCherry-only (acidic) ird5 puncta in starved conditions (Fig. 6e, f). mCherry-only puncta exclusively co-localized with the lysosomal marker Cathepsin-L (Fig. 6g). These results suggest that ird5 is targeted to the autophagosomes to undergo lysosomal degradation. This was further confirmed by the accumulation of mCherry-ird5 when larvae were fed with the lysosomal inhibitor chloroquine (Fig. 6h). Interestingly, we observed that adult flies fed with Bortezomib exhibited a reduction of mCherry-ird5 protein levels (Supplementary Fig. 6). This observation reinforces the hypothesis that ird5 is degraded by autophagy as impairment of the UPS enhances autophagic degradation34. We then expressed mCherry-ird5 along with GFP-Kenny wild-type or its LIR-defective counterpart. We observed that the LIR motif in Kenny was not required for mCherry-ird5 and GFP-Kenny interaction (Fig. 7a, b and Supplementary Fig. 7). An in vitro GST pull-down experiment confirmed the LIR-independent direct interaction between ird5 and Kenny (Fig. 7c)24. In order to examine whether the targeting of the whole IKK complex to the autolysosome is dependent on Kenny’s LIR motif, we probed fat bodies from starved larvae with an anti-Cathepsin-L antibody. While we observed a clear co-localization of the three proteins when GFP-KennyWT is expressed (Fig. 7d), the co-localization between mCherry-ird5 and the lysosomes was lost when GFP-KennyF7A/L10A is expressed (Fig. 7e). Additionally, we observed that starvation-induced mCherry-ird5 accumulation is more prominent upon kenny RNAi-mediated knockdown (Supplementary Fig. 8). Finally, we observed that the size of the starvation-induced mCherry-ird5 puncta was reduced when key RNAi was co-expressed, compared to control RNAi (Fig. 7f–h), suggesting less autophago-lysosomal targeting. Collectively, these results show that Kenny is a selective autophagy receptor for the degradation of the IKK complex by autophagy.
The IMD pathway is controlled by selective autophagy
In order to examine the physiological significance of IKK complex accumulation in autophagy-depleted flies, we analyzed the expression level of the IMD target gene Dpt. We observed that there was a significant systemic upregulation of Dpt in Atg8a and Atg7 mutant flies (Fig. 8a, one-way ANOVA test ***P < 0.001, ****P < 0.0001) that can be partially rescued by re-expressing mCherry-Atg8a in the fat body (Supplementary Fig. 9). Furthermore, the upregulation of Dpt was accompanied with nuclear localization of Relish in fat body cells (Fig. 8b–d). However, although we could detect nuclear translocation of Relish in autophagy mutant gut cells, no significant upregulation of the Dpt gene expression could be detected in isolated guts (Supplementary Fig. 10, one-way ANOVA, P > 0.05), suggesting that the systemic deregulation of the IMD pathway is predominatly controlled by the fat body. To assess whether the deregulation of the IMD pathway in autophagy-deficient flies is related to the accumulation of Kenny protein, we created Atg8a;kenny double mutant flies. We observed that the absence of kenny expression in autophagy-deficient flies abrogates the constitutive activation of Dpt observed in Atg8a mutant flies (Fig. 8e). Together, these results show that the upregulation of the IMD pathway when autophagy is blocked depends on Kenny.
In order to test whether the presence of commensal bacteria is responsible for the upregulation of Dpt in autophagy-deficient flies, we examined its expression in flies reared in germ-free (axenic) conditions. Interestingly, no upregulation of Dpt was observed in axenic Atg8a and Atg7-depleted flies compared to their conventionally reared siblings (Fig. 8a). Together these results indicate that selective autophagic degradation of IKK complex prevents constitutive activation of the IMD pathway in response to commensal microbiota.
Autophagy mutant flies exhibit hyperplasia in the gut
In order to examine the physiological effect of systemic upregulation of the IMD pathway in conventionally reared autophagy-deficient flies, we examined cell proliferation rates by immunostaining posterior midguts for phospho-Histone H3 (pH3), a specific marker for mitotic cells. We observed that Atg8a and Atg7 mutant flies exhibited higher numbers of pH3-positive cells compared to wild-type control flies (Fig. 8f–h, l). Interestingly this phenotype was not observed in guts from axenic Atg8a and Atg7 mutant flies (Fig. 8i–k), suggesting that persistent deregulation of the IMD pathway as a result of stimulation from commensal bacteria induces a hyperplasia phenotype in autophagy mutant flies. Additionally, we observed that there is no increased pH3 staining in Atg8a;kenny mutants (compared to Atg8a mutants), indicating that Kenny-mediated upregulation of AMPs is causative of gut hyperplasia in autophagy mutant flies (Supplementary Fig. 11). Finally, we performed clonal expression of GFP-KennyWT and GFP-KennyLIR mutant in adult midgut. We observed that pH3-positive cells were present only in cells expressing GFP-KennyLIR mutant and not in cells expressing GFP-KennyWT (Supplementary Fig. 12).
To understand the physiological relevance of intestinal dysplasia caused by autophagy depletion, we analyzed gut function in wild-type and autophagy mutant flies. Feeding flies with food containing blue dye revealed that flies with dysplastic guts defecated significantly less (Fig. 8m, one-way ANOVA, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001), indicating increased retention of ingested food. In addition, we observed that rearing Atg8a mutant flies in axenic conditions increased their median lifespan (Supplementary Fig. 13). Together these results show that an uncontrolled systemic immune response to commensal bacteria in Drosophila promotes a dysplastic dysfunctional gut that contributes to organismal death.
Mammalian IKKγ has lost its LIR motif during evolution
In order to explore the conservation of Kenny LIR motif in arthropods, we used BLAST analysis. We found that Kenny’s LIR motif is conserved in other Drosophila species, in the common house fly (Musca domestica), mosquito (Aedes aegypti), butterfly (Papilio xuthus), and silk moth (Bombyx mori) suggesting that there is conservation between Diptera and Lepidoptera (Supplementary Fig. 14). To examine whether Kenny’s LIR motif is evolutionarily conserved in mammals, we tested the interaction of its human homolog IKKγ/ΝΕΜΟ with mammalian Atg8-family proteins. Interestingly, we found that IKKγ/NEMO does not interact with any of mammalian Atg8-family proteins (LC3A/B/C, GABARAP/L1/L2), which is consistent with the fact that we could not identify any predicted functional LIR motif in its sequence (Supplementary Fig. 15). In addition, using UAS-GFP-NEMO transgenic flies, we observed that unlike GFP-Kenny, human NEMO was unable to co-localize with mCherry-Atg8a-positive autophagosomes (Supplementary Fig. 16).
To understand how the functionality of the LIR motif in IKKγ may have been lost during evolution, we developed a deterministic mathematical model, in which different host types, with and without IKKγ LIR motifs, compete in the presence of a pathogen and a proxy for a member of the microbiota (Fig. 9). Mammalian pathogens have been observed to produce factors which interact directly with NEMO/IKKγ and promote the degradation of NEMO/IKKγ by autophagy35. We therefore allowed the pathogen in our model to exist as two variants: one encoding a protein with a functional LIR motif and one without.
Our results have shown that Drosophila Kenny/IKKγ can be directly targeted for degradation in a lysosome by interacting with Atg8a via its LIR motif. However, NEMO/IKKγ in mammals has been shown to be ubiquitinated itself and therefore could also interact with selective autophagy receptors via ubiquitin tags29,36,37. Mammals are known to possess a greater range of selective autophagy receptors than flies—thus the major mechanism by which mammalian IKKγ is likely to be degraded during infection probably involves ubiquitin tagging of IKKγ followed by autophagy. We therefore considered 4 hypothetical host types within our model: host type 1, in which IKKγ lacks a LIR motif and cannot be tagged for autophagic degradation by ubiquitination; host type 2, in which IKKγ posseses a LIR motif and cannot be tagged for autophagic degradation by ubiquitination; host type 3, in which IKKγ lacks a LIR motif and can be tagged for autophagic degradation by ubiquitination, and host type 4, in which IKKγ possesses a LIR motif and can be tagged for autophagic degradation by ubiquitination. Type 2 hosts are Drosophila-like, and type 3 hosts are human-like.
Full details of the mathematical model and its parameters are found in the Methods section, but the following biological assumptions were applied at all times: (1) A host with neither a LIR motif on IKKγ nor the specific molecular machinery to target IKKγ for autophagic degradation using a ubiquitin tag (host type 1) suffers excess mortality due to the stimulation of the innate immune response by the microbiota. (2) Maintaining a regulatory system capable of appropriately degrading IKKγ using ubiquitination is inherently more costly than regulating IKKγ by LIR induced degradation, due to the proteins required to ubiquitinate IKKγ and the need for appropriate selective autophagy receptors.
To explore the possible evolutionary reasons why the LIR motif may have been lost during mammalian evolution, we considered the following biologically plausible effects: (Effect I) If the host can up or downregulate the degradation of IKKγ by ubiquitin tagging, this may afford greater precision in regulating immunopathology during infection, and result in reduced mortality during infection. (Effect II) As in Effect I, but the advantage of being able to precisely regulate the rate of degradation of IKKγ by ubiquitin tagging may be reduced if a LIR motif is present on IKKγ, constantly signaling it for autophagy. (Effect III) If the pathogen encodes a LIR motif, this may enhance the degradation of IKKγ during infection and attenuate the innate immune response. The consequences of this (both of which could be advantageous to the pathogen) may be (IIIa) a slower recovery rate from infection, or (IIIb) reduced host mortality during infection. We assumed that the pathogen is only able to manipulate the host in this way if host IKKγ lacks a LIR motif (host types 1 and 3).
Figure 9 illustrates the consequences for host and pathogen of applying parameter values capturing these different effects. In the absence of Effects I or II, or any pathogen LIR Effect (IIIa or IIIb), host type 2 dominates, in the presence of a non-LIR encoding pathogen (Fig. 9a, top left hand panel). This is a Drosophila-like scenario (Fig. 9b). Host type 1 cannot succeed in this scenario because it suffers excess mortality from the stimulation of innate immune responses by the microbiota, and host types 3 and 4 cannot succeed because they bear the extra cost of maintaining pathways to autophagically degrade IKKγ by ubiquitin tagging.
As soon as we make additional assumptions about host or pathogen advantage during infection, it becomes possible to obtain a human-like scenario in which host type 3 dominates. If we make the assumption that the host is best able to regulate immunopathology during infection if IKKγ is regulated using ubiquitin tagging, and that this advantage is greatest if IKKγ is not being targeted for degradation by a LIR motif (Effect II), host type 3 will dominate (Fig. 9a, bottom row). Alternatively, if there exists a pathogen which encodes LIR, for which infection mortality is specifically reduced in hosts where IKKγ lacks LIR (Effect IIIb), theoretically this alone could overcome the cost of maintaining pathways to degrade IKKγ by ubiquitin tagging and allow host type 3 to dominate (Fig. 9a, top right hand panel). There are therefore at least two biologically plausible mechanisms by which host-pathogen co-evolution could result in the loss of the LIR motif from IKKγ.