Botulinum neurotoxins (BoNTs) are bacterial toxins responsible for the neuroparalytic disease of botulism. BoNTs are highly potent toxins, with estimated human lethal doses ranging from 0.1 to 100 ng/kg, depending on the route of exposure. Clinical botulism can result from ingestion, inhalation or injection of preformed toxin, or following infection with toxin-producing strains of Clostridium (Pirazzini et al., 2017). Although botulism is rare in humans, mass casualty events can be caused by accidental or deliberate exposure to BoNT (Wein and Liu, 2005; Villar et al., 2006). The widespread availability of BoNT-producing bacterial strains, extreme lethality, and lack of specific treatments have resulted in classification of BoNTs as Tier 1 select agents, which are at greatest risk of deliberate misuse with most significant potential for mass casualties (Arnon et al., 2001).

Botulinum neurotoxins cause prolonged but reversible muscle denervation without direct trauma to motor neurons. Flaccid muscle paralysis results from a complex set of molecular actions, involving selective uptake into motor neurons, translocation into the cytosol, and proteolytic cleavage of presynaptic proteins necessary for synaptic exocytosis (Simpson, 2004). The active form of the neurotoxin is a heterodimer protein consisting of a 100 kDa heavy chain (HC) and 50 kDa light chain (LC) that remain attached through electrostatic interactions and a single disulfide bond (Pirazzini et al., 2017). The C-terminal of HC mediates highly selective and efficient binding to endosomal and ganglioside receptors on the presynaptic membrane of peripheral neurons (Bandyopadhyay et al., 1987). Following neuronal uptake via endocytosis, the N-terminal domain of HC facilitates translocation of LC across the endosomal membrane to the presynaptic cytosol (Shone et al., 1987; Fischer and Montal, 2007; Winner et al., 2020), where LC specifically targets and cleaves SNARE proteins essential for the fusion of synaptic vesicles. As the concentration of cleaved SNARE proteins at the presynaptic membrane increases, the motor nerve terminals cannot reliably elicit muscle contraction, causing muscle weakness that progresses to flaccid paralysis (Bradford et al., 2018). Some BoNT serotypes cause muscle paralysis that can last for weeks or longer, triggering secondary morphological changes, such as NMJ degeneration and muscle atrophy, which can contribute to delayed recovery (Borodic et al., 1994; Williams et al., 2013).

Clinical symptoms of botulism emerge 12–36 h after exposure to BoNT, resulting from peripheral blockade of neurotransmission at neuromuscular junctions (NMJs) and autonomic nerve terminals (Sobel, 2005). At lethal doses, toxic signs typically include generalized muscle weakness, descending flaccid paralysis, and death from respiratory failure. Post-exposure prophylaxis with equine-derived antitoxin can prevent botulism symptoms, but only if given shortly after exposure. Approximately 70% of botulism patients given antitoxin within 48 h after symptomatic emergence still require mechanical ventilation for survival (Yu et al., 2017; Richardson et al., 2019). Once botulism symptoms develop, survival from a lethal exposure requires sustained administration of mechanical ventilation until natural recovery of respiratory function. Respiratory paralysis can continue for weeks or longer, requiring chronic ventilation with an increased risk of life-threatening comorbidities (Sobel, 2005; Touman and Stratakos, 2018).

Neuromuscular junctions are specialized synapses that exhibit robust forms of short-term and long-term plasticity (Ouanounou et al., 2016). Although neuromuscular junctions between phrenic motor neurons and diaphragm nerve fibers are the main clinical targets of botulism (Sobel, 2005), few studies have characterized neurophysiological mechanisms of diaphragm intoxication and recovery from intoxication. This is mainly due to the difficulty in administering doses of BoNT that cause respiratory toxic signs without also causing respiratory failure (Emanuel et al., 2019). Although isolated diaphragm muscle preparations have been used to characterize the acute effects of BoNT intoxication on phrenic neurotransmission, the short useful life of these preparations (6–8 h) precludes their use in studying longer-term responses ex vivo (Stanley and Drachman, 1983; Dolly et al., 1987; Bradford et al., 2018; Vazquez-Cintron et al., 2020). Alternatively, studies to understand mechanisms of paralysis and recovery have mainly focused on local administration of paralytic doses of BoNT to skeletal muscles of the limb followed by functional characterization over time (Morbiato et al., 2007; Rogozhin et al., 2008). However, diaphragm fibers have specialized features of excitation- contraction coupling that are mediated, in part, through expression of the diaphragm-specific RyR3 isoform of the sarcoplasmic reticulum calcium release channel (Conklin et al., 2000; Rossi et al., 2001). Diaphragm muscles are highly specialized to contract continuously and rhythmically (Polla et al., 2004) with fibers that are functionally and structurally distinct from limb skeletal muscle (Hughes and Whaler, 1962; Stuelsatz et al., 2012). Diaphragm muscle fibers exhibit relatively small training-induced changes in aerobic capacity or fiber type (Dempsey, 1986; Powers et al., 2011) and atrophy develops slowly in paralyzed diaphragm muscles, manifesting over weeks to months vs. days in paralyzed limb skeletal muscles (Welvaart et al., 2011). These differences suggest the response of diaphragm muscles to reduced activity following BoNT intoxication may be mechanistically distinct from limb skeletal muscle.

Here we characterized longitudinal changes in neuromuscular transmission at diaphragm neuromuscular junctions (NMJs) following systemic intoxication with sub-lethal doses of BoNT/A. We chose to use BoNT/A in these studies for several reasons. BoNT/A is responsible for approximately half of natural botulism cases in the United States and thus represents an important clinical target (Centers for Disease Control and Prevention [CDC], and U. S. Department of Health and Human Services, 2021). BoNT/A is also the most persistent serotype, producing long-lasting paralytic effects that are amenable to long-term neurophysiological assessments (Foran et al., 2003). Finally, BoNT/A is the active component in most BoNT-based pharmaceuticals (Brashear, 2008). We anticipate that characterization of endogenous mechanisms of neurophysiological recovery from BoNT/A paralysis will inform treatment strategies for botulism symptoms caused by natural or iatrogenic exposures. Mice challenged by systemic administration of BoNT/A developed physiological, functional and neurophysiological evidence of respiratory paralysis. Recovery of diaphragm muscle function preceded recovery of neurotransmission in individual end plates, suggesting a role for compensatory processes in response to low-amplitude synaptic transmission. Multiple forms of synaptic plasticity were identified that increased the probability of muscle contraction, providing a mechanism for uncoupling of end plate recovery from physiological recovery. These studies illustrate several compensatory mechanisms activated in response to BoNT/A paralysis, including modulation of the Ca²⁺ sensitivity of evoked neurotransmission. Furthermore, they identify endogenous mechanisms that may be targeted pharmacologically to treat diaphragm weakness caused by botulism and other neuromuscular diseases.

With the goal of developing a sublethal model of systemic botulism, mice were challenged by intraperitoneal administration of 0.55 LD₅₀ (n = 8), 0.70 LD₅₀ (n = 16), or 0.85 LD₅₀ (n = 8) BoNT/A and toxic signs were monitored over time (Figures 1A–C; Vazquez-Cintron et al., 2020). Intoxication produced respiratory signs of botulism, including abdominal paradox and altered respiratory pattern, in 75% (0.55 LD₅₀), 62.5% (0.7 LD₅₀), and 87.5% (0.85 LD₅₀) of mice, respectively. There were no significant differences in the overall proportion of mice that developed toxic signs among the three doses (χ² = 1.70, p = 0.43) nor in the proportion that progressed to mild (χ² = 0.13 p = 0.94) or severe toxic signs (χ² = 0.14, p = 0.93). There was a trend toward increasing mortality at higher toxin doses, however, differences in mortality were not significant at these group sizes (χ² = 4.5, p = 0.11). Collectively these data suggest that 0.55–0.85 LD₅₀ produce generally similar outcomes in terms of toxic presentation, albeit with a trend towards increased mortality at higher doses.

Among the different challenge doses, surviving mice could be classified into two groups based on the duration of respiratory signs: strong responders (n = 15) exhibited respiratory signs lasting ≥3 days after injection, whereas weak responders (n = 12) developed respiratory signs that resolved within 3 days or did not develop respiratory signs at all (Figure 1D). There were no apparent differences among doses in duration of toxic signs (χ² = 1.63 p = 0.44) so the different doses were aggregated for subsequent analysis. Strong responders exhibited toxic signs for an average of 10 days (95% CI: 8–12 days; Figure 1E) and underwent acute weight loss after intoxication (Figure 1F). Weak responders exhibited transient symptoms of botulism that rapidly resolved without weight loss (Figures 1E, F). In comparison, control mice challenged with BoNT/A plus antitoxin exhibited neither toxic signs of botulism nor weight loss, confirming toxic signs resulted from BoNT toxemia (Figures 1E, F). Given the lack of difference among the challenge doses, subsequent studies used exclusively 0.7 LD₅₀ BoNT/A challenge.

Voluntary running activity has been used to monitor the physiological effects of intoxication in mice (Keller, 2006; Kutschenko et al., 2016; Schwartz et al., 2021). To further interrogate the apparent differences between strong and weak responders, mice were challenged with BoNT/A or BoNT/A plus antitoxin and nocturnal running activity was monitored for 20 days (Supplementary Figure 2). Weak responders (n = 24) and antitoxin-treated mice (n = 8) showed no reductions in running activity. In strong responders (n = 26), running distances declined by 85.8 ± 4.1% within 2 days after injection and gradually returned to baseline over the next 2 weeks, paralleling changes in toxic signs (Supplementary Figures 2A–G). Strong responders exhibited shorter distances per running bout (Supplementary Figure 2H) and fewer bouts per night (Supplementary Figure 2I), consistent with reduced respiratory capacity and/or skeletal muscle weakness (Amann, 2012). Taken together, these results suggested strong responders suffered systemic botulism with respiratory involvement. Further studies focused on strong responders, as these mice exhibited clinical and physiological evidence of severe respiratory toxemia.

To directly confirm respiratory weakness in strong responders, nerve-elicited contraction strengths were measured in phrenic nerve-hemidiaphragm preparations isolated from naïve mice, strong responders and mice treated with BoNT/A plus antitoxin. In strong responders, twitch contraction strengths significantly declined to 17.3 ± 18.0% of naïve values at 3 days after intoxication and progressively recovered to baseline between 7 and 21 days (Figures 2A, B). Longitudinal changes in sub-tetanic and tetanic muscle contraction strengths closely paralleled twitch contractions, with significant decreases in contraction strength that recovered to baseline by 21 days (Figures 2A, C). No differences were apparent between naïve diaphragms and diaphragms from antitoxin-treated mice (p = 0.99). The half-maximal force-frequency was not significantly changed during the recovery from intoxication [F_(5,29) = 2.1; p = 0.09, one-way ANOVA; Table 1], suggesting BoNT/A intoxication does not affect use-dependent facilitation of diaphragm muscle contraction.

Because muscle fiber contraction is an “all-or-none” response to threshold release of acetylcholine. measurements of contraction strength can obscure graded changes in neurotransmission. To characterize neurophysiological changes during recovery from intoxication, nerve-evoked end plate potentials (EPPs) were recorded in diaphragms isolated from naïve mice and strong responders at 3, 7, 14, 21, and 28 days after BoNT/A injection. In naïve end plates, a large EPP was observed after each stimulus of a ten-stimulus train (0.2 Hz), producing an EPP success rate of 100% (Figures 3A–C). By 3 days after intoxication, the EPP success rate declined to 22.5%, with 93.2% of end plates suffering one or more stimulation failures per train. Although success rates progressively improved over time, they remained significantly depressed at 28 days, with 17.5% of end plates exhibiting at least one neurotransmission failure. Consistent with decreased success rate, EPP amplitude was reduced to 2.8% of naïve end plates by 3 days after intoxication and remained significantly depressed through 28 days (Figures 3D, E). EPPs were infrequent at early time points (Figure 3C) and, when present, usually involved ≤2 quanta (Figure 3E). Histogram analysis of EPP amplitudes revealed a subpopulation of end plates that remained stubbornly intoxicated through 28 days (Figure 3E). Incidentally, we also observed the significant depolarization of muscle fiber resting membrane potential at 3, 7, and 14 days (Supplementary Figures 3A, B), which may allow muscle fibers to contract more readily in response to decreased acetylcholine release. These data reveal that recovery of neurotransmission remains incomplete at 28 days, despite the apparent reversal of respiratory toxemia and restoration of diaphragm muscle contraction strengths (Figure 2). Furthermore, they illustrate multiple compensatory mechanisms that are activated in the NMJ during the recovery from intoxication.

To understand the hysteresis between neurophysiological recovery (>28 days) and functional recovery (<21 days) in strong responders, we measured two fundamental parameters of neurotransmission that directly reflect the degree of motor neuron intoxication: miniature EPP (mEPP) frequency and quantal content (QC) (Del Castillo and Katz, 1954). Because spontaneous release is a stochastic phenomenon directly correlated with the availability of functional release complexes, mEPP frequency serves as an agnostic marker of active site availability in intoxicated nerve terminals (Beske et al., 2016; Bradford et al., 2018). Alternatively, QC approximates the number of quanta that fuse during an action potential and thus represents the average release competency during evoked release. The mEPP frequency decreased to 3.2 ± 9.5% of the naïve value at 3 days after intoxication and partially recovered to 52.0 ± 65.3% at 28 days (Figures 4A–C). In comparison, QC was similarly reduced to 4.6 ± 20.1% of naïve values at 3 days but recovered to 76.6 ± 64.0% of naïve end plates at 28 days. Both metrics remained significantly depressed vs. naïve endplates; however, QC recovery was significantly improved compared to mEPP frequency at 28 days (p < 0.0001; Figures 4D–F). Consistent with accelerated recovery, QC histograms contained a mixture of recovered and stubbornly impaired subpopulations at 7, 14, 21, and 28 days while mEPP frequency histograms were unimodal until 28 days (Figures 4B, E). Although quantal size declined by 33.7 ± 38.3% within 3 days and remained significantly depressed through 28 days (Supplementary Figures 3C, D), quantal size underwent a monotonic recovery. Consequently, the bimodal recovery of EPP amplitude and QC cannot be attributed to changes in quantal size.

Evoked release is tightly coupled to voltage-gated calcium channel (VGCC) activation. In contrast, mEPPs can be triggered by several mechanisms, including release of internal calcium stores, calcium-regulated signaling or local activation of individual VGCCs (Williams and Smith, 2018; Lee et al., 2022). This raised the possibility that differential regulation of VGCCs contributes to accelerated recovery of QC vs. mEPP frequency. To test this hypothesis, QC and mEPP frequency were compared between diaphragms treated with vehicle or selective antagonists of P/Q-type (ω-agatoxin IVA, ATX), N-type (ω-conotoxin GVIA, CTX) or L-type VGCCs (nimodipine, NIM) (Sano et al., 1987; Protti and Uchitel, 1993; Flink and Atchison, 2002). Selective antagonists were tested on naïve diaphragms and diaphragms prepared 21 days after intoxication, when the difference between recovery of mEPP and QC was largest (Figure 4). Whereas block of P/Q-type VGCC reduced QC under all conditions, the effect was larger in intoxicated end plates (1.6 ± 2.5% of vehicle) than naïve end plates (6.3 ± 6.9% of vehicle; p < 0.0001), suggesting P/Q-type VGCCs play a larger role in intoxicated end plates (Figures 5A, B). Block of N-type VGCC had context-specific effects on evoked release, increasing QC to 129.9 ± 49.1% in naïve end plates but decreasing QC to 67.5 ± 102.8% in intoxicated end plates (Figures 5C, D). Block of L-type VGCC had no effect on QC in naïve or intoxicated end plates (p ≥ 0.21, Figures 5E, F). Selective VGCC antagonists were also tested on mEPP frequencies. Although block of P/Q-type VGCC reduced mEPP frequency in naïve end plates, the effect was small and not significant between naïve vs. intoxicated end plates (Figures 5G, H). Block of N-type (Figures 5I, J) and L-type (Figures 5K, L) VGCCs did not significantly alter mEPP frequencies in naïve or intoxicated end plates. Control studies showed vehicle treatment did not affect QC or mEPP frequencies in naïve end plates (Figures 5M–P). Taken together, these data indicate that P/Q and N-type VGCCs increase evoked release but not spontaneous release in intoxicated end plates.

The inherent difficulty of studying functional responses of respiratory muscles to BoNT intoxication has limited research into the neurophysiological mechanisms involved in recovery from paralysis. Here, we establish a model of botulism involving respiratory toxemia with low mortality. Recovery of respiratory function was monitored by measuring toxic signs (abdominal paradox and respiratory pattern), running activity, diaphragm function, and synaptic physiology at multiple time points after challenge with doses of BoNT/A that produce a high incidence of toxic signs with low mortality.

Despite the relatively low dose (compared to intramuscular challenges), we found a relatively large portion of intoxicated endplates remained impaired 28 days after intoxication. Recovery of BoNT/A-intoxicated muscles involves extensive growth of new endplate sprouts that are capable of synaptic transmission unlike the original parent endplate (De Paiva et al., 1999). However, rapid endplate outgrowth after intoxication occurs only in slow twitch fibers, and synaptic sprouts take weeks to develop in fast twitch fibers (Duchen, 1971). Diaphragm muscles consist of a mix of slow twitch and fast twitch muscle fibers (Polla et al., 2004), suggesting that stubbornly intoxicated fibers might be fast twitch fibers that do not develop outgrowths until weeks after intoxication.

The most striking finding in these studies was the enhanced recovery of evoked release compared to spontaneous release, suggesting compensatory responses to neuromuscular paralysis that enhanced respiratory function despite prolonged inhibition of synaptic transmission. Several forms of synaptic plasticity were identified that likely contribute to these compensatory responses, including differential recruitment of VGCCs, post-synaptic modulation of resting membrane potential and changes in quantal size. These studies point to potential therapeutic targets and help explain the effectiveness of proposed treatments that directly or indirectly enhance VGCC activity (Adler et al., 1996; Bradford et al., 2018; Vazquez-Cintron et al., 2020; Machamer et al., 2022).

Because this study was designed to characterize the progression of neuromuscular function recovery in animals displaying respiratory signs of botulism, the cutoff for inclusion in the strong responder group was 3 days, which was the first time point chosen to quantify diaphragm function using ex vivo assays. However, toxic signs resolved between 3 and 5 days in approximately 50% of strong responders whereas the remaining 50% resolved after 10 days. This raises the possibility that restricting studies to mice that exhibit toxic signs beyond 5 days would further improve the characterization of compensatory changes during recovery. The variable duration of toxic signs is consistent with clinical data (Sobel, 2005), while evidence of long-lasting neurophysiological deficits without associated physiological signs is consistent with single fiber electrode recordings in clinical patients given therapeutic doses of BoNT by intramuscular injection (Kouyoumdjian et al., 2020).

The key mechanistic difference between evoked release and spontaneous release is the identity of the vesicle fusion trigger and the associated proteins. Under physiological conditions, evoked release is primarily triggered by Ca²⁺ binding to synaptotagmin-1 and synaptotagmin-7 in response to action potential (AP) induced activation of VGCCs (Rizo and Rosenmund, 2008; Jahn and Fasshauer, 2012). Conversely, spontaneous vesicle fusion is loosely coupled to Ca²⁺ influx and the identity of the Ca²⁺ sensor and the mechanisms driving vesicle fusion remain uncertain (Xu et al., 2009; Groffen et al., 2010; Kavalali, 2015). AP-induced increases in cytosolic Ca²⁺ transients can potentiate spontaneous vesicle fusion, but with substantially lower sensitivity than in evoked release (Hubbard et al., 1968; Beske et al., 2017), and activity-dependent potentiation of VGCCs increases evoked release, while having little effect on spontaneous release (Frank et al., 2006; Jeans et al., 2017; Böhme et al., 2018). Thus, enhancement of AP-induced VGCC Ca²⁺ currents could contribute to the accelerated recovery of evoked release without significantly affecting spontaneous release.

Although the C-terminal nine residues cleaved by BoNT/A (producing SNAP-25₁₉₇) are not required for association of SNAP-25 with ternary SNARE complexes (Banerjee et al., 1996), their cleavage destabilizes the SNARE complex and reduces the efficiency of the fusogenic configuration, thereby inhibiting vesicle release (Huang et al., 1998). Indeed, release complexes containing SNAP-25₁₉₇ remain competent to effect vesicle fusion, but only at elevated Ca²⁺ concentrations (Trudeau et al., 1998; Weber et al., 2010; Beske et al., 2017). Thus, it is possible that vesicles associated with full-length SNAP-25 but not SNAP-25₁₉₇ contribute to spontaneous fusion events, whereas the increased Ca²⁺ levels associated with evoked release are required to recruit SNAP-25₁₉₇ (Beske et al., 2018), thus contributing to the difference in recovery between evoked release and spontaneous release in motor neurons intoxicated by BoNT/A.

Vesicle fusion is mediated by three high voltage-activated VGCCs distinguished by the alpha subunit (Dolphin, 2016): P/Q-type (CaV2.1), L-type (CaV1), and N-type (CaV2.2). Under physiological conditions, P/Q-type VGCCs are the principal VGCC used to trigger vesicle fusion in adult mammalian motor neurons. However, during the early stages of development (Rosato Siri and Uchitel, 1999), or after physical, chemical or genetic, synaptic transmission becomes sensitive to N-type, L-type, and R-type channel blockers injury (Katz et al., 1996; Fratantoni et al., 2000; Urbano et al., 2001, 2003; Flink and Atchison, 2002). We found inhibition of P/Q-type and N-type channel activity decreases evoked release in intoxicated diaphragms, indicating that intoxicated endplates are more sensitive to the loss of P/Q- and N-type VGCCs and suggesting that a homeostatic increase in P/Q-type and N-type channel activity can promote evoked release in BoNT/A-intoxicated motor neurons. Indeed, selective VGCC recruitment in intoxicated NMJs may contribute to the strong effects of aminopyridines and selective VGCC agonists at BoNT-intoxicated synapses (Beske et al., 2017, 2018; Bradford et al., 2018). Similar homeostatic increases in synaptic transmission occur in response to postsynaptic receptor blockade and are mediated by increased P/Q-type VGCC currents (Wang et al., 2004; Jeans et al., 2017). Although the mechanisms underlying homeostatic compensatory increases in QC in mammalian NMJs have not been fully characterized (Wang et al., 2016, 2018), numerous studies have identified mechanisms of homeostatic QC increases at the Drosophila NMJ (Dickman et al., 2012; Frank, 2014). In contrast to the loss of synaptic vesicle fusion in BoNT/A intoxicated endplates, most of these studies used pharmacological or genetic methods to inhibit synaptic transmission (DiAntonio et al., 1999; Frank et al., 2006). However, reducing presynaptic innervation (Davis and Goodman, 1998) or muscle excitability is sufficient to induce compensatory increases in quantal content (Paradis et al., 2001). Several retrograde signaling pathways communicate the loss of postsynaptic function (Haghighi et al., 2003; McCabe et al., 2003; Goold and Davis, 2007; Frank et al., 2009), many of which converge to increase calcium flux through Cacophony (Frank et al., 2006, 2009), the Drosophila homolog of mammalian N and P/Q-type VGCCs (Smith et al., 1996; Brooks et al., 2003), or increase coupling between Cacophony and the synaptic vesicle release machinery (Frank et al., 2020). Other homeostatic mechanisms may also contribute to compensatory plasticity; for example, it has been shown in primary rat glutamatergic neurons that knock-down of SNAP-25 increases VGCC current density without affecting VGCC subtype proportions or the expression or localization of P/Q-type channels (Condliffe et al., 2010). Thus, although BoNT/A is not known to change the level of expression or localization of SNAP-25, it is possible that cleavage of SNAP-25 alters regulation of VGCC function.

Blockade of N-type channels has previously been reported to have no effect on neuromuscular synaptic transmission (Wright and Angus, 1996; Santafé et al., 2000; Giovannini et al., 2002). This discrepancy with our data could indicate a small effect size, such that contributions of N-type channels to neurotransmission are only detectable in experiments with many recordings, such as conducted here. Alternatively, the discrepancy might be attributable to divergent animal behavior because of differential housing conditions. Naïve mice in our study were given access to free-spinning running wheels for 12 days prior to experimentation. It has been shown that mice housed with running wheels have increased rates of exercise, respiration and diaphragm activity (Tallis et al., 2017), which can elicit activity-dependent changes in motor neuron synaptic structure and function (Nishimune et al., 2014). Given that N-type channel activity both promotes synaptic transmission in BoNT/A-paralyzed diaphragms and inhibits synaptic transmission in diaphragms from exercising mice, N-type channels may have conditional effects at the neuromuscular junction that extend beyond directly promoting vesicle fusion, such as regulating bidirectional activity-dependent synaptic plasticity. Indeed, Ca²⁺ signaling through VGCCs is an essential component of bidirectional synaptic plasticity in motor neurons (Liang et al., 2003; Catterall and Few, 2008), and N-type channels could differentially regulate synaptic transmission by initiating Ca²⁺ signaling cascades that modify release downstream of P/Q channel activity. Although N-type channels are present at rodent NMJs, they appear to be expressed in the terminal Schwann cell (TSC), a glial cell covering the presynaptic terminal, and not in motor neurons (Day et al., 1997). The function of TSCs has primarily been characterized during neuromuscular reinnervation following nerve injury (Barik et al., 2016; Santosa et al., 2018). However, TSCs also mediate bidirectional plasticity based on the characteristics of cytoplasmic Ca²⁺ waves (Todd et al., 2010), consistent with the observation that VGCCs are expressed in many types of Glia including schwann cells, and VGCC-dependent calcium signaling in Glia is essential for maintaining changes in synaptic efficacy (Verkhratsky et al., 1998; Fiacco and McCarthy, 2006; Letellier et al., 2016). Future work should directly test the hypothesis that N-type channels contribute to bidirectional activity-dependent plasticity by modulating Ca²⁺ signaling in TSCs.

Consistent with other models of BoNT/A intoxication (Sellin and Thesleff, 1981; Kim et al., 1984; Molgó et al., 1989), we observed a decrease in mEPP amplitudes in diaphragm muscle fibers 3 days after BoNT injection, followed by a partial recovery through 28 days. Recovery of mEPP amplitude could enhance muscle fiber function by reducing the number of quanta required to trigger muscle contraction. Postsynaptic changes in receptor activity and localization are believed to be the primary locus of quantal size regulation (Turrigiano and Nelson, 2004), but multiple presynaptic mechanisms regulating mEPP amplitude have also been identified (Wang et al., 2005; Fong et al., 2010; Goh et al., 2011; Morsch et al., 2018). Pharmacologically targeting these regulatory pathways to increase mEPP amplitude is anticipated to promote motor neuron recovery, especially in stubbornly impaired end plates. Although inhibiting acetylcholinesterase in the synaptic cleft has been shown clinically to promote muscle function in botulism patients (Young and Halstead, 2014; Boerner et al., 2018), we are unaware of any studies focusing on modulation of mEPP amplitudes to treat botulism. Recently, compounds targeting presynaptic cannabinoid receptors have been found to regulate mEPP amplitudes and have been shown effective at restoring neuromuscular transmission in myasthenic mice (Morsch et al., 2018), suggesting they may have therapeutic potential for treating botulism.

We observed a depolarizing shift in muscle fiber RMP following intoxication, consistent with other studies investigating the loss of synaptic transmission as a consequence of BoNT intoxication (Colméus et al., 1982) or denervation (Mathers and Thesleff, 1978; Bray et al., 1982). The RMP is set by the permeability of K⁺ channels, the activity of Na⁺/K⁺ pumps, and to a lesser extent the permeability of Na⁺ and Cl^– channels (Fauler et al., 2012). Multiple types of disuse atrophy share a common transcriptional response (Lecker et al., 2004; Sacheck et al., 2007), and long-term disuse upregulates voltage-dependent Na⁺ and ClC-1 chloride channels (Desaphy et al., 2001; Pierno et al., 2002) and down-regulates the ATP-sensitive K⁺ channels (Tricarico et al., 2010). Similar changes likely occur because of motor neuron paralysis from BoNT intoxication. However, among the few studies that have identified changes in postsynaptic transcriptional programming after BoNT intoxication, none have identified changes in gene expression that directly regulate RMP (Shen et al., 2006; Mukund et al., 2014). Considering the speed of RMP reduction after BoNT/A intoxication, local signaling mechanisms that are independent from transcription may contribute to changes in RMP. In denervated rat diaphragms, RMP is modulated by the loss of glutamatergic and cholinergic anterograde signaling (Urazaev et al., 1995, 1997; Proskurina et al., 2018), and recent work has identified a pathway by which AChRs directly modify RMP through altering Na,K-ATPase pump activity (Krivoi et al., 2006; Heiny et al., 2010; Kravtsova et al., 2016). More work is needed to determine the functional consequences of these changes on muscle excitability. However, depolarization of the muscle fiber membrane at voltages closer to the voltage-gated Na⁺ channel threshold increases its intrinsic excitability. These studies suggest the mechanisms regulating RMP after BoNT intoxication could be multifaceted and evolve during recovery. The large number of druggable membrane proteins involved in regulating membrane potential offer fertile targets for pharmacological interventions to promote muscle contraction.

In conclusion, we describe a reproducible, sublethal model of botulism well-suited for evaluating post-symptomatic pharmacotherapies aimed at enhancing the recovery of respiratory function. We identified multiple forms of endogenous plasticity that compensate for low-amplitude synaptic transmission in diaphragm NMJs. These compensatory mechanisms exploit the malleability of release probability, quantal size and membrane potential of muscle fibers to drive membrane potential closer to muscle contraction threshold during evoked synaptic transmission. The effectiveness of these endogenous mechanisms is demonstrated by the rapid recovery of diaphragm function, respiration, and motility despite a subpopulation of diaphragm end plates that remain stubbornly impaired. Treatments that synergize with these changes could be effective at all stages of botulism.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The experimental protocol was approved by the United States Army Medical Research Institute of Chemical Defense (USAMRICD) IACUC. This study was conducted in accordance with the local legislation and institutional requirements.

JM, EV-C, MS, KP, AB, CO, and PM contributed to conception, design, and execution of the study. JM, EV-C, and PM wrote sections of the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version.