Hsp40 overexpression in pacemaker neurons delays circadian dysfunction in a Drosophila model of Huntington's disease Open Access

Huntington's disease (HD) is a neurodegenerative disease (ND) due to a dominant mutation in the huntingtin gene (HTT), leading to the expansion of the glutamine (Q) amino acid repeat tract in the huntingtin protein (HTT) beyond a threshold of 35-40 polyglutamine (polyQ) repeats. It shares several features with other NDs (Gusella and MacDonald, 2006; Bates et al., 2015), such as a typical middle-age onset, specificity of brain areas affected, motor and cognitive impairments, psychiatric disturbances, a progressive worsening with age, and decline in the quality of life and longevity. The presence of disease protein aggregates, aberrant proteostasis, synaptic toxicity, oxidative stress and neurodegeneration are shared pathophysiological mechanisms.

Circadian and sleep disruptions are now recognised as early symptoms of many NDs, including HD (Goodman et al., 2011; Morton et al., 2014; Lebreton et al., 2015; Bellosta Diago et al., 2017). The mammalian clock centre suprachiasmatic nucleus (SCN) is affected in HD mice, including molecular clock disruptions and reduction in the vasoactive intestinal peptide, a clock output neuropeptide (Morton et al., 2005; Maywood et al., 2010; Kudo et al., 2011; van Wamelen et al., 2013). Emerging evidence supports bi-directional crosstalk between the circadian and neurodegenerative axes, with circadian function impacting the aetiology and progression of NDs (Hood and Amir, 2017; Leng et al., 2019; Carter et al., 2021; Voysey et al., 2021). Improving clock function and sleep in HD mice have been neuroprotective (Pallier et al., 2007; Maywood et al., 2010; Ouk et al., 2017; Wang et al., 2017; Whittaker et al., 2018), whereas clock disruptions worsen ND (Krishnan et al., 2012; Lauretti et al., 2017; Kim et al., 2018; Sharma and Goyal, 2020).

Given the beneficial effects of circadian improvement on neurodegeneration, we aimed to uncover modifiers of expanded HTT (expHTT)-induced circadian arrhythmicity that could also serve as modifiers of neurodegenerative phenotypes. We have previously established and characterised a circadian model of HD in Drosophila melanogaster (Sheeba et al., 2008, 2010; Prakash et al., 2017), flies expressing an expHTT with 128 polyQ (HTT-Q128) in a subset of the pacemaker neurons, the ventral lateral neurons (LNv). The LNv express a critical circadian output neuropeptide, Pigment dispersing factor (Pdf), and are composed of the approximately four small LNv (sLNv) and four to five large LNv (lLNv) (Helfrich-Förster, 1995; Renn et al., 1999). Pdf and the sLNv are essential for locomotor activity/rest rhythms in constant darkness (DD) (Renn et al., 1999; Grima et al., 2004; Stoleru et al., 2004; Shafer and Taghert, 2009). Most flies expressing expHTT in the Pdf⁺ LNv (Pdf>Q128) exhibited disrupted behavioural activity rhythms in DD, a loss of Pdf from sLNv soma, loss of Period (Per) and its oscillations from LNv, and the presence of expHTT inclusions (Sheeba et al., 2008; Prakash et al., 2017). In these Pdf>Q128 flies, we carried out a genetic screen for modifiers that could rescue circadian behavioural arrhythmicity. These genes are grouped under different categories based on function (Table S1), and the expressed proteins assist in neuronal function and are modifiers of neurodegeneration (Steffan et al., 2001; Gunawardena et al., 2003; Shulman and Feany, 2003; Sang et al., 2005; Wyttenbach and Arrigo, 2009; Zhang et al., 2009; Sutton et al., 2013; Kampinga and Bergink, 2016; Menzies et al., 2017; Metaxakis et al., 2018). Co-expression of expHTT in the LNv with candidates from the Heat shock protein (Hsp) or autophagy group significantly improved the rhythmicity of flies compared to that of their expHTT-only expressing counterparts (Table S1). In the Hsp group, these included Hsp23, Hsp40 and HSP70 homologues. The tim>Q128,HSP70 also had better rhythmicity than tim>Q128. The central HSP70 and co-chaperone Hsp40 were chosen for further analysis.

Hsps play a central role in cellular proteostasis, aiding in protein folding, trafficking and degradation, and preventing aberrant interactions and disaggregation (Hartl and Hayer-Hartl, 2009; Kim et al., 2013; Nillegoda et al., 2018). Hsp40 and Hsp70 colocalise with mutant HTT aggregates (Jana, 2000; Muchowski et al., 2000; Kim et al., 2002; Scior et al., 2018), and their levels reduce with age, coinciding with proteostasis decline and middle-age HD onset (Hay, 2004; Ben-Zvi et al., 2009; Taylor and Dillin, 2011; Brehme et al., 2014; Hipp et al., 2019). Hsp upregulation alleviates proteotoxic stress and HD symptoms (Jana, 2000; Hansson et al., 2003; Branco et al., 2008; Labbadia et al., 2012; Brehme and Voisine, 2016; Kakkar et al., 2016), whereas their reduction aggravates neurodegenerative phenotypes in cell culture and animal models (Tagawa et al., 2007; Wacker et al., 2009; Hageman et al., 2010; Jiang et al., 2012; Scior et al., 2018). Therefore, the emergence of Hsps as modifiers in our screen is not surprising. However, the role of Hsps in circadian rehabilitation is relatively unexplored.

We investigated the role of Hsp40 and HSP70 as modifiers of expHTT-induced circadian neurodegenerative phenotypes in Drosophila. Of the two Hsps, Hsp40 emerged as the more potent modifier in delaying expHTT-induced phenotypes. Its overexpression postponed the loss of rhythmic locomotion over a substantial duration and the loss of Per and its oscillations from sLNv. Notably, there was a rescue of Pdf loss from sLNv and a decrease in the visible expHTT inclusion load favouring a new feature – a spot-like form of expHTT. HSP70 overexpression rescued rhythmicity and lowered expHTT inclusion numbers only at an early age, without rescuing Pdf or Per in the sLNv or affecting inclusions as the predominant form of expHTT in the LNv. Co-expression of Hsp40 and HSP70 in Pdf>Q128 led to a synergistic improvement in the consolidation of activity rhythms. Overall, the present study establishes a role for Hsps as suppressors of expHTT-induced circadian impairments in a Drosophila circadian HD model.

Flies expressing expHTT in LNv are arrhythmic immediately upon entering DD at 25°C (Fig. 1A, bottom left). Overexpressing Hsp40 or HSP70 in Pdf>Q128 flies delayed the onset of this arrhythmicity (Fig. 1A, bottom middle and right). Most flies co-expressing HTT-Q128 with Hsp40 were rhythmic during the early age window (henceforth, AW1) and mid-age window (henceforth, AW2), their mean rhythmicities being comparable to those of controls expressing HTT-Q0 and significantly higher than those of Pdf>Q128 (Fig. 1B). However, rhythmicity of Pdf>Q128,Hsp40 during a later age window (henceforth, AW3) declined significantly more than during the earlier AWs and when compared to their age-matched controls and was like that of Pdf>Q128. Despite a drop in rhythmicity compared to AW1, most Pdf>Q128,Hsp40 flies remained rhythmic during AW2. Pdf>Q128,HSP70 had significantly higher rhythmicity than Pdf>Q128 in AW1 and AW2 (Fig. 1B). However, it was like controls only in AW1, beyond which it declined. Notably, in AW2, although ∼50% of Pdf>Q128,HSP70 remained rhythmic, their rhythmicity was significantly lower than that of Pdf>Q128,Hsp40. In AW3, like Pdf>Q128,Hsp40 and Pdf>Q128, Pdf>Q128,HSP70 also had poor rhythmicity. Whereas the rhythmicity of Pdf>Q128,Hsp40 was like that of background controls across AW1 and AW2 (Fig. 1C, left; Fig. S2A, middle column), that of Pdf>Q128,HSP70 was control-like only during AW1 (Fig. 1D, left; Fig. S2A, right column). Unlike the sharp fall in rhythmicity of Pdf>Q128,HSP70 in AW2, Pdf>Q128,Hsp40, despite having a progressive reduction in rhythmicity with age, showed a significant fall in mean rhythmicity only in AW3 (Fig. 1B,C, left, D, left). Thus, the rescue of rhythmicity by Hsp40 lasted longer than that by HSP70.

In AW1, the rhythmic flies of Pdf>Q128,Hsp40 had robust rhythms comparable to those of most controls and with significantly higher robustness than Pdf>Q128 (Fig. 1C, middle). In AW2, the robustness of rhythmicity of Pdf>Q128,Hsp40 dropped lower than that of both parental controls and was comparable to that of Pdf>Q128. Overall, the overexpression of Hsp40 in Pdf>Q128 flies rescued both rhythmicity and rhythm robustness in AW1. In AW1, the rhythmic Pdf>Q128,Hsp40 and Pdf>Q0,Hsp40 flies had longer periods than other controls (Fig. 1A; Fig. S2A,B). The Pdf>Q128,Hsp40 flies had significantly better activity consolidation than Pdf>Q128 flies across ages (5-8 days and 11-14 days) (Fig. 1C, right). However, this improved consolidation was comparable to that of controls at 6-7 days and 12-15 days. Thus, overexpression of Hsp40 in Pdf>Q128 flies rescues both rhythmicity and rhythm strength at an early age. Activity rhythms persist until the middle age, postponing arrhythmicity onset by more than 2 weeks. These results indicate that Hsp40 is a potent suppressor of expHTT-induced circadian behavioural arrhythmicity.

The rhythmic flies of Pdf>Q128,HSP70 exhibited weaker rhythms than those of controls across AWs, and robustness in older AWs was lower than that in AW1 (Fig. 1D, middle). Controls Pdf>Q0,HSP70 and Pdf>HSP70 mostly had longer periods than rhythmic Pdf>Q128,HSP70 and Q128,HSP70 across AWs (Fig. 1A; Fig. S2A,C). The activity consolidation of Pdf>Q128,HSP70 was significantly better than that of Pdf>Q128 across 4-5 days and 6-13 days, and was comparable to that of controls at most of these ages (Fig. 1D, right). Thus, the overexpression of HSP70 in Pdf>Q128 flies improves their early-age rhythmicity and activity consolidation. In contrast to the rescue with Hsp40 overexpression, rhythm rescue with HSP70 overexpression at an early age did not extend to the middle age. Hence, HSP70 is a partial suppressor of expHTT-induced circadian behavioural arrhythmicity and is less efficient than Hsp40.

Previous studies show that co-expression of Hsp40 and Hsp70 has a synergistic effect of providing a greater effect on neurodegenerative features than expressing Hsp40 or Hsp70 individually (Chan et al., 2000; Jana, 2000; Kobayashi et al., 2000; Muchowski et al., 2000; Sittler et al., 2001; Bailey et al., 2002; Bonini, 2002; Rujano et al., 2007). Therefore, we asked whether overexpression of both Hsps provides a greater rescue (e.g. sustained robust rhythms across AWs) than expressing each alone. In AW1, flies expressing both Hsps in the presence of expHTT, i.e. Pdf>Q128,Hsp40,70 were mostly rhythmic, comparable to Pdf>Q128,Hsp40, Pdf>Q128,HSP70 and control Pdf>Q0,Hsp40,70 and significantly better than Pdf>Q128 (Fig. 2A,B, left). However, in AW2, the percentage rhythmicity of Pdf>Q128,Hsp40,70, like that of Pdf>Q128,HSP70, dropped significantly compared to AW1, while remaining higher than that of Pdf>Q128, but lower than that of Pdf>Q128,Hsp40 and Pdf>Q0,Hsp40,70 (Fig. 2B, left). In AW3, the rhythmicity percentage of Pdf>Q128,Hsp40,70 declined further and, like for single Hsp overexpression, was comparable to that of Pdf>Q128. In AW1, the rhythmic Pdf>Q128,Hsp40,70 flies exhibited robust rhythms comparable to those of control Pdf>Q0,Hsp40,70 and the single rescue Pdf>Q128,HSP70 (Fig. 2B, middle). Its period was similar to that of most other genotypes across age (Fig. 2A; Fig. S2D). Both Pdf>Q128,Hsp40,70 and Pdf>Q128,HSP70 had weaker rhythms than Pdf>Q128,HSP70 in AW1 and AW2 and than Pdf>Q0,Hsp40,70 in AW2 (Fig. 2B, middle). Interestingly, the extent of activity consolidation ‘r’ of Pdf>Q128,Hsp40,70 was significantly greater than that of Pdf>Q128 and both Pdf>Q128,Hsp40 and Pdf>Q128,HSP70 during most of the early age, while remaining comparable to that of Pdf>Q0,Hsp40,70 and Pdf>Hsp40,70 (Fig. 2B, right; Fig. S2E). Overexpression of both Hsp40 and HSP70 (Pdf>Hsp40,70) leads to a higher ‘r’ and, by extension, better-consolidated activity rhythms than those of experimental genotypes of Pdf>Q128, Pdf>Q128,Hsp40 and Pdf>Q128,HSP70 in early and middle ages, and also from control Q128,Hsp40,70 at early ages (Fig. S2E). This enhanced consolidation upon expressing both the Hsps in LNv is also reflected in the significantly higher ‘r’ of Pdf>Q0,Hsp40,70 and Pdf>Q128,Hsp40,70 than that of the other experimental genotypes. Thus, although overexpression of both Hsp40 and HSP70 in Pdf>Q128 did not have a synergistic effect on percentage rhythmicity per se, there was a synergistic improvement in early-age activity consolidation.

We then investigated whether overexpression of Hsp40 or HSP70 in Pdf>Q128 also rescues LNv cellular features. As described previously (Sheeba et al., 2010; Prakash et al., 2017) and as is also shown here, Pdf>Q128 flies had a loss of Pdf from the sLNv soma from an early age, whereas Pdf in lLNv soma was unaltered (Fig. 3, top panel set; Fig. 4A). In contrast, flies overexpressing Hsp40, the Pdf>Q128,Hsp40, showed significantly higher Pdf⁺ sLNv soma numbers than Pdf>Q128 and were indistinguishable from control Pdf>Q0,Hsp40 across ages (Fig. 3, middle panel sets; Fig. 4A, left). The shapes of the frequency distributions of Pdf⁺ sLNv numbers for Pdf>Q128,Hsp40 across age were left-skewed, like those of controls, with most hemispheres having four to five sLNv, and differed significantly from the right-skewed distribution of Pdf>Q128 (Fig. 4B). In contrast, at 3 days and 9 days, the Pdf⁺ sLNv soma numbers of Pdf>Q128,HSP70 were diminished, like those of Pdf>Q128 and significantly lower than those of control Pdf>Q0,HSP70 and Pdf>Q128,Hsp40 (Fig. 3, bottom panel sets; Fig. 4A, left). Mirroring the mean Pdf⁺ sLNv soma numbers was the shape of their distributions at both ages: Pdf>Q128,HSP70 was like Pdf>Q128 and different from controls (Fig. 4C). The Pdf⁺ lLNv soma numbers were comparable for all the genotypes across age (Fig. 3, ‘>’; Fig. 4A, right). Thus, overexpression of Hsp40, but not HSP70, completely rescues Pdf⁺ sLNv numbers. The sustained circadian rhythm rescue in Pdf>Q128,Hsp40 accompanied by the rescue of the circadian output neuropeptide Pdf in the soma of pacemaker neurons sLNv suggest Hsp40 as a disease modifier effective in restoring cellular function as well as associated behaviour. The lack of rescue of Pdf⁺ sLNv soma in Pdf>Q128,HSP70 at an early age despite the persistence of circadian activity rhythms suggests an unconventional mode of rhythm restoration by HSP70 in the absence of somal Pdf in the sLNv.

Hsps are molecular chaperones that are known to interfere with various stages of aggregation and modify the nature, conformation and solubility of expHTT inclusions (Barral et al., 2004; Wyttenbach and Arrigo, 2009; Lotz et al., 2010; Arrasate and Finkbeiner, 2012). We used immunocytochemistry and light microscopy to determine whether Hsp overexpression modifies the expHTT forms detected in the LNv of Pdf>Q128. As detailed in the Materials and Methods section, the expHTT forms present in each sLNv or lLNv were categorised into different types based on their appearance. Interestingly, Pdf>Q128 with overexpressed Hsp40 had an additional expHTT form that has not been observed in these flies and instances of which seem unreported in the literature. Visually and qualitatively, this form of HTT-Q128 appeared as a compact oval and was excluded from the cytoplasmic Pdf staining (Fig. 3, middle-right panel sets, see ‘◄’; Fig. 5A,B). We refer to this as the ‘Spot’ form of expHTT. Per at circadian time (CT)23 was mainly nuclear, and this compact Spot expHTT form appeared in the vicinity of nuclear Per and might be peri-nuclear (Fig. 5A,B). The Spot form was also restricted to a single structure per LNv. Further, the spots appeared to be present throughout the circadian cycle, and, when Per oscillations were examined, a similar number of expHTT spots was observed at both CT23 and CT11.

Only the Pdf>Q128,Hsp40 showed the presence of expHTT spots (Fig. 3). Within each hemisphere, Spot expHTT was present in ∼75% of sLNv (three of four) at 3 days and ∼50% of sLNv (two of four) at 9 days and 16 days (Fig. S3A, top). At 3 days and 9 days, nearly every hemisphere of Pdf>Q128, Hsp40 had at least one sLNv with Spot expHTT, which decreased to∼75% at 16 days (Fig. 5C, left). In lLNv of Pdf>Q128,Hsp40, Spot expHTT was absent at 3 days, detected in ∼60% lLNv per hemisphere (two to three of four) at 9 days as Spot accompanied by uniform diffuse expHTT staining (Diff+Spot) and in ∼50% lLNv at 16 days as a distinct Spot (Fig. S3A, bottom). Across samples of Pdf>Q128,Hsp40, most of the hemispheres showed the presence of Spot in at least one lLNv (as Diff+Spot or Spot) at 9 days and 16 days (Fig. 5C, left). On average, sLNv of older flies had significantly bigger spots (∼25 µm) than those of 3 days flies (∼17 µm), and in lLNv, a similar trend was seen, with larger spots at 16 days (∼20 µm) than at 9 days (∼12 µm) (Fig. 5C right). The spots in the sLNv were larger than those in age-matched lLNv (Fig. 5C, right).

Comparing the between-hemisphere distribution of predominant expHTT forms in LNv, it is apparent that expHTT appearing as puncta-like shiny specks of varying sizes, referred to as inclusions (Inc), dominated at 3 days and 9 days in both the sLNv and lLNv of Pdf>Q128 and Pdf>Q128,HSP70 (Fig. 3, Fig. 6A,B). In contrast, in the LNv of Pdf>Q128,Hsp40, non-inclusion forms of expHTT like Diff and Spot dominated over exclusively Inc (Fig. 6A,B). At both ages, the overall distributions of expHTT forms in sLNv and lLNv of Pdf>Q128,Hsp40 differed significantly from those of Pdf>Q128 and Pdf>Q128,HSP70 (Fig. 6A). We then compared the relative proportion of hemispheres in various pairwise category combinations. Specifically, at 3 days, the proportion of hemispheres dominated by Inc expHTT in sLNv relative to Spot forms was significantly higher in Pdf>Q128 and Pdf>Q128,HSP70 than that in Pdf>Q128,Hsp40, which mostly had hemispheres dominated by Spot expHTT and to a lesser extent Spot+Inc in the sLNv (Fig. S3B, top). At 9 days, nearly 50% of Pdf>Q128,Hsp40 hemispheres still had Spot-enriched sLNv either as Spot or Spot+Inc expHTT, while a similar proportion of hemispheres also had Inc-enriched sLNv (Fig. S3B, bottom). It is of note that, by 9 days, more than 50% of Pdf>Q128 and Pdf>Q128,HSP70 had no Pdf⁺ sLNv, with a mean number of approximately one, whereas nearly all Pdf>Q128,Hsp40 had four Pdf⁺ sLNv (Fig. S3B, mean Pdf⁺ LNv numbers are indicated at the bottom of each bar).

The proportion of hemispheres with Inc-enriched lLNv relative to other forms of expHTT was significantly higher in Pdf>Q128 and Pdf>Q128,HSP70 than Pdf>Q128,Hsp40 at both 3 days and 9 days (Fig. S3C). Pdf>Q128,Hsp40 at these ages mostly favoured hemispheres dominated by lLNv enriched with non-Inc forms. At 3 days, most of the hemispheres of Pdf>Q128,Hsp40 had Diff expHTT in lLNv, and, by 9 days, Spot expHTT appeared, giving rise to hemispheres predominated by mostly non-Inc expHTT forms in lLNv, namely Diff+Spot and Diff+Spot+Inc (Fig. S3C). Thus, the inclusion form of expHTT predominates over other forms in LNv across age in Pdf>Q128 and Pdf>Q128,HSP70, whereas in Pdf>Q128,Hsp40, diffuse and spot forms are prevalent.

To track the progress of these distinct forms of expHTT over a more extended duration in the presence of Hsp40, in a separate experiment with Pdf>Q128 and Pdf>Q128,Hsp40, we quantified cellular phenotypes up to 16 days. Like the previous experimental results (Fig. 6A,B), across age, the relative proportions of hemispheres with different expHTT forms in sLNv and lLNv differed significantly between genotypes (Fig. 6C,D). Most of the Pdf>Q128 hemispheres had Inc-enriched sLNv across age and Inc-enriched lLNv at 9 days and 16 days; in Pdf>Q128,Hsp40, Spot expHTT in various combinations dominated the LNv (Fig. 6C,D). In Pdf>Q128, as the flies aged, there was a significant reduction in the proportion of hemispheres predominating in Diff or diffuse expHTT with a few puncta-like inclusions (Diff+Inc) expHTT in lLNv relative to those predominating in Inc expHTT (Fig. S4A). In hemispheres of Pdf>Q128,Hsp40, Spot-enriched sLNv were present across age (Fig. S4B), and, with age, the proportion of hemispheres with predominantly Spot-enriched sLNv relative to Inc-enriched diminished (Fig. S4B, left). Pdf>Q128,Hsp40 showed a significant change across age in the proportion of hemispheres dominated by Diff-enriched lLNv relative to those dominated by lLNv enriched by other expHTT forms (Fig. S4C, top row). Post-3 days, Diff expHTT steeply declined in lLNv, making way for Diff+Spot, Diff+Spot+Inc and Inc at 9 days and Spot, Spot+Inc and Inc at 16 days (Fig. S4C, top row). From 9 days to 16 days, the relative proportions of hemispheres of Diff+Spot- (and Diff+Spot+Inc)-enriched lLNv to that of Inc-enriched lLNv decreased significantly (Fig. S4C, middle row, first and second panels). Concomitantly, the relative proportions of hemispheres of Spot (and Spot+Inc)-enriched lLNv to that of Inc-enriched lLNv increased significantly (Fig. S4C, middle row, third and fourth panels). Thus, the overexpression of Hsp40 in Pdf>Q128 flies decreases the expHTT inclusions in LNv, in favour of expHTT spots. HSP70 overexpression, on the other hand, did not decrease expHTT inclusions in LNv.

In summary, Hsp40 overexpression improves LNv health by reducing inclusions of expHTT in favour of a new form of expHTT, the ‘Spot’, and preserving Pdf⁺ sLNv. The Spot expHTT might be a relatively less toxic form of expHTT, given the control-like Pdf⁺ sLNv numbers of Pdf>Q128,Hsp40. Also, Pdf>Q128,HSP70 and Pdf>Q128 were nearly indistinguishable in the dominance of expHTT inclusions in LNv, suggesting that the mechanisms mediating early-age rhythm restoration upon HSP70 overexpression might not involve mitigation of visible inclusions.

We quantified the number and size of expHTT inclusions found in and around the LNv. At both 3 days and 9 days, Pdf>Q128,Hsp40 had significantly fewer inclusions than Pdf>Q128 and Pdf>Q128,HSP70 (Fig. 6E, left). At 3 days, Pdf>Q128,HSP70 also had fewer inclusions than Pdf>Q128, but not at 9 days. Both Pdf>Q128,Hsp40 and Pdf>Q128,HSP70 exhibited an increase in inclusion number with age. Altogether, these results indicate that overexpression of Hsp40 or HSP70 in Pdf>Q128 reduces expHTT inclusion numbers, with Hsp40 having lasting effects.

The mean inclusion size of Pdf>Q128,Hsp40 was higher than that of Pdf>Q128 and Pdf>Q128,HSP70 at 3 days, which is likely a reflection of including the relatively large-sized expHTT spots among inclusions during quantification (Fig. 6E, right). Surprisingly, at 9 days, Pdf>Q128 had smaller inclusions than those of Pdf>Q128,Hsp40 and Pdf>Q128,HSP70. Inclusion size of Pdf>Q128 and Pdf>Q128,HSP70 increased with age.

In summary, co-expression of expHTT with Hsp40 in the LNv decreases the proportion of hemispheres with inclusion-enriched LNv across age, with a concomitant increase in the proportion of hemispheres enriched in a hitherto unreported Spot form of expHTT in LNv and a decrease in the expHTT inclusion numbers. All the above observations, taken together, will be henceforth referred to as a decrease in the ‘inclusion load’. Thus, Hsp40 overexpression in Pdf>Q128 flies reduces the inclusion load of the LNv.

Per, a central clock protein, is lost from the soma of LNv in Pdf>Q128 flies (Prakash et al., 2017), also seen here, with Pdf>Q128 having significantly fewer Per⁺ sLNv soma at 3 days and 9 days and almost none at 16 days (Fig. 7A, leftmost column, B, top; Fig. S4D, left). We addressed whether the neuroprotective effect of Hsp40 overexpression on Pdf>Q128 flies extends to loss of Per and its molecular oscillations in the LNv. Pdf>Q128,Hsp40 showed the presence of Per⁺ sLNv and lLNv soma at 3 days and 9 days, with control-like numbers (Fig. 7A, left panel sets, B) and their frequency distributions were left-skewed like for Pdf>Q0,Hsp40 and differed from Pdf>Q128 (Fig. 7C, left and middle columns). However, unlike the rescue of Pdf in sLNv soma, Per rescue in the LNv soma was not sustained up to 16 days, by which time Pdf>Q128,Hsp40 had significantly fewer Per⁺ sLNv and lLNv than those of control and was comparable to Pdf>Q128 (Fig. 7B; Fig. S4D). The shape of the frequency distribution of Per⁺ sLNv and lLNv soma numbers in Pdf>Q128,Hsp40 changed from a control-like left-skew at 9 days to a Pdf>Q128-like shape at 16 days (Fig. 7C, left and middle panel sets). Pdf>Q128,HSP70 did not show rescue of Per⁺ sLNv soma across age. Its mean numbers and distribution shapes were comparable to those of Pdf>Q128 and significantly differed from those of Pdf>Q0,HSP70 and Pdf>Q128,Hsp40 (Fig. 7A, right panel sets, B, top, C, right panel sets). Per⁺ lLNv soma numbers of Pdf>Q128,HSP70 were comparable to those of Pdf>Q128 at 3 days and 9 days, control-like at 3 days and significantly reduced at 9 days (Fig. 7B, bottom). The shape of the Per⁺ lLNv soma distribution of 9-day-old Pdf>Q128,HSP70, like that of Pdf>Q128, differed from the left-skewed distribution of Pdf>Q0,HSP70 (Fig. 7C, bottom right).

Because Pdf>Q128,Hsp40 showed control-like Per⁺ sLNv soma numbers at 3 days and 9 days, Per oscillations in LNv were assessed at these ages. At 3 days, Pdf>Q128 did not show Per oscillations in sLNv; Pdf>Q128,Hsp40 showed a significant oscillation in Per levels like Pdf>Q0,Hsp40 (Fig. 7A, top-left panel sets, D, top left). The Per intensity in sLNv of Pdf>Q128 was significantly lower than for the other two genotypes at CT23. However, at 9 days, despite having control-like sLNv numbers and Per in the sLNv, Per oscillation was absent in sLNv of Pdf>Q128,Hsp40, with intensity at CT23 being significantly diminished compared to that of Pdf>Q0,Hsp40 (Fig. 7A, bottom-left panel sets, D, top middle). At 3 days, Per oscillation was seen in lLNv of Pdf>Q128, Pdf>Q128,Hsp40 and Pdf>Q0,Hsp40 (Fig. 7A, top-left panel sets, D, bottom left). At 9 days, Per oscillations were absent from lLNv of both Pdf>Q128,Hsp40 and control Pdf>Q0,Hsp40 (Fig. 7A, bottom-left panel sets, D, bottom middle), as is reported for wild-type flies (Shafer et al., 2002; Veleri et al., 2003). Per oscillations in LNv of Pdf>Q128,HSP70 were like Pdf>Q128, with no Per oscillation in sLNv at 3 days and a significant oscillation in lLNv (Fig. 7A, top-right panel sets, D, right column). HSP70 overexpression did not rescue Per in sLNv, even in young Pdf>Q128 flies. Thus, in young expHTT-expressing flies, Hsp40 overexpression restores both Per numbers and Per oscillations in the sLNv. This study is the first thus far to report rescue in circadian molecular oscillations accompanying the restoration of behavioural rhythms observed in these flies, underscoring the effectiveness of Hsp40 as a potent circadian modifier in HD.

Hsp40 overexpression in LNv of expHTT flies leads to the rescue of sLNv circadian clock output, molecular oscillations and their associated behavioural rhythms in young flies. The rescue of behavioural rhythms and the clock output Pdf is long-lasting. There is also a considerable reduction in the expHTT inclusion load. This sustained rescue at multiple levels posits Hsp40 as a potential therapeutic candidate for improving circadian health under neurodegenerative conditions.

The role of Hsps, known modifiers of neurodegeneration, in HD-associated circadian disturbances is relatively unexplored. In this study, we show that the overexpression of the co-chaperone Hsp40 in circadian pacemaker neurons of Drosophila delays expHTT-induced circadian behavioural arrhythmicity over extended durations and suppresses circadian neurotoxicity. Overexpression of the central chaperone HSP70 also mitigated expHTT-induced circadian behavioural rhythm disruptions in young flies but did not rescue cellular phenotypes. The rescue upon Hsp40 overexpression was more robust, pronounced and sustained. In young flies, Hsp40 overexpression seemed to restore the functionality of LNv, particularly the central pacemaker sLNv. Evidence for sLNv functionality is the restoration of circadian proteins, the core-clock protein Per, its oscillations and the output neuropeptide Pdf in the sLNv soma, and lowered expHTT inclusion load, leading to an overall improvement in the LNv circuit-associated behavioural rhythms. These flies continued to be behaviourally rhythmic up to 16 days but with lowered robustness, with the presence of Per in LNv and control-like Pdf⁺ sLNv soma numbers and diminished inclusion load, but without Per oscillations. The persistence of activity rhythms in the absence of Per oscillations in sLNv suggests two conclusions. First, that Hsp40-mediated rescue at the circadian output level seems sufficient for behavioural rhythm rescue. Second, the Per oscillations in the sLNv might be dispensable for rhythm sustenance. Two recent studies that support this reasoning show that Per in LNv does not seem necessary for the persistence of free-running activity rhythms but is vital for rhythm strength (Delventhal et al., 2019; Schlichting et al., 2019). In relatively older flies, despite the presence of nearly four Pdf⁺ sLNv soma (16 days) and reduction in the inclusion load, the flies were arrhythmic during AW3 (16-23 days). Thus, the Hsp40 neuroprotection does not seem sufficient as the flies age, contributing to a deterioration of LNv health. The inadequacy of Hsp40 expression to extend protection to LNv for extended durations suggests two conclusions. First, the restoration of Pdf⁺ sLNv does not guarantee sustained free-running rhythms in the absence of Per. Our previous results show that ∼20% of 7-day-old arrhythmic Pdf>Q128 flies had at least one to two Pdf⁺ sLNv. Also, the Pdf levels in the sLNv dorsal projections were oscillating and functional in synchronising downstream circadian neurons (Prakash et al., 2017). The previous results and present observations of behavioural arrhythmicity in older flies despite Pdf rescue in sLNv indicate that sLNv Pdf, in the absence of Per, is insufficient for rhythmic activity. Second, over time, neuroprotective benefits offered by Hsp40 can be overwhelmed upon HD progression, probably by the age-related burden on LNv proteostasis, rendering the cells vulnerable to expHTT toxicity. Therefore, sustained rhythm rescue might require supplementing Hsps with further enhancement of proteostasis via proteasomal or autophagic upregulation.

HSP70 overexpression, in contrast, showed a rescue in only early-age rhythms and activity consolidation, albeit of lowered robustness, a decrease in early-age inclusion number, without the rescue of Pdf⁺ sLNv numbers or Per oscillations in sLNv or alterations to inclusions being the prevalent form of expHTT in LNv. The rhythmic flies of Pdf>Q128,HSP70 have poor rhythm robustness and can be attributed to the absence of Per rescue in the LNv. However, the persistence of behavioural rhythms with HSP70 overexpression in the absence of Pdf rescue in the sLNv soma is intriguing. It suggests that the presence of Pdf in sLNv is not necessary for behavioural rhythm restoration. Other studies in Pdf>Q128 have reported rhythm rescue with only marginal Pdf restoration in sLNv soma upon Atx2 or Hop downregulation (Xu et al., 2019b; Xu et al., 2019a). Together with ours, these reports suggest that other mechanisms might drive circadian behavioural rhythms without canonical circadian cellular proteins. Possible intersections of Hsp onto improving LNv function and output in orchestrating rhythmicity are circadian oscillations in arborisations of the sLNv termini and secondary molecular loop components, non-Per driven clocks, LNv membrane properties, neuronal firing, synaptic strength and network-level communication (Edgar et al., 2012; Beckwith and Ceriani, 2015; Yao et al., 2016; Rey et al., 2018; Bulthuis et al., 2019).

Circadian disturbances in HD stem from perturbations to the circadian organisation's input, molecular oscillator and output components (Fifel and Videnovic, 2020; Colwell, 2021). In R6/2 HD mice, free-running activity rhythms are disrupted, and the SCN molecular oscillations are impaired in vivo, while persisting in organotypic slices, suggesting that the inputs to and outputs from the central clock are affected rather than the molecular clock (Morton et al., 2005; Pallier et al., 2007). Dysfunctional intrinsically photosensitive retinal ganglion cells, reduction of VIP immunostaining in the SCN, and disrupted rhythms in SCN electrophysiology, cortisol, melatonin, body temperature, heart rate and metabolic outputs (Smarr et al., 2019; Fifel and Videnovic, 2020; Colwell, 2021) provide evidence for circadian disturbances in HD mice at levels of clock input and output, thus also affecting molecular clockwork in vivo, resulting in overt behavioural and peripheral rhythm disturbances. Although the HD flies used in this study had a subset of their clock neurons targeted, they exhibited a definite circadian disturbance in the overt behavioural rhythms, molecular oscillations, and circadian output neuropeptide Pdf, recapitulating the central clock and circadian output impairments seen in vivo in HD mice. Such parallels between model systems suggest the possibility of finding mammalian counterparts to the Hsp40-mediated rescue of circadian disturbances. Whether the benefits of Hsp40 treatment extend to other circadian rhythms remains to be elucidated.

Hsps dilute the presence of aggregate-prone proteins by interfering with the aggregation pathway by delaying nucleation, fibril elongation or redirecting the pathway towards less-toxic versions, sequestrating intermediates into cellular compartments or organelles and targeting for degradation (Kampinga and Bergink, 2016; Mannini and Chiti, 2017; Hipp et al., 2019). The effects of Hsps on aggregation vary, depending on a host of factors such as the definition of aggregates, their nature and conformation, the cellular context, the stage of aggregation, age and disease stage, quantification method and model system. Indeed, with upregulation of Hsps (Hsp40 and Hsp70) in HD models, there is evidence for differential effects: many show a decline in aggregation (Jana, 2000; Zhou et al., 2001; Hay, 2004; Guzhova et al., 2011; Labbadia et al., 2012; Maheshwari et al., 2014; Scior et al., 2018), some show no effect (Wyttenbach et al., 2000; Karpuj et al., 2002; McLear et al., 2008), and one study shows an increase in aggregation (Wyttenbach et al., 2000).

The present study defines the visible puncta-like clumped appearance of HTT-Q128, as detected under a fluorescence light microscope using immunocytochemistry, as an inclusion. This definition excludes detecting many species below the resolution limit and does not distinguish based on solubility and other biochemical features. Hence, the inferences are limited to the size range and gross features of particles detected via an epifluorescence scope. However, this does not take away the validity of the effect of Hsp40 overexpression on expHTT inclusions and its impact on LNv function at the cellular and behavioural stages. In Pdf>Q128 flies, Hsp40 overexpression in the LNv leads to a reduction in the number of visible expHTT inclusions, a decline in the dominance of inclusion form of expHTT, and the appearance and dominance of expHTT spots. Accompanying them are improvements in LNv pacemaker function, as evidenced by the re-establishment of circadian molecular and behavioural rhythms. Thus, a decreased inclusion load and the dominance of expHTT spots could lead to enhanced functionality of the LNv.

A pictorial representation of the locomotor behaviour and the LNv cellular phenotypes comparing Pdf>Q128 with Pdf>Q128,Hsp40 is shown (Fig. 8). A clear pattern for expHTT forms in LNv with age emerges. For the toxic Pdf>Q128 and relatively less-toxic Pdf>Q128,HSP70, Diff+Inc expHTT in lLNv at an early age gives way to Inc only at later ages. In the neuroprotective Pdf>Q128,Hsp40, across ages, Spots are present in the sLNv, dominating over Inc at 3 days, whereas Inc dominates at later ages. In the lLNv of Pdf>Q128,Hsp40, Diff expHTT dominates at 3 days, giving way to combinations of diffuse, spot and inclusion at 9 days, and then to non-diffuse expHTT (Spot and Spot+Inc) at 16 days. In Pdf>Q128,Hsp40, the continued presence and domination of Spot expHTT in LNv is associated with intact Pdf⁺ sLNv and behavioural rhythmicity of most Pdf>Q128,Hsp40 flies up to 16 days. Together, these results indicate an association between specific expHTT forms predominating in LNv and LNv health, namely diffuse and spot forms with healthy LNv and rhythmic activity, and inclusions with poor LNv function and arrhythmicity.

HSP70 overexpression in Pdf>Q128 flies rescues early-age rhythms and reduces expHTT inclusion numbers, but with the dominance of inclusion as the main expHTT form in LNv, suggesting that HSP70-mediated improvements to LNv health are via inclusion-independent mechanisms. HSP70 serves aggregation-independent neuroprotective roles such as inhibiting apoptosis (Beere, 2004; Kennedy et al., 2014), combating inflammation (Borges et al., 2012; Dukay et al., 2019), reducing reactive oxygen species and oxidative stress (Kalmar and Greensmith, 2009; Wyttenbach and Arrigo, 2009), and supporting synaptic function (Deane and Brown, 2016; Gorenberg and Chandra, 2017). Studies supporting such aggregation-independent neuroprotection by Hsp40 and Hsp70 in HD are reported (Zhou et al., 2001; Wyttenbach, 2002; Borrell-Pages et al., 2006; Wacker et al., 2009). HSP70 versatility in enhancing neuronal health and function can be attributed to the circadian rhythm improvements in the absence of an effect on the inclusion load and Pdf restoration in the sLNv.

Upon overexpression of Hsp40 in Pdf>Q128 flies, a new form of expHTT with a spot-like appearance located close to the nucleus and seemingly overlapping with the nuclear Per was observed. We discuss the possible significance of the Spot expHTT. In eukaryotes, aggregate-prone proteins are often sequestered into specialised cellular compartments that are thought to be neuroprotective and are typically membraneless, sometimes referred to as sequestrosomes, or into membrane-bound organelles (Sontag et al., 2014; Tan and Wong, 2017; Johnston and Samant, 2021). A few examples of such spatially sequestered quality control sites are the cytoplasmic Q-bodies or stress foci, cytoplasmic p62 bodies or aggresome-like induced structures, peri-nuclear juxta-nuclear quality control compartment, intra-nuclear quality control compartment, peri-nuclear aggresomes and peri-vacuolar insoluble protein deposit (Tan and Wong, 2017; Johnston and Samant, 2021). Hsps are known to participate in such compartmentalisations (Nollen et al., 2001; Specht et al., 2011; Escusa-Toret et al., 2013; Miller et al., 2015). The aggresomes are mostly juxta-nuclear, membrane-free inclusions carrying ubiquitinated misfolded proteins formed at the microtubule organising centre (Johnston et al., 1998; Kopito, 2000; Johnston and Samant, 2021). There is evidence for colocalisation of Hsp40 and Hsp70 with aggresomes (García-Mata et al., 1999; Junn et al., 2002; Gamerdinger et al., 2011; Zhang and Qian, 2011).

Another observation here is that the average size of an expHTT Spot in the small LNv is significantly larger than those in the large LNv. This sizeable expHTT Spot in the sLNv could reflect a greater expHTT burden and toxicity in the vulnerable sLNv or a longer duration of expression of HTT and Hsp40 owing to their earlier appearance than the lLNv during development (Helfrich-Forster, 1997). The presence of three to four Pdf⁺ sLNv in nearly every hemisphere of Pdf>Q128,Hsp40 across age parallels with the presence of at least one sLNv per hemisphere having Spot expHTT (proportion of hemispheres with at least one Spot⁺ sLNv: 3 days, 100%; 9 days, ∼97%; 16 days, ∼79%). Such co-occurrences indicate that the appearance of expHTT spots might be protective. Thus, Hsp40 might modify the nature of expHTT inclusions, and the spots could represent a less reactive and relatively benign form of expHTT, contributing to an enhancement in LNv health and function.

Many studies in PolyQ models have described a synergistic effect upon co-expression of Hsp40 and Hsp70 on the toxicity: expression of both offered better protection than either alone (Chan et al., 2000; Jana, 2000; Kobayashi et al., 2000; Muchowski et al., 2000; Sittler et al., 2001; Bailey et al., 2002; Bonini, 2002; Rujano et al., 2007). The current study shows a synergistic improvement in young HD flies' circadian rhythms in daily activity consolidation, but not in percentage rhythmicity or robustness. The synergistic effect seems subtle, evident with a daily readout like ‘r’, but not with a 7 days overt readout of rhythmicity. Over time, there was a decline in rhythm robustness of Pdf>Q128 flies expressing both Hsps than those expressing Hsp40 alone, suggesting that co-expression of multiple Hsps could become detrimental over time. In another study, such co-expression eliminated the survival benefit of HSP70-only expression, enhancing cell death (Ormsby et al., 2013). Some of the drawbacks of Hsp co-expression and Hsp overexpression, like their pro-carcinogenic effect and generation of seeding-competent nuclei, call for caution when targeting central Hsps for therapy (Jäättelä, 1995; Nylandsted et al., 2002; Tittelmeier et al., 2020b).

In our study, Hsp40 emerges as a superior suppressor of most expHTT-induced phenotypes examined and in terms of the duration of rescue. There is substantial support for Hsp40 being a more effective HD neurotoxic modifier. Among the HSP40, HSP70 and HSP110 chaperone families, the DNAJB class of HSP40 emerged as the most potent protector against PolyQ toxicity (Hageman et al., 2010), and, in R6/2 mice, Hsp70 suppressed HD only moderately (Hansson et al., 2003; Hay, 2004; Popiel et al., 2012), while Hsp40 members had better success (Labbadia et al., 2012; Kakkar et al., 2016). Hsp40 also prevented the secretion of expanded PolyQ proteins from cultured cells (Popiel et al., 2012), thus likely preventing their cell-to-cell transmission, an emerging concern in NDs. Findings from the present study and other studies (Chai et al., 1999; Zhou et al., 2001; Rujano et al., 2007; Ormsby et al., 2013) show that Hsp40 reduces aggregation more often than Hsp70. Hsp40 is rate limiting in the suppression and reversal of expHTT aggregation by disaggregases (Rujano et al., 2007; Scior et al., 2018), and some members can act without requiring Hsp70 (Hageman et al., 2010; Kuo et al., 2013; Månsson et al., 2013; Kakkar et al., 2016). A study of genetic modifiers of different NDs, including HD in different model organisms, revealed Drosophila DnaJ-1 (and its mammalian ortholog Dnajb4) as a modifier across many NDs (Na et al., 2013). The significant role of the DNAJB protein family in synaptic health and neuronal proteostasis, and their diversity in function, distribution and substrate specificity, underscore their usefulness in directed therapy while minimising the side effects (Chuang et al., 2002; Westhoff et al., 2005; Gibbs et al., 2009; Kampinga and Craig, 2010; Gao et al., 2015; Nillegoda et al., 2018; Kampinga et al., 2019; Tittelmeier et al., 2020a).

In an in vivo system, we have shown a neuroprotective role of chaperone Hsp40 in rescuing HD-induced circadian deficits and neurotoxicity at multiple levels and across the temporal scale. Such a multi-level associated functional rescue offers an edge over conventional non-associated cellular and behavioural rescues such as to make better cause–effect inferences due to reduced off-target effects, rigorously test the robustness and versatility of the modifying treatment, and serve as proof-of-principle evaluations. Also, although most of the candidates screened are known modulators of neurotoxicity, only two groups of proteins emerged as potent suppressors of circadian disruption, uncovering a gap in establishing circadian-specific neuroprotective agents.

There is ample evidence for clock-controlled regulation of proteostasis components, including chaperones (Desvergne and Friguet, 2017; Ryzhikov et al., 2019; Wang et al., 2020), with both Hsp40 and Hsp70 isoforms showing rhythmic gene expression across taxa (Li et al., 2017). In the Drosophila LNv, mRNA transcripts of the Hsp40 isoforms, DnaJ-1 and the DnaJ-H are present, and, in the lLNv, DnaJ-H gene expression is cyclic (Kula-Eversole et al., 2010; Ma et al., 2021), while Hsp70 transcripts cycle in the LNv (Abruzzi et al., 2017; Ma et al., 2021). Further, Hsp40 gene transcripts have been classified under the experimentally identified circadian genes (Li et al., 2017). The converse, i.e. proteostasis affecting the molecular clock via post-translational modifications and autophagy, is also prevalent (Mehra et al., 2009; Stojkovic et al., 2014; Toledo et al., 2018; Juste et al., 2021). However, very few studies have assessed the role of Hsps in circadian maintenance and its deterioration in NDs, especially in animals. In Drosophila, the Hsp70/Hsp90-organising protein (Hop) improved rhythmicity in an HD model (Xu et al., 2019b), Hsp70 expression overcame arrhythmicity due to Gal4-overexpression in the LNv (Rezaval et al., 2007), Hsp90 disruptions led to the loss of activity rhythms without affecting molecular oscillations (Hung et al., 2009), and Hsps were indirectly implicated in circadian behaviour (Benbahouche et al., 2014; Means et al., 2015). In mouse fibroblasts, Hsp90 is required for circadian rhythmicity, while Hsf1 and endoplasmic reticulum Hsp70 strengthen rhythms post-stress (Tamaru et al., 2011; Schneider et al., 2014; Pickard et al., 2019). The present findings of a relatively novel role for the Hsps in protecting against ND-induced circadian dysfunction and the above studies encourage further research on Hsps in circadian function. Given that circadian and sleep disturbances occur early and are pre-manifest in HD (Soneson et al., 2010; Morton et al., 2014; Lazar et al., 2015; Bellosta Diago et al., 2017), treatments targeting HSPs could impact HD's early stages in postponing symptoms and provide a meaningful therapeutic impact. An ageing population worldwide has increased the prevalence of NDs (Gitler et al., 2017; Lassonde et al., 2017; Béjot and Yaffe, 2019). Given the pivotal roles of proteostasis and circadian health in NDs, studying the involvement of molecular chaperones in circadian maintenance will significantly improve our understanding of ND progression and treatment.

The UAS-HTTpolyQ lines used in this study, namely w;+;UAS-HTT-Q0A;+ (without Q repeats) and w;+;UAS-HTT-Q128C;+ (with 128 polyQ repeats), were a gift from Troy Littleton, Massachusetts Institute of Technology (Lee et al., 2004). They contain the first 548 amino acids of the human HTT gene. The UAS males were crossed with virgin females of either w;PdfGal4:+ or w¹¹¹⁸:+:+ (BL 5905) to generate fly lines expressing HTT-polyQ in the Pdf⁺ LNv neurons designated Pdf>Q0 and Pdf>Q128 or the UAS background controls designated Q0 and Q128, respectively. In some instances, a broader circadian driver, the w;timGal4:UAS-GFP was used for the screen. For the genetic screen, the fly lines used are listed in Table S1. The genes were either overexpressed or downregulated in the Pdf>Q128 background (and in a few cases, tim>Q128 background) (Table S1). A recombinant line of w;PdfGal4;+ and w;UAS-HTT-Q128/+;+ was generated, denoted as w;Pdf-Q128;+, and used for the screen. The sample size for the screen varied between 16 and 32 flies per genotype.

For most of this study, the fly lines of focus were generated using the two Hsp UAS lines, namely w;+;UAS-DnaJ1k (Bl 30553) and w;+;UAS-HSPA1L (Bl 7054). The human HSP70 homolog used in this study was HSPA1L (Bl 7454) (https://flybase.org/reports/FBgn0029163.html), as a recent analysis revealed that, among the various chaperone families, the HSP70 family most frequently provided considerable proteotoxic relief in a variety of protein misfolding diseases (Brehme and Voisine, 2016). The next potent proteotoxic suppressor was the Hsp40 family, among which DnaJb4 and DnaJb6 were the most potent polyQ disease modifiers (Brehme and Voisine, 2016). The dHdj1 (DnaJ1k or DnaJ-1 or Hsp40 or CG1058) (Bl 30553) used in this study is a Drosophila ortholog of members of the human HSP40 Class B family with varying degrees of homology (DNAJB5, DNAJB4 and DNAJB1) (https://flybase.org/reports/FBgn0263106.html).

The UAS-Hsp lines were used to generate the w;PdfGal4/UAS-Q128;UAS-DnaJ1k/+ or w;PdfGal4/UAS-Q128;HSPA1L/+ lines, which would respectively overexpress Hsp40 or HSP70 in the Pdf⁺ LNv neurons, also expressing HTT-Q128. The above-generated lines are referred to as Pdf>Q128,Hsp40 and Pdf>Q128,HSP70 throughout the text. Their UAS control lines are referred to as Q128,Hsp40 and Q128,HSP70 and their driver control lines as Pdf>Hsp40 and Pdf>HSP70. Their corresponding Q0 lines are Pdf>Q0,Hsp40 and Pdf>Q0,HSP70. All other relevant background controls were also used. Owing to space constraints, in some figures (Fig. 6A,C; Fig. S3B,C), Pdf>Q128, Pdf>Q128,Hsp40 and Pdf>Q128,HSP70 are abbreviated as Q128, H40 and H70, respectively. Flies co-expressing both Hsp40 and HSP70 along with HTT-Q128 in the LNv are referred to as Pdf>Q128,Hsp40,70, and their Q0 counterparts as Pdf>Q0,Hsp40,70. Crosses were maintained under 12 h:12 h light: dark cycles (LD), with ∼200 lux intensity of light phase, at 25°C. Flies were moved to DD 25°C after 2 days post-eclosion for behavioural and immunocytochemical assays. All flies and crosses were maintained on a standard cornmeal medium.

Most of the locomotor activity setup, assay conditions and analyses performed are described in a previous study (Prakash et al., 2017). Briefly, the activity rhythms of 3-day-old virgin male flies were recorded in 7 mm glass tubes using the Drosophila Activity Monitoring system 2 (DAM2) from TriKinetics (Waltham, MA, USA). Activity recordings were carried out in an incubator (Percival DR36VL) at 25°C in DD for at least 21 days (3-23 days) with ∼80% humidity. The data obtained were analysed using the Chi-Square periodogram in the CLOCKLAB software (Actimetrics, Wilmette, IL, USA). The periodogram threshold was set at P=0.01 (Pfeiffenberger et al., 2010). A fly was considered rhythmic if its periodogram amplitude was above the threshold and confirmed by visual inspection of the actogram by a single-blind analyser. The various activity rhythm features such as rhythmicity, rhythm strength or rhythm robustness and period were calculated over three 7 days windows to view progressive changes with age. The three temporal windows (AWs) spanning 21 days were designated as age window 1 (AW1; 3-9 days), age window 2 (AW2; 10-16 days) and age window 3 (AW3; 17-23 days).

Additionally, to track daily changes in activity/rest rhythms, the extent of activity consolidation ‘r’ was also calculated using a MATLAB code as previously described (Prakash et al., 2017) with a few modifications. ‘r’ represents the extent to which activity data points are consolidated over a circadian cycle of activity. The activity time series for a genotype was obtained at a resolution of 20 min bins. This time series was divided into cycles of length determined by the period (T) of that genotype, thereby identifying each circadian cycle in the time series. Thus, each ‘day’ used to calculate ‘daily’ ‘r’ was obtained as a modulo T value. On each ‘day’, activity counts were imagined as unit vectors for which direction represented the timepoint (t) at which the count occurred within the cycle. Because counts were clustered into 20 min intervals, the data can be represented as vectors with a constant angular separation of 20*2π/T radians and magnitudes corresponding to the number of activity counts in each interval. R was calculated as the magnitude of the mean of these activity vectors. The rectangular coordinates of the mean vector were obtained (Zar, 2010) using X=Σ A_t.cosθ_t/Σ A_t and Y=Σ A_tsinθ_t / Σ A_t, where A_t represents activity counts at a given timepoint t and θ_t represents the vector angle associated with the timepoint. The magnitude of this vector ‘r’ was calculated as √X²+Y². The greater the magnitude of ‘r’, the better the degree of consolidation, with most activity occurring over a few closely spaced timepoints. The lower the magnitude of ‘r’, the poorer the consolidation, with activity spread over time. Given that sometimes period changes were observed across age for a fly, daily cycles were identified separately for each 7 days AW using the mean period values for the corresponding AW. For arrhythmic flies, the cycle length was determined using the mean period of the genotype for that AW. For a fly dying in the middle of an AW, the period of that fly in the prior AW (if any) was used to determine cycle length. The daily ‘r’ was averaged across flies to obtain the mean daily ‘r’.

At least three independent activity runs were carried out for overexpression of Hsp40 with its Q0 and UAS controls, all giving similar results. For overexpression of HSP70, three independent activity runs were carried out with its Q0 controls, giving similar results. At least one experiment was carried out with all possible genetic controls for both sets of Hsp40 and HSP70 overexpression experiments. No statistical tests were carried out to determine the minimum required sample sizes. However, as recommended (Kostadinov et al., 2021), one full DAM2 monitor accommodating 32 flies per genotype was set up at the start of the experiment. The average percentage rhythmicity across multiple runs is plotted for Hsp40 and HSP70 overexpression experiments. In addition, the percentage rhythmicity, period, robustness and extent of activity consolidation ‘r’ values of a representative run with all relevant controls for each of the above overexpression experiments are plotted. For synergistic effects, a single run was carried out. There were fly deaths in AW3; therefore, AW3 analyses had fewer samples. When the fly numbers went below ten (e.g. a few instances in AW3), those genotypes were excluded from statistical analyses of robustness and period for that AW. Owing to fewer surviving flies across genotypes at later ages in the synergistic effect experiment, the ‘r’ value's statistical analyses were restricted to 16 days of age. In cases in which very few flies (n<10) were rhythmic, like Pdf>Q128 number across AWs in the HSP70 overexpression and synergistic effect experiments and during AW3 in Hsp40 overexpression experiments, or Pdf>Q128,Hsp40 during AW3 in the Hsp40 overexpression experiments or most of the experimental genotypes during AW3 in the synergistic effect experiments, those genotypes were excluded from the between-genotype statistical analyses of period and rhythm robustness for that AW. Indicated in Table S2 are the numbers of surviving flies in AW2 and AW3 across experiments.

Datasets were first tested for normality using Shapiro–Wilk's test and then for variance homogeneity using Levene's test. Across experiments, the data comparing period or activity consolidation ‘r’ between genotypes for an AW or age did not satisfy the ANOVA assumption of normality, despite transformations. So was the case for rhythm robustness in the Hs40 overexpression experiment. Therefore, the non-parametric Kruskal–Wallis test of ranks followed by multiple comparisons of mean ranks was used. For the HSP70 overexpression experiment, datasets comparing rhythm robustness between genotypes for an AW were normally distributed post-transformation, but variances were not always homogenous. Hence, Welch's ANOVA was used, followed by Games-Howell post-hoc test. For the synergistic effect experiment, the data comparing robustness between genotypes for an AW were normally distributed and their variances homogeneous. A one-way ANOVA followed by unequal N HSD tests were carried out. Friedman's test for repeated measures was used to compare robustness, period and ‘r’ between AWs or ages for a genotype. Then, Wilcoxon matched-pairs tests (or Conover Test for ‘r’) with Bonferroni correction [or Benjamini–Hochberg (BH) procedure to decrease the false discovery rate (FDR) for ‘r’; FDR set at 5%] on the pairwise P-values were used. A m×n Fisher's exact test, followed by multiple 2×2 Fisher's exact tests with BH procedure on all relevant comparisons, were used (using R) to compare the proportions of rhythmic flies between genotypes for an AW. Cochran Q test on the dichotomous variable rhythmicity (rhythmic and arrhythmic categories) was used to compare the proportion of rhythmic flies between AWs for a genotype, followed by multiple 2×2 McNemar's tests on the dependent samples and Bonferroni correction on pairwise P-values. For comparing the mean rhythmicity of multiple independent runs between genotypes or between ages, a repeated-measures ANOVA followed by Tukey's HSD was performed post-arcsine conversion of the square-root-transformed data.

Statistical analyses were executed using STATISTICA^TM 7.0 (https://statistica.software.informer.com/7.0/) and R (https://www.r-project.org/). Welch's ANOVA was performed using a Microsoft Excel template from http://www.biostathandbook.com/onewayanova.html (McDonald, 2014), McNemar's test using SciStatCalc (https://scistatcalc.blogspot.com/2013/11/mcnemars-test-calculator.html) and Friedman's test followed by Conover test for ‘r’ using ASTATSA (https://astatsa.com/FriedmanTest/).

The dissections and immunocytochemistry procedures performed were as described previously (Prakash et al., 2017). Adult fly brain tissues were dissected in 1× PBS at different ages, fixed with 4% paraformaldehyde at room temperature, blocked and stained with the appropriate primary antibodies, followed by secondary antibodies, and then mounted on slides using 70% glycerol in 1× PBS. Primary antibodies used for double staining were mouse anti-HTT (1:500; MilliporeSigma MAB2166) and rabbit anti-Pdf (1:30,000; Michael Nitabach, Yale University) (Nitabach et al., 2006), and for triple staining, rabbit anti-Per (1:20,000; Jeffrey C Hall, Brandeis University) (Houl et al., 2008), mouse anti-HTT (1:500) and rat anti-Pdf (1:3000) (Jae Park, Vanderbilt University) (Park et al., 2000). Rabbit anti-Per was pre-absorbed onto per⁰¹ embryos at 1:100 dilution. Secondary antibodies (1:3000) were from Invitrogen: anti-rabbit Alexa Fluor 488, anti-mouse Alexa Fluor 546, anti-mouse Alexa Fluor 647 and anti-rat Alexa Fluor 555.

For characterising the Per⁺ and Pdf⁺ LNv soma numbers, adult fly brain dissections were carried out in parallel at different ages. Because with Pdf>Q128,Hsp40, a sustained rescue for two AWs was seen in behaviour, dissections were carried out on 3-day-old, 9-day-old and 16-day-old flies corresponding to the beginning of AW1, the transition of AW1-AW2 and end of AW2, respectively. With Pdf>Q128,HSP70, rhythm rescue was restricted to AW1. Hence, dissections were carried out at 3 days and 9 days, corresponding to the beginning and end of AW1, respectively. For characterising HTT status in LNv, adult fly brain dissections were carried out in parallel for the five genotypes (Pdf>Q128, Pdf>Q0,Hsp40, Pdf>Q128,Hsp40, Pdf>Q0,HSP70 and Pdf>Q128,HSP70) at 3 days and 9 days. Many of the Hsp-expressing flies had periods longer than 24 h. Therefore, the mean period values of the respective genotypes were considered to calculate the CT for dissections to detect Per oscillations in the LNv. For quantifying Per oscillations in DD, LD-reared flies were dissected at CT23-24 (CT23) and CT11-12 (CT11) at different ages: all the five genotypes at 3 days and Pdf>Q0,Hsp40 and Pdf>Q128,Hsp40 also at 9 days. These samples were co-stained with anti-Pdf to aid in identifying LNv and anti-HTT. The sample sizes were determined empirically (Table S3). There was no blinding during sample preparation.

The number of Pdf⁺ and Per⁺ LNv and the form of expHTT in the LNv were noted on manual observation of the samples using a Zeiss Axio Observer Z1 epifluorescence microscope with a 63×/oil 1.4 NA objective and without blinding. The collected data were then cross-verified with images captured using a 40×/oil 1.3 NA objective as a z-stack of 1 µm intervals. The Pdf-stained sLNv and lLNv were distinguished based on their anatomical location, size and staining pattern. For quantification of Per intensity and expHTT inclusions, the lamp intensity and exposure time were kept constant across samples for an experiment. The Per intensity was calculated from the images described previously in a single-blind manner (Prakash et al., 2017). National Institutes of Health imaging software ImageJ was used for image analysis and quantification (Schneider et al., 2012). For representative images, confocal z-stacks were captured using a Zeiss LSM 880, keeping the photomultiplier tube gain gain, offset and laser power below saturating pixels. In the representative images, brightness/contrast adjustments were applied to the whole image for better visualisation of the LNv, especially the sLNv, as they showed less intense and sparser Pdf staining than the lLNv.

The numbers of sLNv and lLNv with different forms of expHTT were noted in each hemisphere. The expHTT forms appearing in each LNv soma were classified based on their appearance under the light microscope as follows: uniform diffuse expHTT staining (Diff), expHTT appearing as puncta-like shiny specks of varying sizes referred to as inclusions (Inc), diffuse expHTT with a few puncta-like inclusions (Diff+Inc) and without expHTT staining (No HTT) (Fig. S1A, top row). With overexpression of Hsp40, we also observed a hitherto unreported form of expHTT, oval with a compact appearance, henceforth referred to as ‘Spot’ expHTT or spot-like expHTT (Spot) (Fig. S1A, bottom row). Also observed less frequently was an LNv with an expHTT spot and the canonical inclusions, giving the spot a distorted appearance. Hence, such forms of expHTT were included under the Inc category. If the spot appeared amidst a diffused expHTT distribution, mostly seen in lLNv of young flies, it was designated as Diff+Spot (Fig. S1A, bottom row). Two sets of information can be gathered from such an exercise of labelling expHTT types per LNv. One is at the level of cells, namely the proportion of sLNv or lLNv with different forms of expHTT within each hemisphere. This comparison (intra-hemisphere) gives an idea of the within-hemisphere variation in LNv expHTT distribution. Our experimental replicates are at the level of hemispheres and not cells, so the within-hemisphere expHTT diversity in cells is only qualitative information. The second set of information is at the level of hemispheres. Each hemisphere was allotted a particular category depending on the most predominant expHTT form found in the sLNv (or lLNv) (Fig. S1B). The categorisation of each hemisphere (inter-hemisphere) based on predominant expHTT form in the LNv (sLNv or lLNv) was as follows: predominantly diffuse distribution (Fig. S1B, top row, left), an equal distribution of diffused and inclusions (Diff+Inc) (Fig. S1B, top row, middle and right), predominantly had Diff+Spot (Fig. S1B, second row, left), predominantly had spots (Spot) (Fig. S1B, second row, middle and right), a near equal mix of diffuse, inclusions and spots (Diff+Spot+Inc) (Fig. S1B, third row, left), an equal distribution of spot and inclusions (Spot+Inc) (Fig. S1B, third row, middle and right), or predominantly had inclusions (Inc) (Fig. S1B, fourth row, middle and right). Hemisphere-level categorisation just described is at the level of experimental replicates. It, therefore, makes room for quantitative statistical analysis and allows for comparisons between genotypes and among ages of the relative proportions of hemispheres enriched in one form of expHTT in sLNv or lLNv versus other forms of expHTT. Upon such a hemisphere-level categorisation, hemispheres in which sLNv or lLNv were without expHTT never dominated, so the ‘No HTT’ category does not exist. The arrangement order used in the figures describing various expHTT forms is based on observations made post-hoc as to the appearance and predominance of various expHTT forms in LNv over time (Fig. S1C). For example, in most Pdf>Q128 samples, sLNv show Diff expHTT as larvae, and young adults mostly exhibit Diff+Inc in lLNv, followed by Inc as they age. Young Pdf>Q128,Hsp40 flies mostly show Diff expHTT in their LNv. With age, different combinations of Diff, Spot and some Inc appear and dominate (mostly non-Inc forms), and exclusively Inc becomes more prominent only in much older flies.

The expHTT inclusion number and size were quantified using ImageJ. Maximum-intensity projection images were converted to 8-bit images. Their backgrounds were subtracted (rolling ball radius of 10 pixels), an unsharp mask filter was applied (radius of 1 pixel, mask weight of 0.7) and further processed to sharpen the image and find edges, and then the intermodes threshold was applied. The area in and around the LNv was then marked. The analyse particles tool with size specification of 1 to infinity (in µm) was used to obtain measures of inclusion number and the size of inclusions for each hemisphere. A lower limit of 1 µm was set for size to avoid false positives and background specks. The quantification method did not distinguish between expHTT Inc and Spots, resulting in Spots being included in the inclusion number and size quantifications.

For comparing Pdf⁺ or Per⁺ LNv numbers between genotypes or between ages, the Kruskal–Wallis test of ranks, followed by multiple comparisons of mean ranks, was used. The change in the shape of frequency distributions of Pdf⁺ and Per⁺ LNv numbers between genotypes was assessed using Kolmogorov–Smirnov tests with α=0.05, followed by Bonferroni correction on pairwise P-values. For comparing inclusion numbers between genotypes or across age, multi-factor ANOVA followed by Tukey's HSD post-hoc test was used on transformed data. Because the variances were not homogenous for inclusion size comparisons, the transformed datasets, primarily normal, were subjected to Welch's ANOVAs followed by Games-Howell post-hoc tests (McDonald, 2014). The datasets were either transformed (where required), to analyse the status of Per oscillations between the timepoints CT23 and CT11 for a genotype or the Per intensities at a timepoint between genotypes, or directly analysed using one-way ANOVA, followed by Tukey's HSD post-hoc test wherever necessary. To compare the proportion of hemispheres predominated by different expHTT forms between genotypes for an age or between ages for a genotype, m×n Fisher's exact tests were used. Wherever necessary, multiple specific 2×2 Fisher's exact test sets with BH procedure on all relevant comparisons were applied (using R). The proportion of hemispheres with Spot⁺ LNv between ages was compared using multiple 2×2 Fisher's exact tests followed by Bonferroni corrections. Spot sizes between ages for sLNv or lLNv were analysed by one-way ANOVA (on transformed data, if required), followed by Tukey's HSD tests. For Spot sizes between sLNv and lLNv, a factorial ANOVA with LNv and age as fixed factors was carried out on transformed data, followed by Tukey's HSD.

OriginPro 8 (https://www.originlab.com/origin), Sigma Plots 11.0 (https://systatsoftware.com/sigmaplot/) and Adobe InDesign 3.0 (https://www.adobe.com/uk/products/indesign.html) were used for making figures.

We sincerely thank Troy Littleton, Massachusetts Institute of Technology, USA; Michael Nitabach, Yale University, USA; Jeffrey C Hall, Brandeis University, USA; and Jae Park, Vanderbilt University, USA, for sharing reagents. Special thanks to Florence Maschat, Université de Montpellier, France for the UAS-dhtt81aa and UAS-dhtt-620aa lines (Table S1) (Mugat et al., 2008), and Norbert Perrimon, Harvard Medical School, USA for the UAS-dhtt line (Zhang et al., 2009). We thank Sunil Kumar from the Jawaharlal Nehru Centre for Advanced Scientific Research Confocal Imaging Facility for his assistance. We thank Ankit Sharma for his comments on the manuscript and help with experiments, and Rutvij Kulkarni and Abhilash Lakshman for statistical analyses. We thank Sushma Rao for help with dissections, and Rajanna N. and Muniraju for their assistance.