Classical Complement Pathway Inhibition in a “Human-On-A-Chip” Model of Autoimmune Demyelinating Neuropathies

Abstract

Chronic autoimmune demyelinating neuropathies are a group of rare neuromuscular disorders with complex, poorly characterized etiology. Here we describe a phenotypic, human-on-a-chip (HoaC) electrical conduction model of two rare autoimmune demyelinating neuropathies, chronic inflammatory demyelinating polyneuropathy (CIDP) and multifocal motor neuropathy (MMN), and explore the efficacy of TNT005, a monoclonal antibody inhibitor of the classical complement pathway. Patient sera was shown to contain anti-GM1 IgM and IgG antibodies capable of binding to human primary Schwann cells and induced pluripotent stem cell derived motoneurons. Patient autoantibody binding was sufficient to activate the classical complement pathway resulting in detection of C3b and C5b-9 deposits. A HoaC model, using a microelectrode array with directed axonal outgrowth over the electrodes treated with patient sera, exhibited reductions in motoneuron action potential frequency and conduction velocity. TNT005 rescued the serum-induced complement deposition and functional deficits while treatment with an isotype control antibody had no rescue effect. These data indicate that complement activation by CIDP and MMN patient serum is sufficient to mimic neurophysiological features of each disease and that complement inhibition with TNT005 was sufficient to rescue these pathological effects and provide efficacy data included in an investigational new drug application, demonstrating the model’s translational potential.

Graphical Abstract

This manuscript presents a novel in vitro human-on-a-chip system to 1) investigate multi-focal motor neuropathy and chronic inflammatory demyelinating polyneuropathy patients’ sera mediated changes in peripheral motoneuron conduction velocity due to complement deposition. 2) A novel C1s inhibitor’s ability to prevent complement deposition and reverse the conduction block. 3) An IND to the FDA has been filed utilizing these results.

Introduction

Rare diseases, defined in the US as diseases with a prevalence of fewer than 200,000 individuals, remain an underrepresented area of biomedical and pharmaceutical research. While the prevalence of any single rare disease is low, there are over 7,000 identified rare diseases worldwide. It is estimated that 25 million individuals have a rare disease diagnosis in the United States.^[1] There are more than 600 rare neurodegenerative diseases usually characterized by a chronic, progressive clinical course and neuronal loss with regional specificity.^[2] Although medical and social concerns raised by rare diseases have been increasingly recognized by the public and pharmaceutical industry over the past two decades, most rare diseases still lack a cure or effective treatment strategy.^[3] This is due, in part, to a lack of robust animal models for most rare diseases and the poor predictive capability of the existing animal models.

Multifocal motor neuropathy (MMN) and chronic inflammatory demyelinating polyneuropathy (CIDP) are rare, autoimmune neuropathies that present clinically with muscle weakness, reduced or absent spinal reflexes, and electrophysiological conduction block.^[4] While these diseases share certain diagnostic features, there are distinct clinical, electrophysiological and biomarker characteristics that distinguish MMN from CIDP. Specifically, MMN patients show asymmetric muscle weakness that preferentially affects the distal regions of upper limbs.^[5] Additionally, MMN patients do not develop sensory deficits; sensory nerve action potentials (SNAPs) are normal. Serologically, IgM autoantibodies against the ganglioside GM1 (monosialotetrahexosylganglioside) are often present with anti-asialo-GM1, GD1a or GM2 less frequently observed.^[6] In contrast, CIDP patients exhibit symmetric muscle weakness, and both proximal and distal limbs are affected. Additionally, CIDP patients present with proximal and distal sensory deficits including reduced or absent sensory nerve conduction numbness, tingling and gait ataxia.^[7] Serologically, autoantibodies are common, but frequently are not anti-GM1 antibodies.^[4] Despite these differences, the similarities in immune system hyperactivity including inflammation and autoantibody production, peripheral nerve demyelination, nerve conduction block and responsiveness to intravenous immunoglobulin (IVIg) treatment warrant investigations into a single therapeutic approach to treat both diseases.^[4]

Like most rare diseases, few models exist to study peripheral demyelinating neuropathies. Current methods to induce experimental autoimmune neuritis (EAN) focus on immunizing animals, including mice, rats and rabbits, with peripheral nerve homogenate or purified GM1 ganglioside or myelin proteins or peptides and then assaying for signs of neuropathy over several weeks.^[8–10] Interestingly, while anti-GM1 IgM antibodies can be detected in rats and rabbits, no rats developed peripheral neuropathy and only a subset of rabbits developed disease characteristics.^[8–9] These data suggest a more complex mechanism is involved in the pathogenesis of MMN and CIDP than can be captured using current animal models. Other studies have observed reduced motor amplitude and temporal dispersion on nerve conduction studies, indicators of demyelination, after acute intraneural injection of diseased-patient sera into rats.^[11] While these models have been valuable in the identification of important cytokines and costimulatory signals as well as the mechanism of antigen recognition and nerve injury, they have been criticized for failing to translate to successful identification of therapeutic targets.^[10] Mouse models have been developed for the evaluation of complement inhibitors but have not shown clinical correlations.^[12] Induced pluripotent stem cell-derived motoneurons have also been used to investigate the role of MMN patient serum pathogenicity.^[13] In the study, Harschnitz et al. showed that MMN patient serum contained anti-GM1 IgM antibodies that were sufficient to activate the complement system and cause structural damage and neurite loss in motoneurons.^[13]

Animal models have been indispensable in biomedical and pharmaceutical research. However, ethical considerations and their limitations in mimicking human physiology and disease mechanisms have led scientists to develop more advanced human-based models for research and drug discovery.^[14] Moreover, due to the poor reproducibility of human disease conditions in animals, these models have shown limited success in generating highly efficacious and safe drugs to treat human diseases.^[15] Human-on-a-chip (HoaC) systems represent the convergence of cell biology, engineering, surface chemistry and electronics that facilitate investigations into organ system physiology and organ-organ interactions in a microfluidic platform.^[15] Further, the ability of HoaC systems to monitor changes in physiological processes such as force generation, barrier formation and maintenance, and electrical activity in response to drug treatment makes them ideal models to study efficacy and safety of novel therapeutics without cell death.^[15] In our lab, these systems have demonstrated the ability to predict known toxicity profiles for established drugs such as terfenadine and cyclophosphamide^[16] and tamoxifen^[17] indicating their potential for preclinical evaluation of novel compounds to treat disease.^[18] Additionally, we have shown multi-organ HoaC systems can maintain organ physiology for up to 28 days in serum-free medium^[19] as well as reproduce the basic responses of the innate immune system.^[20] In the context of rare diseases, bioengineered HoaC systems hold promise for modeling complex pathophysiology, uncovering disease mechanisms, and determining the therapeutic efficacy and safety of novel compounds using clinically relevant assays, especially for biopharmaceuticals.^[21]

In this study, we describe the development of a novel HoaC model of peripheral motoneuron conduction velocity. The system, composed of human induced pluripotent stem cell (iPSC)-derived motoneurons and human Schwann cells cultured on microelectrode arrays (MEAs) fitted with tunnels to direct axonal outgrowth over the electrodes, was treated with anti-GM1 polyclonal IgG and purified human complement proteins to establish and characterize the first HoaC disease model of EAN. The HoaC system was then treated with serum from MMN and CIDP patients establishing two rare peripheral demyelinating neuropathy models. Finally, the disease effects were rescued by treating the MMN and CIDP systems with TNT005, a murine monoclonal antibody (mAb) that inhibits C1s, a key protease of classical complement pathway. The study presented here represents the first rare disease-on-a-chip model of motoneuron conduction velocity to study acquired inflammatory demyelinating neuropathies and provides evidence that C1s inhibition is sufficient to rescue neuronal defects characteristic of those seen in patients. Data generated with this model provide support for testing C1s inhibition in clinical trials of autoimmune demyelinating neuropathies, such as CIDP (NCT04658472), and suggests great potential for microphysiological systems use for translational research leading to IND generation.

Results

Characterization of complement degrading protein expression on human Schwann cells and iPSC-derived motoneurons

To establish a “human-on-a-chip” disease system to study the autoimmune peripheral demyelinating neuropathies MMN and CIDP, the expression of cell surface complement regulatory proteins on human primary SCs and human iPSC-derived MNs was examined. As expected, SCs exhibited the characteristic spindle-like morphology in serum-free medium and stained positive for S100b at day 14 (Figure 1A–B). SCs expressed the four membrane-bound complement activation regulators: complement receptor type 1 (CD35), membrane cofactor protein (CD46), decay accelerating factor (CD55), and (MAC-inhibitory protein (CD59). Over 90% of the cells expressed CD35, CD46 and CD55, whereas 86.6% of cells expressed CD59 at 14 days in culture (Figure 1B and Table 1). These findings confirm previous work investigating complement regulatory protein expression on human SCs.^[22] Similarly, MNs displayed the characteristic branched morphology indicating axonal dendritic processes in serum-free medium and stained positive for neurofilament-medium (NF-M) (red) and microtubule-associated protein 2 (MAP2) (green) at day 14 (Figure 1C). Additionally, MNs expressed all four complement regulators, but at reduced levels compared to SCs (Figure 1D and Table 1). Lastly, as the ganglioside GM1 has been implicated in the pathology of MMN, its expression by SCs and MNs was examined by flow cytometry and found to be on 99.4% of SCs and 84.8% of MNs (Table 1).

Figure 1. Characterization of Schwann cells and iPSC-motoneurons complement degrading protein expression.

A) phase image of day 14 primary human Schwann cells (left) and immunostaining for S100b (right), B) percentage of CD35+ cells (left), percentage of CD46+ cells (middle left), percentage of CD55+ cells (middle right), percentage of CD59+ cells (right), C) phase image of day 14 human iPSC-motoneurons (left) and immunostaining for NF-M (red) and MAP2 (green) (right), D) percentage of CD35+ cells (left), percentage of CD46+ cells (middle left), percentage of CD55+ cells (middle right), percentage of CD59+ cells (right). Scale bars = 50 μM.

TABLE 1.

Characterization of Schwann cells (hSC) and iPSC-motoneurons (hMN) complement degrading protein and ganglioside GM1 expression.

CD Marker	% Positive hSC	% Positive hMN
CD35	91.2	64.9
CD46	97.2	45.5
CD55	98.8	60.5
CD59	86.7	15.3
GM1	99.4	84.8

Diagnosis and characterization of patient serum samples

Serum samples from 15 patients with an autoimmune demyelinating neuropathy were obtained for autoantibody characterization. As shown in Table 2, ten CIDP patient samples and five MMN patient samples were received for this study. The first 10 serum samples (3 MMN and 7 CIDP) received were evaluated for the presence of IgM and IgG autoantibodies. All 10 samples evaluated were positive for either anti-GM1 IgM or IgG, while 6 were positive for both antibody isotypes. We did not observe any disease-specific differences in autoantibody expression in the 10 sample we evaluated (Supplemental Figure 1). These serological observations agree with others’ findings and underly the complex nature of these rare autoimmune diseases.^[23–24]

TABLE 2.

Patient serum sample ID, diagnosis, and basic demographics including age and treatments at the time of biosample collection. Abbreviations: CYC = cyclosporine; IVIG = intravenous immunoglobulins; MMF = mycophenolate mofetil; mo = months; MTX = methotrexate; Pred = prednisone; RTX = rituximab.

Sample ID	Diagnosis	Sex	Race	Age (years)	Active Treatment
ND132	Probable CIDP	Male	Caucasian	51	IVIG, MMF
ND137	Definite CIDP	Female	Caucasian	67	MMF
ND140	Probable CIDP	Male	Caucasian	55	MTX
ND143	Definite CIDP	Male	African-American	61	IVIG
ND146	Probable CIDP	Female	Caucasian	66	Pred, MMF, RTX 9mo prior
ND41	MMN*	Female	Caucasian	64	IVIG
ND47	MMN*	Female	Caucasian	55	IVIG
ND51	Probable CIDP	Female	Caucasian	51	IVIG
ND71	Definite CIDP	Female	Caucasian	59	CYC, Pred
ND76	Probable CIDP	Male	Caucasian	71	None
ND142	MMN*	Male	Caucasian	57	IVIG
ND15	MMN*	Male	Caucasian	51	IVIG
ND30	MMN*	Male	African-American	47	IVIG
ND35	Definite CIDP	Female	African-American	68	MMF
ND45	Probable CIDP	Male	Caucasian	63	IVIG, MMF

^*MMN patients were all tested clinically for anti-MG1 antibodies. ND15 and ND47 tested positive.

Patient serum autoantibody binding to SCs and MNs

To investigate the ability of serum-derived autoantibodies from CIDP and MMN patients to bind to SCs and MNs, we established monocultures of each cell type, grew them for 14 days, and then followed standard immunocytochemistry protocols to assess the presence of SC- and MN-binding IgM and IgG autoantibodies in patient sera (see the methods section for detailed protocol). Following incubation in patient sera, autoantibodies were detected on the surface of SCs and MNs for all diseased-patient serum samples evaluated (Figure 2). In general, SCs showed a stronger anti-IgG signal (red stain) than anti-IgM signal (green stain), although regions of green and yellow, indicating colocalization, (white arrowheads) were observed (Figure 2A). For MNs, a positive anti-IgG and anti-IgM signal was observed in all samples evaluated; in some regions, autoantibody colocalization was observed (Figure 2B). We did not observe a correlation between autoantibody binding to SCs and MNs indicating the possibility of additional cell-type specific autoantibodies in the patient serum. Importantly, no autoantibodies were observed on the surface of SCs or MNs incubated in the pooled normal human serum (NHS) controls (Figure 3C).

Figure 2. Patient serum auto-antibody binding on human SCs and iPSC-MNs.

A) patient serum auto-antibody binding to human primary Schwann cells, B) patient serum auto-antibody binding to human iPSC motoneurons. Scale bars = 100 μM. anti-IgM (green); anti-IgG (red); DAPI (blue); Colocalization (yellow).

Figure 3. Patient serum-mediated complement C3b and C5b-9 deposition on human SCs and iPSC-MNs.

A) complement deposition on human primary Schwann cells as a result of auto-antibody complement activation, B) complement deposition on human iPSC motoneurons as a result of auto-antibody complement activation. Arrow heads indicate colocalization. C) Alexa Fluor 488nm and 568nm secondaries only (far left), negative control − normal human serum + complement only in Schwann cells (second left) and motoneurons (center), positive control − anti-GM1 + complement in Schwann cells (second from right) and motoneurons (far right). Scale bars = 100 μM. anti-C3b (green); anti-C5b-9 (red); DAPI (blue); Colocalization (yellow).

Patient serum-mediated complement deposition onto SCs and MNs

After establishing that the diseased-patient serum contained anti-GM1 autoantibodies and that the autoantibodies could be detected bound to the membranes of SCs and MNs, we investigated whether the bound autoantibodies activated the complement cascade in the model. SC and MN monocultures sensitized with a commercially available anti-GM1 IgG antibody and exposed to human complement exhibited surface C3b and C5b-9 staining by immunohistochemistry, demonstrating the sensitivity of the system to detect complement deposition (Figure 3C). In SC (Figure 3A) and MN (Figure 3B) monocultures, C5b-9 deposition (red) on the plasma membranes was clearly observed for all patient samples tested. While most images showed some regions of C3b (green) and C5b-9 colocalization (yellow), most of the detected signals were C5b-9. This observation was similar in both SC and MN monocultures. Importantly, no autoantibody-mediated complement deposition onto SCs or MNs treated with the NHS+complement control was observed (Figure 3C).

Effects of TNT005 on patient serum-mediated complement deposition on MN+SC cocultures

To continue developing the peripheral demyelinating neuropathy disease-on-a-chip model, we tested the effects of TNT005, a mAb inhibitor of C1s, the key classical complement pathway protease, on patient sample-mediated complement deposition in MN+SC cocultures. To validate the effectiveness of the approach, MN+SC cocultures were treated with a commercially available anti-GM1 antibody+complement (positive control) in the presence of either TNT005 or a mouse IgG2a isotype control (IC) molecule. TNT005, in contrast to the IC, completely inhibited complement deposition in this system (Supplemental Figure 2). For cocultures treated with patient serum+complement+IC, deposition of complement fragments C3b and C5b-9 driven by all samples tested was observed (Figure 4). While there were occasional regions of colocalization, most of the observed signal was anti-C3b. In contrast, in cocultures treated with patient serum+complement+TNT005, an almost complete abolishment of both C3b and C5b-9 deposition was observed (Figure 5). We determined the percent reduction in fluorescence intensity between IC-treated and TNT005-treated conditions and found a greater than 90% reduction in patient autoantibody-mediated complement deposition for all samples except for ND71 and ND47 (89.1% and 89.3% reduction in C5b-9) (Figure 6).

Figure 4. Effects of IgG2a IC treatment on patient serum-mediated C3b and C5b-9 deposition on iPSC-MN-Schwann cell co-cultures.

C3b and C5b-9 deposition in the presence of IC. Scale bars = 100 μM. anti-C3b (green); anti-C5b-9 (red); DAPI (blue); Colocalization (yellow).

Figure 5. Effects of TNT005 treatment on patient serum-mediated complement deposition on iPSC-MN-Schwann cell co-cultures.

Blocked C3b and C5b-9 deposition in the presence of TNT005. Scale bars = 100 μM. anti-C3b (green); anti-C5b-9 (red); DAPI (blue); Colocalization (yellow).

Figure 6. Relative effects of IC and TNT005 on complement deposition.

Percent reduction in complement C3b and C5b-9 deposition on motoneuron+Schwann cell co-cultures in TNT005- and IC-treated systems (n=15, r=2) (n=2).

Effects of TNT005 on patient serum-mediated changes in motoneuron physiology

To understand whether patient sera mediated complement activation was sufficient to recapitulate the deficits in neuronal function observed clinically in patients with CIDP and MMN, we developed a HoaC system capable of measuring MN spontaneous action potential (AP) frequency and conduction velocity (CV) (Figure 7). Figure 7A shows a cartoon of the system indicating the cell-plating layout and the axon guidance microtunnels over the electrodes. The PDMS microtunnels and the assembled system with aligned microtunnels are shown in Figure 7B–C. Motoneuron axons growing through the microtunnels over the electrodes can be observed in Figure 7D.

Figure 7. Human-on-a-chip system design for conduction velocity measurements on microelectrode arrays (MEAs).

A) cartoon of the system design indicating plating layout, B) image of the PDMS microtunnels, C) image of the assembled MEA+microtunnel system, D) image of two rows of tunnels with axons growing across the electrodes, E) high magnification of the image in panel D. Scale bar = 50 μm. MEAs = microelectrode arrays. PDMS = Polydimethylsiloxane.

The HoaC system was characterized using NHS+complement (negative control) and anti-GM1+complement (positive control) (Figure 8A–B). Figure 8A–B shows representative graphs depicting conduction frequency histograms overlays for pre-compound addition (blue bars) and post-TNT005 addition measurements (salmon bars); the dashed lines indicate the average CV for each condition. Figure 8C indicates the average AP waveform shape (red line) overlaid on the individual waveforms (blue lines) collected from a NHS+complement+TNT005 treated system during the recording window. Spontaneous AP frequency and CV were measured before and after treatment and are expressed as baseline normalized parameters (see statistical analysis). Review of the data indicated a right skewed distribution for patient sample data and therefore the patient data was analyzed using a non-parametric evaluation and data is displayed as the median [Q1;Q3]. In the NHS+complement+IC condition, the baseline-normalized median spontaneous AP frequency was 1.98 [1.63;2.32] Hz and the median CV was 0.91 [0.83;0.99] m/s (Figure 9A). In the NHS+complement+TNT005 treated condition, the median spontaneous AP frequency was 2.45 [1.84;3.06] Hz and the median CV was 1.18 [1.17;1.20] m/s. The differences in AP frequency and CV in the isotype-treated compared to TNT005-treated conditions were not statistically significant (Figure 9A). These data indicate that NHS±TNT005 treatment does not reduce spontaneous AP frequency or CV.

Figure 8. Effects of TNT005 and IC on action potential generation and conduction velocity in motoneurons exposed to normal human serum or anti-GM1+complement serum influence.

A) effects of IC (left) and TNT005 (right) on normal human serum-dosed system physiology, B) effects of IC (left) and TNT005 (right) on cells sensitized with anti-GM1 antibody and exposed to complement, C) representative composite waveform (average waveform (red) and individual traces (blue)) from electrode recordings in the tunnel system.

Figure 9. Summary of the effects of IC and TNT005 on normal human serum, anti-GM1 antibody and diseased patient serum influence on action potential generation and conduction velocity.

A) effects of IC and TNT005 in the presence of normal human serum, B) effects of IC and TNT005 in the presence of anti-GM1 antibodies+complement, C) effects of IC and TNT005 in the presence of diseased-patient serum. (* p <0.05 as determined by student t-test between isotype treated and TNT005 treated conditions with α=0.05. ** p ≤ 0.001, as determined by Wilcoxon Signed Rank Test between isotype treated and TNT005 treated conditions with α=0.05). (Panel A n=2, R=2, Panel B n=2, R=2, Panel C n=2, r=20)

For systems treated with anti-GM1 antibodies+complement+IC, compared to baseline, the median spontaneous AP frequency decreased to 0.26 [0.25;0.27] Hz and the median CV decreased to 0.22 [0.21;0.24] m/s (Figure 9B). In contrast, systems treated with anti-GM1 antibodies+complement+TNT005 had a median spontaneous AP frequency of 1.53 [1.46;1.61] Hz and the TNT005 prevented the reduction of the CV induced by Ab/complement for a value of 0.98 [0.94;1.02] m/s, which was similar to pre-dose values (Figure 9B). These data indicate that anti-GM1 antibodies are sufficient to reduce spontaneous AP frequency and CV in the presence of the TNT005-IC; however, treatment with the TNT005 inhibitor is sufficient to rescue the anti-GM1-mediated effects on AP frequency and CV. For systems treated with diseased-patients sera+IC, compared to baseline, the median spontaneous AP frequency decreased to 0.36 [0.21;0.52] Hz and the median CV decreased to 0.42 [0.23;0.86] m/s compared to 1.09 [0.94;1.97] Hz and 1.06 [0.89;1.49] m/s when treated with patients sera+TNT005 (Figure 9C). These data indicate that the TNT005 inhibitor was sufficient to rescue motoneuron pathophysiology mediated by CIDP and MMN patients sera. There were no observable differences between the effects of MMN patient serum on motoneuron physiology compared to CIDP patient serum (data not shown).

Discussion

Human-on-a-chip, microphysiological system (MPS) technology has advanced rapidly recently with new platforms mimicking essential elements of human physiology appearing regularly in the literature and commercial marketplace.^{[17, 25–26]} Currently, systems that reproduce critical elements of the physiology of the lung, liver, kidney, gut, nervous system, heart, and skeletal muscle have been established.^[27–31] By mimicking key aspects of human physiology or disease pathophysiology, HoaC systems offer an alternative to conventional preclinical models for drug screening such as animal models and static-well, cell culture-based assays. In this study, we established a novel HoaC system composed of previously validated iPSC-MNs and primary SCs that were co-cultured on an MEA-microtunnel platform.^[32–33] After system validation, a rare disease model for autoimmune demyelinating neuropathies was developed using serum from CIDP and MMN patients. In response to patient-serum treatment, it was possible to observe clinically relevant neurophysiological hallmarks of these diseases in the system, including reduced or blocked conduction velocity and decreased action potential firing frequency. Finally, the efficacy of TNT005, a novel antibody that inhibits the classical complement pathway, was evaluated for its ability to rescue patient serum-induced neuronal dysfunction. TNT005 was sufficient to block complement deposition on MN and SC monocultures and cocultures and rescued the functional neurophysiologic deficits observed when cells were exposed to patient serum.

The complement system, composed of recirculating, soluble proteins, is an effector system of the innate and adaptive humoral immune system. In addition to the major role it plays in the recognition and elimination of pathogens and non-self material, the complement system plays important roles in the removal of senescent cells and tissue remodeling (e.g., synaptic pruning).^[34] Patient serum treatment of SC and MN cultures resulted in autoantibody binding to cell plasma membranes and subsequent classical complement pathway-mediated deposition of C3b and C5b-9 (Figures 2 and 3). Several studies have investigated the role of complement deposition on human glial cells including SCs and astrocytes.^{[22, 35]} In a study conducted by Koski et al., the authors confirmed SC expression of CD46, CD55 and CD59 and observed C3b deposition on SCs when incubated with human complement + complement-fixing antibodies to peripheral nerve myelin. The authors also observed poly-C9 formation on myelin segments which was increased when the SCs were treated with anti-CD59 antibodies. While their data indicated the importance of the complement system regulators, they also showed that complement regulation is insufficient in the presence of antibodies against peripheral nerve myelin. Their findings support our observations that complement regulating proteins expressed on the SCs are insufficient to block complement deposition induced by CIDP and MMN patient autoantibodies. In a study by Spiller et al. using human astrocytes, the authors observed expression of CD46, CD55 and CD59 which was sufficient to block complement-mediated cell lysis. However, when CD59 was blocked they observed >90% complement-mediated cell lysis of astrocytes. These data suggest a mechanism by which we were able to observe significant complement deposition on SCs and MNs without observing cell death, while still observing patient serum-mediated functional deficits. Taken together, these data support the hypothesis that these functional systems provide a sensitive platform for evaluating the efficacy and safety of novel therapeutics as they more closely approximate the clinical presentation.

The etiology of chronic autoimmune neuropathies remains elusive.^[24] The data described here supports the hypothesis that CIDP and MMN patient autoantibodies play a direct role in disease pathogenesis through activation of the classical complement pathway; the importance of humoral factors in axonal demyelination is well-established.^[36] While we assayed the diseased-patient serum samples for anti-GM1 IgG and IgM antibodies, other autoantibodies are likely to play an important role in disease pathology (Supplemental Figure 1).^[37] However, our model does not represent a complete human immune system. This is important because the complement system interacts with both innate and adaptive immune system cell types in addition to its autonomous lytic immune activities.^[34] Specifically, we did not model the cellular immune components believed to be important in autoimmune neuropathies including monocytes and T lymphocytes (T-cells).^[38] Specifically, activated monocytes and T-cells have been identified in patient nerve biopsies after crossing the blood-nerve barrier.^{[36, 39]} While the exact role of T-cells and monocytes in the pathogenesis of CIDP and MMN has yet to be established, they undoubtedly contribute to disease pathology. A clinical study conducted by Marsh et al. using alemtuzumab to target and deplete circulating monocytes and T-cells demonstrated at least partial disease remission in four of seven CIDP patients treated.^[40] However, toxicity concerns have limited this approach. Therefore, further investigation of the role of monocytes and T-cells in autoimmune neuropathies as well as therapies aimed at limiting their role in disease are warranted. Future studies could investigate the role of monocytes and T-cells in CIDP and MMN disease pathology by adding the appropriate cell type to the HoaC system. Additional studies could be undertaken to investigate the role of the blood-nerve barrier in autoimmune neuropathy pathogenesis.

In this study, we showed that the anti-complement C1s monoclonal antibody, TNT005, was effective at blocking CIDP and MMN patient serum-induced reductions in spontaneous AP frequency and conduction velocity (Figure 9). Since the initial description of natural complement inhibitors in 1902 with the discovery that cobra venom factor had anti-complement activity, scientists have worked to develop anti-complement therapeutics.^[41] Synthetic molecules like nafamastat mesylate (FUT-175) that effectively inhibited C1r, C1s, C3 convertase and C5 convertase were not complement-specific and inhibited several other plasma proteases.^[42–43] The approach of using antibodies to block one of the key complement pathway proteins led to the development of the first monoclonal antibody, N19–8, which was shown to inhibit C5 cleavage and block terminal complement activation.^[44–45] Eculizumab, a humanized anti-C5 monoclonal antibody, has been approved by the FDA for the treatment of neuromyelitis optica as well as myasthenia gravis. More recently, monoclonal antibodies targeting the C1 complex components have been developed and are being tested in early phase clinical trials, including ANX005 targeting C1q.^[34] These studies confirm the efficacy of complement targeting therapies for the treatment of acute and chronic diseases where complement activation plays a pathological role. Interestingly, eculizumab was tested in a small, open label trial of ten patients with MMN, but demonstrated little benefit in clinical response or in improvement of electrophysiological parameters.^[46] Currently, there are no approved therapies directed at inhibiting the complement pathway for autoimmune demyelinating neuropathies, including CIDP or MMN.^[47] The data described here indicate that TNT005 is effective for inhibiting CIDP and MMN patient sera mediated complement activation on MNs and SCs (Figure 5). Further, TNT005 effectively rescues neuronal dysfunction induced by CIDP and MMN patient serum in our HoaC conduction velocity model (Figure 9). These data suggest that proximal inhibition of the classical pathway at the level of C1s vs terminal inhibition at C5 may be a more effective therapeutic approach for inhibiting complement mediated pathology in CIDP and MMN. In a study by Harschnitz et al., iPSC-MNs were treated with MMN patient serum. The authors showed that anti-GM1 IgM autoantibodies damaged the MNs by complement-dependent and complement-independent mechanisms.^[13] The data presented here support the authors’ findings indicating that autoantibodies from demyelinating neuropathy patients are sufficient to recreate important aspects of the clinical phenotype in vitro and support clinical investigations of complement therapeutics in patients with CIDP and MMN. Data generated with this model provide support for testing C1s inhibition in clinical trials of autoimmune demyelinating neuropathies, such as CIDP (NCT04658472).

Human-on-a-chip disease systems hold great potential for modeling rare diseases and enhancing the predictive power of novel therapeutics. In this study we demonstrate the efficacy of a novel therapeutic to block disease pathology using patient samples in a clinically relevant functional HoaC model and illustrate the utility of MPS technology in the drug discovery process especially where alternative models are lacking, not feasible due to lack of species cross-reactivity with target, or absent. Further, one major hurdle in the development of safe and effective therapeutics for rare diseases is the absence of representative models of disease; while much has been learned from animal models of autoimmune neuropathies, they have been criticized for failing to translate to successful identification of novel therapeutic options.^[48] The HoaC model described here indicates that MPS systems, when properly designed, can be used to mimic the clinical hallmarks of human diseases and serve a very valuable tool for establishing efficacy of available and novel therapies. These systems may also be a better predictor for clinical outcomes as they provide data on functional parameters used to monitor disease progression clinically. MPS technology aims to bridge the gap between animal models and patients by serving as models for investigating rare disease molecular mechanisms as well as platforms for drug discovery and data generation for drug approval. Importantly, these studies provide a novel system for testing the role of complement in nervous system disease and demonstrate the potential for classical complement inhibitors, for which none are currently approved in CIDP or other autoimmune demyelinating neuropathies, to prevent autoimmune mediated loss-of-function and nerve damage.

Materials and Methods

Microelectrode Arrays (MEA)

MEAs (Axion Biosystems, 60MEA200/30IR–TI–W/O) contained 59 titanium nitride electrodes with a diameter of 30μm, that are organized in an 8×8 array with 200 μm pitch in both directions.

PDMS tunnels preparation

A tunnel mold was created on a 4” silicon wafer utilizing two layers of photolithographically patterned SU-8. The first layer was produced with SU-8 2002, spun to a thickness of 4 μm thick, and bakes according to the manufacturer’s recommendations. A photomask containing the pattern for the tunnels (trapezoidal from 30 to 100 μm wide) was used to cure the SU-8 using an ABM contact mask aligner at an exposure of 105 mJ/cm² at 365nm. Another layer of SU-8 2150 was spun to a thickness of 400 μm and baked according to manufacturer’s recommendations. A second mask containing cell chambers was used to cure the second layer of SU-8 using an exposure of 550 mJ/cm² at 365nm. The two layers were immersed in SU-8 developers for 30 minutes and rinsed with isopropyl alcohol and water, and dried.

Polydimethylsiloxane (PDMS) barriers were prepared by mixing Sylgard 184 (Ellsworth Adhesive #2065622) silicone elastomer with curing agent (10:1 ratio) in a 50mL conical tube and hand-mixed until opaque. The PDMS was centrifuged at 300g for 5 minutes (5 brake/5 acceleration) and poured onto the tunnel mold in the lid of a 145 mm Petri dish. The PDMS and mold was placed into an 60°C oven for 24 hours, after which the PDMS was removed from the wafer. Because the mold contained 12 copies of each tunnel mold, the individual tunnel chambers were cut out and soaked in 70% isopropyl alcohol to remove residual uncured PDMS monomer. On the day of plating, the barriers were removed from IPA, sterilized in 100% ethanol and allowed to airdry until the PDMS appeared clear (~1.5 hours).

MEA and coverslip surface preparation

The MEAs and coverslip surfaces were modified using silane chemistry as previously described.^[49] Briefly, the surfaces were modified with 3-(Trimethoxysilylpropyl)diethylenetriamine silane (DETA) in toluene to produce an amine-containing, hydrophilic surface. Next, PDMS microtunnels were hand-aligned over the MEA electrodes with the smaller tunnel aperture facing the ground electrode of the MEA using a phase contrast microscope. Once aligned, gentle pressure on the microtunnels fixed them to the MEA surface. Next, the MEAs and coverslips were coated with entactin-collagen-laminin (ECL matrix) (EMD Millipore 08–100) + 1 mg/mL fibronectin (Sigma FC010) diluted in PBS for 2 hours in a 37°C, 5% CO₂ incubator.

Human iPSC-motoneuron culture

Human induced pluripotent stem cells (iPSCs) (line ND41866) were purchased from the Coriell Institute for Medical Research and denoted as passage 0 (P0). The Coriell Institute maintains and is responsible for all IRB approval and informed consent for all lines deposited and purchased from their repository. The cells were passaged up to P10 using NIH recommended protocols. Cells from P6–10 were used for motoneuron (MN) differentiation. MNs were differentiated from iPSCs by following previously published protocols with the modification of replacing Component C with 0.1 μM LDN193189 (Tocris #6053) and 6 μM SB431542 (Tocris #1614).^[50] Differentiated MNs were maintained in human motoneuron medium (hMN) as previously described.^[51] Vials of MNs were cryopreserved for use in future experiments. Following retrieval of MNs from cryostorage, the vial of cells was thawed in a water bath with gentle swirling. The cells were transferred to a 50 mL conical tube and 9mL of cold PBS was added to the cells dropwise. The cells were pelleted at 250g for 5 minutes in a 4°C centrifuge. The PBS was aspirated and the MNs were suspended in 1mL of hMN medium and a cell count was performed. The cells were plated on 15mm round coverslips (Sigma CLS284518) at 200 cells/mm² suspended in 200 μL of hMN medium for controls. For MEAs, 50,000 cells were suspended in 100 μL of hMN medium and plated on the somal side of the MEA. The cultures were maintained in a 37°C, 5% CO₂ incubator. The following day, a full medium change was performed to remove nonadherent cells. A 50% medium change was performed daily for the duration of the experiments.

Schwann cell culture

Human primary Schwann cells (SCs) were cultured according to the vendor’s protocol (Sciencell 1700). Briefly, cryopreserved SCs were thawed in a water bath with gentle swirling. The cells were immediately transferred to a T-75 flask containing prewarmed Schwann cell medium (Sciencell 1701) and put into a 37°C, 5% CO₂ incubator. The SCs were expanded and passaged one time prior to use in experiments. For monoculture experiments, SCs were plated on ECL+fibronectin-coated 15 mm round coverslips at 200 cells/mm² suspended in 200 μL of hMN medium. For coculture with MNs on coverslips or MEAs, SCs were plated at 200 cells/mm² suspended in 100 μL of hMN medium. The following day, a full medium change was performed to remove nonadherent cells. A 50% medium change was performed daily for the duration of the experiments.

MMN and CIDP study population

This study was approved by the Duke University and University of North Carolina Chapel Hill Institutional Review Boards (Pro00084018, Pro00075115) and written informed consent was obtained from all subjects. Serum samples were collected from patients meeting established criteria for MMN and definite or probable CIDP.^[52–53] Antibody testing for anti-GM1 antibodies was performed for all patients with MMN as part of their clinical evaluation.

Isolation and storage of patient serum

Following venipuncture, peripheral blood was collected in a serum separator tube (BD Vacutainer) and placed on ice for 30 minutes to allow for coagulation. Serum was then placed in room temperature centrifuge and spun at 1200g for 10 minutes. Total recovered serum was divided evenly into five cryovials and immediately stored at −80°C. Processing of each serum was completed within one hour of drawn time.

Biopharmaceutical Compounds

TNT005 is a mouse IgG2a monoclonal antibody that shows binding and functional specificity to human C1s. TNT005 was discovered from a mouse hybridoma library generated through standard immunization techniques using activated human C1s protein as an immunogen (EMD Millipore, Billerica, MA). Purified TNT005 was obtained from hybridoma culture supernatant using Protein A chromatography.

Flow cytometry

At day 14, human iPSC-MNs were analyzed for expression of the complement regulatory factors CD35 (complement receptor type 1), CD46 (complement regulatory protein), CD55 (decay accelerating factor) and CD59 (MAC-inhibitory protein). Flow cytometry buffer (FC buffer) was prepared by adding 1% bovine serum albumin (Sigma A2058) and 0.1% sodium azide (Sigma S2002) to 1x PBS (Sigma P5493). iPSC-MNs were lifted from coverslips using accutase (Sigma A6964) at room temperature and transferred to a 15 mL conical tube. The accutase was neutralized with an equal volume of cold FC buffer and the cells were pelleted in a Beckman-Coulter centrifuge at 300g for 5 minutes at 4°C. The supernatant was aspirated, and the cells were 200 μL suspended in cold FC buffer. Next, 20 μL of FcR blocking reagent (Miltenyi Biotec 130–059-901) was added to the cell suspension and incubated for 30 minutes on ice to block the FcR. Next, a cell count was performed and 1.0×10⁶ iPSC-MNs were transferred to microcentrifuge tubes for primary antibody incubation and the volume of each tube was adjusted to 100 μL. Primary antibodies were added at a concentration of 5 μg/mL for all antibodies and incubated for 1 hour on ice. The following tubes were prepared: 1) unstained, 2) anti-CD35-FITC (ThermoFisher 11–0359-42), 3) anti-CD46-PE (ThermoFisher MA1–19649), 4) anti-CD55-FITC (ThermoFisher MA5–16596), and 5) anti-CD59-APC (ThermoFisher MA1–19463). All tubes were incubated with DRAQ7 Live/Dead dye (Abcam ab109202). Following antibody incubation, the cells were pelleted, washed 3x in FC buffer, and finally resuspended in 120 μL of FC buffer for analysis. The cells were analyzed for marker expression using a CytoFlex flow cytometer (Beckman-Coulter B53006) and the data analyzed using CytExpert software.

Enzyme-linked Immunosorbent Assay (ELISA)

MMN and CIDP patient sera samples were analyzed using an anti-GM1 IgG and IgM ELISA kit (Buhlmann Labs EK-GM1-GM). Samples were analyzed following the vendor supplied protocol. Following assay completion, the 96-well plate was analyzed using a BioTek Gen5 2.06 plate reader for colorimetric reaction. Data was presented as a percentage ratio of provided calibrator sample.

Diseased-patient sera immunoglobulin deposition assay

Monocultures and cocultures of iPSC-MNs and SCs were established as described. At day 14, patient serum was added to the medium of monocultures and cocultures diluted 1:10 and returned to the incubator for 90 minutes. Following the incubation period, cultures were washed 3x with PBS in preparation for immunocytochemistry. For these experiments, cultures were examined for the binding of diseased patient IgG and IgM antibodies to the surfaces of the iPSC-MNs and SCs by using human anti-IgG and anti-IgM secondary antibodies (see Immunocytochemistry methods for details).

Complement deposition assay

Monocultures and cocultures of iPSC-MNs and SCs were established as described. At day 14, patient serum and purified human complement (Sigma-Aldrich S1764) was added to the medium of monocultures and cocultures diluted 1:10 and returned to the incubator for 90 minutes. Following the incubation period, cultures were washed 3x with PBS in preparation for immunocytochemistry. For these experiments, cultures were examined for complement deposition by immunostaining for C3b and C5b-9 deposition (see Immunocytochemistry methods for details).

Immunocytochemistry, confocal microscopy, and image analysis

At the end of the experiment, iPSC-MN coverslips were washed 1x with cold PBS and fixed with cold 4% paraformaldehyde diluted in PBS for 5 minutes. The paraformaldehyde solution was aspirated, the cells were rinsed 3x with PBS, and then permeabilized with a solution of 0.1% Triton X-100 in PBS (permeabilization solution) for 30 minutes. Finally, the iPSC-MNs were blocked with permeabilization solution + 5% donkey serum (blocking solution) for 60 minutes. The cells were then incubated overnight at 4°C with primary antibody in blocking solution. The primary antibodies used were 1) anti-NF-H (Millipore ab5539), 2) anti-S100 (ThermoFisher 710363), 3) mouse anti-C3/C3b (Abcam ab11871), 4) rabbit anti-C5b-9 (Abcam ab55811) and rabbit anti-GM1 IgG (Abcam ab23943). Following three, five-minute washes with PBS, the coverslips incubated with unconjugated primary antibodies were incubated for 2 hours at room temperature in the dark with a secondary antibody conjugated to Alexa-488 (ThermoFisher A32723) or Alexa-568 (ThermoFisher A10037). For the diseased-patient sera immunoglobulin deposition assay, anti-human IgG Alexa Fluor 568 (Invitrogen A-21090), and 2) anti-human IgM Alexa Fluor 488 (Invitrogen A-21215) were used to visualize bound disease-patient antibodies. Following the secondary antibody incubation, surfaces were washed 3x in PBS and incubated with 3 mM 4′ −6-Diamidino-2-Phenylindole (DAPI) in PBS for 10 min, in the dark at room temperature for nuclei staining. DAPI solution was removed and the cells were washed 3X with PBS. Finally, coverslips were mounted on glass slides using a Hard-Set Mounting with DAPI (Vector Labs H-1400). Fluorescence images were collected using 10X or 40X objectives and 10X magnification of an Axioskop 2 mot plus upright spinning disk confocal microscope (Carl Zeiss), connected to XCite 120 Fluorescence Illumination system (EXFO) beam, a multi-spectral laser scanning and Volocity software (Perkin Elmer). The level of complement protein fluorescence pixel intensity was quantified from confocal microscopy images using ImageJ software.^[54]

MEA recording of iPSC-MN action potential (AP) frequency and conduction velocity (CV)

Motoneuron electrical activity was measured noninvasively by connecting the MEA chips to a temperature-controlled (37°C) MEA head stage connected to an Axion Biosystems amplifier. The activity of MN+SC HoaC systems was recorded at 25kHz using Axion’s Integrated Studio (AxIS) software. The data was then imported into Matlab (The MathWorks) for further analysis (see below). First, baseline spontaneous activity of the HoaC systems was recorded for five minutes (pre-dose measurement). Next, systems were dosed with respective conditions and returned to the incubator for 90 minutes. Following the 90-minute incubation, systems were reconnected to the Axion Biosystems head stage and recorded (post-dose measurement). Following the post-dose measurement, systems received a full medium change to remove the dosed compounds and then returned to the incubator for observation.

Data analysis

MEAs were obtained from Multichannel systems and consisted of 59 embedded surface electrodes and one large internal reference electrode. Each 30 μm electrode was spaced 200 μm apart in an 8×8 grid and measured extracellular changes in voltage produced during action potentials of nearby neurons. Signals were filtered and analyzed offline with Python 2.7 scripts, using the NumPy, SciPy, Matplotlib, and pandas libraries. Raw signals were filtered with a 2nd order Butterworth 200 Hz high-pass filter, then spikes were detected and extracted by scanning the signals for time points when the magnitude of the signal exceeded 5X the standard deviation of its baseline noise. Axonal conduction velocity was assessed by comparing spike times across each row of MEA electrodes. Spike times were analyzed to identify conduction events, where a conduction event is defined as a trio of successive spike times occurring across three adjacent electrodes. Conduction events are identified as such if the spike times occur one after another with no interruptions - i.e., if three electrodes designated A, B, and C are being evaluated, a conduction event would be deemed to have occurred when electrode C fired after electrode B and electrode B fired after electrode A. This was taken to be the electrical signature generated by an action potential traveling down an axon over a series of electrodes. For each conduction event, conduction velocity was calculated by taking the spike time difference between each pair of electrodes, dividing the electrode spacing (200 μm) by that time difference, then averaging the resulting velocities for each electrode pair. To ensure that noise events did not impact our calculations, conductions between electrodes that were slower than 0.05m/s or faster than 5.0m/s were excluded from the analysis. Further, the voltage traces of the spikes in each conduction event were aggregated and presented for user review, allowing the user to manually discard conduction events that appeared to consist primarily of noise events. The frequency of conduction events was calculated by taking the overall number of conduction events and dividing that number by the recording time.

Statistical analysis

The statistical analysis for the data presented in Figures 6 and 9 are listed below:

Figure 6 calculations: Data was collected across two replicates by measuring pixel intensity of sera incubated coverslips using image J. The background intensity was determined by following the same procedure for secondary antibody only conditions. The mean pixel intensity of the background was subtracted from the intensity of isotype treated (represented in figure 4) and TNT005 treated (represented in Figure 5) conditions across all 15 patient samples. The % inhibition was calculated by subtracting the TNT005 intensity from the isotype control intensity, dividing by the isotype control intensity and then subtracting that from 1. The data is presented as the mean for each sample across those two replicates.

Figure 9 calculations: Data was collected as individual replicates across two individual experiments. Calculations for CV and frequency were performed as described in the methods section. The post-treatment recording values were divided by the pre-treatment recording values to determine the baseline-normalized fold change. Since the data for NHS and anti-GM1 has equitable distribution an individual student t-test (α=0.05) was performed to compare treated vs. untreated. The patient sera dosed data does not have equal variances and does not have gaussian distribution so therefore required non-parametric analysis and data is presented as the median with error bars representing the interquartile range. Statistical significance was determined using a Wilcoxon signed rank tests (α=0.05) to determine differences between treated and untreated parameters. In the comparison of TNT005 and IC in the presence of normal human serum (Figure 9A), there was no observable difference between the TNT005 and IC with respect to frequency and conduction, and even with the small sample size, shows that the TNT005 does not decrease either of these metrics. The small sample size limits the statistical determination of the precise limit in the differences between the TNT005 and IC groups with these two metrics.