Lazarillo-related Lipocalins confer long-term protection against type I Spinocerebellar Ataxia degeneration contributing to optimize selective autophagy

The expression of human ApoD is boosted upon various neurodegenerative conditions (for reviews see [25-27]), but SCA1 is still unexplored. We therefore tested whether the Drosophila Lipocalins change their expression in SCA1-affected flies.

To assay expression of the Drosophila Lipocalins in a poly-Q-based neurodegenerative disease, we have used the GAL4/UAS system to express a pathogenic version of human Ataxin 1 with an expanded glutamine tract (hATXN1^82Q) in Drosophila retinal photoreceptors using the gmr:GAL4 driver. In this model of SCA1 [30] photoreceptors accumulate nuclear inclusions of the human protein and start degenerating during late pupal stage when flies develop at 25°C. We first tested GLaz expression. Figure 1A shows that its expression increases 3-fold in the photoreceptors degeneration model of SCA1 when assaying mRNA levels in 5–6 days old fly heads.

Since these results demonstrate that GLaz, like ApoD in vertebrates, is part of the natural response of the tissue to the neurodegenerative insult, we characterized its expression in the Drosophila eye, searching for cells that might influence the tissue response to the neurodegenerative condition. To assay for GLaz expression we used two independent insertions of a fusion protein construct (glaz:GLaz-GFP[FX] and glaz:GLaz-GFP[F2] [12]). These reporters show GFP-positive cells in the retina (Figure 1B). The GLaz signal does not co-localize with the neuronal marker 22C10 (Figure 1Bb), suggesting that support retinal cells (arrows in Figure 1B) and retinal basal glia (arrowhead in Figure 1B) are the source of this Lipocalin.

To test whether GLaz is able to protect cells that are degenerating due to the expression of hATXN1^82Q, we either expressed the Lipocalin in the degenerating photoreceptors (using the same controller to drive the expression of the poly-Q protein and GLaz), or in the cells naturally expressing the GLaz protein under the control of its native promoter (using a glaz:GLaz-GFP construct).

We first demonstrated that the expression of the UAS:GLaz transgene does not directly cause an alteration in the amount of hATXN1^82Q produced in photoreceptors by quantifying the hATXN1^82Q immunolabeling signal in eye imaginal discs of 3^rd instar larva (Additional file 1). This time point was chosen to avoid variation in the number of of hATXN1^82Q-expressing cells, since cell death caused by neurodegeneration is not present yet [30].

GLaz produces a significant rescue of photoreceptor degeneration when co-expressed with hATXN1^82Q (Figure 1C-F, Additional file 2). Two UAS:GLaz and two EP insertions upstream of GLaz were used to drive the expression of the Lipocalin under the control of gmr:GAL4. The adult eye external morphology was inspected (Figure 1C; Additional file 2A) and the degree of surface regularity analyzed (Figure 1D) and quantified (Figure 2B) by estimating a regularity index according to the variance of intensity maxima (See Methods section; [31]). Histochemistry of paraffin sections (Figure 1E) shows severe atrophy of retinal cells in the hATXN1^82Q-expressing flies, while the retinal epithelial pattern is reasonably preserved in the presence of gain-of-function constructs of GLaz. The degeneration observed in hATXN1^82Q-expressing photoreceptors of 3 days old flies is already extensive, and no further deleterious effect is observed when native GLaz expression is eliminated in the null mutant background (Additional file 3A). The expression of an unrelated control protein (LacZ) does not rescue the degeneration phenotype of the SCA1 model (Additional file 2C, Figure 2B). Specificity was further tested by simultaneously over-expressing GLaz in photoreceptors (with the UAS:GLaz2) and decreasing the expression of GLaz with a UAS:GLazRNAi construct. The rescue is abolished in this situation (Additional file 3B, Figure 2B). Moreover, the rescue observed with GLaz gain-of-function is dependent on temperature (due to the temperature sensitivity of the GAL4/UAS system) and the transgene expression levels. The GLaz-triggered rescue at 25°C shows a similar extent when comparing flies with either one or two copies of the UAS:hATXN1^82Q transgene located in the 1^st chromosome (Additional file 2D, Figure 2B). However, a lesser rescue was obtained either at 29°C or with a 3^rd chromosome UAS:hATXN1^82Q insertion line that shows a higher expression level (results not shown).

Human Ataxin 1 was detected by immunohistochemistry in the retina sections (Figure 1F). Nuclear inclusion bodies of the extended poly-Q protein are evident in the photoreceptors expressing hATXN1^82Q as previously described [30]. The expression of GLaz does not fully abolish the formation of these protein aggregates (arrows in Figure 1F) in spite of the preserved tissue integrity.

Taken together our data show that GLaz ameliorates the hATXN1^82Q-provoked degeneration, but there seems to be a limited range of the disease severity where GLaz can exert a beneficial effect when expressed by the degenerating photoreceptors.

Since the Lazarillo-related fly Lipocalins are secreted proteins [9,32], we further tested if GLaz expressed with its native spatiotemporal pattern would be able to protect photoreceptors when accessing the degenerating cell from the extracellular space. We tested the effect of GLaz gain-of-function by expressing, in wild type background, the GLaz-GFP fusion protein under the control of the native GLaz promoter (glaz:GLaz-GFP [F2] or glaz:GLaz-GFP[FX] insertion lines) on the pathogenic human poly-Q Ataxin 1-generated photoreceptor degeneration (Figure 2A,B and Additional file 4). The rescue is present from early stages of degeneration in the pupal eyes, 5 days after pupa formation (APF), to aged adult flies (30 days old) (Additional file 4). These results show that the rescue does not require Lipocalin expression in the degenerating cell, and that the GLaz protecting effect can be exerted when the source of protein is a different cell. The support and basal glial cells in the retina are the closest natural source of GLaz in this case (Figure 1B). Therefore GLaz can act as an extrinsic factor protecting the photoreceptors in a non-cell-autonomous manner. This paracrine neuroprotective mechanism is stable enough to maintain retinal rescue in aged flies (30 days old flies, Figure 2A and Additional file 4).

GLaz-GFP protein localization is evident in cells undergoing degeneration in 5–6 days old flies (Figure 2Ca,b). When analyzing mildly degenerated regions of retinas at 4d-APF, no GLaz-GFP signal is detected in the hATXN1^82Q-expressing photoreceptors (Figure 2Cc). However, more degenerated regions of retina showed GLaz protein signal co-localizing with the photoreceptor marker 4C5 (Figure 2Cd). These results suggest a potential incorporation of the extracellular protein into the degenerating photoreceptors that resembles the internalization of exogenous ApoD observed in cultured astrocytes [18].

The Drosophila Lipocalin NLaz is expressed by neurons in the fly CNS and PNS [32]. The native expression of NLaz increases 2-fold in retinal SCA1 model fly heads (Figure 3A). However, when NLaz expression was assayed by either a direct reporter construct (nlaz:GFP[R2]) or by a nlaz:GAL4 construct driving the expression of GFP (UAS:CD8-GFP), none of the NLaz reporters showed expression in the fly retina (not shown).

Nevertheless, co-expressing NLaz with hATXN1^82Q in retinal photoreceptors under the control of the gmr:GAL4 driver protects cells from degeneration (Figure 3B and Additional file 2B). The rescue level obtained with NLaz was similar to that with GLaz, as quantified by the degree of surface regularity (Figures 2B and 3D, 66-84% rescue with GLaz, 80% rescue with NLaz). Interestingly, the rescue attained by each Lipocalin is comparable to that obtained when both GLaz and NLaz are over-expressed in photoreceptors (Figure 3C,D, 80% rescue), suggesting that both proteins could be acting through the same mechanism and no additive effect is obtained.

Since GLaz is natively expressed in cells within the niche of the degenerating photoreceptors in the SCA1 model, we focused on this Lipocalin for the subsequent search of the cellular mechanism causative of neuroprotection.

Terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL) analysis reveals that photoreceptor cells die by apoptosis upon hATXN1^82Q expression (Figure 4A). GLaz over-expression causes a reduction in the number of apoptotic cells detected in the fly retina (Figure 4A). To have a quantitative estimate of the cell death rescue, we measured the levels of hATXN1 mRNA by qRT-PCR under two assumptions: (i) at pupal stage, hATXN1 mRNA levels are similar when degeneration is not yet present (hATXN1 protein levels show no difference, Additional file 1), and (ii) only intact photoreceptors would contribute significantly to the pool of detectable mRNA transcribed from the human transgene once the degeneration has started. GLaz does increase the levels of human Ataxin 1, both when co-expressed in photoreceptor neurons and when expressed by the nearby support cells and basal glial retinal cells (Figure 4B), in agreement with the decrease observed in the number of TUNEL-positive cells.

Among the genes found to modify SCA1-induced neurodegeneration in Drosophila, two genes involved in lipid management were uncovered [30,33]: Glutathione S transferase S1 (GstS1), a homologue of the mammalian hematopoietic prostaglandin D synthase (HPGDS), and Vibrator (Vib) a phosphatidylinositol transfer protein, homologue of PITPN-alpha in vertebrates. Over-expression of GstS1 or Vib suppresses the neurodegeneration caused by the poly-Q human Ataxin 1, and the loss of function of GstS1 enhances the phenotype [30,33].

Although GstS1 is classified as part of the constitutively expressed oxidative stress defensive enzymes [34], no transcriptional regulation upon neurodegeneration has been described for GstS1 or Vib. We explored the putative transcriptional regulation of these two genes in our system. While no modification of Vib transcript levels was seen in the SCA1 model either alone or combined with over-expression of GLaz, a 4-fold induction of the GstS1 transcript was obtained upon hATXN1^82Q-induced photoreceptor degeneration (Figure 4C). This induction was significantly reduced when GLaz is over-expressed (Figure 4C).

The direct cause of GstS1 up-regulation under SCA1-induced neurodegeneration remains to be explored. However, since GLaz controls lipid peroxide levels [12], we propose a mechanistic link between the lower levels of oxidized substrates upon GLaz gain-of-function and the decreased GstS1 transcription. Thus, the ectopic expression of poly-Q hAtaxin 1 results in the induction of both GLaz and GstS1 genes, and they contribute to an efficient lipid peroxide clearance. Other lipid management genes, like Vib, are not induced even though Vib is also a modifier of SCA1 neurodegenerative phenotype.

To address whether GLaz over-expression modifies any of the protein quality control systems altered in the SCA1fly model, we analyzed by immunohistochemistry and immunoblot the expression pattern of ubiquitinated proteins (Figure 5). Abundant ubiquitinated protein aggregates are evident upon neurodegeneration (Figure 5Ab,d) when compared to controls (driver alone, not shown, or GLaz over-expression in wild type background, Figure 5Aa). This accumulation is greatly reduced when GLaz is co-expressed with hATXN1^82Q (Figure 5Ac,e,f). A quantitative analysis by immunoblot demonstrates that over-expressing GLaz, either in the degenerating photoreceptor or in the native GLaz expression domain, results in a significant decrease of ubiquitinated protein levels (Figure 5B).

A defense mechanism against lipid peroxides in a ROS-promoting environment, and a mechanism controlling proteostasis seem unconnected processes at first glance. However, links between these mechanisms are starting to emerge. GstS1 is oxidized while catalyzing the conjugation of lipophilic substrates with reduced glutathione, and this fact has another consequence: the induction of autophagy by the Jun-N-terminal Kinase (JNK) when this kinase is released from the inhibitory interaction of the reduced form of GstS1 protein [34]. On the other hand, ubiquitinated proteins are known to be degraded by autophagy in neurons [35-37]. Thus, we hypothesized that the protective effects of proteins like Lipocalins, influencing both lipid oxidation and proteostasis, might be related to the membrane-mediated process of autophagy, potentially adding another link between these processes.

In order to understand the protective mechanism of Lipocalins we studied the endogenous expression of the Drosophila Atg8a gene, a homologue of LC3 in vertebrates [38], as a first approximation to the levels of autophagy activity. Transcription of Atg8a is known to be strongly up-regulated upon starvation [39] and to show a strong correlation with autophagy activity. Although autophagy initiation might not be dependent on Atg8a up-regulation, this transcriptional response contributes to replenish the protein consumed by autophagic activity [5]. By measuring the transcription levels of the Atg8a gene we can therefore assess whether autophagic activity is increased in the SCA1 model of photoreceptor degeneration. A twofold Atg8a induction is observed (Figure 6A). This indication of increased autophagic activity is in agreement with the observed induction of GstS1 (Figure 4C), which upon oxidation would contribute to JNK-based autophagy induction, and supports the view that inducing autophagy might be a general response in poly-Q based pathologies of different etiologies [3]. Neurodegeneration-triggered Atg8a induction in the fly head is similar in extent to the induction obtained upon starvation in this tissue (Figure 6A), although quantitatively smaller than the changes observed in larval and fat body tissues [39]. If flies expressing hATXN1^82Q are starved, Atg8a transcription is further increased, suggesting that different mechanisms of autophagy induction are acting in an additive manner.

When GLaz is over-expressed in the photoreceptors or in retinal support cells and basal glial cells, the SCA1-dependent Atg8a induction is significantly reduced (Figure 6B). Thus, a paradox arises from the data presented so far (Figures 5 and 6), since GLaz gain-of-function reduces accumulation of ubiquitinated proteins, but at the same time reduces the levels of an indicator of autophagic activity induction (Atg8a) and an oxidative stress defensive enzyme that affects autophagic activity (GstS1). When autophagy is further stimulated by feeding SCA1 model flies with rapamycin (a TOR signalling inhibitor; see [40] for a review), two important consequences are observed: 1) more autophagic induction results in higher levels of ubiquitinated proteins, instead of enhanced clearance, and 2) GLaz over-expression reduces the accumulation of ubiquitinated proteins (Figure 6C). The first observation suggests that autophagy promotion alone is not beneficial in the SCA1 neurodegeneration model, leading instead to autophagic stress [6,7] that ultimately makes ubiquitinated proteins clearance less efficient. Although rapamycin treatment, in addition to autophagy induction, might have wider effects on protein homeostasis, it is clear that the combination of rapamycin treatment with GLaz over-expression results in an enhanced clearance of ubiquitinated proteins (Figure 6C).

These data could support different scenarios: (i) The GLaz antioxidant properties contribute to maintain induction of autophagy and other antioxidant systems at an appropriate level where these mechanisms are most effective. (ii) GLaz is not directly acting on the induction or initiation steps of autophagy, but rather promoting its resolution, making the process to progress more efficiently. (iii) The GLaz contribution to the protein quality control systems is independent of autophagy (e.g., by potentiating proteasomal function).

To discern between the alternatives stated above we monitored the accumulation of p62, a protein acting as a decision point for the degradation of aggregasomes (aggregates of misfolded proteins too large to be degraded by the proteasome system). p62 docks aggregasomes to newly formed autophagic membranes, thus controlling their entrance in the autophagosome [41-43]. If autophagy is working properly, p62 should also be degraded in the process. Therefore, measurement of endogenous p62 can be used to quantitatively assess the flow of autophagy of protein aggregates [44,45] serving as a readout of autophagic degradation efficiency [5]. The hypothesis that GLaz might be promoting the resolution of autophagy predicts a reduction in p62 accumulation. Levels of p62 show a 30-40% decrease upon GLaz over-expression (Figure 7A). Moreover, p62 protein levels increase upon rapamycin-induced autophagy in both, wild type control flies (Figure 7B) and in flies undergoing photoreceptor neurodegeneration (Figure 7C), as expected by the known transcriptional up-regulation of p62 in autophagy-inducing conditions [39]. Over-expression of GLaz reduces p62 levels in the SCA1 model with or without further induction of autophagy by rapamycin (Figure 7A,C). The effect of GLaz over-expression on autophagy markers does not support its role in clearance of aggregate-prone proteins by an autophagy-independent proteasome mechanism.

However, we need to discern whether p62 protein level changes are accompanied by transcriptional changes upon GLaz over-expression. We then measured the p62 transcript levels by qRT-PCR (Figure 7D) and found no significant changes upon hATXN1^82Q or GLaz over-expression alone, while a 25% decrease is observed by over-expressing GLaz in the SCA1 model compared to the neurodegenerative condition. These results show that GLaz over-expression results in both p62 mRNA and protein levels decrease in the SCA1 model.

In basal conditions, when no neurodegeneration-triggered autophagy activity is present, GLaz null mutant flies show higher levels of p62 protein (Figure 8A), while over-expressing GLaz under basal conditions does not change p62 protein levels (Figure 8B). Transcript levels of p62 slightly decrease in GLaz null mutants, but do not change significantly by GLaz gain-of-function in the absence of neurodegeneration (Figure 8C). Atg8a transcript levels increase in GLaz null mutant flies, but no changes in Atg8a mRNA levels are observed in flies over-expressing GLaz (Figure 8D). These results suggest, on the one hand, the existence of a GLaz-dependent effect on basal autophagy progression, since the absence of GLaz expression results in increased Atg8a expression and accumulation of p62 without a parallel increase in p62 transcription. On the other hand, increasing GLaz expression levels does not affect, by itself, the expression of genes monitoring autophagic activity or the accumulation of p62 protein in the absence of a deteriorating cellular context.

We used four genetic manipulations to decrease autophagy activity in our SCA1 Drosophila model: Atg1^K38Q, a kinase-negative form of Atg1 that inhibits starvation- and rapamycin-induced autophagy [46] by impairing the induction of autophagosome formation; Atg4a^C98A, a dominant-negative form of Atg4a suppressing autophagy by defective processing of Atg8a/LC3 [47]; RNAi silencing of Atg8a (JF02895 line) [Dr. Juhász personal communication] and of p62 (KK105338 line) [47], that reduce Atg8a or p62 up-regulation respectively and impair the replenishment of both proteins when they are consumed by autophagic activity. The four constructs were expressed with the UAS/GAL4 system in the neurodegenerating photoreceptors with or without over-expression of GLaz (Figure 9).

All forms of autophagy reduction tested produce a significant rescue of photoreceptor degeneration (Figure 9), suggesting that in this neurodegenerative condition, autophagy must be contributing to neuronal death. These data, together with the observed accumulation of ubiquitinated proteins upon autophagy induction in the SCA1 model flies, suggest that hATXN1^82Q expressing cells are suffering autophagic stress [6,7]. GLaz over-expression under all four conditions of autophagy down-regulation results in a rescue of the rough eye phenotype, that is either similar or even slightly better than when autophagy is not genetically modified. No significant additive effect of both rescuing manipulations is observed, thus reflecting epistatic relationship between GLaz and autophagy control genes.