J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp-2017-317562 on 15 February 2018. Downloaded from Neuromuscular

Research paper PFN2 and GAMT as common molecular determinants of axonal Charcot-Marie-Tooth disease Manisha Juneja,1,2 Abdelkrim Azmi,3 Jonathan Baets,2,3,4 Andreas Roos,5,6 Matthew J Jennings,5 Paola Saveri,7 Chiara Pisciotta,7 Nathalie Bernard-Marissal,8,9 Bernard L Schneider,9 Catherine Verfaillie,10 Roman Chrast,11,12 Pavel Seeman,13 Angelika F Hahn,14 Peter de Jonghe,2,3 Stuart Maudsley,3 Rita Horvath,5 Davide Pareyson,7 Vincent Timmerman1,2

►► Additional material is Abstract Molecular genetic studies have revealed that for published online only. To view Background charcot-Marie-Tooth type 2 (CMT2) the axonal CMT alone, already more than 140 please visit the journal online neuropathy is characterised by a vast clinical and genetic mutations in 26 genes have been identified and the (http://dx.​ ​doi.org/​ ​10.1136/​ ​ 2 jnnp-2017-​ ​317562). heterogeneity complicating its diagnosis and therapeutic list keeps growing every year. The diagnosis and intervention. Identification of molecular signatures that prognosis for a given individual involves careful For numbered affiliations see are common to multiple CMT2 subtypes can aid in assessment of medical history, physical examina- end of article. developing therapeutic strategies and measuring disease tion, neurological examination, nerve conduc- outcomes. tion testing as well as a detailed family history Correspondence to and genetic testing.3 However, the presence of Dr Vincent Timmerman, Methods a proteomics-based approach was performed Peripheral Neuropathy on lymphoblasts from CMT2 patients genetically tremendous genetic diversity renders its diagnosis Research Group University of diagnosed with different gene mutations to identify and prognosis rather challenging. Very recently, Antwerp, Antwerpen 2610, differentially regulated proteins. The candidate proteins potential biomarkers for CMT1 have been reported Belgium; vincent.​ ​timmerman@​ from plasma and skin biopsies derived from CMT1 uantwerpen.be​ were validated through real-time quantitative PCR and western blotting on lymphoblast samples of patients and patients.4 5 However, no studies have been reported Received 6 November 2017 controls, motor neurons differentiated from patient- to date describing biomarkers for CMT2. Revised 24 January 2018 derived induced pluripotent stem cells (iPSCs) and sciatic Thus, it remains essential to identify hallmarks Accepted 30 January 2018 nerves of CMT2 mouse models. of axonal degeneration that are specific for CMT2. Published Online First Moreover, experimental studies have so far focused 15 February 2018 Results proteomic profiling of patient lymphoblasts resulted in the identification of profilin 2 (PFN2) on the clinical aspects and/or pathomechanisms for and guanidinoacetate methyltransferase (GAMT) as single genes.6 Given the genetic heterogeneity of the commonly downregulated proteins in different genotypes disease and the fact that it is a rather rare disorder compared with healthy controls. This decrease was of the peripheral nervous system, this ‘single gene’ also observed at the transcriptional level on screening approach may prove impractical clinically. Here, we 43 CMT2 patients and 22 controls, respectively. A intend to identify commonalities among different progressive decrease in PFN2 expression with age axonal CMT subtypes, which may reveal novel was observed in patients, while in healthy controls its common biomarkers, and therapeutic targets that http://jnnp.bmj.com/ expression increased with age. Reduced PFN2 expression will be of benefit for a larger group of CMT patients. was also observed in motor neurons differentiated from CMT2 patient-derived iPSCs and sciatic nerves of CMT2 Materials and methods mice when compared with controls. However, no change in GAMT levels was observed in motor neurons and Participants and clinical characterisation CMT2 mouse-derived sciatic nerves. This was a multicentre study involving axonal CMT Conclusions We unveil PFN2 and GAMT as molecular and distal hereditary motor neuropathy (dHMN) on 13 July 2018 by guest. Protected copyright. determinants of CMT2 with possible indications of patients obtained through neuromuscular centres in the role of PFN2 in the pathogenesis and disease Belgium (18 patients), Italy (17 patients) and the UK progression. This is the first study describing biomarkers (8 patients). The study included overall 43 CMT2/ distal dHMN patients (19 men and 24 women; that can boost the development of therapeutic strategies median age, 54 years; range, 17–78 years) and 22 targeting a wider spectrum of CMT2 patients. healthy controls (18 from Belgium and 4 from Italy, 10 men and 12 women; median age, 57.5 years; range, 31–71 years) (table 1). We included dHMN patients who have a predominant motor neurop- Introduction athy without clear evidence of sensory involvement Charcot-Marie-Tooth (CMT) disease or hereditary at clinical and electrophysiological examination motor and sensory neuropathy (HMSN) is one of because (1) there is a clear genetic and pheno- To cite: Juneja M, Azmi A, the most frequent inherited peripheral neuropa- typic overlap between axonal CMT and dHMN; Baets J, et al. J Neurol thies that is classified into a demyelinating (CMT1) (2) both clinical entities are difficult to differen- Neurosurg Psychiatry and an axonal form (CMT2) based on where the tiate from each other; (3) they can be caused by 2018;89:870–878. primary deficit occurs.1 the same mutation in genes like GARS, HSPB1 and

870 Juneja M, et al. J Neurol Neurosurg Psychiatry 2018;89:870–878. doi:10.1136/jnnp-2017-317562 J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp-2017-317562 on 15 February 2018. Downloaded from Neuromuscular

Table 1 Patients and families involved in this study Family CMT* subtype Sex Origin Gene involved Mutation Age at sampling Pure motor or/and sensory involvement F1 CMT2I Male Italy MPZ p.Ser44Phe 72 Motor and sensory F1 Control Female Italy NA NA 61 NA F2 CMT2N Female Italy AARS p.Thr385Ile 35 Motor and sensory F2 CMT2N Female Italy AARS p.Thr385Ile 61 No info F3 dHMN Female Italy GDAP1 p.His123Arg 23 Motor only F3 CMT2K Female Italy GDAP1 p.His123Arg 54 Motor and sensory F4 CMT2I Male Italy MPZ p.Ser44Phe 40 Motor and sensory F5 dHMN Female Italy MFN2 p.Hys165Leu 42 Motor only F6 CMT2K Female Italy GDAP1 p.Arg120Trp 68 Motor and sensory F6 Control Male Italy NA NA 71 NA F7 CMT2K Male Italy GDAP1 p.Arg226Ser 65 Motor and sensory F7 CMT2K Female Italy GDAP1 p.Arg226Ser 50 Motor and sensory F8 dHMN Female Italy BSCL2 p.Asn88Ser 51 Motor only F8 dHMN Female Italy BSCL2 p.Asn88Ser 50 Motor only F9 CMT2I Male Italy MPZ p.Pro70Ser 60 Motor and sensory F9 CMT2I Female Italy MPZ p.Pro70Ser 69 Motor and sensory F9 Control Female Italy NA NA 41 NA F10 dHMN Male Italy BSCL2 p.Asn88Ser 40 Motor only F11 dHMN Male Italy MPZ p.Pro70Ser 50 Motor only F12 CMT2K Female Italy GDAP1 p.Arg120Trp 70 Motor and sensory F12 Control Male Italy NA NA 68 NA F13 dHMN Male UK AARS p.Arg329His 57 Motor only F13 CMT2N Male UK AARS p.Arg329His 34 Motor and sensory F14 CMT2O Female UK DYNC1H1 p.Arg3078Gln 46 Motor and sensory F15 CMT2N Female UK AARS p.Arg329His 78 Motor and sensory F15 CMT2N Female UK AARS p.Arg329His 55 Motor and sensory F16 CMT2N Male UK AARS p.Arg329His 48 Motor and sensory F17 dHMN Female UK GARS p.Lys510Gln 52 Motor only F17 CMT2D Female UK GARS p.Lys510Gln 29 Motor and sensory F18 dHMN Female France HSPB1 p.Arg127Trp 57 Motor only F19 dHMN Male Belgium HSPB8 p.Lys141Asn 57 Motor only F19 dHMN Female Belgium HSPB8 p.Lys141Asn 59 Motor only F19 dHMN Female Belgium HSPB8 p.Lys141Asn 65 Motor only F19 dHMN Male Belgium HSPB8 p.Lys141Asn 61 Motor only F20 CMT2B Male Austria RAB7 p.Val162Met 45 Motor and sensory

F21 CMT2E Female Belgium NEFL p.Pro8Arg 51 Motor and sensory http://jnnp.bmj.com/ F21 CMT2E Female Belgium NEFL p.Pro8Arg 68 Motor and sensory F22 CMT2B Male Austria RAB7 p.Leu129Phe 77 Motor and sensory F22 CMT2B Male Austria RAB7 p.Leu129Phe 43 Motor and sensory F23 CMT2A Female Belgium MFN2 p.Arg94Gln 58 Motor and sensory F23 CMT2A Male Belgium MFN2 p.Arg94Gln 36 Motor and sensory F24 dHMN Male Austria HSPB1 p.Pro182Leu 30 Motor only F25 dHMN Male Germany HSPB1 p.Thr139Ala 55 Motor only on 13 July 2018 by guest. Protected copyright. F26 CMT2K Female Finland GDAP1 p.His123Arg 73 Motor and sensory F26 CMT2K Female Finland GDAP1 p.His123Arg 48 Motor and sensory F26 CMT2K Female Finland GDAP1 p.His123Arg 17 Motor and sensory F27 dHMN Male Austria HSPB1 p.Gly84Arg 44 Motor only F28 NA Male Utah NA NA 39 NA F29 NA Female Utah NA NA 41 NA F30 NA Female Utah NA NA 55 NA F31 NA Male Venezuelan NA NA 57 NA F31 NA Male Venezuelan NA NA 31 NA F31 NA Female Venezuelan NA NA 51 NA F32 NA Female Utah NA NA 67 NA F32 NA Male Utah NA NA 67 NA F33 NA Female Utah NA NA 45 NA

Continued

Juneja M, et al. J Neurol Neurosurg Psychiatry 2018;89:870–878. doi:10.1136/jnnp-2017-317562 871 J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp-2017-317562 on 15 February 2018. Downloaded from Neuromuscular

Table 1 Continued Family CMT* subtype Sex Origin Gene involved Mutation Age at sampling Pure motor or/and sensory involvement F33 NA Male Utah NA NA 68 NA F34 NA Male Utah NA NA 72 NA F35 NA Female Utah NA NA 41 NA F36 NA Female Utah NA NA 63 NA F36 NA Female Utah NA NA 58 NA F37 NA Female Utah NA NA 61 NA F38 NA Female Spain NA NA 35 NA F38 NA Male Spain NA NA 32 NA F38 NA Male Spain NA NA 58 NA F39 dHMN Female Canada HSPB1 p.Gly84Arg 40 Motor only F40 dHMN Female Czech Republic HSPB8 p.Lys141Asn 25 Motor only *As per OMIM nomenclature. All these patients have genetically defined CMT2. NA, not applicable.

HSPB87; and (4) dHMN patients may exhibit a sensory pheno- 4).11 Moreover, we differentiated four healthy control iPSC lines type on disease progression as previously described.8 9 Bearing (age-matched) for the comparative studies. in mind that CMT2 represents a subgroup of rare inherited neuropathies, our cohort is, to the best of our knowledge, the Mass spectrometry largest cohort that includes 43 axonal CMT/dHMN patients We performed quantitative proteomics using iTRAQ (isobaric harbouring different CMT causal mutations. The patient pheno- mass-tag labelling for relative and absolute quantification) (Sciex) type was defined based on the following clinical and electrodi- as per manufacturer’s instruction, followed by tandem liquid agnostic inclusion criteria: (1) progressive distal muscle wasting chromatography mass spectrometry Q-Exactive Plus benchtop and weakness with reduced or absent reflexes in both upper and Orbitrap (LC-MS/MS) (ThermoScientific). The analysis of the lower limbs; (2) motor axonal neuropathy on nerve conduction raw data was achieved by Proteome Discoverer V.2.0 software studies (NCS); and (3) for CMT2, lower limb sensory loss and (ThermoScientific) using Sequest HT as search engine against sensory abnormalities on NCS (decreased lower limb sensory the human UniProt/SwissProt database. The identified proteins nerve action potential amplitudes) (online supplementary table were delivered in a list containing ratios of expression levels over 2). This study was performed according to ethical guidelines and control sample. These ratios were converted into log scale. To diagnostic procedures issued by the local research ethics commit- identify differentially regulated proteins, the mean and the 95% tees. Informed written consent was obtained from all patients confidence limits (2 SD) were calculated. Proteins outside 2 SDs involved in this study. were considered to be significantly altered.

Sample collection, cell culturing and maintenance Blood samples were collected at the respective hospitals and shipped RNA extraction and real-time quantitative PCR to us in vacutainers coated with lithium heparin. From these Total RNA was isolated using the Roboklon kit (Genematrix) according to the manufacturer’s instructions. One microgram of samples, lymphocytes were collected and Epstein–Barr virus-trans- http://jnnp.bmj.com/ formed to obtain immortalised lymphoblast cell lines. These lines RNA was reverse transcribed by using M-MLV Reverse Tran- scriptase system (Life Technologies) according to the manu- were cultivated at 37°C and 6% CO2 in Gibco RPMI 1640 medium supplemented with 15% fetal calf serum. All cell culture media and facturer’s protocol. The cDNA was amplified by qPCR using supplements were purchased from Life Technologies and checked SYBR Green dye chemistry using the ViiA 7 Real-Time PCR for mycoplasma infection by our cell-core facility. system (Applied Biosystems) using primers for profilin2 (NM_053024.3; fwd: 5′-​CGGC​AGAG​CTGG​TAGA​GTCTT- 3′; rev: 5′-​TAGC​AGCT​AGAA​CCCA​GAGTC-3′), GAMT Human iPSC generation and differentiation into motor on 13 July 2018 by guest. Protected copyright. (NM_000156.5; fwd: 5 -TGGCACACACACCAGTTCA-3 ; neurons ′ ′ rev: 5′- ​AAGGCGTAGTAGCGGCAGTC-3′), ChAT (S56138.1; Primary human fibroblast cultures were obtained from skin fwd: 5′-​GTTGTAGCAGGCACCATACC-3′; rev: 5′- ​AAAC​ biopsies of five CMT2 patients after informed consent. Human CTAC​CTGATGAG​ ​CAACC-3′) and HB9 (NM_005515.3; iPSC lines were generated and characterised by the Stem Cell 10 fwd: 5′-​CTTCTGTTTCTCCGCTTCCT-3′; rev: 5′- CACCTC- Institute, KU Leuven, Belgium as previously described. The GCTCATGCTCAC-3′). Data analysis was performed with different iPSC clones were validated for their pluripotency and the qbase+ software (Biogazelle). The RT-qPCR data were the presence of the disease-associated CMT2 mutation. The normalised by geometric averaging of multiple internal control iPSC lines were differentiated into spinal motor neurons (iNeu- genes as described previously.12 rons) using a published protocol.11 The differentiated motor neurons were collected on day 30 of the protocol and were vali- dated by confirming the expression of choline acetyltransferase Protein extraction and western blotting from human and (ChAT) and homeobox protein (HB9) by real-time quantitative mouse samples PCR (RT-qPCR; online supplementary figure 4). In addition, Hspb8K141N/K141N (knock-in, KI), Hspb8+/+ (wild type, WT) these motor neurons were further validated by immunostainings mice were housed under the care of the Animal Facility Interfac- as previously described using anti-ChAT (Millipore: AB144P) ulty Unit at the University of Antwerp accredited by the Associ- and anti-Islet1 (Abcam: ab20670; online supplementary figure ation for Assessment and Accreditation of Laboratory Animals.

872 Juneja M, et al. J Neurol Neurosurg Psychiatry 2018;89:870–878. doi:10.1136/jnnp-2017-317562 J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp-2017-317562 on 15 February 2018. Downloaded from Neuromuscular

All Hspb8 mouse experiments were carried out with approval variance and Bonferroni post hoc multiple comparison test or of the Ethical Committee for Laboratory Animals (University of Dunnett’s multiple comparison test. All reported P values were Antwerp). B6;D2-Tg (Eno2-MFN2*R94Q) L51Ugfm/J (Mito- two-tailed, with a P value <0.05 indicating statistical signifi- charch1 or CMT2A Tg) mice were purchased from Jackson cance. All calculations and statistical analyses were performed Laboratories and were housed in the animal facility of Ecole with GraphPad Prism V.5.01. Polytechnique Fédérale de Lausanne. All experiments on MFN2 mice were done in accordance with Swiss legislation and the European Community Council directive (86/609/EEC) for the Results care and use of laboratory animals and were approved by the Identification of differentially expressed proteins among Veterinarian Office of the canton of Vaud and a local ethics various CMT2 patients committee. All experiments were performed on adult female An iTRAQ-based quantitative proteomics was carried out to Hspb8 KI and WT mice and on adult male WT or CMT2A Tg compare the proteome of lymphoblasts from five selected mice as previously described.13 14 CMT patients with different phenotypes (mean age: 51 years, For total protein extraction, cells or sciatic nerves from patient IDs: F19/M/61, F20/M/45, F21/F/51, F26/F/48 and these mice were lysed with RIPA buffer for 30 min on ice. F18/F/57; each of the five patients carrying a mutation in a Protein concentration was quantified with the Bicinchoninic different gene: HSPB8_K141N, RAB7_V162M, NEFL_P8R, Acid Protein Assay Reagent (Pierce), according to the manu- GDAP1_H123R, HSPB1_R127W), with pooled samples from facturer’s instructions. Western blotting was performed as six age-matched and sex-matched unaffected controls (mean described previously15 using the following antibodies: anti-PFN2 age: 56 years). Using Proteome Discoverer V.2.0, 16 643 (Sigma-Aldrich: HPA035611), anti-GAMT (Bethyl laboratories: peptides and 4600 proteins were identified from the mass A304-182A), anti-GAPDH (GeneTex: GTX627408) and anti-β- spectrometry run. The significantly altered proteins from each actin (Sigma-Aldrich: A5441). patient compared with their respective control were anal- ysed (online supplementary table 1) and compared with other Statistical analysis patients using Vennplex program.16 We found no differentially Comparison between the two groups was performed by a regulated proteins common to all genotypes. However, two Mann–Whitney U test. Spearman correlation coefficients were proteins, profilin 2 (PFN2) and guanidinoacetate methyltrans- used to describe the association between two variables. To ferase (GAMT), were found to be downregulated in three out evaluate the ability of proteins to predict the disease clinically, of five genotypes analysed (figure 1A). These three patients receiver operating characteristic (ROC) curves were plotted. (F19/M/61, F20/M/45 and F21/F/51) showed an early onset of Multiple comparisons were performed by one-way analysis of the symptoms within the first two decades along with a severe http://jnnp.bmj.com/ on 13 July 2018 by guest. Protected copyright.

Figure 1 Proteomic analysis of lymphoblasts derived from CMT2/dHMN patients. (A) Vennplex analysis of significantly differentially regulated proteins (mean±2 SD, 95% CI) among patients with HSPB8, NEFL and RAB7 mutations as compared with unaffected control individuals. Pooled samples from age- matched and sex-matched individuals (three men; Con1 and three women; Con2) were used as controls. (B) Profile plot generated by the Perseus program indicating the log fold change in the expression of the differential proteome. Red and blue lines represent the expression profile ofP FN2 and GAMT, respectively. Grey lines represent the rest of the proteins. (C) Western blot showing PFN2 and GAMT expression in protein lysates of lymphoblasts obtained from different patients with CMT gene mutations. β-Actin is used as a loading control. (D) Quantification of western blot normalised toβ -actin using ImageJ software.

Juneja M, et al. J Neurol Neurosurg Psychiatry 2018;89:870–878. doi:10.1136/jnnp-2017-317562 873 J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp-2017-317562 on 15 February 2018. Downloaded from Neuromuscular phenotype. A clear downregulation of PFN2 and GAMT PFN2 and GAMT expression levels as potential biomarkers for proteins was also demonstrated by creating a profile plot using CMT2 the Perseus program (figure 1B). To assess the clinical significance and diagnostic power of PFN2 To validate the proteomics data, proteins were isolated and GAMT, we plotted ROC curves.18 The area under the from a different passage of the patient’s lymphoblast cell lines curve (AUC, a measure of diagnostic power) of PFN2 for the entire along with control samples. We also included a lymphoblast cohort was 0.708 and that of GAMT was 0.692 (figure 3A, E). cell line from a CMT2A patient (MFN2_R94Q, F23/F/58) As expected, the AUC for PFN2 was significantly increased when as CMT2A is the predominant CMT2 subtype character- classifying patients based on their CMT2 or dHMN phenotype ised by mutations in mitofusin 2 (MFN2).17 Western blot- (AUC=0.7441), age >50 years (AUC=0.8318) and particularly ting clearly suggested a decrease in both PFN2 and GAMT for CMT2 patients ≥50 years of age (AUC=0.9167), showing protein expression in the patient’s lymphoblasts when a significant association of PFN2 expression with age and the compared with the control lines (figure 1C, D), confirming presence of both motor and sensory deficits (figure 3B–D). No the proteomics data. significant difference in AUC for GAMT level was found when classifying patients based on their phenotype (AUC=0.705; data not shown) or age-based clustering (AUC=0.704, figure 3F), PFN2 and GAMT expression is transcriptionally regulated which was comparable to that of the entire patient cohort (0.69). Next, we studied if the downregulation detected on protein level Nevertheless, AUC was highest for GAMT (0.733, figure 3G) when patients were older than 50 years and had both motor and also corresponded to the change in transcript levels. The decrease sensory deficits. in protein expression of PFN2 and GAMT in the MFN2 patient Of note, the Spearman coefficient indicated a positive described above further motivated us to screen our all in-house correlation between GAMT and PFN2 expression, r=0.78 for lymphoblasts (Belgian cohort). Remarkably, RT-qPCR analysis healthy individuals and r=0.55 for all the samples included in of the Belgian cohort demonstrated that the mRNA expression this study (online supplementary figure 3). Taken together, our of PFN2 and GAMT was significantly downregulated in the results suggest that a decrease in PFN2 expression over time lymphoblast lines from CMT2 patients (n=10) when compared might be an important indicator of CMT2 disease progres- with samples from the healthy individuals described above in the sion, particularly for patients showing both motor and sensory proteomics study (n=6) (P=0.0047 and P=0.0002, respectively; deficits. In contrast, though GAMT appears to be a marker figure 2A, F). for axonal CMT, its relationship with disease progression still This prompted us to expand our CMT cohort to anticipate needs to be explored. if these two proteins have relevance as molecular determinants for CMT2. In our extended CMT2 patient cohort (n=43), the expression of PFN2 and GAMT transcripts was significantly Relevance of PFN2 and GAMT in the pathogenesis of axonal downregulated in the patient group compared with the control CMT2 group comprising 22 unaffected individuals (P=0.005 and We subsequently investigated the expression of PFN2 and GAMT P=0.01, respectively; figure 2B, G). in differentiated motor neurons obtained from patient-derived Furthermore, we examined the association of PFN2 and iPSC lines. We selected motor neurons, as motor neuron involve- GAMT transcripts with several parameters representing ment was common to both dHMN and CMT2 phenotypes. Consis- disease and patient subgroups. No gender-associated differ- tent with the lymphoblast data, we observed a twofold increase ences were observed for PFN2 and GAMT expression in PFN2 expression (figure 4A; P=0.003) in two older controls (online supplementary figure 1). The CMT2 patients with (derived from individuals of 50 and 51 years old, provided by both motor and sensory symptoms showed a severe and signif- KU Leuven) compared with two younger controls (both 24 years http://jnnp.bmj.com/ icant downregulation in PFN2 (P=0.003; figure 2C), whereas old, provided by KU Leuven) suggesting the progressive increase PFN2 downregulation did not reach statistical significance in in PFN2 with age. Thus, on comparing neurons from the CMT2 patients presenting only motor deficits (dHMN) at the time patients (F23/F/58; MFN2_R94Q and F21/F/51; NEFL_P8R) of sample collection. Additionally, Spearman’s rank correla- with older controls, we observed a similar twofold reduction in tion coefficient (r) suggested a positive correlation between PFN2 expression as observed in the lymphoblasts (P=0.002). No significant change was observed for younger patients (dHMN age and PFN2 expression, and an age-dependent progression

patients: F40/F/25; HSPB8_K141N and F39/F/40; HSPB1_G84R) on 13 July 2018 by guest. Protected copyright. of PFN2 in the control group (online supplementary figure compared with younger controls. However, for GAMT, no change 2A). Strikingly, patients older than 50 years showed a decline in expression was observed (figure 4B). in the PFN2 expression while a steep increase was observed Our data suggest a systemic decrease in PFN2 expression in healthy controls older than 50 years (online supplementary in patient-derived lymphoblasts and neurons pinpointing figure 2B). Thus, we stratified patients based on age using a to its possible role in the pathogenesis. However, the major mean cut-off of 50 years and found a significant and progres- limitation with iPSC technology is that the motor neurons sive increase in PFN2 expression in lymphoblasts from healthy are cultured in an isolated environment unlike the situation control individuals (P=0.02; figure 2D); however, this increase in brain and spinal cord and thus may not reflect the real was not observed in patients. Our data demonstrated a signif- pathology of a specific disorder. icant decrease in PFN2 expression in patients with age ≥50 Therefore, we also investigated PFN2 expression in the sciatic years as compared with healthy age-matched controls, which nerve of a knock-in mouse model for HSPB8 (K141N) mimicking becomes even stronger if we consider only CMT2 patients, CMT2L/dHMN13 and in a transgenic mouse expressing a mutated with a sensorimotor phenotype (P=0.0004 and P<0.0001, (R94Q) human mitofusin 2 mimicking CMT2A.14 We found respectively; figure 2D, E). an increase in PFN2 levels over time in the healthy littermate For GAMT, not much of a difference was seen on classifying animals (2-month-old presymptomatic vs 12-month-old symp- our cohort into CMT2 and dHMN populations (figure 2H–J). tomatic mice) while a decrease in PFN2 expression was detected

874 Juneja M, et al. J Neurol Neurosurg Psychiatry 2018;89:870–878. doi:10.1136/jnnp-2017-317562 J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp-2017-317562 on 15 February 2018. Downloaded from Neuromuscular http://jnnp.bmj.com/

Figure 2 mRNA expression levels of PFN2 and GAMT in the CMT2/dHMN patient cohort. (A, F) Relative quantities (ΔΔCT) of PFN2 and GAMT mRNA levels in lymphoblasts that were already available in-house (Belgium cohort). This included 10 CMT2/dHMN patients and six control samples. (B, G) Relative quantities (ΔΔCT) of PFN2 and GAMT mRNA levels in the entire cohort including 43 CMT2/dHMN patients and 22 control samples. (C, H) PFN2 and GAMT mRNA levels in 27 CMT2 patients and 16 dHMN patients. (D, I) mRNA expression levels of PFN2 and GAMT in patients classified based on age (cut-off, 50 years; 24 patients belong to ≥50 years age group and 19 patients belong to ≤50 years age group whereas 14 controls belong to ≥50 years age group on 13 July 2018 by guest. Protected copyright. and 8 controls belong to ≤50 years age group). (E, J) Finally, expression levels of PFN2 and GAMT in patients classified based on both motor and sensory symptoms as well as age, that is, 15 patients with age ≥50 and with both motor and sensory symptoms, compared with data from 14 controls with age ≥50. All samples were run in duplicates. ns, not significant. in postsymptomatic HSPB8 KI mice compared with age-matched genetic heterogeneity thereby complicating its diagnosis and healthy littermate controls (n=3 for each group, figure 4C, D). therapeutic intervention. Differential proteomic analysis In 5-month-old CMT2A transgenic mice (n=3 for each group), led to the identification of PFN2 and GAMT as commonly we observed a tendency towards decrease in PFN2 protein expres- downregulated proteins in the patients compared with sion in the sciatic nerve compared with healthy littermate controls age-matched and sex-matched controls. Interestingly, the further supporting the significance of PFN2 in the pathogenesis of proteins and also their transcripts were found to be signifi- the disease (figure 4C, D). cantly altered over controls as determined by RT-qPCR in our multicentre CMT2/dHMN cohort. Subsequent valida- Discussion tion studies on patient-derived iNeurons and sciatic nerve CMT2 and dHMN are inherited axonal sensorimotor from CMT2 mice further suggested towards involvement of or pure motor neuropathies exhibiting a vast clinical and PFN2 in the CMT2 pathogenesis (figure 4E).

Juneja M, et al. J Neurol Neurosurg Psychiatry 2018;89:870–878. doi:10.1136/jnnp-2017-317562 875 J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp-2017-317562 on 15 February 2018. Downloaded from Neuromuscular

Figure 3 Receiver operating characteristic (ROC) curves for assessing the diagnostic power of PFN2 and GAMT. (A) ROC curve for CMT diagnosis using PFN2 expression in the overall patient versus healthy control group. (B) CMT2 patients with both motor and sensory deficits versus healthy control group. C( ) Patients ≥50 years of age versus healthy controls ≥50 years of age. (D) CMT2 patients with both motor and sensory deficits ≥50 years of age versus healthy controls ≥50 years of age. (E) ROC curve for CMT diagnosis using GAMT expression in the overall patient versus healthy control group. (F) Patients ≥50 years age versus healthy controls ≥50 years age. (G) CMT2 patients ≥50 years age versus healthy controls ≥50 years age. Dotted lines indicate identity (%) and straight lines indicate sensitivity (%). AUC, area under the curve.

It was intriguing that three out of five lymphoblast samples possible that besides PFN2, other actin-associated proteins are showed decrease in PFN2 and GAMT expression in the initial involved in the disease mechanism leading to an aberrant cyto- proteomic analysis (F26/F/48 for the GDAP1 mutant and skeletal dynamics, which warrants further investigation. F18/F/57 for the HSPB1 mutant did not show this reduction), GAMT belongs to the creatine biosynthetic pathway that though we were able to extend this decrease to 43 different converts guanidinoacetate into creatine. GAMT is crucial for the genetically defined CMT patients. The F18 patient also had an cellular phosphate-associated energy production and storage, additional carpal tunnel syndrome, which has been shown to particularly in tissues with higher metabolic demands like the upregulate several proteins including actin-binding proteins,19 brain. GAMT deficiency was found to be implicated in a rare 25 thus possibly explaining why we did not observe the change. cause of epilepsy. In this study, the difference in expression http://jnnp.bmj.com/ However, for GDAP1 mutations, the results were individual-spe- of GAMT in controls versus patients was rather subtle (though cific (F26) as other GDAP1 patients recruited in the extended significant) as compared with PFN2. This could be due to the cohort clearly exhibited the decrease in PFN2 and GAMT. cellular model (lymphoblasts) we have used in this study. We Profilins are actin dynamics regulators, ubiquitously found in have noticed that unlike for PFN2, lymphocytes show higher mammals and are essential for the actin polymerisation in cells. expression of GAMT over lymphoblasts (data not shown). Two isoforms, PFN1 and PFN2, are present in the brain and However, lymphoblasts were used in this study as they offer

play a pivotal role in neurogenesis and synapse formation by a continuous source of patient cells with comparable molec- on 13 July 2018 by guest. Protected copyright. interacting with various proteins.20 21 It must be pointed out that ular and functional characteristics to their parent lymphocytes we detected PFN2 expression as a whole and did not distinguish thereby eliminating the need of resampling.26 Concomitantly, its two isoforms. PFN2 has been recently described as a poten- no differences in GAMT expression was observed in iPSC-de- tial diagnostic biomarker and therapeutic target for oesophageal rived neurons from CMT2 patients and in mouse sciatic nerves squamous cell carcinoma.22 Intriguingly, an increase in PFN2 compared with healthy controls. Of note, creatine kinase (CK) expression has been described in an intermediate spinal muscular levels in serum are routinely measured in clinics for neuromus- atrophy (SMA) mouse model of axonal degeneration suggesting cular disorders that include muscle weakness and myopathies.27 affected actin dynamics.23 However, both in SMA and CMT, Unfortunately, we do not have data on CK levels in our patient dysregulated actin-associated proteins eventually alter actin cohort. It remains likely that the elevated levels of CK in these dynamics and thus probably lead to the collapse of neuronal patients give rise to reduced GAMT levels and is indicative of networks. The decrease in PFN2 in neurons derived from iPSCs muscle weakness in CMT patients. However, it remains inter- as well as in mouse models suggests its key role in the pathogen- esting to investigate the expression of GAMT in patient’s serum esis of CMT besides its role as a prognostic or predictive marker. or in the muscle and liver samples where GAMT is primarily This is in line with other studies demonstrating the defect in expressed. actin-dependent processes as causative of several neurodevelop- To investigate the feasibility of using PFN2 and GAMT as mental and neurodegenerative diseases.24 However, it remains a clinical biomarker of CMT2, ROC curves were plotted to

876 Juneja M, et al. J Neurol Neurosurg Psychiatry 2018;89:870–878. doi:10.1136/jnnp-2017-317562 J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp-2017-317562 on 15 February 2018. Downloaded from Neuromuscular

Figure 4 PFN2 and GAMT expression in motor neurons. (A) Relative quantities (ΔΔCT) of PFN2 and (B) GAMT transcript expression in the motor neurons http://jnnp.bmj.com/ differentiated from control and patient-derived fibroblasts. ‘Control young’ indicates motor neurons derived from healthy individuals (24 years old) whereas ‘control old’ indicates neurons from healthy individuals which are above 50 years old. ‘Patient young’ and ‘patient old’ represent patients either younger than or older than 50 years, respectively. (C) Western blot showing PFN2 expression in sciatic nerve lysates of 2-month-old and 12-month-old HSPB8 KI mice and 5-month-old MFN2-R94Q transgenic (Tg) mice along with their respective wild-type (WT) littermates. α-Tubulin is used as a loading control. (D) Quantification of theP FN2 bands normalised to α-tubulin using ImageJ software. Values are expressed as mean±SEM, n=3. (E) Work flow diagram summarising the project. on 13 July 2018 by guest. Protected copyright. determine the specificity, sensitivity and AUC score. We acquired symptoms and 13 patients have mild symptoms) as measured by similar or even higher sensitivity, specificity and AUC values CMT examination score (CMTES) scoring (CMTES ≤8, mild; compared with the traditional Alzheimer disease biomarkers CMTES ≥9, moderate; CMTES>16, severe). The CMTES Aβ42 and tau (AUC=0.755 and 0.78, respectively)28 on clas- scoring was not available for the Belgian cohort. Thus, it was sifying patients based on their age (>50 years) and phenotype difficult to correlate the expression of these two genes with (both motor and sensory symptoms involved) underscoring the disease severity. However, this F6 patient showed 63% and 40% importance of PFN2 as a novel biomarker for CMT2. It would reduction in PFN2 and GAMT expression, respectively, when be interesting to screen other non-genetic progressive axonal compared with the mean value of age-matched controls. Of neuropathies to determine whether our markers are specific for note, two Belgian patients in family F23 (age 58 and 36, both CMT2 or if these are markers for all axonal neuropathies. In wheelchair bound and with a severe disease phenotype) showed addition, we found that PFN2 expression gradually increases a strong 60% and 88% reduction in PFN2, and 40% reduction with age in healthy participants, in contrast to CMT2 patients, in in GAMT when compared with the mean value of all controls which PFN2 expression progressively decreases with age. Unfor- irrespective of age, linking the decrease with severity. Addition- tunately, among our Italian and UK cohorts, only one patient ally, since CMT is an age-dependent progressive disease, ageing (F6) displays a severe phenotype (11 patients display moderate and involvement of both motor and sensory symptoms positively

Juneja M, et al. J Neurol Neurosurg Psychiatry 2018;89:870–878. doi:10.1136/jnnp-2017-317562 877 J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp-2017-317562 on 15 February 2018. Downloaded from Neuromuscular contribute towards disease severity. Thus, PFN2 and GAMT © Article author(s) (or their employer(s) unless otherwise stated in the text of the expression regulation can be disease-severity dependent and may article) 2018. All rights reserved. No commercial use is permitted unless otherwise help to measure disease outcomes for a clinical trial independent expressly granted. of the age of the patients. Indeed, future studies on severe cases would reinforce the conclusions from these findings. References Taken together, we initiated this study based on our hypothesis 1 saporta MA, Shy ME. Inherited peripheral neuropathies. Neurol Clin that pathophysiological pathways encompassing multiple forms 2013;31:597–619. 2 Rossor AM, Polke JM, Houlden H, et al. Clinical implications of genetic advances in of CMT2/dHMN are likely to implicate similar underlying mech- Charcot-Marie-Tooth disease. Nat Rev Neurol 2013;9:562–71. anisms. This led to the identification of candidates PFN2 and 3 pareyson D, Marchesi C, Diagnosis MC. Diagnosis, natural history, and management GAMT implicated as ‘signature’ molecules in multiple CMT2 of Charcot-Marie-Tooth disease. Lancet Neurol 2009;8:654–67. subtypes. To our knowledge, this is the first study involving 4 soldevilla B, Cuevas-Martín C, Ibáñez C, et al. Plasma metabolome and skin proteins in Charcot-Marie-Tooth 1A patients. PLoS One 2017;12:e0178376. multiple CMT2/dHMN causative genes at once and reporting 5 Fledrich R, Mannil M, Leha A, et al. Biomarkers predict outcome in Charcot-Marie- biomarkers for CMT2. Our results highlight that measuring Tooth disease 1A. J Neurol Neurosurg Psychiatry 2017;88:941–52. PFN2 and GAMT may help clinicians to monitor disease activity 6 Bouhy D, Timmerman V. Animal models and therapeutic prospects for Charcot-Marie- and progression, and may guide potential therapeutic strategies. Tooth disease. Ann Neurol 2013;74:391–6. 7 Reilly MM. Classification and diagnosis of the inherited neuropathies. Ann Indian Acad Neurol 2009;12:80–8. Author affiliations 8 capponi S, Geroldi A, Fossa P, et al. HSPB1 and HSPB8 in inherited neuropathies: 1P eripheral Neuropathy Research Group, University of Antwerp, Antwerp, Belgium study of an Italian cohort of dHMN and CMT2 patients. J Peripher Nerv Syst 2Institute Born Bunge, Antwerp, Belgium 2011;16:287–94. 3VIB Center for Molecular Neurology, University of Antwerp, Antwerp, Belgium 9 Frasquet M, Chumillas MJ, Vílchez JJ, et al. Phenotype and natural history of inherited 4Neuromuscular Reference Centre, Department of Neurology, Antwerp University neuropathies caused by HSJ1 c.352+1G>A mutation. J Neurol Neurosurg Psychiatry Hospital, Antwerpen, Belgium 2016;87:1265–8. 5Institute of Genetic Medicine, Newcastle University, Newcastle upon Tyne, UK 10 Takahashi K, Tanabe K, Ohnuki M, et al. Induction of pluripotent stem cells from adult 6Leibniz-Institut für Analytische Wissenschaften -ISAS- e.V., Dortmund, Germany human fibroblasts by defined factors. Cell 2007;131:861–72. 7Unit of Rare Neurodegenerative and Neurometabolic Diseases, Department of 11 Guo W, Naujock M, Fumagalli L, et al. HDAC6 inhibition reverses axonal Clinical Neurosciences, C. Besta Neurological Institute IRCCS Foundation, Milan, Italy transport defects in motor neurons derived from FUS-ALS patients. Nat Commun 8Aix Marseille University, INSERM, MMG, U1251, Marseille, France 2017;8:861. 9Brain Mind Institute, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, 12 Vandesompele J, De Preter K, Pattyn F, et al. Accurate normalization of real-time Switzerland quantitative RT-PCR data by geometric averaging of multiple internal control genes. 10Stem Cell Institute, KU Leuven, Leuven, Belgium Genome Biol 2002;3:research0034.1. 11Department of Neuroscience, Karolinska Institutet, Stockholm, Sweden 13 Bouhy D, Juneja M, Katona I, et al. A knock-in/knock-out mouse model of HSPB8- 12Department of Clinical Neuroscience, Karolinska Institutet, Stockholm, Sweden associated distal hereditary motor neuropathy and myopathy reveals toxic gain-of- 13DNA Laboratory, Department of Child Neurology, 2nd Medical School, Charles function of mutant Hspb8. Acta Neuropathol 2018;135:131–48. University and University Hospital Motol, Prague, Czech Republic 14 cartoni R, Arnaud E, Médard JJ, et al. Expression of mitofusin 2(R94Q) in a transgenic 14Department of Clinical Neurological Sciences Centre, University Hospital, Western mouse leads to Charcot-Marie-Tooth neuropathy type 2A. Brain 2010;133:1460–9. University, London, Ontario, Canada 15 Ydens E, Demon D, Lornet G, et al. Nlrp6 promotes recovery after peripheral nerve injury independently of inflammasomes. J Neuroinflammation 2015;12:143. Correction notice This article has been corrected since it was published Online 16 cai H, Chen H, Yi T, et al. VennPlex--a novel Venn diagram program for comparing First. A typo in Supplementary Figure 4 has been corrected. and visualizing datasets with differentially regulated datapoints. PLoS One Acknowledgements We are grateful to Elias Adriaenssens and Delphine Bouhy 2013;8:e53388. for the assistance with Perseus program and mice experiments, respectively. We 17 Timmerman V, Strickland AV, Züchner S. Genetics of Charcot-Marie-Tooth appreciate the assistance of Dr Boglarka Bansagi for seeing some of the patients (CMT) disease within the frame of the Human Genome Project success. Genes and Ingrid Emmerson for taking and processing the blood samples. 2014;5:13–32. 18 hajian-Tilaki K. Receiver operating characteristic (ROC) curve analysis for medical Contributors MJ designed and performed the experimental work and wrote diagnostic test evaluation. Caspian J Intern Med 2013;4:627–35. the manuscript. AA and SM assisted in the proteomics analysis and wrote the 19 Gingery A, Yang TH, Passe SM, et al. TGF-β signaling regulates fibrotic expression and http://jnnp.bmj.com/ corresponding sections. NBM, BLS and RC provided tissues of the Mfn2 transgenic activity in carpal tunnel syndrome. J Orthop Res 2014;32:1444–50. mouse model and wrote the corresponding sections. CV generated the iPSC lines. JB, 20 Wang YY, Wu HI, Hsu WL, et al. In vitro growth conditions and development affect AR, MJJ, PS, CP, PS, AFH, PdJ, RH and DP provided patient samples and summarised differential distributions of RNA in axonal growth cones and shafts of cultured rat the clinical tables. VT coordinated the design of this study, wrote and edited the hippocampal neurons. Mol Cell Neurosci 2014;61:141–51. manuscript. 21 Murk K, Wittenmayer N, Michaelsen-Preusse K, et al. Neuronal profilin isoforms are Funding This work was supported in part by the Fund for Scientific Research (FWO- addressed by different signalling pathways. PLoS One 2012;7:e34167. Flanders), the ’Association Belge contre les Maladies Neuromusculaires’ (ABMM), 22 cui XB, Zhang SM, Xu YX, Yx X, et al. PFN2, a novel marker of unfavorable prognosis, the Medical Foundation Queen Elisabeth (GSKE) and the EU FP7/2007_2013 under is a potential therapeutic target involved in esophageal squamous cell carcinoma. J on 13 July 2018 by guest. Protected copyright. grant agreement number 2012-305121 (NEUROMICS). JB is supported by a Senior Transl Med 2016;14:137. Clinical Researcher mandate of the Research Fund—Flanders (FWO). AR received 23 Bowerman M, Anderson CL, Beauvais A, et al. SMN, profilin IIa and plastin 3: a support by the Ministerium für Innovation, Wissenschaft und Forschung des Landes link between the deregulation of actin dynamics and SMA pathogenesis. Mol Cell Nordrhein-Westfalen. RH is a Wellcome Trust Investigator (109915/Z/15/Z), who Neurosci 2009;42:66–74. receives support from the Medical Research Council (UK) (MR/N025431/1) and 24 Kevenaar JT, Hoogenraad CC. The axonal cytoskeleton: from organization to function. the European Research Council (309548). NBM, BLS and RC received support from Front Mol Neurosci 2015;8:44. the Neuromuscular Research Association Basel (NeRAB). NBM and BLS are also 25 stern WM, Winston JS, Murphy E, et al. Guanidinoacetate methyltransferase (GAMT) supported by ERANET E-Rare FaSMALS (grant 31ER30_160673). PS is supported by deficiency: a rare but treatable epilepsy. Pract Neurol 2017;17:207–11. the Ministry of Health of the Czech Republic (AZV 16-30206 and DRO 00064203). 26 hussain T, Mulherkar R. Lymphoblastoid cell lines: a continuous in vitro source of cells to study carcinogen sensitivity and DNA repair. Int J Mol Cell Med 2012;1:75–87. Competing interests None declared. 27 Moghadam-Kia S, Oddis CV, Aggarwal R. Approach to asymptomatic creatine kinase Patient consent Obtained. elevation. Cleve Clin J Med 2016;83:37–42. 28 craig-Schapiro R, Kuhn M, Xiong C, et al. Multiplexed immunoassay panel identifies Ethics approval University of Antwerp. novel CSF biomarkers for Alzheimer’s disease diagnosis and prognosis. PLoS One Provenance and peer review Not commissioned; externally peer reviewed. 2011;6:e18850.

878 Juneja M, et al. J Neurol Neurosurg Psychiatry 2018;89:870–878. doi:10.1136/jnnp-2017-317562