Endoplasmic reticulum (ER) lumenal indicators in Drosophila reveal effects of HSP-related mutations on ER calcium dynamics

Genes for ER-shaping proteins are among the most commonly mutated in Hereditary Spastic Paraplegia (HSP). Mutation of these genes in model organisms can lead to disruption of the ER network. To investigate how the physiological roles of the ER might be affected by such disruption, we developed tools to interrogate its Ca2+ signaling function. We generated GAL4-driven Ca2+ sensors targeted to the ER lumen, to record ER Ca2+ fluxes in identified Drosophila neurons. Using GAL4 lines specific for Type Ib or Type Is larval motor neurons, we compared the responses of different lumenal indicators to electrical stimulation, in axons and presynaptic terminals. The most effective sensor, ER-GCaMP6-210, had a Ca2+ affinity close to the expected ER lumenal concentration. Repetitive nerve stimulation generally showed a transient increase of lumenal Ca2+ in both the axon and presynaptic terminals. Mutants lacking neuronal reticulon and REEP proteins, homologs of human HSP proteins, showed a larger ER lumenal evoked response compared to wild type; we propose mechanisms by which this phenotype could lead to neuronal dysfunction or degeneration. Our lines are useful additions to a Drosophila Ca2+ imaging toolkit, to explore the physiological roles of ER, and its pathophysiological roles in HSP and in axon degeneration more broadly.


Introduction
Hereditary Spastic Paraplegia (HSP) is a genetically heterogeneous disorder, with over 80 loci and 60 genes identified. It also shows phenotypic heterogeneity in manifestations such as age of onset, and the presence of other symptoms (Blackstone 2018). Despite this heterogeneity, there is evidence for some common mechanisms of cellular pathophysiology.
The endoplasmic reticulum (ER) is implicated in these mechanisms, with HSP mutations affecting ER proteins with a variety of roles including lipid metabolism, ER membranecontact-site function, and ER architecture (Blackstone 2018). Gene products that shape the tubular ER network are among the most commonly mutated in HSP, and mutation of these genes in model organisms leads to physical disruption of the ER network (O'Sullivan et al. 2012, Fowler and O'Sullivan 2016, Summerville et al. 2016, Yalçın et al. 2017, Lindhout et al. 2019. These studies point to the importance of a continuous and connected tubular ER network throughout axons and presynaptic terminals, and suggest disruption to this network as a common thread in HSP pathogenesis. Although changes in synaptic structure and function are found in these mutants (Summerville et al. 2016, Li et al. 2017, Lindhout et al. 2019, the direct physiological consequences of altered ER architecture remain elusive. It is imperative to understand the roles of ER architecture and the consequences of disrupting it, to understand how this might result in a neurodegenerative cascade. We therefore need tools to study the physiological functions of axonal and presynaptic ER that depend on its architecture. Recently, de Juan-Sanz et al (2017) developed sensitive Genetically Encoded Calcium Indicators (GECIs) targeted to the ER lumen, to examine the role of the ER in Ca 2+ handling and signaling in synaptic activity in mammalian cultured neurons. Here we have expressed one of these probes (ER-GCaMP6-210) under GAL4 control in Drosophila motor neurons, and targeted two other Ca 2+ sensors with higher Ca 2+ affinities, CEPIA3-ER and CEPIA4-ER (Suzuki et al. 2014) to the Drosophila ER lumen. We show ER-GCaMP6-210 as the most effective reporter of the three by consistently producing the largest changes in fluorescent intensity, and characterize its localization and its reporting of ER behavior in axon and presynaptic terminals. We then use it to show perturbation of Ca 2+ handling in Drosophila in which ER organization is disrupted by loss of HSP-related ER-modeling proteins.

Gene Constructs
The pUASTattB vector (Bischof et al. 2007), was previously modified to include attR sites and the ccdB gene to make it compatible with the Gateway cloning system (Invitrogen) (Moreau et al. 2014). We further modified this vector, increasing the number of GAL4binding sites from 5 to 17, to generate p17xUASTattB, and increase GAL4-dependent gene expression (Supplementary File 1). To generate p17xUASTattB-ER-GCaMP6-210, we PCRamplified a fragment encoding GCaMP210-KDEL fused to a calreticulin signal peptide, from Addgene plasmid ER- GCaMP6-210 (de Juan-Sanz et al. 2017) (Table 1), while adding attB1 and attB2 sites to its 5' and 3' ends to enable it to integrate into p17xUASTattB using the Gateway cloning system (Supplementary File 1). To generate p17xUASTattB-CEPIA3-ER and p17xUASTattB-CEPIA4-ER, we PCR-amplified the CEPIA coding region from Addgene plasmids pCMV CEPIA3mt and pCMV CEPIA4mt (Suzuki et al. 2014) (Table 1), while adding attB1 and a BiP signal sequence to the 5' end, and HDEL and attB2 to the 3' end (BiP and HDEL ER-targeting as in (Summerville et al. 2016)), and then integrating it into p17xUASTattB (Supplementary File 1).

Drosophila Genetics
p17xUASTattB-ER-GCaMP6-210, p17xUASTattB-CEPIA3-ER and p17xUASTattB-CEPIA4-ER were injected into Drosophila embryos and integrated at landing site attP86Fb (Bischof et al. 2007) using phiC31 integrase-mediated integration, by the University of Cambridge Genetics Department injection service. Other fly stocks used can be found in Table 1.
HL3 was then replaced with PBS and larvae fixed for 10 min in PBS with 4% formaldehyde.
Fixed preparations were mounted in Vectashield (Vector Laboratories), and images were collected using EZ-C1 acquisition software (Nikon) on a Nikon Eclipse C1si confocal microscope (Nikon Instruments, UK). Images were captured using a 40x/1.3NA oil objective.
Segmental nerves were cut with dissection scissors and then drawn by suction into a heatpolished glass pipette, ~10 μ m internal diameter (Macleod et al. 2002), which was connected to a stimulator and train generator (DS2A-MK.II Constant Voltage Stimulator and DG2A Train/Delay Generator (Digitimer, Welwyn Garden City, UK)) to deliver suprathreshold electrical pulses (~5 V). Larval NMJ responses were recorded in abdominal segments A4-A6. After 10 s of baseline image acquisition, 400-μs impulses were delivered for 2 s at a range of frequencies. The interval between different stimulation frequencies was ~60 seconds, or until ER lumen fluorescence returned to baseline levels (whichever was longer).

Image analysis and Figure Preparation
Confocal microscopy images of fixed samples were collected using EZ-C1 acquisition software on a Nikon Eclipse C1si confocal microscope. Images were analyzed and processed using ImageJ Fiji (https://fiji.sc) (Schindelin et al. 2012). Wide-field live images were opened in ImageJ Fiji, and channels were individually bleach-corrected (simple ratio, background intensity=0). Fluorescent intensity time courses were obtained for an ROI traced around the entire NMJ, or axon segment, in each frame in a given data set using Time Series Analyzer V3 plugin, saved as txt files and fed into R scripts written by the authors to obtain the Δ F/F data (Supplementary File 2). Graphs were generated using GraphPad Prism 8. Figures were made using Adobe Illustrator.

Statistical analysis
Data were analyzed in GraphPad Prism, using 2-way ANOVA, unpaired two-tailed Student's t-tests; graphs show mean ± SEM, or Mann-Whitney U tests (for data not normally distributed); graphs show median with interquartile range. Pearson's correlation was used to examine relationships between datasets. Sample sizes are reported in figures.

The ER lumenal Ca 2+ evoked response is more delayed and sustained compared to the cytoplasmic response
To monitor Ca 2+ flux in the ER lumen, we generated transgenic flies that encoded ER-GCaMP6-210 (de Juan-Sanz et al. 2017), a GCaMP6 sensor fused to a signal peptide and ER retention signal, with a Ca 2+ affinity optimized for the relatively high [Ca 2+ ] in ER lumen.
ER-GCaMP6-210 localization in presynaptic terminals largely overlapped with the ER marker tdTomato::Sec61β (Fig. 1A), being more restricted than cytoplasmic jRCaMP1b ( Fig.   1B), consistent with localization to the ER lumen. To test the response of the sensor to neuronal activation we electrically stimulated the nerve. The cytoplasmic Ca 2+ response quickly reached a peak in the first couple of seconds following stimulation, and decayed back to resting levels over several seconds. In contrast, the ER-GCaMP6-210 lumenal response increased more slowly, reaching a peak five or more seconds following stimulation, with a slow return to resting levels, in some cases over minutes ( Fig. 2A (Dana et al. 2016), compared to GCaMP6f (0.4s) (Chen et al. 2013) at 10 Hz. In around 10% of nerves, usually at higher stimulation frequencies, we alternatively (or sometimes additionally) observed small, sharp decreases in ER lumenal fluorescence immediately following stimulation. This decrease was followed by a quick return to baseline fluorescence, or by a slow increase in fluorescence as described above. This short-term decrease in fluorescence was observed more frequently in the axon than at the NMJ. There did not appear to be anything notably different in the cytoplasmic flux in these instances, compared to when the ER flux showed only an increase in fluorescence ( Fig. 2A; Fig. S1).
Together this data suggests a slower and somewhat less predictable ER Ca 2+ flux, compared to the cytoplasmic flux in axons and NMJs of motor neurons.

ER lumenal sensor with affinity closest to predicted ER lumenal [Ca 2+ ] shows the strongest response
Since the high surface/volume ratio of the small ER tubules of axons might potentially make them more susceptible than other ER regions to leakage of Ca 2+ to the cytosol, we generated transgenic flies carrying two additional ER lumenal sensors with higher Ca 2+ affinities. The highest affinity sensor, CEPIA3-ER (K d =11 µM), showed very small or no responses at multiple NMJs of seven larval preparations, even when a robust cytoplasmic response was elicited (not shown); CEPIA4-ER (K d =59 µM) usually showed a small response at most frequencies. ER-GCaMP6-210 (K d =210 µM) produced the most consistent response, and the magnitude of its response increased with increasing stimulation frequency ( Fig. 2B-D The axonal cytoplasmic evoked response is smaller compared to NMJ response, whereas the ER lumenal evoked response is consistent across locations As we were interested to see if we could observe Ca 2+ signals along the length of the axon, we used FMR-GMR27E09-GAL4 to record in the two Type Is motor neurons per hemisegment where it drives expression: RP2 and RP5, which together innervate several NMJs (Perez-Moreno and O'Kane 2019). We compared several axonal and NMJ locations within a muscle hemisegment (Fig. S2A). Motor neuron RP2 (also known as ISN-Is) innervates dorsal muscles, and we recorded from its NMJ boutons at muscle 1, and from two axonal locations: the "entry" point at which the axon starts to cross the muscles as it emerges from the ventral nerve cord, and more laterally as it passes over muscle 4. Motor neuron RP5 (also known as ISN b/d) innervates ventral muscles, and we recorded from its NMJ at muscle 6/7, and from one axonal location, the "entry" point, as defined for motor neuron RP2 (Fig. S2A,B; Supplementary Video 2). Dpr-GMR94G06-GAL4 drives expression in the aCC Type Ib motor neuron (also known as MN1-Ib), which innervates muscle 1 (Perez-Moreno and O' Kane 2019). The responses of both cytosolic and lumenal sensors at the aCC muscle 1 NMJ boutons were comparable to those in RP2 at the same muscle (Fig.   3A,B), suggesting a comparable action potential response between the two neuronal types.
Fluorescence of the sensors driven by Dpr-GMR94G06-GAL4 in the axons of the Type Ib aCC neuron was too dim to obtain reliable evoked responses, which may be due to a relatively small diameter of its axon, and hence relatively low amounts of sensors at this subcellular location.
The evoked cytoplasmic NMJ and axonal Ca 2+ responses were both comparable between the two Type Is motor neurons, although axonal responses were consistently smaller than NMJ responses whereas ER lumenal evoked responses were consistent across all NMJ and axonal locations (Fig. 3C). This did not appear to be due to alterations in ER volume as resting fluorescence was consistent across locations (Fig. S2C). Both cytoplasmic and ER lumenal responses reached peak fluorescence on similar timescales across locations (Fig.   3D), suggesting that the observed responses were due to Ca 2+ flux across the plasma membrane, and not a propagation of Ca 2+ down the axon. The ER lumenal response appeared in general not to be influenced by the cytoplasmic response; most locations showed no correlation of peak Δ F/F between ER lumen and cytosol, apart from RP2 axon at entry point (Fig. 3E,F, Table 2). This suggests a more complex transfer of Ca 2+ within and between intracellular compartments, than simply a flow of Ca 2+ from cytosol to ER lumen.

Rtnl1 -ReepA -ReepBmutants display an increased ER lumenal evoked response compared to WT
Rtnl1 -ReepA -ReepBmutant larvae have a disrupted axonal ER network, with larger and fewer tubules, and occasional fragmentation (Yalçın et al. 2017). To test whether these ER changes had functional consequences, we used Dpr-GMR94G06-GAL4 to examine the cytoplasmic and ER lumen evoked responses at muscle 1 NMJ in these mutants. Mutant NMJs showed an increased evoked response compared to WT in the ER lumen, but not in the cytoplasm (Fig. 4A-B). There was no significant difference in lumen resting fluorescence between mutant and WT (Fig. 4C).
There was no correlation between cytoplasmic and ER lumenal evoked responses in either mutant or WT at 40 Hz (Fig. S3), supporting the finding in Fig. 3 of a more complex intracellular transfer of calcium. It has been reported that ER lumen [Ca 2+ ] affects presynaptic cytoplasmic Ca 2+ flux (de Juan-Sanz et al. 2017); however we found no correlation here between resting ER fluorescence and peak cytoplasmic fluorescence, in either WT or mutant stimulated at 40 Hz (Fig. S3). The time-to-peak in the cytoplasm or ER lumen was not significantly different between mutant and WT at 40 Hz (Fig. S4). The time for decay to half-maximal in the cytoplasm was also not significantly different between mutant and WT at 40 Hz (Fig. S4). This parameter was not measured in the ER lumen due to the slow temporal dynamics of this flux.
These data suggest a significant increase in ER lumenal evoked responses in Rtnl1 -ReepA -ReepBmutants compared to WT, with no change in ER lumenal resting fluorescence levels, no corresponding change in the cytoplasmic evoked response, and no change in the temporal dynamics of the response.

Discussion
Direct measurement of ER lumenal Ca 2+ dynamics is essential for reporting on the physiological and pathophysiological responses of the ER network. An ER lumenal UAS-GECI, ER-GCaMP6-210, provided the most robust responses in Drosophila motor neurons (Fig. 2), in keeping with results in mammalian hippocampal neurons (de Juan-Sanz et al. 2017). This sensor was also recently successfully cloned and expressed in Drosophila, and used to measure Ca 2+ efflux from the ER in mushroom body Kenyon cells (Handler et al. 2019).
In mutant Drosophila lacking various ER-shaping HSP protein homologs, we previously found decreased ER levels compared to WT, or occasional discontinuity of the ER network (Yalçın et al. 2017). As sustained discontinuity could potentially disrupt long-range ER communication along the axon, we wanted to investigate if we could record Ca 2+ fluxes in WT axons. We analyzed both cytoplasmic and ER lumenal Ca 2+ fluxes, at three axonal recording positions across two neurons. The amplitude of the ER lumenal evoked response was similar in axons and at the NMJ; however cytoplasmic evoked responses were smaller in axons compared to the NMJ. Even though Ca 2+ would be fluxing across the ER membrane in response to stimulation, it appears that the two signals do not influence each other in most cases (Fig. 3).
While we observe Ca 2+ signals along the length of motor axons, we do not find any indication of propagation via a Ca 2+ wave. Propagation speeds for Ca 2+ waves in dendrites have been measured at 60-90 µm/s (Nakamura et al. 2002), with a rapid lumenal Ca 2+ diffusion through dendritic ER at ~30 µm/s (Choi et al. 2006). Here we find no evidence for any spatial differences in the time to peak fluorescence between different axon positions and the NMJ, both for ER lumen and cytoplasmic evoked responses (Fig. 3); in a typical video frame width of 100 µm (e.g. Fig. S1, S2) we should easily detect propagation at these speeds with our frame rate of 10 Hz. The instantaneous spread of Ca 2+ responses that we observe is more consistent with them being a consequence of the action potentials generated by stimulation, which propagate so fast that we cannot resolve their spread using our optical recording.
Detection of Ca 2+ waves that propagate along the ER membrane, or through the lumen would probably require recording of ER Ca 2+ fluxes in the absence of action potentials.
Although the cytoplasmic evoked responses of Rtnl1 -ReepA -ReepBmutants were normal, the ER lumenal evoked responses were significantly larger than WT (Fig. 4). This did not appear to be due to increased resting ER lumenal [Ca 2+ ] since resting fluorescence levels showed no significant difference between genotypes, for either ER or cytoplasmic sensors.
In the simplest interpretation, the presynaptic ER lumen appears to take up more Ca 2+ on stimulation in the Rtnl1 -ReepA -ReepBmutant compared to WT. Since cytoplasmic Ca 2+ responses are unaffected in mutants, increased cytosolic Ca 2+ cannot be an explanation for the increased ER response in mutants, implying that the increased ER response is due to intrinsic changes in Ca 2+ -handling properties of the mutant ER, consistent with the roles of reticulon and REEP proteins in directly shaping ER architecture. Furthermore, increased uptake of Ca 2+ (Opas et al. 1991) although the protein-folding chaperone role of calreticulin is more likely to place its main function in rough rather than smooth ER. However, altered lumenal Ca 2+ responses could also affect physiology of organelles that obtain Ca 2+ from the ER, including mitochondria and endo/lysosomal organelles (Raffaello et al. 2016)]. These effects could endure well beyond the original train of action potentials, due to the slow kinetics of Ca 2+ uptake and release by the ER, with lumenal [Ca 2+ ] not always returning to baseline during our recordings. Furthermore aberrant ER Ca 2+ handling has been postulated as a trigger for neurodegeneration, so what we are observing may well reflect deleterious changes in cellular health (Stutzmann and Mattson 2011). In support of our findings, mutation or knockdown of genes with a role in ER shaping, including atl, Rtnl1 and VAP, also affect aspects of presynaptic function including neurotransmitter release and synaptic vesicle density, and the cytoplasmic presynaptic Ca 2+ response (Summerville et al. 2016, De Gregorio et al. 2017, Lindhout et al. 2019. Taken together, this data strongly suggests a role for ER in modulating presynaptic function, and highlights the importance of having an intact, functional ER network.
In conclusion, we have demonstrated the use of ER lumenally targeted GECIs for use in Drosophila, successfully recording fluxes at both the NMJ and in the axon. We have demonstrated the utility of these sensors in identifying functional differences between WT and Rtnl1 -ReepA -ReepBmutant flies, whose axonal ER network it is known to be affected.
We anticipate these will be useful additions to a Drosophila Ca 2+ imaging toolkit in the future, to further our understanding of HSP, and neurodegeneration more broadly.    were plotted for the NMJ (E) and axonal (F) compartments of RP2 and RP5. A positive correlation was found only for RP2 axon entry but not for other locations (Table 2). One outlier in RP2 axon (muscle 4) jRCaMP1b was excluded from graphs (C) and (F) (value 0.29), but included in analysis in (F).