Muscle MRI will be performed on a 3-T MR machine (Trio; Siemens, Erlangen, Germany). Each subject will lie in a supine position comfortably, supplied with ear plugs. A high resolution T1 weighted scan and Short-T1 Inversion Recovery series (STIR) of the four limbs muscle were obtained in axial and coronal plane. MR spectroscopy in interested muscle was also sampled to evaluate the composition of fat, water, lactate, and other studied molecules.

Nerve conduction study will be performed with a Nicolet Viking IV Electromyographer (Madison, WI) in all patients following standardized methods recommended by the Consensus Development Conference on Standardized Measures in Diabetic Neuropathy. Studied nerves include sural, peroneal, tibial, median and ulnar (motor and sensory) nerves.Autonomic functions will be assessed by the SSR and RRIV with established protocol by using Nicolet Viking IV Electromyographer (Madison, WI).

Quantative sensory test will be performed with a Thermal Sensory Analyzer and Vibratory Sensory Analyzer (Medoc Advanced Medical System, Minneapolis, MN). The procedure is the same as previously described 25. Briefly, the machine delivers to the patient a stimulus of constant intensity which is pre-set by the algorithm. By adjusting the intensity of stimulus (increase or decrease the intensity by a fixed ratio) according to the response of the subject (i.e. whether the subject perceives the stimulus or not), sensory thresholds of the warm, cold and vibratory modalities will be measured.

A skin specimen of 3 mm in diameter will be taken with a biopsy punch from the lateral side of the distal leg under 2% lidocaine local anaesthesia 26. No suturing is required, and the wounds are covered with a piece of gauze. Wound healing takes 7~10 days, similar to a typical abrasion wound. Informed consent will be obtained from each patient before the skin biopsy. The intraepidermal nerve fiber density and sweat gland nerve innervation will be examined.

Two muscle specimens with 5 x 5 x 5 mm were collected in an open muscle biopsy or needle biopsy at studied muscles under 2% lidocaine local anaesthesia. The wound was about 2-3 cm long, and suture was required. Would healing usually takes 10-14 days. First specimen was undergoing snap freezing fixation in a longitudinal axis perpendicular to the cork with the liquid nitrogen and isopentane. The second specimen was divided into two equiponderous tissues, and one was freezed in liquid nitrogen for DNA and protein analysis. The another one was treated with RNAlater solution in 4°C overnight for RNA analysis. Sural nerve biopsies or superficial peroneal nerve were obtained from a standard site posterior to the lateral malleolus under local anesthesia. The nerves were then fixed in 5% glutaraldehyde in 0.1 M phosphate buffer (PB) at 4 °C overnight. All samples were stored in the -80°C refrigerator for further analysis.

The DNA, RNA, and protein of the tissues (blood, muscle, nerve and skin) were retrieved and stored in the -80°C refrigerator. The next generation sequencing (whole exon sequencing or whole genome sequencing) and RNA-seq would be performed by NGS & Microarray Core lab in National Taiwan University or another professional team. The protein analysis would be performed by the Proteomics & Protein Function Core Lab in National Taiwan University.

The methods of the measurement will follow the standards set by Department of Laboratory Medicine of National Taiwan University Hospital.

A contact heat evoked potential stimulator (Medoc, Ramat Yishai, Israel) will be used for delivering heat stimulation. Stimuli will be delivered repeatedly to the same stimulation site and the inter-stimulus interval will be randomly set to around 18~22 s. CHEP will be recorded using a Nicolet Bravo evoked potential system (Nicolet Biomedical, Madison, WI). The recording electrode was placed at the Cz and P3 of international 10-20 system. The impendence of all recording electrodes was kept below 3 kΩ. The evoked potentials were filtered with a bandpass filter at 0.1~30 Hz. Recording was triggered by the onset of each stimulus, and the sweep time was 1500 ms.

Nerve excitability studies will be undertaken on the median, tibial, peroneal and sural nerves as per previously detailed protocols. Skin temperature will be monitored at the site of stimulation and was maintained at >32°C. Stimulation and recording will be controlled by automated computerized system (QTRAC; Institute of Neurology, London, U.K.) and the stimulus current will be administered using an isolated linear bipolar constant-current stimulator (DS5; Digitimer, Welwyn Garden City, U.K.). Responses will be amplified (ICP511 AC amplifier, Grass Technologies, West Warwick, USA) with electronic noise removed (Hum Bug 50/60 Hz Noise Eliminator, Quest Scientific Instruments, North Vancouver, Canada).

In simple rTMS protocols, individual stimuli are spaced apart by identical interstimulus intervals (ISI). The stimulation protocol was in accordance with published safety recommendations. The patterned rTMS protocols included (1) Theta burst stimulation, (2) Repetitive paired-pulse TMS. All the above stimulation protocols or paradigms will follow the international guideline of the use of TMS in clinical practice and research (Rossi et al., 2009; Fitzgerald and Daskalakis, 2012; Groppa et al., 2012; Steeves et al., 2012).

fMRI will be performed on a 3-T MR machine (Trio; Siemens, Erlangen, Germany). Each subject's head will be positioned comfortably inside a receive-only 8-channel birdcage head coil, supplied with ear plugs, heavily padded and secured with a strap across the forehead in order to minimize head motion. All data will be processed using SPM2 (Wellcome Department of Cognitive Neurology, London UK) 36 implemented on MATLAB (Mathworks Inc. Sherborn, MA).