An in vivo Caenorhabditis elegans model for therapeutic research in human prion diseases

Abstract

Human prion diseases are fatal neurodegenerative disorders that include sporadic, infectious and genetic forms. Inherited Creutzfeldt-Jakob disease due to the E200K mutation of the prion protein-coding gene is the most common form of genetic prion disease. The phenotype resembles that of sporadic Creutzfeldt-Jakob disease at both the clinical and pathological levels, with a median disease duration of 4 months. To date, there is no available treatment for delaying the occurrence or slowing the progression of human prion diseases. Existing in vivo models do not allow high-throughput approaches that may facilitate the discovery of compounds targeting pathological assemblies of human prion protein or their effects on neuronal survival.

Here, we generated a genetic model in the nematode Caenorhabditis elegans, which is devoid of any homologue of the prion protein, by expressing human prion protein with the E200K mutation in the mechanosensitive neuronal system.

Expression of E200K prion protein induced a specific behavioural pattern and neurodegeneration of green fluorescent protein-expressing mechanosensitive neurons, in addition to the formation of intraneuronal inclusions associated with the accumulation of a protease-resistant form of the prion protein. We demonstrated that this experimental system is a powerful tool for investigating the efficacy of anti-prion compounds on both prion-induced neurodegeneration and prion protein misfolding, as well as in the context of human prion protein. Within a library of 320 compounds that have been approved for human use and cross the blood–brain barrier, we identified five molecules that were active against the aggregation of the E200K prion protein and the neurodegeneration it induced in transgenic animals.

This model breaks a technological limitation in prion therapeutic research and provides a key tool to study the deleterious effects of misfolded prion protein in a well-described neuronal system.

Introduction

Prion diseases are fatal neurodegenerative disorders that occur in humans and animals. The neuropathological hallmarks of transmissible spongiform encephalopathies include spongiosis, glial proliferation and neuronal loss. The only known specific molecular marker of prion diseases is the abnormal isoform (PrP^Sc) of the host-encoded prion protein (PrP^c), which accumulates in the brain of infected subjects and forms infectious prion particles.¹ Although this transmissible agent lacks a specific nucleic acid component, several prion strains have been isolated.^1,2 They are characterized, notably, by differences in disease duration, PrP^Sc distribution pattern and brain lesion profile at the terminal stage of the disease.

This group of fatal diseases is highly heterogeneous and includes, in humans, sporadic, infectious and genetic forms, among which a limited number of prion strains have been identified so far.³ The familial Creutzfeldt-Jakob disease (CJD) due to the E200K mutation of the PrP gene (PRNP) is the most common form of inherited prion disease.^4,5 The phenotype resembles that of sporadic CJD at both the clinical and neuropathological levels with a median disease duration from first symptoms to death of 4 months.^6-8

A few anti-prion compounds, such as quinacrine and pentosan polysulphate, have shown therapeutic activity in vivo and in vitro, but none has proved effective in humans.^9–13 A major issue in therapeutic research is the variability of the anti-prion effect between strains. Identifying drugs that are effective towards human prions is thus a key objective.^3,14,15 In addition, existing in vivo models of infectious and inherited forms do not allow high-throughput approaches that may facilitate the discovery of anti-prion compounds targeting human PrP pathological assemblies or their effects on neuronal function and survival. It is thus a prerequisite to establish an in vivo experimental model that allows the screening of medium- to large-size libraries of compounds active against human prions. Here, we generated a genetic model in the nematode Caenorhabditis elegans, which is devoid of any PrP homologue, by expressing human PrP with the E200K mutation in the mechanosensitive neuronal system. This model has a fully described cell lineage and well identified neuronal networks with precisely known functional connectivity for each of its 302 neurons.^16–18 In addition, most of C. elegans genes have counterparts in humans.^19,20,C. elegans is increasingly used as a screening model to identify novel therapeutic targets and compounds including anti-microbial agents.^21,22 We chose to express PrP in mechanosensitive neurons of the nematode because their dysfunction leads to a specific phenotype that can be quantified.²³ Compared with their counterparts expressing a similar level of wild-type human PrP, we observed that E200K PrP-expressing animals showed PrP aggregates and cell loss in green fluorescent protein (GFP)-tagged targeted neurons and formed protease-resistant PrP. This model provided a unique tool to assess, in a non-a priori manner, the effect of therapeutic compounds by monitoring the number of GFP-expressing mechanosensitive neurons in human PrP-expressing transgenic worms. First, we validated this approach using a well-studied molecule in prion diseases of direct clinical interest (i.e. doxycycline) and another one that had shown some efficacy in an experimental model of scrapie (i.e. astemizole). Then, we screened a library of 320 drugs that are known to cross the blood–brain barrier and have been FDA approved to facilitate, after further validation steps, clinical application of the most promising compounds in such a rare but devastating disease.

Materials and methods

Generation of transgenic lines

We generated transgenic lines, a control GFP-expressing line (named control^GFP) and two double transgenic lines co-expressing the full length PRNP (wild-type or c200E>K variant; named PrP_Wt and PrP_E200K, respectively) and GFP by microinjection following standard transformation techniques.²⁴ For microinjection, to generate the control^GFP line (NB13; nbIs13[mec-7p::gfp]) we used the pNB13 plasmid (40 ng/µl) and, for the two PrP lines, PrP_Wt (NB2; nbIs2[mec-7p::PRNP^Wt; mec-7p::gfp]) and PrP_E200K line (NB8; nbIs8[mec-7p::PRNP^E200K; mec-7p::gfp]), we used a plasmid mix composed of pNB13 + pNB2 (for PrP_Wt) or pNB8 for (PrP_E200K) at a final concentration of 40 ng/µl and 60 ng/µl, respectively. Injected animals were isolated to select the F₁ progeny transformants according to the presence of the GFP signal in the mechanosensitive system under a Leica MZ16 stereomicroscope. To obtain genetically stable transgenic lines, GFP-expressing larval stage L4 animals were isolated and exposed to gamma irradiation (42 Gy; CellRad Faxitron, Edimax). Lines showing stable and full transmission of the gfp transgene expression were cloned and backcrossed three times within the N2 wild-type background.

RNAi feeding

Synchronized larval stage L1 animals were cultured in plates containing nematode growth medium (NGM), seeded with the HT115 (DE3) Escherichia coli strain transformed with the PRNP^RNAi vector producing the dsRNAi (PRNP^RNAi) against the PRNP transgene (for sequence see Supplementary material). The empty vector L4440 (Addgene) was used as a control. The production of dsRNAi was induced by the addition of 1 mM of isopropyl β-d-1-thiogalactopyranoside (IPTG) and worms were chronically exposed to dsRNAi for 3 days at 25°C as described.²⁵ The knockdown PRNP transgene effect was quantified using a quantitative reverse transcription (qRT)-PCR technique.

Quantification of the PrP signal

The optical density (OD) was measured on the PrP western blots using a GS-800 calibrated densitometer and Quantity One software (Bio-Rad Laboratories).²⁶ The unspecific background was subtracted. For each experimental condition, the PrP immunosignal before and after proteinase K (PK) digestion was analysed. The ratio between the PrP signal after and before PK treatment for each condition was calculated and defined the percentage of PrP^res (% of PrP^res = OD of PrP_E200K after PK treatment / OD of PrP_E200K without PK treatment) × 100. Then, we defined the value of the treatment effect as follows: (% of PrP^res after treatment with molecule X / mean of % of PrP^res in untreated PrP_E200K controls of corresponding experiments) × 100.

Immunostaining

C. elegans strains were synchronized and young adult animals were prepared using a variation of the freeze-crack method.²⁷ Worms were washed by centrifugation in M9 and placed on slides (Superfrost^®) previously coated with poly-l-lysine hydrobromide (Sigma-Aldrich) for 15 min at 60°C. After cutting the head of the animals, 10 × 10 mm coverslips were placed gently on the top of the drop and the slides were placed on dry ice for 30 min. Coverslips were removed and the nematodes were fixed in 4% paraformaldehyde diluted in methanol for 15 min. The slides were incubated in a series of alcohol solutions of decreasing concentrations (90, 70, 50, 30 and 10%) and then incubated in a PBS solution with 0.01% Triton^TM X-100. Preparations were incubated in a blocking solution containing PBS with 0.1% Triton^TM X-100 and 5% skimmed milk for 40 min. Slides were then washed with PBS containing 0.01% Triton^TM X-100. For the PrP-aggregates analysis, slides were pretreated with a solution of 3 M guanidine thiocyanate (Sigma-Aldrich) for 5 min at room temperature. Samples were incubated with 50 µl of primary antibodies diluted in the blocking solution overnight at 4°C. The monoclonal anti-prion antibody Sha31 (1:200, Bertin Bioreagents) that recognizes the human protein sequence within amino acids 145–152 was used. After washing, worms were incubated in a solution containing a secondary anti-mouse antibody (Alexa Fluor, Red-548, Life Technologies) for 2 h at room temperature. Preparations were stained with DAPI (Sigma Aldrich). Slides were mounted with 8 µl of Gold Antifade (Life Technologies).

Microscopy images and analysis

Images sections of posterior lateral microtubule (PLM) neurons from slide-mounted young adult animals were captured with an inverse confocal microscope (Leica TCS SP8) using a ×63 oil objective. Stacks of 20 images of 0.228 µm steps were acquired per GFP-tagged PLM neuron. For PrP cluster quantification, positive neurons presenting PrP immunopositive intracellular inclusions were identified and scored. Numbers and volumes of inclusions were assessed using ImageJ software. For each stack, individual inclusions were analysed with the plugin 3D object counter²⁸ using the following parameters: size filter 2–95, 256 voxels and the threshold set to 100.

Neuronal loss assays

Quantification of the cell bodies of GFP-tagged mechanosentive neurons was performed in young adult-stage hermaphrodites raised at 20°C, fixed in 4% PFA and then mounted on slides. To compare PrP_Wt and PrP_E200K lines (50 animals per experiment with three independent experiments), all types of mechanosensitive neurons were assessed by manual counting under an epifluorescence microscope (Zeiss Apotome.2 imaging system) using a ×63 magnification lens. In a first step of drug screening (n = 320 compounds), PLM neurons of treated animals were analysed using an automated system (Arrayscan XTI, Thermo Fisher Scientific) after having dropped worms in 96-well microplates. Duplicates of 25 worms per condition were analysed. In a second step, the potential neuroprotective effect of the preselected compounds (n = 17, 50 animals per condition and per experiment with three independent experiments) was assessed by manual counting under an epifluorescence microscope with a ×63 objective.

Mechanosensitive functional touch test

Touch response assays were performed on young adult synchronized hermaphrodites raised at 20°C.²³ Animals were rinsed and transferred into a clean plate containing NGM without food. Touch tests involved scoring for reflex locomotor response to a light touch at the tail using a fine hair. This reflex involves the PLM neurons. This is the reason for focusing mainly on this type of mechanosensitive neuron. Each animal was tested once and scored as responsive or unresponsive. The reflex response was calculated as the percentage of positive responders among the tested population. We used the wild-type N2 strain as positive control and a mechanosensitive defective strain mec-7 (e1343) as a negative control.

Motility assay

Young adults were scored on their motility in liquid media at 20°C. Numbers of swimming and motionless animals were assessed by direct counting under a Leica MZ165 stereomicroscope and the percentage of motionless animals was calculated.

Lifespan

Worms were synchronized, hatched overnight in M9 at 20°C, transferred into solid media and grown at 20°C until they reached the L4 developmental stage. They were picked individually and transferred into new plates containing NGM (10 animals per plate with two plates per line). The lifespan of the animals was measured daily by counting the total living worms per plate that were transferred in fresh NGM plates every day.

Drug testing

A library of FDA-approved compounds (n = 320) described to have a pharmacological bio-distribution in the SNC was purchased from Prestwick (for more details, see Supplementary Table 1). Lyophilized molecules were solubilized in 1% DMSO, diluted in ddH₂O-DMSO at 1% to a final concentration of 1 µM, 10 µM and 100 µM and stocked at −80°C.

Pharmacological test in liquid media

After bleaching, synchronized L1 larvae stage were grown in 900 µl of culture liquid media (1 worm/µl) in 24 well plates without antibiotics and chronically incubated in presence of 9 µl of the drugs of interest at various final concentrations (1, 10 and 100 µM) during 72 h at 25°C. Doxycycline and astemizole compounds were added directly to obtain different final concentrations (1, 10, 100 and 300 µM and 0.2, 1, 5 and 15 µM, respectively).

Statistical analysis

Quantitative data are expressed as the mean ± standard error of the mean (SEM; error bars) and were independently repeated at least three times. Statistical comparisons of the data between groups were analysed for significance by one-way ANOVA. Comparisons were performed using Tukey’s or Dunnett’s post hoc test and Student’s t-test when required. Differences were considered as statistically significant at P < 0.05. Statistical analyses were conducted using Prism 6 (GraphPad Software).

For C. elegans strains, cultures, synchronization of worms, plasmid constructs, PCR amplification, RNA extraction, qRT-PCR, protein extraction, western blotting, PK treatment and PNGase F incubation, see Supplementary material.

Data availability

All the data supporting the findings of this study are available upon reasonable request by a qualified researcher.

Results

Generation of PRNP transgenic C. elegans lines

The C. elegans wild-type reference line N2 was used to generate lines expressing human PrP. Transgenes were composed of the promoter sequence of mec-7 (mec-7p) fused to wild-type PRNP or to PRNP with the E200K mutation and co-injected with a reporter GFP transgene (Supplementary Fig. 1A–C). We obtained 10 and 22 stable lines expressing PRNP_E200K and PRNP_Wt transgenes, respectively. As expected,²⁹mec-7p specifically drove GFP expression into the six neurons of the mechanosensitive neuronal system including one pair of neurons in the posterior part of the animal (PLMs), one pair of neurons in the anterior part [anterior lateral microtubule cells (ALMs)] and two individual neurons [anterior and posterior ventral microtubule cells (AVMs and PVMs)] (Fig. 1A). These observations validated the accurate targeting of the mechanosensitive system by using mec-7p in our transgenes. The presence of PRNP transgenes was assessed using PCR amplification. Migration on agarose gel of the PCR products showed a unique band corresponding to a specific amplicon of 750 bp, consistent with the expected size of each PRNP transgene (Supplementary Fig. 2A). PCR products were sequenced to validate the entire PRNP transgene for both PRNP genotypes, notably the presence of a methionine at codon 129 in both lines and of a lysine at codon 200 in the E200K line (Supplementary Fig. 2B). To select lines expressing similar levels of PrP transgenes, a quantitative analysis of PRNP mRNAs with qRT-PCR using two different sets of primers was performed. Two lines (designated PrP_Wt and PrP_E200K), expressing a similar amount of mRNA^Wt-PRNP and mRNA^E200K-PRNP as well as similar amounts of mRNA^gfp transcripts, were selected to allow comparisons (Supplementary Fig. 2C). Total PrP expression, as controlled by western blot without PK treatment of the samples, was comparable in PrP_Wt and PrP_E200K lines (Supplementary Fig. 2D).

Staining and quantification of intraneuronal PrP inclusions in transgenic nematodes

The PrP immunostaining, performed after guanidine pretreatment, was analysed using confocal microscopy. PrP was detectable in GFP-tagged mechanosensitive neurons only, in both transgenic lines, with a fluorescent staining reinforcement at the periphery of the cell body, suggesting that human PrP was correctly expressed at the plasma membrane of the targeted neurons (Fig. 1B). Moreover, in the PrP_E200K line, intraneuronal cytoplasmic clustering of PrP signal was more frequently observed than the PrP_Wt line, suggesting the presence of PrP inclusions that are usually associated with PrP aggregation and accumulation in mammalian prion diseases (Fig. 1B). These PrP clusters were observed in cell bodies and axons of mechanosensitive neurons.

PrP^res detection in transgenic nematodes

To determine the biochemical properties of the human PrP variants expressed in worms, we performed a western blot study with PNGase F and PK treatments (Fig. 1C and D). In PrP_Wt and PrP_E200K lines, we observed a tri-band pattern ranging from ∼27 kDa to 31 kDa (Fig. 1C, left). These bands corresponded to the different glycoforms of PrP expressed in nematodes, as shown by the deglycosylation experiment using PNGase F, which led to the detection of a single PrP band migrating at 27 kDa (Fig. 1C). To assess whether E200K PrP expressed in nematodes shares some biochemical properties with PrP^res, as detected in the brain of patients with E200K CJD,³⁰ we incubated worm homogenates from both lines with increasing concentrations of PK (Fig. 1D). We observed an increased PK resistance in PrP_E200K animals, compared with those expressing PrP_Wt.

PrP aggregates and loss of PLM neurons in PrP_E200K nematodes

We investigated the number of PLM neurons in the PrP_E200K line (Fig. 2A), observing a severe and significant loss compared with the PrP_Wt line (Fig. 2B). We confirmed that these alterations resulted from the expression of PrP_E200K by invalidating PRNP transgene expression using RNAi feeding (Fig. 2C and D). Decreasing PRNP_E200K transgene expression by 90%, as shown by qRT-PCR (Fig. 2D), fully restored the number of PLM neurons per animal in the PrP_E200K line (Fig. 2A and B). Of note, the significant loss of other mechanosensitive neurons (ALM, AVM and PVM neurons) was also observed in the PrP_E200K animals (Supplementary Fig. 3).

To estimate the presence of PrP inclusions in the mechanosensitive neuronal system, we quantified the number of PLMs showing condensed PrP immuno-signal spotted with the PrP specific Sha-31 antibody in >150 animals from each PrP line (Fig. 3A). The proportion of PLMs with aggregates increased by 120% in the PrP_E200K line as compared with the PrP_Wt line (Fig. 3B). The number and the volume of the PrP aggregates observed in the PLM cellular body was investigated using confocal microscopy at high magnification (Fig. 3A). First, we observed a significant increase in the number of aggregates per neuron in the PrP_E200K line (Fig. 3C). Second, we categorized the volume of PrP-immunopositive neuronal inclusions in different subgroups. The volume distribution between the PrP_E200K and PrP_Wt lines was significantly different, with a 21.9% increase in the proportion of small inclusions ranging from 1 to 5 µm³ and a substantial decrease for those >20 µm³ in PrP_E200K animals (Fig. 3D).

Phenotypical characterization of transgenic nematodes

To study the functional impact of the deleterious effect of PrP_E200K expression on mechanosensitive neurons, we used specific phenotypical tests exploring the mechanosensitive neuronal system.²³ The motility in liquid media (Fig. 4A) and touch reflex response (i.e. touch test assay; Fig. 4B) were scored. We observed a significant alteration of mobility and touch response in PrP_E200K line compared with the PrP_Wt line (Fig. 4B). The invalidation of PRNP expression by RNAi feeding fully restored the functional scores in PrP_E200K animals (Fig. 4A and B). In addition, to test the extensive impact of PrP expression, we measured the survival of the different PrP lines. We observed a significant decrease in the survival rate in the PrP_E200K line that was abolished by RNAi feeding (Supplementary Fig. 4).

Validation of the PrP_E200K line for therapeutic research using astemizole and doxycycline treatment

To validate the usefulness of the PrP_E200K line in therapeutic research, we first assessed the efficacy of two known anti-prion compounds, astemizole and doxycycline, which have shown an activity against prion propagation.^14,31,32 The effect of treatment with both molecules in the behavioural mechanosensitive tests was assessed. In the PrP_E200K line, astemizole and doxycycline treatments induced a significant and concentration-dependent improvement of motility and a significant increase in the mechanosensitive touch reflex response (Fig. 5A, B and Supplementary Fig. 5). We found that astemizole at 5 µM and doxycycline at 100 µM were the optimal concentrations (Supplementary Fig. 5), which were subsequently used in further experiments. The loss of PLM neurons in the PrP_E200K line was significantly reduced by astemizole and doxycycline treatments compared with the untreated worms (Fig. 5C). Treatment with streptomycin, used as an antibiotic control, had no effect (Supplementary Fig. 6). PrP immunostaining of treated animals showed that the number of PrP inclusions in PLM neurons decreased in worms treated with astemizole, but not with doxycycline, as illustrated in Fig. 5D and quantified in Fig. 5E. In astemizole-treated PrP_E200K animals, we observed a significant decrease in the number of PrP aggregates per neuron, compared with the animals treated with doxycycline. In PrP_E200K animals treated with astemizole, the size distribution of the PrP inclusions showed a significant decrease in the number of objects with a volume >5 × 10⁻³ µm³ compared with the untreated animals (Fig. 5F). On the contrary, in doxycycline-treated PrP_E200K animals, we observed a significant increase in the number of objects with a volume >5 × 10⁻³ µm³ (Fig. 5F), consistent with what has been observed in CJD prion-infected mammalian neurons.¹⁴ We then assessed the effect of astemizole and doxycycline treatment on the detection of PK-resistant PrP_E200K (Fig. 5G). Although astemizole treatment at 5 µM did not induce a decrease in PrP^res detection in PrP_E200K animals, the PrP^res signal was abolished in animals treated with doxycycline at a 100 µM concentration. These results at the functional, cellular and biochemical levels validated the usefulness of our C. elegans model to investigate the anti-prion effect of pharmacological compounds in a human PrP context.

Screening of a library of FDA-approved compounds that cross the blood–brain barrier

We took advantage of our C. elegans prion model associated with a PrP_E200K-induced neurotoxicity that allows large-scale analyses using an automated imaging system to screen the efficacy of 320 pharmaceutical compounds from an FDA-approved library described to cross the blood–brain barrier. First, the absence of neuronal toxicity of the selected compounds in the PrP_Wt line was investigated (Fig. 6). Five molecules showed a neurotoxic effect, which was further confirmed (Supplementary Fig. 7), and these were excluded from further investigation. Using the PrP_E200K line (n = 24 animals per condition), 17 molecules were identified to have a significant protective effect against PrP_E200K toxicity on PLM GFP-tagged neurons (Fig. 6).

Treatment with retained compounds was neuroprotective and restored neuronal functionality

We used a larger number of animals and a different counting method with an epifluorescence microscope to manually assess the number of GFP-expressing PLM neurons (Fig. 7A). We confirmed the neuroprotective effect of the 17 retained compounds with effective concentrations ranging from 1 to 100 µM. To assess the functionality of the protected neurons, we further studied the effect of treatment on motility and touch test reflex responses. In the PrP_E200K line, all molecules (except one—pheniramine maleate—for the touch assay) restored significantly, at least in part, the behaviours associated with the mechanosensitive system (Fig. 7B and C).

Some selected compounds modified intraneuronal PrP aggregates and reduced PrP^res accumulation

We next studied the effect of the selected compounds on the formation of PrP aggregates observed in PLM neurons in the PrP_E200K line. Using immunodetection of human PrP and confocal cell imaging, we quantified the number and the volume of PrP aggregates detected in the cell body of PLM neurons (Fig. 8A–C). Among the 17 selected molecules, nine induced a significant decrease of either the number (n = 7; pheniramine maleate, domperidone, tenoxicam, citalopram hydrobromide, butamben, indoprofen and baclofen) or the volume (n = 3; baclofen, naloxone hydrochloride and cinnarizine) of intraneuronal PrP aggregates. Then, the effect of these nine compounds on the accumulation of PrP^res was assessed in PrP_E200K line using a western blotting approach (Fig. 8D). We observed a decreased in PrP^res signal in animals treated with five compounds (tenoxicam, indoprofen, cinnarizine, butamben and naloxone) as illustrated in Fig. 8D (western blot) and 8E (semi-quantification of PrP^res western blot signal using an optical densitometry methodology). This effect, as for the effects of doxycycline and astemizole, was not due to a decrease in the expression of the PrP transgene in treated animals (Supplementary Fig. 8).

Discussion

Although infectious CJD is a less frequent form of human prion disease, most of the in vivo and in vitro models that have been set up to explore prion mechanisms and for therapeutic research are infectious. One of the reasons for this is the absence of relevant models of sporadic CJD, the most common form of human prion disease. Inherited prion diseases represent nearly 15% of human cases and some genetic models have been developed using transgenic or knock-in mice.³³ For example, transgenic mice overexpressing a mouse–human chimeric PrP with the E199K mutation develop an encephalopathy at 5–6 months of age leading to terminal condition at about 12–15 months of age.³⁴ Although such mouse models are useful to assess the efficacy of a specific treatment,³⁵ they are time-consuming, their potential for experimental genetic intervention is limited and they are not compatible with medium to high throughput screening of therapeutic compounds. Here, we provide the first simplified genetic model of E200K-associated CJD expressing the human PrP and recapitulating the cardinal features of prion diseases at the neuronal levels (i.e. neuronal dysfunction and loss leading to a modified phenotype, PrP aggregation and the formation of PK-resistant form of the prion protein). We demonstrated that this in vivo model is useful for the search of anti-prion compounds and identified within a library of FDA-approved compounds known to penetrate the brain several molecules showing an effect on both abnormal PrP accumulation and prion-induced neuronal death.

We focused on the E200K mutation for several reasons. First, familial CJD associated with E200K mutation (fCJD-E200K) is the most frequent form of inherited prion diseases. Second, the clinico-pathological phenotype of fCJD-E200K is similar to that observed in typical cases of sporadic CJD. Lastly, the disease duration in fCJD-E200K is shorter than in other genetic forms of prion diseases (such as fatal familial insomnia and, notably, Gerstmann-Sträussler-Scheinker disease) suggesting a higher toxicity of E200K PrP. Indeed, in a previous study, we expressed in a specific neuronal population of C. elegans, a human PrP carrying an expansion in the octarepeat domain of the protein that is associated with Gerstmann-Sträussler-Scheinker disease and long disease duration. In this model, although animals showed a dysfunction of the targeted neurons, no neuronal loss was observed.³⁶

Another group developed a C. elegans model of mammalian prion-cell toxicity by expressing soluble mouse prion proteins, devoid of a GPI anchor, in the cytoplasm of muscle cells.³⁷ However, it remains unclear how such PrP in muscles can accurately model a disease that almost exclusively targets the CNS. In addition, one of the major advantages of our model is the expression of human PrP sequence that is key to developing CJD-relevant therapeutic research.

We chose to target the mechanosensitive system because its neurons can easily be quantified and their function monitored using C. elegans functional testing.²⁹ Human PrP produced by C. elegans mechanosensitive neurons showed some of the biochemical properties of the protein to be detected in mammals. Indeed, we observed the conventional tri-band pattern on western blot corresponding to glycosylated forms of PrP as confirmed by the results of PNGase treatment. Compared with the control line expressing similar levels of PrP, the PrP_E200K-expressing line showed a higher propensity to form PrP aggregates in the soma and axon of the PLM neurons. This result is consistent with previous studies in a C. elegans model of protein aggregation using yeast-prion sequences.^38,39 Aggregate formation in PrP_E200K line was associated with a partial resistance to proteolysis confirming that invertebrate models, including C. elegans and Drosophila, are appropriate to generating a protease-resistant PrP core when a mutant PrP is expressed.^40,41

In PrP_Wt animals, we did not observe significant alterations of the mechanosensitive neuronal system, suggesting that the expression of wild-type PrP by itself is not sufficient to alter neuronal homeostasis in our experimental conditions. In contrast, a similar level of PrP_E200K induced the dysfunction of the mechanosensitive system associated with a specific neuronal death occurring 3 days after hatching. To date, such neuronal loss had never been observed in invertebrate models of prion diseases, notably in C. elegans lines expressing a Gerstmann-Sträussler-Scheinker disease-associated insertional mutation and in Drosophila.^36,37,42 Thus, our model may provide a useful tool to explore further the mechanisms involved in prion-induced neurodegeneration and to identify novel therapeutic targets.

To validate the PrP_E200KC. elegans line as a genuine in vivo model for the search of anti-prion compounds, we assessed the effect of two molecules, doxycycline and astemizole, previously described as active on both PrP accumulation and prion-induced neurodegeneration.^14,31 Doxycycline has been shown to bind PrP, to inhibit CJD prions propagation in cell cultures and to prolong the survival of scrapie-infected hamsters.^14,43,44 In PrP_E200KC. elegans line, doxycycline treatment significantly reduced the loss of mechanosensitive neurons and PrP^res formation. Thus, our results support the efficacy of doxycycline against E200K prions. We observed that doxycycline treatment induced an increase in the volume of PrP aggregates in PrP_E200K neurons. Interestingly, we previously observed a similar effect of doxycycline on PrP aggregates in mouse neuronal cultures infected with CJD prions.¹⁴ These observations are consistent with a mechanism proposed to explain the effect of cyclines on amyloid-β peptides and β2-microglobulin and of polythiophenes on PrP with the formation of amorphous aggregates that may sequester harmful PrP species and limit prion propagation.^45–47 Yet, a controlled clinical trial failed to show a significant effect of doxycycline treatment at a daily dose of 100 mg in patients with sporadic and genetic CJD.³² In the brain of treated patients, doxycycline reached low micromolar concentrations, suggesting that it may be effective at higher doses. Indeed, our results suggest that doxycycline effect is dose-dependent and increases from 10 to 100 µM, as observed in CJD-infected neuronal cultures.¹⁴ To note, the possibility that doxycycline could be effective in the presymptomatic stage is currently being tested in a controlled trial in presymptomatic carriers of the D178N mutation of PRNP (EudraCT 2010-022233-28) in Italy. Astemizole, an antagonist of histamine H₁ receptors, inhibits prion replication in cell cultures and prolongs the lifespan of Rocky Mountain Laboratory (RML) prion-infected mice.³¹ In the PrP_E200KC. elegans line, astemizole also showed anti-prion effects. In addition, astemizole treatment induced a decrease in the number and size of intraneuronal PrP aggregates, which is consistent with the proposed mechanism of preventing prion protein aggregation.⁴⁸ Why the detection of PrP^res is unchanged in astemizole-treated PrP_E200K animals remains to be investigated further. Taken together, these results clearly supported the relevance of our genetic approach in the nematode to model mammalian prion diseases and its usefulness in the search for novel anti-prion compounds.

In a repositioning drug strategy, we then evaluated the anti-prion activity of compounds issued from an FDA-approved library that are able to cross the blood–brain barrier in the PrP_E200KC. elegans line. To our knowledge, this is the first large in vivo screening assay relying on functional assessment and neuronal and biochemical studies in a model of prion diseases. It is also worth noting that this study was performed in worms that express human PrP, neutralizing the issue related to the species dependence of anti-prion activity.

Among 320 molecules, we identified 17 compounds showing a significant neuroprotective effect against the toxicity induced by the presence of human PrP_E200K. All of the 17 selected molecules improved the results of the mechanosensitive functional test and restrained the proportion of paralysed animals in liquid media. Among the 17 selected compounds, eight did not reduce the number or the volume of neuronal PrP aggregates. One plausible explanation is that these eight compounds had only neuroprotective properties in our model without affecting mutant prion protein aggregation. Nine molecules decreased the number and/or volume of PrP intraneuronal inclusions in the PrP_E200K line. Among these, five molecules induced a prominent decrease in PrP^res accumulation. The five candidates originated from four pharmaceutical classes, including: (i) anti-histaminic (cinnarizine); (ii) anti-inflammatory (tenoxicam and indoprofen); (iii) anaesthetic (butamben); and (iv) opioid antagonist (naloxone hydrochloride).

A role for the histamine-associated pathway has been suggested in neurodegenerative diseases, notably in microglia-mediated neuroinflammation.^49,50 Our results are consistent with those obtained with astemizole, another antagonist of H₁ receptors, in cellular and experimental mouse models of scrapie infection.³¹ It is striking that different approaches in various species produced converging data supporting the potential of anti-H1 molecules in the treatment of prion diseases. Because worms are devoid of H1 metabolic receptor homologues and do not produce histamine, it can be hypothesized that the anti-prion effect shared by both molecules is related more to their structural similarities than to a specific anti-H1 effect. Of note, cinnarizine has also anti-D2 properties and is able to block calcium channels. However, C. elegans mechanosensitive neurons do not express D2/D2 like receptor and calcium channels.⁵¹ In addition, flumarizine, another antagonist of D2 receptors, had no effect in the PrP_E200K line, indicating that the mechanism is independent of the dopamine pathway.

Tenoxicam, a piroxcam analogue derived from thieno-thiazine, and indoprofen belong to the non-steroidal anti-inflammatory drugs (NSAIDs) family. Both of these drugs are known to limit inflammation by specifically inhibiting the activity of both cyclooxygenase isoforms (COX-1 and COX-2).⁵² The involvement of COX-2 in CJD has been proposed.⁵³ COX-2 immunoreactivity is observed in microglial cells in the brain of scrapie-infected mice⁵⁴ and an increased production of prostaglandins PGE₂ is associated with COX-2 alterations in the hippocampal region.⁵⁵ In patients with variant CJD, the level of PGE₂ in the CSF is increased.⁵⁶ No COX homologue has been identified in the nematode C. elegans,⁵⁷ suggesting that the effects of indoprofen and tenoxicam involve either a direct effect on PrP^res formation and toxicity or a different cellular pathway.

Butamben (n-butyl p-aminobenzoate) is an ester-linked topical anaesthetic that blocks the sodium and potassium channels. To date, there are no data suggesting that butamben has anti-prion activity. As observed for COX-2 inhibitors and anti-H1 compounds, the main target of butamben is absent in nematodes and the other analogues tested showed no effect, supporting the idea that the effect of this compound in PrP^res formation relies on another molecular mechanism.

Naloxone is a selective competitive antagonist of the opiate receptors, used in medication for example as an antidote in opioid overdose. In C. elegans, while the opioid mammalian system is not present, naloxone reverses opioid invertebrate behaviour⁵⁸ and inhibits the feeding reflex⁵⁹ suggesting the presence of specific molecular targets in the nematode. Whether the anti-prion effect of naloxone is direct or involves such unknown pathway remains to be investigated.

Our study has limitations that are inherent to the C. elegans model. First, although we expressed a human prion protein and the gene homology between humans and nematodes, we cannot exclude that species-specific cofactors involved in the aggregation and toxicity of prion protein are not present in our model. This may prevent the identification of new compounds that target such cofactors in humans. Moreover, if the identified compounds do not act directly on the prion protein, their molecular targets in C. elegans may not necessarily have counterparts in humans. Furthermore, the transposition of active concentrations and tissue distribution of compounds from worms to mammals are obviously not easy to anticipate. However, we studied molecules that cross the blood–brain barrier in humans with known pharmacokinetics. The mechanisms involved in the anti-prion effect of the molecules identified in the PrP_E200K line remain to be established. Because of these limitations, a direct perspective is to validate the anti-prion effect of the most promising compounds we identified in mammalian models of human prion diseases such as transgenic mice overexpressing mutant PrP³⁴ or human PrP and inoculated with brain extracts from patients with sporadic and inherited CJD.⁶⁰

To conclude, we provide the first experimental system allowing the in vivo screening of large libraries of compounds in the prion field. Besides breaking a technological limitation in therapeutic research in prion diseases, it will be key to studying misfolded PrP propagation in a well-described neuronal system. Indeed, our observation of the reduced survival of PrP_E200K animals suggests that PrP_E200K is able to diffuse and trigger cell dysfunction and death outside the mechanosensitive system. In the near future, taking advantage of C. elegans models of prion propagation will be useful to exploring the still unsolved questions such as the mechanisms involved in inter-neuronal prion propagation and the prion strain phenomenon.

Acknowledgements

We thank the Caenorhabditis Genetics Center for providing strains. We acknowledge the technical service from the Brain Institute platforms, in particular: Delphine Boutelier and Yannick Marie (IGenSeq); David Akbar and Patrick Michel (Celis); and Dominique Langui, Aymeric Millecamps, Claire Lovo and Basile Gurchenkov (icm.Quant). We acknowledge Vincent Galy for technical help.

Funding

This work was supported by the CJD foundation and Santé Publique France.

Competing interests

N.B. has received research support from Ehtnodyne outside the scope of this study. S.H. has received research support from Institut de Recherche Servier, LFB Biomedicaments and MedDay Pharmaceuticals outside the scope of this study. N.B. and S.H. have a patent method for treating prion diseases (PCT/EP 2019/070457) pending. International patents (EP 21 305 659 and EP 21 305 660), which includes some of the data herein, were filed by N.B., S.L. and S.H.

Supplementary material

Supplementary material is available at Brain online.

References

 
• CJD

  Creutzfeldt-Jakob disease

 
• PK

  proteinase K

 
• PLM

  posterior lateral microtubule

Author notes