Amplifying the heat shock response ameliorates pathology in mouse and human models of ALS and FTD.

36 Amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD) are now widely considered to 37 be part of a disease spectrum with the identification of common pathological features and genetic 38 causes. However, despite these advances, there remains no effective therapy for these conditions. In 39 this study we demonstrate that mice expressing mutant valosin containing protein (VCP) develop an 40 ALS/FTD-like phenotype in the spinal cord and brain, and treatment with arimoclomol, a 41 pharmacological amplifier of the cytoprotective heat shock response ameliorates this phenotype. 42 Moreover, the beneficial effects of arimoclomol are seen in both fibroblasts and iPSC-derived motor 43 neurons from patients. Importantly, we show the pathological changes targeted by arimoclomol in our 44 experimental models are present in post-mortem FTD patient tissue. Together with previous data 45 demonstrating the efficacy of arimoclomol in SOD1-ALS models, our findings suggest that 46 arimoclomol may have therapeutic potential not only in non-SOD1 ALS but also for the treatment of 47 FTD. arimoclomol in mVCP mice and VCP patient cells were also found to be present in in post-mortem brain tissue from patients with FTD, confirming the relevance of the findings in our experimental models. These results provide further evidence that amplification of the HSR by treatment with arimoclomol may be a beneficial therapy for both non-SOD1 ALS as well as FTD.


INTRODUCTION
to activation of the transcription factor Heat Shock Factor 1 (HSF-1) 13

160
Since mVCP mice have been reported to also display spinal cord and brain pathology which is similar 161 to that observed in MSP, and which reflects ALS and FTD 26 , respectively, in this study we examined 162 the potential beneficial effects of arimoclomol on the ALS pathology and FTD pathology that is present 163 in these mice. We observed a significant improvement in the pathology present in both the spinal cord 164 and brain of mVCP mice following treatment with arimoclomol. Furthermore, we also examined the 165 effect of augmenting the HSR by arimoclomol treatment in human cellular models, including mutant 166 VCP patient fibroblasts and iPSC-derived motor neurons, and found that arimoclomol rescues key 167 degenerative features in these mVCP patient cells in vitro. Importantly, the pathological features that 168 were improved by arimoclomol in mVCP mice and VCP patient cells were also found to be present in 169 in post-mortem brain tissue from patients with FTD, confirming the relevance of the findings in our 170 experimental models. These results provide further evidence that amplification of the HSR by 171 treatment with arimoclomol may be a beneficial therapy for both non-SOD1 ALS as well as FTD.    extensor digitorum longus (EDL) muscles in both hindlimbs were exposed and dissected free of other 205 tendons before being attached by silk thread to isometric force transducers (Dynamometer UFI 206 Devices, Welwyn Garden City, UK). The sciatic nerve was exposed and sectioned, and all branches 207 were cut except for the deep peroneal nerve that innervates the EDL muscles. Isometric contractions 208 were elicited by stimulating the sciatic nerve using square-wave pulses of 0.02-ms duration at 209 supramaximal intensity, using silver wire electrodes. The number of motor units in the EDL muscles             To assess the percentage of cells with morphological nuclear abnormalities, fibroblasts were stained 326 with DAPI to outline the nuclear region of fibroblasts. A series of ten images were taken at 20x 327 magnification, from one coverslip for each control or patient cell line, using a Leica DMR microscope.

328
The number of cells with nuclear abnormalities were manually counted using ImageJ and the  LGlutamine, N2 supplement, non-essential amino acids, B27 supplement, β-mercaptoethanol (all from 341 Life Technologies) and insulin (Sigma). Treatment with small molecules from day 0-7 was as follows:

400
Treatment with arimoclomol significantly improved motor neuron survival in mVCP mice, and 612 ± 401 18 motor neurons survived representing an improvement of 17% (p=0.029) when compared to 402 untreated mVCP mice. Importantly, arimoclomol had no effect on motor neuron survival in wtVCP 403 mice, so that 708 ± 24 motor neurons survived compared to 725 ± 9 in untreated wtVCP mice.

404
Furthermore, in untreated mVCP mice, motor neurons which survived at 14 months, had an abnormal 405 morphology, with small, compacted cell bodies, contrasting with the large polygonal shape of motor 406 neurons observed in control animals (Fig. 1C). We therefore assessed the total soma area of motor 407 neurons in the sciatic pool in each cohort of mice. As can be seen in Fig. 1E, there was a clear  (Fig. 2C and 2D). Importantly, this indicates 429 that the endogenous response to cell stress can and has been triggered in the mVCP motor neurons, 430 although this was not sufficient to prevent the development of pathology. In mVCP mice treated with 431 arimoclomol, an amplifier of the HSR, an even greater increase in HSP70 expression was observed,

432
to approximately 3-fold that of controls ( Fig. 2C and 2D). In addition, in arimoclomol-treated mVCP 433 mice, GFAP-positive astroglia were also found to express increased levels of HSP70 in the spinal 434 cord (Fig. 2E), which likely reflects an additional cytoprotective response to mVCP-induced stress.

436
VCP is known to play an essential role in autophagy 10 , and dysfunctional autophagy has been 437 implicated in the pathogenesis of ALS and MSP. For example, MSP patients expressing mVCP show 438 evidence of disrupted autophagy, with accumulation of the key autophagic markers p62 439 (Sequestosome 1) and LC3 within myofibres 10, 32 . We therefore examined the pattern of expression 440 of these two autophagic markers in the spinal cord of mVCP mice. As can be seen in Figure 3, we 441 observed an increase in p62 expression in both the white and grey matter of the spinal cord compared 442 to control animals ( Fig. 3A and B), with p62 aggregates clearly visible in motor neurons (Fig. 3B i-443 magnified inset). Co-labelling with fluoromyelin suggested that the intense p62 staining observed in 444 mVCP spinal cord sections was associated with oligodendrocytes ( Fig. 3A ii). Closer examination of

472
In contrast, TDP-43 expression in the cortex of arimoclomol-treated mVCP mice was similar to control animals, as observed in the spinal cord, and no cytoplasmic mislocalisation associated with nuclear 474 clearance of TDP-43 was apparent in the brains of the arimoclomol-treated group (n=3 mice, Fig. 4A).

476
We next examined the expression of ubiquitin in the brain of mVCP mice. Similar to our findings in the 477 spinal cord, in mVCP mice we observed ubiquitin-positive intracellular aggregates in the brain, 478 although these were not limited to the cortex. No ubiquitin-positive aggregates were detected in control 479 animals, or in mVCP mice treated with arimoclomol (Fig. 4B).

500
(A232E) mouse muscle 41 . It is therefore possible that a similar phenomenon also occurs in the brain 501 of mVCP mice, leading to the aggregates observed in this study. Two additional markers of stress 502 granules, FMRP and G3BP were also associated with protein aggregates in the brain of mVCP mice 503 ( Supplementary Fig. 1 B). In control animals and mVCP mice treated with arimoclomol, no aggregates 504 positive for any of the tested stress granule markers were observed ( Supplementary Fig. 1 B).

506
Furthermore, as observed in the spinal cord, immunostaining of mVCP mouse cortex also revealed 507 the presence of p62 and LC3-positive aggregates, with intense cytosolic LC3 staining ( Fig. 5A and B).

508
No LC3 staining was observed in control tissue or arimoclomol treated mVCP mice.

509
Similar to the findings in the spinal cord of mVCP mice, brain tissue from these animals also showed 511 an increase in the expression of HSP70 compared to wtVCP and non-transgenic WT controls, 512 indicative of stress-induced activation of the HSR. In mVCP mice treated with arimoclomol, HSP70 513 expression was found to be further increased, suggesting that the HSR has been augmented by 514 arimoclomol treatment (Fig. 5C). Interestingly, as noted in the spinal cord, HSP70 expression was 515 significantly pronounced in glial cells in the brain, which are known to have a robust stress induced 516 HSR.

518
Pathology observed in mVCP mice reflects that observed in ALS/FTD patient brain tissue.

519
To verify whether the pathological features observed in the CNS of mVCP mice, which were improved 520 by treatment with arimoclomol, have relevance to the human disease, we examined post-mortem brain

526
Since autophagy is a common pathway that may be disrupted in FTD brain regardless of specific 527 disease aetiology, we examined human FTD brain for the presence of autophagy-related proteins 528 which were abnormally present in mVCP mice. Globular, juxtanuclear p62-positive cytoplasmic 529 inclusion bodies were present in cortical brain sections of the FTD cases (Fig. 6A) and were similar to 530 those seen in mVCP mice (Fig. 5A), suggesting a generalised disruption in protein homeostasis in 531 FTD patients regardless of subtype; no p62 immunostaining was observed in brain sections from 532 healthy controls. Strong p62 staining was also observed in the FTD-MAPT brain (Fig. 6A), similar to 533 that reported in the brain of Alzheimer's disease patients early in pathogenesis within neurofibrillary 534 tangles 43 . In FTD-TDPA brain, intense p62 staining was observed in neural processes in addition to  Fig. 2). It is likely that total LC3 I level in samples with less 544 cytoplasmic staining are also above basal levels but not detected in the soluble lysates due to 545 aggregation. No significant difference in LC3 II, which closely correlates to the number of 546 autophagosomes, was detected by immunoblot.
Although TDP-43 mislocalisation in FTD brain is a well-established phenotype, we demonstrate 548 extensive cytoplasmic mislocalisation of TDP-43 in the brain of all four FTD patients in which p62 and 549 LC3 are also aggregated. Such mislocalisation was not observed in samples from healthy controls 550 (Fig. 6C). Our findings thus reveal that the pathological phenotypes identified in the mVCP mouse 551 model are indeed present in the human disease. Moreover, these phenotypes are not limited to 552 mutations in VCP, therefore expanding the relevance of our findings of the beneficial effects of 553 arimoclomol to non-VCP FTD patients.

555
While the ability to mount a HSR under conditions of cell stress is present throughout life, this 556 cytoprotective mechanism is thought to become less effective in later life, likely contributing to the 557 age-related increase in susceptibility to degenerative diseases 45 . In our study, compared to age-558 matched control post-mortem samples, HSP70 expression was indeed significantly lower in the cortex 559 of 2 out of 4 of the FTD patients examined (Fig. 6D). However, a reduced HSR is not a prerequisite

578
Fibroblasts derived from patients with the most prevalent MSP-associated VCP mutation, R155H, 579 have been reported to display some evidence of a pathological phenotype when grown in culture 47 .

580
In this study, we assessed the distribution of a number of ALS/FTD-associated proteins in fibroblasts 581 expressing either the R155H VCP mutation, obtained from three patients with a mild disease 582 phenotype or the R93C VCP mutation, from a patient with a more severe disease phenotype. We 583 observed an increase in ubiquitin, p62 and TDP-43 immuno-reactive aggregates in the cytoplasm of mVCP fibroblasts from all patients. The majority of p62-positive aggregates were located adjacent 585 to the nucleus (Fig. 7A), whilst ubiquitin-positive aggregates were present either as large, distinct 586 aggregates (Fig. 7B i) or as smaller, diffuse aggregates (Fig. 7B ii) that were more evenly dispersed 587 throughout the cytoplasm. Although this dispersed expression of ubiquitin was also noted in a small 588 number of age-matched control fibroblasts (data not shown), large aggregates of ubiquitin or p62 were 589 not detected in control cells or in patient fibroblasts after treatment with arimoclomol (10µM, 24-hour 590 treatment).

593
However, abnormal TDP-43 expression patterns were noted in mVCP fibroblasts, with differences 594 observed between patient lines, potentially reflecting phenotypic differences between mild and severe 595 cases. In some mVCP fibroblasts, TDP-43 was retained in the nuclei of cells in which cytoplasmic 596 aggregates were present (Fig. 7C ii). However, several cells from the more severely affected mVCP 597 R93C line showed evidence of TDP-43 nuclear depletion, even in the absence of cytoplasmic 598 aggregates (Fig. 7C iii). This pattern was not seen in the milder R155H line, control cells or in R93C 599 cells treated with 10µM arimoclomol. Indeed, quantification of nuclear TDP-43 levels shows that 600 arimoclomol treatment prevents the loss of TDP-43 in the nuclei of R93C cells (Fig. 7D).

602
The pathological changes observed in mVCP fibroblasts were also associated with an increase in the 603 expression of HSP70 (Fig. 7E), which was almost 2-fold higher than that observed in control cells,

604
indicating the presence of cell stress. Furthermore, we also observed a significant increase in the 607 Surprisingly, these cells were not undergoing apoptosis, as assessed by TUNEL staining for DNA 608 double-strand breaks ( Supplementary Fig. 3). Quantification of the number of aberrant nuclei in mVCP 609 fibroblasts in the absence and presence of increasing concentrations of arimoclomol revealed a dose-610 dependent reduction in the number of disrupted nuclei, with a statistically significant difference 611 observed at 50µM of arimoclomol (Fig. 7F iii).

613
In order to confirm our results in a more relevant cellular model of ALS, we also investigated the effects 614 of arimoclomol on iPSC-derived motor neurons from patients with VCP mutations (3 individual patient 615 lines). TDP-43 cytoplasmic mislocalisation has been previously reported in iPSCs from patients with 616 mutated VCP 46 . As expected, we frequently observed distinct cytoplasmic TDP-43 in mVCP IPSC-617 derived motor neuron cultures, with many cells also exhibiting nuclear loss of TDP-43 (Fig. 8A,   618 magnified image of cell). Importantly, mislocalised TDP-43 was rarely seen in the mVCP cultures 619 treated with 50 µM arimoclomol and was absent from healthy control cells. motor neurons, either as many small bodies dispersed throughout the cell or as one large distinct 622 globular aggregate within the cytoplasm (Fig. 8B and D). Quantification of cells with ubiquitin-positive 623 aggregates revealed a 22% increase in mVCP neurons compared to controls, with a significant 624 reduction (down to 11%) in arimoclomol treated cultures (Fig 8C). No statistical significance was 625 identified between control and arimoclomol treated mVCP groups. These findings corroborate our 626 observations in mVCP mice, where abnormal ubiquitin and p62 accumulation, possibly linked to 627 impairment of autophagy or proteasomal degradation, was present in both the spinal cord and brain.

628
More importantly, these data clearly show that arimoclomol treatment leads to a reduction of 629 cytoplasmic ubiquitin aggregates in a specific neurological cellular system with VCP mutation.

640
We have previously demonstrated the beneficial effects of augmenting the HSR with arimoclomol on 641 the ALS phenotype of mutant SOD1 mice 11,12,13 as well as the IBM-like muscle pathology that is 642 present in mutant VCP mice 27 . In the present study, we expand these findings to examine the effects 643 of arimoclomol on the spinal cord and brain pathology which is also part of the MSP phenotype of 644 mVCP mice, and which we show to be similar to pathological characteristics of human ALS/FTD.

683
We also examined two human cellular models of mutant VCP disease, which provide more clinically

708
In parallel to the experimental data, we also confirm that the same key pathological characteristics of 709 ALS/FTD observed in the mouse and recapitulated in the human cell models of mVCP disease, and 710 which were ameliorated by treatment with arimoclomol, are present in cortical brain samples of FTD 711 patients. Our results show that TDP-43 pathology was present in all FTD patient samples examined.

712
In addition, both p62 and LC3 were found to be accumulated in FTD brains across the subtypes 713 examined, either as soluble proteins detectable by immunoblot or as aggregates revealed by 714 histochemistry. These findings suggest that impaired clearance of neurotoxic proteins by autophagy 715 may be a common factor in FTD, and not necessarily specific to a particular mutation or aetiology.

717
Although the scale, kinetics and required threshold of the HSR varies between cells and stressors, 718 this vital mechanism routinely keeps cells free of damaged and surplus proteins 58 . As a highly 719 adaptive system, the HSR is able to tailor the type of stress response that is most appropriate for the 720 type of cell involved and for the specific type of stress they are under 59 , coordinating stress-induced 721 transcription of a variety of chaperones and co-chaperones. While we found the HSR to be induced 722 in all the mVCP models we examined, the magnitude of the endogenous HSR mounted was clearly 723 not sufficient to prevent pathology. Furthermore, the HSR, as measured by HSP70 levels, remained 724 unaltered or deficient in FTD patient tissue which share common pathogenic hallmarks with our 725 disease models. It has been long established that attenuation of the HSR is associated with aging and 726 this may be, in part, responsible for the accumulation of aberrant proteins and thus altered regulation 727 of cell death which we see in chronic neurodegeneration 45, 60 . All FTD post-mortem brain samples 728 investigated in this study were from patients of relatively advanced age, ranging from 62-79 years old 729 and were compared to age-matched controls and as such, a lower HSR is not unexpected. Our data therefore suggests that a diminished HSR is not necessarily a precondition for treatment with a HSR 731 amplifier, and pharmacologically correcting any deficiency or amplifying the endogenous HSR using 732 arimoclomol may be an effective approach to alleviate neurodegenerative pathology and therefore 733 delay disease progression in ALS/FTD patients.

735
Importantly, when considering arimoclomol's potential as a therapeutic agent, its ability to act as an 736 amplifier and not an inducer of the HSR; thereby reducing any potential off-target effects, is 737 critical. In all the disease models examined in this study, we demonstrate that amplification of the

758
Arimoclomol is a small molecule HSP amplifier that is clinically well-tolerated and readily crosses the 759 blood-brain barrier. The results of this study thus suggest that arimoclomol treatment may be a 760 disease-modifying therapeutic approach for the treatment of non-SOD1 ALS as well as MSP.       Ubiquitin immunoreactivity in mouse cortex shows ubiquitin-positive aggregates in mVCP brain (red). Scale bar = 20μm. (C) stress granule marker Tia1 colocalised with TDP-43 and ubiquitin in mVCP brain.

Figure 5
Arimoclomol treatment prevents p62 and LC3 aggregation and enhances HSP70 in mVCP mouse brain. Histological images of (A) p62 expression and (B) LC3 expression in mouse brain sections. White arrows indicate protein aggregates. (C) HSP70 expression in mouse brain with and without neuronal marker (β-III tubulin, red) and nuclear marker (DAPI, blue). White arrows indicate glial cells expressing HSP70. Scale bar =10μm. Figure 6