Towards functional restoration for persons with limb amputation: A dual-stage implementation of regenerative agonist-antagonist myoneural interfaces

While amputation has traditionally been viewed as a failure of therapy, recent developments in amputation surgery and neural interfacing demonstrate improved functionality and bidirectional communication with prosthetic devices. The agonist antagonist myoneural interface (AMI) is one such bi-directional neural communication model comprised of two muscles, an agonist and an antagonist, surgically connected in series within the amputated residuum such that contraction of one muscle stretches the other. By preserving agonist-antagonist muscle dynamics, the AMI allows proprioceptive signals from mechanoreceptors within both muscles to be communicated to the central nervous system. Preliminary human evidence suggests that AMIs have the capacity to provide high fidelity control of a prosthetic device, force feedback, and natural proprioception. However, AMIs have been implemented only in planned amputations and require healthy distal tissues, whereas the majority of amputations occur in patients who do not have healthy distal tissues. Through the use of a dual-stage surgical procedure which leverages existent tissues, this study proposes a revision model for implementation of the AMI in patients who are undergoing traumatic amputation or have already undergone a standard amputation. This paper validates the resulting AMI’s physiology, revealing robust viability and mechanical and electrophysiological function. We demonstrate the presence of H-waves in regenerative grafts, indicating the incorporation of the AMI into physiological reflexive loops.


Supplemental Methods:
Numerous electrophysiological, histological, and mechanical assessments were performed to evaluate the function and recovery of AMIs created through a single and dual-stage process. Here, we outline the procedures used to acquire these data.

Insertional EMG to assess reinnervation
To identify the extent of reinnervation in the muscle grafts, insertional EMG was performed. The regenerative muscle grafts were identified by palpation or through a small incision in the hindlimb, thereby exposing the grafts. A monopolar (30-gauge, Natus Medical) needle was placed into the muscle graft and stabilized using an external clamp. This prevented any shifting of the electrode, which could have cause penetration into the biceps femoris. If palpation was used to identify the graft, after insertion of the needle, a small stimulus (100 us, 0.5 mA) was applied on the stimulating needles to ensure placement in the graft and not in the underlying biceps femoris.
Electrophysiology to measure efferent and afferent signals Electrophysiological testing was performed six weeks after the first operation. We assessed the ability of the AMIs to provide stable, isolated, efferent control signals for prosthetic device modulation. We also evaluated their ability to generate natural proprioceptive afferent signals. Skin was incised and pulled back to expose the AMIs or unlinked muscle grafts. 30-gauge bipolar needles were inserted into each muscle and connected to a differential biopotential amplifier (20kS/s sampling frequency, 16-channel amplifier stage with 200x gain, Intan Technologies).
In the first set of tests, the innervating nerve of each muscle was independently stimulated using a hook electrode with pulses generated from an NL800 Current Stimulator (Digitimer) ranging between 0.5 and 12 mA. Pulse width was held constant at 100 us. Efferent signals from both muscles were recorded to map the stimulus-intensity and efferent EMG response. Given the tendon-tendon coaptation, for a given nerve, we expected efferent signals resulting from the agonist, but electrical silence from the antagonist muscle.
In a second set of tests, we performed afferent ENG recordings. The agonist nerve was stimulated using a hook electrode and ENG from the antagonist nerve was recorded. Details of afferent signal recording and processing can be found in Srinivasan et al. (Science Robotics, 2017). Stimulation on the agonist nerve caused contraction of the agonist muscle stretching the antagonist. Consequently, afferent feedback signals were generated in the antagonist and recorded on the antagonist nerve. We analyzed data to reveal gradation in afferent activity proportional to stimulation intensity. All EMG and ENG signals were processed in MATLAB.

Stain generation and excursion
During the aforementioned tests, video recording of the excursing muscle grafts was performed using a Nikon D3200 DSLR camera. This captured the 2D motion of the constructs from an angle that was orthogonal to the plane of the underlying biceps femoris. Frames capturing the resting and maximal contraction points of the antagonist grafts were used to quantify the percent strains and excursion values induced on the agonist at each stimulation amplitude. These values were used to assess whether the AMI's linkage of agonist-antagonist muscles created proportional responses in the antagonist muscle to agonist contraction. Further, strain values were compared to those of physiological, biologically intact muscles to ensure that resulting afferent signals would mimic those naturally received by the central nervous system.

Atrophy
During implantation and the terminal procedure, the muscle grafts were photographed using a Nikon D3200 DSLR camera. The border demarcating the ends of the graft was outlined using surgical marker and used for quantification during data analysis. Furthermore, grafts were weighed upon explant. These data were synthesized to determine the effect of a dual-stage surgical process on the atrophy of the muscle grafts comprising the AMI.  (1) a patient would present for secondary complications of amputation requiring revision surgery. Consistent with the current surgical process, (2) neuromas would be severed and buried in between muscle without a target end organ. These nerves can easily form into painful neuromas. Furthermore, myoelectric control from these nerve endings is challenging because the signals are difficult to source and are often contaminated by surrounding muscles. During revision, the dual-stage process we propose would be employed to create regenerative AMIs using transected peripheral nerves. (3) In the revision surgery, an intraneural dissection would yield numerous terminal motor units. (4) However, the function of each of these nerves would be potentially unknown. The creation of AMIs requires agonist-antagonist pairs. The following dual-stage surgical process enables their identification. (5) During the first revision operation, each terminal motor unit would be placed into a regenerative muscle graft. After a brief period of reinnervation, (6) the function of each graft can be ascertained through patient reporting during ultrasound-guided muscle stimulation or volitional contraction. The grafts can be labeled with radiopaque markers or identified using anatomical landmarks. (7) During a second operation, grafts that are found to be extensor-flexor pairs will be linked to form an AMI. Grafts Patient presents with secondary complications requiring revision.
Pathology is addressed and neuromas are severed.
At this stage, the identity of each terminal nerve branch is unclear. Extensor-flexor pairs are necessary for AMI construction. Thus, the patient undergoes the following dual-stage process.
Intraneural dissection splits nerve into discrete motor units Operation 1: Each terminal nerve branch is placed into a muscle graft.

Identification: The function
of each graft is identified using ultrasound-guided iNEMG and tagged.
Operation 2: AMIs are constructed from identified extensor-flexor pairs. Grafts without a pair remain unlinked. As compared to the current amputation paradigm in which the function of all these transected nerves are lost, the dual stage process provides the opportunity for the functionality of at least some transected nerves to be preserved and utilized for myoelectric control and naturally-generated proprioception. A cross section of muscle near the tendontendon junction is trichrome stained. Myocytes (red) are isotropically aligned and merge healthily with collagenous tendon at the site of coaptation. This suggests that the healing process formed a strong bridge that would enable the translation of tension across the tendon-tendon bridge between the two muscle grafts comprising the AMI.

250um 250um
Supplemental Figure 6. Modular AMI Architecture. Given the complexity of muscle and nerve functions for certain joints, a variety of architectures can be employed to utilize muscle grafts innervated with transected nerves. Pictured is a triad architecture, which could be used for the positioning of numerous regenerative flexor-extensors grafts articulating the same joint in an AMI configuration.