Targeted kinase inhibition relieves slowness and tremor in a Drosophila model of LRRK2 Parkinson’s disease

In a number of Drosophila models of genetic Parkinson’s disease (PD) flies climb more slowly than wild-type controls. However, this assay does not distinguish effects of PD-related genes on gravity sensation, “arousal”, central pattern generation of leg movements, or muscle. To address this problem, we have developed an assay for the fly proboscis extension response (PER). This is attractive because the PER has a simple, well-identified reflex neural circuit, in which sucrose sensing neurons activate a pair of “command interneurons”, and thence motoneurons whose activity contracts the proboscis muscle. This circuit is modulated by a single dopaminergic neuron (TH-VUM). We find that expressing either the G2019S or I2020T (but not R1441C, or kinase dead) forms of human LRRK2 in dopaminergic neurons reduces the percentage of flies that initially respond to sucrose stimulation. This is rescued fully by feeding l-DOPA and partially by feeding kinase inhibitors, targeted to LRRK2 (LRRK2-IN-1 and BMPPB-32). High-speed video shows that G2019S expression in dopaminergic neurons slows the speed of proboscis extension, makes its duration more variable, and increases the tremor. Testing subsets of dopaminergic neurons suggests that the single TH-VUM neuron is likely most important in this phenotype. We conclude the Drosophila PER provides an excellent model of LRRK2 motor deficits showing bradykinesia, akinesia, hypokinesia, and increased tremor, with the possibility to localize changes in neural signaling.


INTRODUCTION
Parkinson's disease (PD) is a progressive neurodegenerative condition characterized by pathological loss of dopaminergic neurons in the substantia nigra. The reduction in dopamine in the projections to the striatum leads to a range of motor disorders, characterized by bradykinesia (slower movements), hypokinesia (a reduction in the amplitude of movements), akinesia (absence of movement altogether), and tremor.
Although the cause of the majority of PD is unknown, a small proportion is inherited (see, for a review, ref. 1 ) with the most common genetic form being due to mutations in LRRK2 (Leucine-Rich Repeat Kinase 2).
This gene is translated into a large, multi-domain protein, and the pathogenic mutations include G2019S and I2020T, 2 in which the kinase activity is increased, 3 and R1441C, a mutation in which the GTPase activity is thought to be reduced. 4 The excellent genetic toolkit provided by Drosophila led to the creation of fly models of PD. These reflect many features of the disease (loss of dopaminergic neurons, reduced movement, mitochondrial abnormalities, oxidative stress. and visual deficits). 5,6 Many labs have adopted the fly negative geotaxis assay (sometimes called the "startle response assay") as their measure of movement. [7][8][9] In this, the speed at which flies walk up a glass cylinder in response to a sharp tap is recorded. Although PDmimic flies have reduced movement, it is hard to specify exactly where the changes are taking place (response to the startle stimulus or gravity, or effects on the central pattern generator or motor neurons or changes directly affecting the leg muscles themselves). This assay also fails to discriminate between the different possible movement defects (akinesia, hypokinesia, and bradykinesia). We suggest the requirement for another, simpler assay system. This is reinforced by the difficulty of determining which of thẽ 125 dopaminergic neurons in the fly CNS 10,11 are important in the negative geotaxis response. Although dopaminergic innervation of the mushroom body by 15 "PAM" neurons plays a major role in this negative geotaxis response, 12 the subsequent neuronal pathway is unclear. Further, manipulations of PD-related transgenes often lead to the loss of a relatively small proportion of dopaminergic neurons, with many clusters remaining unaffected. For example, with the LRRK2-G2019S mutation, the protocerebral posterior medial (PPM) cluster dropped from 14 to 12 dopaminergic neurons but the protocerebral anterior lateral (PAL) cluster remained unaffected. 13 Throughout the literature, the multiple processes involved in slowed negative geotaxis combined with the observed small loss of dopaminergic neurons act to obscure the functional relationship. To progress, we need to link a precise measurement of movement with the physiology of a few specific dopaminergic neurons.
An exciting solution to this problem is provided by the discovery that a single dopaminergic neuron strongly modulates the fly proboscis extension response (PER). 14 As a fly walks into a solution containing sucrose, the Gr5a chemosensory cells on its front legs are activated (Fig. 1, step 1). Their axons project to the sub-esophageal zone of the CNS (SEZ; the part signaling the taste response; Fig. 1, step 2). Within the SEZ, the chemosensory inputs activate the interneuronal pathway, 15 leading to the pharyngeal E49 motoneurons, whose action potentials elicit contraction of the M9 muscle (Fig. 1, step 3). This well characterized neuronal pathway results in the reflex extrusion of the proboscis towards the food (Fig. 1, step 4), allowing the fly to ingest the solution. Although the sensory and motor steps in this pathway have been well defined (see for a review ref. 16 or ref. 17 ), the interneuronal steps mostly remain to be described.
One well-defined neuron that modulates the PER is TH-VUM, a single, unpaired neuron in the SEZ, which makes output synapses onto the sense cells and interneurons 14,18,19 (Fig. 1). Strong activity in the TH-VUM leads to contraction of the proboscis muscle, blocking the output of the TH-VUM reduces the probability of a sucrose-induced PER. The frequency of action potentials in the TH-VUM correlates with the length of starvation. 14 Interestingly, the TH-VUM fires steadily in a way reminiscent of mammalian substantia nigra dopaminergic neurons. 20 We have now found that expression of LRRK2-G2019S, the most common cause of genetic PD, in dopamine neurons results in a reduced PER, with bradykinesia, akinesia, and tremor, and that this is rescued by L-DOPA or by kinase inhibitors targeted at LRRK2.

Upregulation of LRRK2 kinase activity in dopaminergic neurons causes akinesia
In order to test the neuronal specificity of the PD-related mutation LRRK2-G2019S, we first expressed this in each of the components of the PER reflex pathway, recording the proportion of a population of starved flies that extended their proboscis in response to a moderate (100 mM) sugar stimulus (Fig. 2a, b). Strikingly, when we expressed LRRK2-G2019S in the dopaminergic neurons with TH-GAL4 (tyrosine hydoxylase GAL4), the proportion of young flies responding was about half that of control genotypes (no transgene expressed, χ 2 -post-hoc test, p < 0.001; from 76-35%; Fig. 2a). The same result was seen in a second sample, where the proportion of TH>G2019S flies responding was also less than half that of the wild type (Fig. 2b). In contrast, there was no effect of G2019S when driven in all neurons (nSyb-GAL4, p = 0.17), or in just the sensory neurons (Gr5a-GAL4, p = 0.11), or solely in proboscis motoneurons (E49-GAL4, p = 0.21).
In order to establish if this was specific for the G2019S mutation, we also tested a second PD-related mutation in which kinase activity is increased (I2020T), a mutation in the GTPase domain (R1441C), a kinase dead form (LRRK2-G2019S-K1906M), and the normal human form (hLRRK2). Of these, only dopaminergic expression of I2020T had any specific effect, reducing the proportion of flies that responded from 61 to 35 % (χ 2 -post-hoc test, p < 0.001; Fig. 2b). We did note a general reduction on all flies containing the hLRRK2 (Fig. 2a, b), due perhaps to the increased expression levels of this gene (Fig. 2c). In turn, the lack of effect of TH>R1441C might be due to the weaker expression (Fig. 2c), based on our hypothesis that the impact of any one LRRK2 mutant on PER is correlated with the level of LRRK2 protein production in the dopaminergic neurons.
To determine if the PER response was age dependent, we tested flies from 3 days to over 18 days, and found none of the genotypes showed any age dependent change (Fig. 2b). Dopaminergic expression of G2019S or I2020T reduced the proportion of flies responding to the sucrose stimulus at all ages.
We found no effect of genotype on mass at 20 days (nontransgene control vs TH>hLRRK2 vs TH>G2019S, F 2,26 df = 0.003, p = 0.99), suggesting the reduced frequency of PER seen in starved flies was not preventing them from increasing their mass when provided with food ad libitum. However, by 28 days, only 57% of flies were alive, with increased mortality in each of the TH>hLRRK2, TH>G2019S, and TH>I2020T compared to the TH/+ control (χ 2post-hoc test, p < 0.01 for each; survival 38, 53, 57, 80%, respectively).
These experiments show that the expression of high kinase forms of LRRK2 reduces the proportion flies showing PER, i.e. they show akinesia.
Rescue of akinesia by L-DOPA and kinase inhibitors Since the standard symptomatic treatment for PD is L-DOPA, we tested if this compound would rescue the PER deficit induced by dopaminergic expression of G2019S or I2020T. In both cases, the administration of L-DOPA to the adult flies raised the proportion of flies that respond to sucrose to control levels (Fig. 3a). This was not due to an increase in general responsiveness, because we saw no effect of L-DOPA on the flies expressing hLRRK2, KD, or R1441C.
As the most common form of genetic PD is caused by the LRRK2-G2019S mutation, with its increased kinase activity, a promising therapy would be to deploy kinase inhibitors targeted at LRRK2. The first of these to be developed was LRRK2-IN-1. 22 Application of 2.5 μM LRRK2-IN-1 partially rescued the frequency of PER responses, from 32 to 47% (p = 0.002, Fig. 3b). Since LRRK2-IN-1 has off-target effects, in vitro 23,24 and in vivo, 6 we next tested the more specific compound BMPPB-32, which also gave a partial rescue, to 44% (Fig. 3b). The proportion of flies showing the PER was the same for LRRK2 and BMPPB-32 (p = 0.077) and both were below the wild-type level (73%).
We conclude that drugs targeted at LRRK2 ameliorate the reduced PER response of TH>G2019S flies. Dopaminergic expression of LRRK2-G2019S slows movement and increases tremor in the PER For TH>G2019S flies showing a PER, recordings were made using a camera with high frame rate (200 per second), and digitized the distance from the eye to the end of the proboscis, to measure the movement during the PER. Individual traces showed that some TH>G2019S flies had very slow PER, taking three or even four times longer than the median wild-type control (Fig. 4a), with the increase in standard error being very noticeable (Fig. 4b). A second mutant, I2020T, also showed a slower PER (Fig. 4bii). For both TH>G2019S and TH>I2020T flies, the speed of the PER is fully rescued to wild type by feeding 2.5 μM BMPPB-32.
Using the independently generated G2019S and hLRRK2 lines, 21 we also found a slower PER with TH>G2019S. In this case the duration of the PER increased from 0.30 ± 0.025 s (TH/+) to 0.57 ± 0.11 s (TH>G2019S) (mean ± SE, Tukey-post-hoc test, p = 0.017). The TH>hLRRK2 was the same as TH/+ (p = 0.81).
Additionally, it appeared that the TH>G2019S traces were more irregular. To assess this quantitatively, each trace was fitted by a smooth curve (a piecewise cubic spline), and the deviation of the actual trace from smoothed determined (Fig. 4ci). This showed that the TH>G2019S flies had a much longer "path", about twice that of the control flies. The proboscis does not move out in a smooth trajectory, but oscillates, showing tremor. No such changes were seen in the TH>hLRRK2 flies; they were the same as the no-transgene cross (Fig. 4cii).
Thus dopaminergic expression of LRRK2-G2019S induces bradykinesia and tremor as well as akinesia.
The single dopamine neuron (TH-VUM) is mainly responsible for akinesia Our next step was to test which dopaminergic neurons are responsible for akinesia in the PER response, focusing on the difference between flies with increased kinase activity (G2019S or I2020T) and the kinase dead (KD) line.
There are eight clusters of dopaminergic neurons in the fly CNS (Fig. 5b), and a range of GAL4 drivers have been developed to target various subsets of these. We started with a GAL4 driver DDC, 25 which has been widely used for studies of negative geotaxis and which gives generalized dopaminergic neuron expression, as well as expressing in serotonergic neurons. As with the TH-GAL4 (Fig. 1), fewer flies expressing G2019S or I2020T with DDC showed the PER compared with those flies expressing LRRK2-KD (Fig. 5). We next tested HL9 (ref. 26 ) which expresses in all the dopaminergic neurons except for the PAL cluster (Fig. 5b)  label a few non-dopaminergic neurons (Fig. 5c). Again, the G2019S and I2020T flies were less responsive than the KD flies. With the C′-GAL4 line, 27 which expresses in all dopaminergic neurons except the PPM3 and PPL1 clusters, we saw the same pattern: G2019S and I2020T flies showed less PER than the KD flies. All these GAL4 lines express in the VUM dopaminergic neurons. The final GAL4 tested, D′ 27 expresses in all dopaminergic neurons except the PAL, PPL2 and does not express strongly in the VUM neurons. Remarkably, with the D′-GAL4, flies expressing G2019S, I2020T, and KD forms of LRRK2 all showed an equal response to sucrose, strikingly different from all the other dopaminergic GAL4s.
We next used a nuclear GFP (eIfGFP; Fig. 5c) to confirm the expression pattern of the GAL4 lines in the SEZ. DDC, HL9, and C′, all showed a group of three GFP-positive neurons located posteriorly plus a fourth cell anteriorly. One of the three posterior cells is the unpaired midline TH-VUM and the other two are descending neurons (left and right DA-DNs), whose activity represents leg movements. 28 The D′>eIfGFP fluorescence pattern is noticeably different: only the anterior cell shows any GFP signal (and that weakly), but the three posterior neurons not being labeled at all. The presence of the three posterior neurons is confirmed by the anti-TH staining.
Consequently, we suggest that the effect of expressing LRRK2-G2019S in dopaminergic neurons is mostly mediated by the single TH-VUM neuron, because the two descending neurons do not show a direct link to feeding behavior. Only when G2019S is expressed in the TH-VUM do we see the reduction in PER, i.e. akinesia. However, we cannot rule out an effect of the PPL2 neurons, which also modulate the feeding system 29 as they are also not targeted by the D′-GAL4.

DISCUSSION
Our principle finding is that expressing LRRK2 forms with increased kinase activity (G2019S, I2020T) in sets of dopaminergic neurons that include the TH-VUM is sufficient to induce akinesia, bradykinesia, and tremor in the fly PER. Although previous work with flies and rodents have identified movement disorders in PD models, our PER assay uniquely identifies the components of the response, in the context of changes mostly due to to a single dopaminergic neuron (TH-VUM).
A key point is that our PER assay demonstrates dopaminergicbradykinesia even at 3 days. In comparison, data from negative geotaxis (startle-induced climbing response) assays of G2019S, I2020T, or R1441C transgenics is more complex, depending on age and genotype. One report shows that while locomotion in DDC>I2020T flies is already compromised at 5 days, 30 another study shows that TH>I2020T flies show little deficit at any age. 31 This may be because DDC-GAL4 includes serotonergic, as well as dopaminergic neurons, and thus more cells compared with TH-GAL4. Using DDC-GAL4, to express LRRK2 in old (>5 weeks) flies, G2019S and R1441C movement is shown to be reduced, 32,33 while younger flies show no deficit. In contrast, our data show movement deficits in young TH>G2019S and TH>I2020T flies, which are maintained over the first few weeks of adult life. Indeed, LRRK2 mutations already start to affect Drosophila larvae, indicating effects at an earlier timepoint. 21,34,35 One disadvantage of working with older flies is that, by 5 weeks, a proportion of flies will have died, potentially those most strongly influenced by the transgene, so that negative geotaxis assays may underestimate the real impact of LRRK2 mutations. Our assay has the advantage of working at 3 days, before flies have started to die, and potentially could be developed so that the same individual fly might be tested at different time points. This would permit comparison of the individual and population responses.
In our visual assays with TH>G2019S, we found that 3-day-old flies showed no detectable visual deficits, though younger flies (1day-old) had overactive vision, and old flies (28-day-old, or visually stressed) had much reduced response. 6,36,37 Overactivity has also been reported in young transgenic LRRK2 rats, followed later by loss of movement. [38][39][40] We have not tested the feeding response of flies less than 3 days old, because these newly emerged flies rest and expand their cuticle, and are not feeding: this makes them unsuitable for PER assay.
However, the PER of our mildly starved 3-day-old TH>G2019S flies is already reduced, and remains well below wild-type levels for at least 18 days. A more pronounced PER deficit might arise in older flies (5 weeks), and/or those kept at 29°C to enhance transgene expression.
Further, the movement deficits in our PER assay on flies starved for 2-3 h are mainly a consequence of expressing G2019S in a single dopaminergic neuron, TH-VUM, rather than the mixed effect of a range of dopaminergic clusters.
In this respect the PER assay differs from both negative geotaxis and our visual assay (three different kinds of dopaminergic neuron are present in the retina). However, we note that longer term modulation of feeding appears to involve other dopaminergic clusters, interacting in the mushroom bodies. 41 Additionally, our PER deficit occurs in young flies in which the TH-VUM is still present, offering the potential to understand the processes by which LRRK2 leads to neuronal silencing in a single identified neuron. We find that both the akinesia and bradykinesia components of the TH>G2019S effect on PER are dependent on the kinase role of LRRK2. We observe no effect of expressing the KD (kinase-dead, G2019S-K1906M) form of LRRK2, although the expression level is stronger than G2019S. In this respect, it resembles the visual assay, where expressing G2019S, but not this KD construct led to retinal neurodegeneration. 37 The TH>R1441C flies also showed no reduction in PER, or in visual degeneration 37 though it is possible that this is because R1441C is not so effectively expressed. The rescue by the specific LRRK2 inhibitor, BMPPB-32, argues that the reduction in PER is a consequence of phosphorylation of substrate (s) by LRRK2. We previously showed this inhibitor was effective in a visual assay, reverting TH>G2019S phenotypes in both young and old flies. 6,36 Another specific inhibitor LDN-73794 prevents loss of DDC>G2019S induced locomotion in old flies. 42 The PER assay has an advantage over the climbing assay (startle response), where degeneration is usually measured at 4-5 weeks, as our flies only need to be fed with the inhibitors for 3 days, reducing compound requirements.
Genetically activating or silencing the TH-VUM respectively increase or decrease the probability of a PER. 14 Thus, our data showing kinase active LRRK2 transgenes in the TH-VUM reduce PER could be explained by a reduction in dopamine release by this neuron. We hypothesize that expressing G2019S in the TH-VUM could lead to either (i) failure of TH-VUM neurites to grow, (ii) a reduction in its tonic firing, (iii) less dopamine synthesis, or (iv) lower probability of release of dopamine onto the reflex pathway. Cultured mammalian neurons, fly motoneurons, and sensory cells all have reduced neurites with G2019S. 23,43,44 While it is possible that G2019S also reduces neuritic branching in the TH-VUM neuron, our data rather favor hypotheses (iii) or (iv) since we found that feeding flies L-DOPA rescued the TH>G2019S loss of PER. Reduced dopamine levels have been reported with DDC>G2019S, and with ubiquitous expression of an increased kinase form of the fly homolog dLRRK. 45,46 In both Drosophila and mammals, L-DOPA can cross the blood-brain barrier, but dopamine cannot. 47 Thus we suggest that uptake of L-DOPA into the TH-VUM leads to increased dopamine levels and release, rescuing the effect of TH>G2019S. Increasing the amount of dopamine released onto the sugar-sensing Gr5a neurons would then rescue the proportion of flies that show the PER (akinesia), while release onto second order, local interneurons, might affect the motoneurons and thence speed (bradykinesia), and tremor of the proboscis extension. Such a dual output onto Gr5a neurons and onto local interneurons is suggested by the fact that 2.5 μM BMPPB-32 fully rescues bradykinesia, but only partially rescues akinesia. Although a number of interneurons in the SEZ with roles controlling proboscis extension, ingestion, and memory have recently been identified (e.g. see refs. [48][49][50], the link between Gr5a sense cells and the E49 motoneurons remains to be established.
PER was performed (A.C.C.) by collecting male flies of known age at the start of the working day, under CO 2 anesthesia, and sticking them ventral side up to card with rubber cement (Fixo Gum). Flies were left to recover for 2-3 h at 25°C. They were presented with a droplet of 100 mM sucrose solution to the legs, and the immediate PER/no PER scored (response in <2 s). Experiments were designed so that each graph plotted here comes from flies scored over three adjacent days, with the genotypes mixed each day, to allow for the small variations in food and environmental conditions. Power calculations indicate that a "medium" effect size, with a sample of 500 flies, and 16 df, would be detected at the 1 % level >98% of the time.
Drugs were fed to adult flies from eclosion until testing. L-DOPA (Sigma) was added to food (final concentration 50 μM). BMPPB-32 and LRRK2-IN-1 (Lundbeck) were dissolved to give a final concentration in the food at 2.5 μM.
PER was filmed using a Mikrotron MC-1362 camera mounted on a Zeiss Stemi microscope. Videos were acquired at 200 frames/s; sample movies for a wild type and TH>G2019S flies are presented in Movies S1 and S2. Only the first PER of each fly was analyzed. In Matlab, the eye and tip of the proboscis were marked and their separation was determined for each frame individually. The analyses of Movies S1 and S2 are shown in Movies S3 and S4, respectively.
Immunocytochemistry was as described recently, 37 using mouse anti-TH (Immunostar, 1:1000) and driving eIfGFP using the required GAL4 line. All data are from male flies, aged 3-5 days. No anti-GFP was used in the data chosen for illustration. The brightness and contrast of the images was adjusted in ImageJ so that the cells could be seen in both color channels, as each GAL4 drove GFP with a different intensity in the VUM neurons. Original images available on request to cje2@york.ac.uk Representative data from at least three preparations are shown.

Statistics
For analysis of the proportion of flies showing a PER, statistical significance was determined using the χ 2 -post-hoc test in the "Fifer" package of R. Confidence limits were determined using the Binomial test in R. Measurements of the speed of the PER were analyzed by ANOVA and Tukey post-hoc tests. For a "medium" effect size, with 26 flies in each of three samples, and the probability of 0.05, the power is 63%. N values for each genotype/treatment are included in Supplementary Table 1.

Data availability statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.