Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative motoneuron disease with presently no cure. Motoneuron (MN) hyperexcitability is commonly observed in ALS and is suggested to be a precursor for excitotoxic cell death. However, it is unknown whether hyperexcitability also occurs in MNs that are resistant to degeneration. Second, it is unclear whether all the MNs within homogeneous motor pools would present similar susceptibility to excitability changes since high-threshold MNs innervating fast fatigable muscle fibers selectively degenerate compared with low-threshold MNs innervating fatigue resistant slow muscle fibers. Therefore, we concurrently examined the excitability of ALS-vulnerable trigeminal motoneurons (TMNs) controlling jaw musculature and ALS-resistant oculomotor neurons (OMNs) controlling eye musculature in a well studied SOD1G93A ALS mouse model using in vitro patch-clamp electrophysiology at presymptomatic ages P8–P12. Our results show that hyperexcitability is not a global change among all the MNs, although mutant SOD1 is ubiquitously expressed. Instead, complex changes occur in ALS-vulnerable TMNs based on motor unit type and discharge characteristics. Firing threshold decreases among high-threshold TMNs and increases in a subpopulation of low-threshold TMNs. The latter group was identified based on their linear frequency–current responses to triangular ramp current injections. Such complex changes in MN recruitment were absent in ALS-resistant OMNs. We simulated the observed complex changes in TMN excitability using a computer-based jaw closer motor pool model. Model results suggest that hypoexcitability may indeed represent emerging disease symptomology that causes resistance in muscle force initiation. Identifying the cellular and molecular properties of these hypoexcitable cells may guide effective therapeutic strategies in ALS.
Amyotrophic lateral sclerosis (ALS) is a devastating disease involving degeneration of brain and spinal MNs leading to muscle atrophy, paralysis, and death (Bruijn et al., 2004). Transgenic mouse models of the human neurodegenerative motoneuron disease, ALS mimic the progressive loss of motoneurons (MNs) and allow investigation of mechanisms underlying disease vulnerability. Motoneuron axons die-back causing muscle denervation and loss of functional motor units before MN loss and overt symptoms (Kanning et al., 2010). Moreover, motor axons of fast fatigable (FF) MUs are the first to retract presymptomatically followed by the fast-fatigue resistant (FR) MUs, whereas the slow (S) MUs are spared until the end stage (Pun et al., 2006). Motoneurons forming the FF MUs typically are larger cells with higher recruitment threshold and lower-input resistance than the fatigue resistant (FR and S) MUs (Kernell, 1966; Burke, 1981). This well known “size principle” is fundamental to the “recruitment order” of motor units of a muscle wherein smaller, low-threshold MUs are recruited first followed by larger higher-threshold MUs resulting in increasing muscle force (Enoka and Pearson, 2013). The observed differential MU (Hegedus et al., 2007) and axonal vulnerability (Pun et al., 2006) in ALS raises the possibility of differential abnormalities in MNs forming fast versus slow MUs. Although membrane hyperexcitability has been reported among prototypical MNs (Pieri et al., 2003; Kuo et al., 2004; Durand et al., 2006; van Zundert et al., 2008; Quinlan et al., 2011), whether this occurs regardless of motor unit size is unknown. Concomitant with the preferential loss of fast MUs, an activity-dependent conversion to fatigue resistant MUs occurs that may increase resilience and survival (Hegedus et al., 2008; Gordon et al., 2010). It is unknown whether compensatory changes occur in MN excitability, and if so, how this would alter muscle function. Additionally, a motor pool can consist of MNs with heterogeneous discharge properties (Cotel et al., 2009) that are altered in various diseases (Bennett et al., 2001b; Bories et al., 2007). Therefore, we hypothesized that ALS vulnerable MNs show presymptomatic abnormalities in membrane properties that may be selective to MU type and MN discharge characteristics.
Second, it is unknown whether the ubiquitous expression of mutant SOD1 alters MN excitability in disease resistant motor nuclei, such as the brainstem oculomotor neurons. The oculomotor neurons (OMNs) and the related oculomotor functions are relatively spared in ALS patients (Radcliffe and Lockhart Clarke, 1862) and in mouse models of the disease (Nimchinsky et al., 2000; Angenstein et al., 2004; Niessen et al., 2006). In contrast, the closely located brainstem trigeminal motoneurons controlling jaw movements are vulnerable to degeneration (Angenstein et al., 2004; Niessen et al., 2006; Ferrucci et al., 2010) and the related masticatory functions are compromised to varying degrees in ALS (Brooks, 1996). The mechanisms underlying the early changes in neuronal excitability, whether causative or adaptive, are not clear. Early homeostatic and compensatory regulation for neuroprotection of the motor pool undoubtedly masks early disease deficits (Saxena et al., 2013). Here we focus on the early adaptive changes in excitability in ALS-vulnerable trigeminal motoneurons (TMNs) and ALS-resistant OMNs in a postnatal (P8–P12) SOD1G93A ALS mouse model (Gurney et al., 1994) to gain insight into the mechanisms underlying disease progression and the basis for differential vulnerability of neuron types.
Materials and Methods
Transgenic mice overexpressing the human SOD1G93A transgene and their WT littermates were used for this study. Experiments were performed on postnatal 8- to 12-d-old mice of either sex. Data presented are from 113 motoneurons with 60 cells from 42 WT animals and 53 cells from 38 age-matched SOD1G93A high expressor mutant animals. All animal protocols were approved by the Institutional Animal Care and Use Committee at UCLA.
The trigeminal motor nucleus was identified in coronal brainstem slices in bright field as an opaque, pale oval region that is medial to the trigeminal principal sensory nucleus and ventrolateral to the periaqueductal gray (Chandler et al., 1994). Of the 75 TMNs presented in this study, 46 were labeled with a red fluorescent dye (AlexaFluor 568 hydrazide) injected from the pipette during patch-clamp recording for post hoc imaging and location analyses. In some instances (26/75 cells), the TMNs were identified in retrogradely labeled motor trigeminal nucleus (Mot V) by injecting a green fluorescent retrograde tracer (AlexaFluor 488 hydrazide) into the jaw closer (JC) musculature 48–72 h before patch-clamp recording (Fig. 1A). Fifteen TMNs with the red dye were identified in the Mot V using post hoc confocal imaging of which nine cases were used for double-labeling with choline acetyl transferase (ChAT) with green AlexaFluor 488 hydrazide to further affirm motoneuron identity (Fig. 1B). In initial experiments, the location of the oculomotor nucleus (Mot III) was confirmed in the ventrolateral corner of the periaqueductal gray by retrogradely labeling the Mot III by injection of a green retrograde tracer (AlexaFluor 488 hydrazide) into the extraocular musculature 48–72 h before slice preparation. The located Mot III sections were 1 mm rostral from the sections containing the trigeminal motor nucleus. In subsequent experiments without retrograde tracing of the Mot III nucleus, patched OMN location was confirmed by injecting a red fluorescent dye (AlexaFluor 568 hydrazide) into the cell during patch-clamp recording followed by post hoc confocal imaging of the cell location in 20/38 cells. In a subset of these cells (5/20), motoneuron identity was further affirmed by double-labeling with ChAT (green AlexaFluor 488 hydrazide) immunohistochemistry (Fig. 1C).
All solutions were bubbled with 95% O2/ 5% CO2 and maintained at pH 7.25–7.3. The brain-cutting solution used during slice preparation was composed of the following (in mm): 126 NaCl, 3 KCl, 1.25 NaH2PO4, 26 NaHCO3, 10 glucose, 1 CaCl2, 5 MgCl2, and 4 lactic acid (Del Negro and Chandler, 1998; del Negro et al., 1999), or 85 NaCl, 2.5 KCl, 1.25 NaH2PO4, 24 NaHCO3, 25 glucose, 75 sucrose, 0.5 CaCl2, and 4 MgCl2 (Gupta et al., 2012). The slice incubation solution was composed of the following (in mm): 124 NaCl, 3 KCl, 1.25 NaH2PO4, 26 NaHCO3, 10 glucose, 2 CaCl2, 2 MgCl2, and 4 lactic acid (Schurr et al., 1988). The external recording solution during patch-clamp recording consisted of the following (in mm): 124 NaCl, 3 KCl, 1.25 NaH2PO4, 26 NaHCO3, 10 glucose, 2 CaCl2, and 2 MgCl2. The pipette internal solution used in current-clamp experiments was composed of the following (in mm): 115 K-gluconate, 9 NaCl, 1MgCl2, 10 HEPES buffer, 0.2 EGTA, 3 K2-ATP, and 1 Na-GTP with a pH between 7.28 and 7.3, and osmolarity between 290 ± 5 mOsm.
P8–P12 mice were anesthetized by inhalation of isofluorane and then decapitated. The brain was then removed and immersed in ice-cold cutting solution. The extracted brain block was mounted on a vibrating slicer (DSK Microslicer, Ted Pella) supported by an agar block. Coronal brainstem sections (250 μm) consisting of Mot V and Mot III were obtained from caudal to rostral order. Slices were then placed into the incubation solution for 40–50 min at room temperature (22–24°C).
Patch pipettes (3–5 MΩ) were pulled using a Brown/Flaming P-97 micropipette puller (Sutter Instruments). Slices were perfused with oxygenated recording solution (2 ml/min) while secured in a glass bottom recording chamber mounted on a Zeiss fixed-stage Axioskop microscope. Neurons were visualized with infrared differential video microscopy. Current-clamp recordings were performed with an Axopatch-1D amplifier (Molecular Devices) with pCLAMP acquisition software (v9.2, Medical Devices) with a sampling rate between 20 and 50 kHz. Signals were grounded by a 3M KCl agar bridge electrode mounted in the recording well.
Data analysis and statistics.
Data were analyzed using Clampfit v9.2 (Molecular Devices), Sigma-Aldrich Plot v9.0, SYSTAT 13 (Systat Software), StatView (SAS Institute), and Microsoft Excel. Cells with resting membrane potential more negative than −50 mV and action potential height ≥70 mV measured from spike voltage threshold to peak amplitude were included in the dataset. Although interneurons were avoided during patch-clamp recording through selection of visibly large neurons when retrograde tracer was not used, acceptable motoneurons had membrane capacitance >70 and <300 pF and input resistance <200 MΩ in agreement with values noted from our identified motoneuron recordings. One of two protocols was used to measure membrane capacitance: (1) 15 ms hyperpolarizing voltage commands from −70 to −50 mV in 5 mV steps, and, (2) 10 sweeps of 10 ms voltage command from −70 to −50 mV. In both cases, membrane capacitance was calculated as the average of the integral of capacitive current across the voltage sweeps. Input resistance was determined from the slope of the linear portion of the steady-state I–V relationship from a −60 mV holding potential in current-clamp mode. Voltage responses at the end of 1 s hyperpolarizing current pulses were used to obtain the I–V relationship. Steady-state mean frequencies in response to 1 s current pulses were obtained by calculating the mean frequency of the last 500 ms of the evoked action potential train. Discharge patterns of cells were classified based on membrane response to triangular ramp current injection of 10 s duration and magnitude between 1.5 and 5 times firing threshold current. Instantaneous firing frequencies were calculated to generate f–I curves. The results are reported as mean ± SD, unless stated otherwise. Statistical tests consisted of Student's t test, two-way repeated-measures ANOVA and χ2 proportionality test, and group differences were considered significant if p < 0.05.
Motor unit prediction based on MN membrane properties.
We predicted motor unit type as slow and fast (fatigable or fatigue resistant) based on membrane properties of motoneurons recorded in this study. Previous studies have demonstrated that mechanically typed spinal MUs can be classified into fast (FF, FR) and slow based on membrane properties, such as rheobase and input resistance of the motoneurons with >90% accuracy (Zengel et al., 1985). Because such measurements are commonly made during patch-clamp recordings, we used predictive k-means clustering algorithm (SYSTAT 13) to classify motor unit types based on rheobase, input resistance, and membrane capacitance of motoneurons as categorical variables. The algorithm is based on iterative assignment of a set of observations into k-clusters by maximizing between-cluster variation and minimizing within-cluster variation. Squared Euclidean distance from the cluster mean was chosen as the parameter that determines cluster variations. Initial seeds (k mean values) were not explicitly specified since the observations were made independent of any associable mechanical properties traditionally used for motor unit typing and hence initial cluster means could not be assumed. Significance of classification was noted by the F-ratio (ratio of between group variance and within group variance) for each categorical variable. Larger F-ratio denoted better statistical power of the categorical variable in distinguishing the underlying clusters in the dataset.
Immunohistochemistry and confocal imaging.
AlexaFluor 568 fluorescent dye in the pipette internal solution filled the cells during whole-cell patch-clamp recording. Slices obtained after recording were placed in a 2% paraformaldehyde PBS solution overnight at 4°C, and then permeabilized in 2% Triton X-100 to increase antibody penetration. Motoneuron identity was confirmed by incubating sections in primary goat antibody against ChAT (1:500). Slices were then incubated in Northern Lights 493 donkey anti-goat secondary antibody (1:200; R&D Systems) and mounted on glass slides (Fisher) using aqueous mounting medium (Daido-Sangyo). Slices were then imaged using a confocal microscope (Zeiss LSM 5) attached to an upright microscope (Zeiss, AxioImager) along with Zen 2008 software (Carl Zeiss MicroImaging). Motoneuron identity was confirmed when the cell that was filled with red dye was double-labeled with green ChAT immunolabeling in the nucleus of interest (Fig. 1). For morphology analyses, z-stack images were obtained at high-magnification (63×) with 1 μm depth resolution. Three-dimensional projection of cells were rendered from z-stacks using Volocity 3D Image Analysis software (PerkinElmer) and a region-of-interest was manually established to measure soma surface area excluding dendrites.
Mouse genotype as homozygous WT or heterozygous SOD1G93A was determined using standard PCR technique using the DNeasy 96 Blood and Tissue Kit (Qiagen) according to the manufacturer's protocols. Briefly, frozen samples (mouse tails ∼0.5 cm) were incubated overnight (16–18 h) in 0.2 ml lysis buffer (80% ATL and 20% proteinase K, DNeasy Kit) at 56°C. 200 μl AL buffer and 200 μl 100% ethanol were added to the lysate before loading onto the DNeasy Mini Spin Column and centrifuged at 6000 rpm for 1 min to allow selective DNA binding to DNeasy membrane. This was followed by two efficient wash steps to remove remaining contaminants and enzyme inhibitors. The supernatant was incubated in AE buffer for 1 min and then centrifuged. The collected sample was used for PCR. The PCR mix for each sample consisted of 10 μl 2× TAG PCR Master Mix, four primers, 0.5 μl each for PCR amplification, 6 μl of double-distilled water and 2 μl of DNA from sample. PCR amplification used the following primers (5′ to 3′): for the wild-type allele, CTAGGCCACAGAATTGAAAGATCT and GTAGGTGGAAATTCTAGCATCATCC; and for the mutant allele, CATCAGCCCTAATCCATCTGA and CGCGACTAACAATCAAAGTGA (Integrated DNA Technologies). The reaction consisted of 3 min at 94°C, 35 cycles of 30 s at 95°, 30 s at 60°C, 45 s at 72°C, and 2 min at 72°C. PCR products were separated on a 1.5% agarose gel, allowing resolution of a 324 bp product for the WT allele and a 236 bp product for the mutant allele.
Motor pool simulation model.
Identifying a single motor pool using retrograde tracer injection is technically challenging in neonates. Nonetheless, our retrograde tracer injection targeted musculature visibly located on either sides of the jaw laterally in the mandibular region (thus excluding jaw openers) that labeled one or more JC motor pools: masseter, temporalis, and pterygoid. All the post hoc identified TMNs were in the lateral aspect of the Mot V corresponding to the dorsolateral subdivisions of the trigeminal motor nucleus consisting of JC motor pools (Limwongse and DeSantis, 1977; Matsuda et al., 1978; Uemura-Sumi et al., 1982). Therefore, we assumed a motor pool consisting of JC MNs for our model simulation. Classical studies suggest that based on the muscle fiber type composition, all the jaw closer muscles are considered fast muscles (Taylor et al., 1973). Particularly in the rat (Hiiemae, 1971), as well as in the mouse (Gojo et al., 2002), ∼70% of the fibers types in all of the above three jaw closer muscles are FF, whereas the remaining (∼30%) represent slow-twitch and FR types. Hence, we assumed a motor pool consisting of 100 MNs of which 70 were MNs corresponding to fast MUs and 30 were MNs corresponding to slow MUs. Each MN was represented by three sets of intrinsic properties: (1) the steady-state frequency–current (f–I) gain functions G1 (gain in the primary range) and G2 (gain in the secondary range; Granit et al., 1966), (2) threshold currents for recruitment (Ithresh) and primary-to-secondary range transition current (Itrans), and (3) recruitment frequency (fthresh) and transition frequency from G1 to G2 (fthresh). All of these values were derived from our present data. Table 3 shows the distribution of these properties for the entire motor pool along with the assumed properties of the muscle. The two main sets of MN properties that were explicitly altered to simulate abnormalities in SOD1G93A TMNs include the cells' recruitment threshold (Ithresh), and f–I gain values (G1 and G2). Changes in Ithresh in turn altered Itrans as given by the following equation: (see Results).
The isometric MU force–frequency transformation was performed as by Heckman and Binder (1991) based on the following equation: Fmax is the maximal tetanic force of the MU, and its range across the 100 MUs was based on the range of isometric contraction forces achieved at the optimal jaw muscle length reported in rabbit masseter muscle units (Turkawski and van Eijden, 2000; van Eijden and Turkawski, 2002) due to lack of more specific data from mice. Tf is a frequency constant related to the speed of contraction of the muscle unit and was estimated using the following equation: where a factor that transforms an exponential to sigmoid function. The above MU force–frequency functions were based on the percentages of maximal force output actually generated by motor units at fthresh and ftrans, assuming nearly 8% of Fmax is generated at fthresh and 80% of Fmax is generated at ftrans (i.e., at the end of the primary range; Kernell, 1979). The steady-state input-output function of the simulated motor pool (i.e., isometric force versus effective MN pool input) is a sigmoidal curve where each value on the curve is a summation of all of the MU forces at that input level (Kernell et al., 1983); i.e., where Fi is the isometric MU force obtained above. Note that the properties of muscle fibers were not modeled explicitly and were assumed not to change significantly at this postnatal stage in the mutant mouse. The motor pool model was constructed and simulated using MATLAB (MathWorks).
Prediction of motor unit types and hyperexcitable shifts in SOD1G93A TMNs forming fast motor units
Studies based on SOD1G93A mice have suggested preferential loss of fast motor units before symptom onset in ALS. Although maturational electrophysiological abnormalities have been demonstrated in motoneurons, such changes have not been associated with motor unit type because measuring mechanical properties of motor units to separate fast versus slow populations in young animals is technically challenging. Therefore, using an advanced statistical clustering approach, here we first predicted motor unit types based on motoneuron membrane properties, and then tested the hypothesis that motoneurons forming fast MUs may present excitability changes that are different from motoneurons forming slow MUs. First, we used k-means cluster analyses to predict the major types of motor units: predicted slow (PS), predicted fast fatigue resistant (PFR) and predicted fast fatigable (PFF) using rheobase, input resistance, and membrane capacitance as clustering variables (see Materials and Methods). Figure 2A shows the results of k-means clustering and the normal curves shown indicate the separation of the three MU populations in both WT and SOD1G93A groups. Note that the rheobase and input-resistance of PS, PFR, and PFF TMNs vary in accordance to the expected variation among low-threshold slow MUs and higher threshold fast MUs in both WT and SOD1G93A groups. Similarly, membrane capacitance was on average lowest in the PS TMNs and progressively larger in PFR and PFF cells, respectively. Among the clustering variables, rheobase yielded the highest F-ratios followed by input resistance and membrane capacitance respectively in both WT and SOD1G93A (Fig. 2, legend). Based on this classification, we then compared the membrane properties of WT and SOD1G93A TMNs among the predicted MU types. We observed that both PFF and PFR SOD1G93A TMNs showed hyperexcitable shifts marked by reduced rheobase and increased input resistance, whereas PS TMNs did not present significant alterations in these properties (Fig. 2B). Membrane capacitance that was least effective in MU clustering was not different between WT and SOD1G93A motor units of all types. Interestingly, there was a significant difference in the proportions of MU types in the WT and SOD1G93A groups (Fig. 2C) despite proportionate age distribution in the two animal groups (Fig. 2D). Steady-state frequency–current (f–I) responses for current injections <600 pA showed hyperexcitability in SOD1G93A PFF TMNs (Fig. 3) wherein WT PFF TMNs did not respond during these lower current injections during steady-state. Note that this is consistent with reduced firing threshold due to reduced rheobase and increased input resistance (Fig. 2B; Table 1). However, for larger current injections >600 pA where PFF TMNs of both genotypes responded, no statistically significant differences existed in the steady-state f–I responses between WT and SOD1G93A TMNs (two-way repeated-measures ANOVA; Fig. 3). Similarly, the steady-state f–I responses were not different among WT and SOD1G93A PS and PFR TMNs (two-way repeated-measures ANOVA; Fig. 3B). Interestingly, we noted qualitative differences in the overall pattern of f–I curve distribution among the three predicted MU types in WT versus SOD1G93A recordings. For instance, over the range of injected current tested (Fig. 3B), the WT MU groups presented distinct primary range f–I curves (lower arm of bilinear fits) as normally expected within a motor pool, whereas the SOD1G93A MU group primary range f–I curves overlapped particularly for the PFR and PFF TMNs.
Discharge patterns of TMNs and their altered distribution in SOD1G93A jaw motor pool
The absence of hyperexcitable shifts among SOD1G93A PS TMNs is consistent with our initial hypothesis that MNs forming fast MUs would show selective disease vulnerable modifications, such as hyperexcitability. However, we sought to determine whether more subtle changes in excitability would be revealed if we examined TMN properties according to discharge patterns. To test this possibility, we classified the TMNs based on their discharge properties in response to a somatic injection of triangular ramp current. As described for spinal motoneurons (Bennett et al., 2001a; Button et al., 2006; Cotel et al., 2009), we observed four qualitatively distinct discharge patterns as shown in Figure 4A–D.
Type I neurons show a “linear” relationship in the injected current–frequency response with discharge frequencies matched during both ascending and descending current ramps (Fig. 4A). Type II neurons show rate adaptation during the descending ramp (Fig. 4B). Type III cells show linearity of frequencies during ascending and descending ramps, but with a prominent sustained discharge during the descending ramp that is below the recruitment (or threshold) current (Fig. 4C). Type IV cells also show a sustained discharge during descending ramp accompanied by a counter-clockwise hysteresis in the f–I relationship (Fig. 4D). All four patterns were found among mutant TMNs as well (data not shown). Although in both WT and mutant Mot V, Types III and IV together were the most abundant and the Type II was the least abundant, proportions of cells with these discharge patterns were significantly different in the mutant motor nucleus (χ2 proportion test: p = 0.0007; Fig. 4E).
Compensatory changes in TMN excitability in SOD1G93A mice according to discharge characteristics
Interestingly, based on the classification according to discharge characteristics, membrane properties were altered among Types I and II, but not in Type III and IV cells in the mutant Mot V compared with WT (Fig. 5A). For example, the Type II cells showed a hyperexcitable shift with a significant reduction in rheobase. However, note that such neurons appear to decrease in proportion in the SOD1G93A Mot V (Fig. 4E). In contrast, the Type I linear cells, that were more abundant in the mutant Mot V (Fig. 4E), emerged as a hypoexcitable group with increased rheobase and more depolarized spike voltage threshold (Fig. 5A,B). On average the steady-state f–I response was also lower in mutant Type I TMNs compared with WT but not statistically significant (two-way repeated-measures ANOVA; Fig. 5C). This suggests that such decreases in discharge frequencies shown in Figure 5C are likely a product of increased firing threshold as observed by rheobase and spike threshold changes. Interestingly, both Type I and Type II cells were distributed among PS type TMNs in the WT group, whereas Types III and IV were largely PFR and PFF cells (Fig. 5D). However, in the mutant, the proportional increase in the Type I cells (Fig. 4E) was partly distributed among PFR TMNs suggesting a possible homeostatic mechanism of excitability compensation in the fast MN group through increases in the proportion of hypoexcitable cells in the jaw motor pools. Concomitantly, the SOD1G93A group also showed a decrease in Type III and IV cells that formed the PFR and PFF groups. Overall, PS TMNs were heterogeneous showing all four discharge types and the proportions of cells with specific discharge properties were different in the mutant.
Motor unit prediction in oculomotor nucleus and lack of changes in SOD1G93A oculomotor motoneuron membrane properties
Given that mutant SOD1 is ubiquitously expressed we tested whether the observed changes in MN excitability based on MU type and MN discharge properties would also be present in the ALS-resistant oculomotor motoneurons of SOD1G93A mutants of the same age range (P8–P12). Using the k-means clustering approach as before, we were able to classify two groups possibly corresponding to slow (PS) and fast (PF) MUs (Fig. 6A). Similar to TMNs, the PS OMNs had lower rheobase, higher-input resistance, and lower-membrane capacitance than PF cells in both WT and SOD1G93A animals (also note F-ratios for the three categorical variables in Fig. 6, legend). The membrane properties of neither group were different in the SOD1G93A OMNs compared with WT (Fig. 6B; Table 2). The proportions of PS and PF OMNs were also similar in WT and SOD1G93A mice (Fig. 6C). The distribution of cells within the P8–P12 age group considered is given in Figure 6D. Qualitative differences were noted in the PS and PF f–I curves of WT and SOD1G93A OMNs that may represent altered homeostasis of the underlying motor pools (Fig. 7). Particularly, the WT PS OMNs presented distinct bilinear f–I relationship that did not occur in the mutant PS OMNs over the range of current injections tested and the latter appears to be less excitable at lower values of current injection. Alternatively, the mutant PF OMNs appear to be more excitable than WT PF OMNs in a somewhat compensating manner (an example in Fig. 7A). However, there was no significant change in the steady-state f–I responses between WT and mutant OMNs among either predicted MU groups (two-way repeated-measures ANOVA). In summary, OMNs could be clustered into two distinct PS and PF motor unit types and neither groups presented significant alterations in membrane properties in the mutant mice compared with WT.
Discharge patterns of OMNs and lack of changes in the SOD1G93A oculomotor nucleus
An examination of the discharge characteristics of OMNs in response to triangular ramp current injections revealed three types of discharge patterns (Fig. 8A–C). Type I corresponds to cells showing a linear f–I relationship on ascending and descending current ramps (Fig. 8A). Type II corresponds to “rate adapting” cells that show frequency adaptation during the descending ramp (Fig. 8B). Type III corresponds to the “sustained linear” pattern with an extended frequency range during the descending ramp below the firing threshold of the ascending ramp phase (Fig. 8C). All three patterns were found among SOD1G93A OMNs as well. Moreover, there was no significant redistribution of the discharge types in the SOD1G93A Mot III compared with WT (Fig. 8D) and we did not observe any subtype specific excitability changes. Additionally, the distribution of the discharge types was similar among PS and PF MU types (Fig. 8E) among WT and SOD1G93A OMNs.
Jaw closer motor pool simulations and predicted force development
In summary, brainstem motoneurons including the trigeminal and oculomotor neurons presented here do not show hyperexcitability in the form of increased f–I gain at P8–P12. However, rheobase and the proportions of cells showing different discharge characteristics show complex changes in the ALS vulnerable TMNs but not in the ALS-resistant OMNs. Specifically, TMNs predicted to form fast MUs show reduced rheobase and PS TMNs with linear discharge patterns show increased rheobase and are increased in proportions. If such opposite changes were to occur in the input thresholds of slow and fast MNs in a jaw motor pool, we hypothesized that there could be alterations in MN recruitment and muscle force development early on. To test such a possibility and to predict changes in the muscle force development, we used a simple motor pool model of a jaw closer motor pool.
The simulation model was constructed based on similar models for spinal motor pools (Heckman and Binder, 1991) that allowed explicit parameterization of recruitment thresholds and steady-state f–I gains. The model enabled simulation of muscle isometric forces due to changes in TMN recruitment thresholds based on observed changes in rheobase values. For simplicity, we only represent slow and fast MNs and combine FR and FF as fast MNs to simulate SOD1G93A changes in recruitment threshold values and further assume no changes in muscle properties in the mutant at this presymptomatic age.
A jaw closer masseter pool consisting of 100 TMNs was constructed to estimate normal (WT) and altered (SOD1G93A) isometric forces. Briefly, each model MN was defined using six experimentally derived parameters including threshold recruitment current (Ithresh) and frequency (fthresh), steady-state primary (G1) and secondary (G2) gains, primary to secondary transition current (Itrans), and frequency (ftrans) derived from our experimental data (see Materials and Methods for detailed definitions of terms). The MN and motor unit properties for every fifth MN are listed in Table 3.
Based on the observed f–I gain and rheobase values in WT and mutant TMNs, we simulated three scenarios as follows:
Simulation of standard motor pool behavior
(1) For the standard condition, the primary f–I gains were set as G1 = 0.1 Hz/pA for all 100 neurons. The secondary f–I gain G2 was set at 0.06 Hz/pA for the first 30 neurons representing the slow MUs, and at 0.05 Hz/pA for the remaining 70 neurons representing the fast MUs. These values approximate our experimental values. The resulting f–I curves for every 10th neuron are shown in Figure 9A (left). To generate Ithresh values for individual MNs, first we constructed a line with a certain slope and offset that resulted in a range of 100–950 pA from the first to the 100th neuron in the pool. This provided a ∼10-fold range of Ithresh across the pool reflecting the rheobase range in our data. Each MN in the pool was assigned Ithresh value along this line with evenly spaced increase in Ithresh between consecutive model neurons (Fig. 9A, middle). The corresponding control isometric muscle force is shown in Figure 9A (right).
Simulation of recruitment threshold changes in SOD1G93A TMNs
(2) We modeled the observed aberration in MN recruitment threshold of fast and slow SOD1G93A TMNs in two steps. First, we compressed the Ithresh range of only the fast MNs to represent the observed reduced threshold for recruitment. We did so, by altering the Ithresh–MN line under standard conditions to a bilinear distribution of Ithresh. The Ithresh values of slow MUs remained unaltered, whereas those of the fast MNs were now sampled along a line with reduced slope resulting in a compression of Ithresh range across the fast MUs as shown by the red trace in Figure 9B (left). The resulting change in the muscle isometric force is shown by the red trace in Figure 9B (right). Note that reduced recruitment thresholds of large TMNs could profoundly enhance the steepness of the sigmoidal force curve. Finally, to model the concomitant Ithresh increases in slow MUs, we increased the slope of the lower arm of the Ithresh–MN line that resulted in an increase in the sampled Ithresh values for slow MU neurons (Fig. 9B, left, blue trace). Note that such reorganization of Ithresh values further simulates an increased proportion of low-threshold smaller neurons. This modification resulted in partial compensation of the isometric force as shown in Figure 9B (right, blue trace).
Simulation of f–I gain changes
(3) Because increased f–I gain has been noted in other ALS-vulnerable brainstem and spinal MNs we tested the potential effect of changes in f–I gains. We simulated opposite changes in f–I gains of fast and slow MU TMNs by increasing the gains of fast MU neurons and decreasing the gains of slow MU neurons as in Figure 9C (left). The corresponding isometric force is shown in Figure 9C (right). Note that the f–I gain changes produce a negligible effect on force development (compare green trace with the black control trace in the figure).
In Figure 9D, we show a comparison between the force deviations due to Ithresh and f–I gain changes detailed above. Note from the blue and green curves in Figure 9D that hypoexcitability in Ithresh of slow MU TMNs can suppress the initiation of force development. However, as input levels increase and larger forces develop, note the hypoexcitability in Ithresh of low-threshold TMNs compensates for the hyperexcitibility of Ithresh in fast MU TMNs. Moreover f–I gain changes considered here resulted in negligible force deviation from standard condition. Finally, when both f–I gain changes and Ithresh changes were simultaneously simulated the resulting isometric force shows a reduction in the dynamic range of the motor pool input–output relationship (Fig. 9E). The distortion in the isometric force curve (Fig. 9E, orange) compared with control curve (Fig. 9E, black) highlights a change in the equilibrium of the motor pool function.
A significant finding of this study shows that ALS-vulnerable SOD1G93A MNs exhibit more complex changes in excitability than previously reported. Specifically, SOD1G93A TMNs predicted to form fast MUs show a hyperexcitable shift in rheobase. However, a surprising observation was that a subset of predicted slow MU TMNs in SOD1G93A cells with distinct discharge properties increased in proportions and showed a hypoexcitable shift in rheobase and voltage threshold for spike initiation. Such excitability changes were absent in the disease-resistant OMNs. The observed changes in SOD1G93A TMNs are expected to alter motor unit recruitment and jaw muscle force development. To test this hypothesis, we constructed a simplified JC motor pool model using computer simulations. Model results suggest that hyperexcitable alterations in recruitment threshold of fast motor units would result in a steeper input–output relationship of the motor pool, suggesting more rapid muscle force development. Furthermore, such rapid force development can be partly compensated by the observed hypoexcitable shifts and increased slow motor unit composition in the mutants. Together these changes could reduce the dynamic range of motor pool input-output function. Moreover, near low levels of pool inputs, hypoexcitability of slow TMNs can suppress force initiation.
Motor unit prediction and distribution of electrophysiological properties
Using k-means clustering approach we were able to successfully identify three clusters potentially corresponding to slow, fast fatigue resistant and fast fatigable TMNs and two clusters potentially corresponding to slow and fast OMNs. The F-ratio indicative of the power of the clustering variable was maximal for rheobase followed by input resistance and capacitance in TMNs consistent with the greater accuracy of MU prediction using rheobase as a discriminant variable in previous studies on spinal MNs (Fleshman et al., 1981; Zengel et al., 1985; but see Gardiner, 1993). A similar trend in F-ratio was also noted in OMN clustering, however, the input resistance was less powerful in MU prediction compared with TMNs.
Although the high-threshold MNs are unambiguously α-motoneurons that innervate the extrafusal muscles, some low-threshold MNs could be gamma MNs that innervate the muscle spindles. Parameters that may distinguish gamma MNs from α MNs, such as higher antidromic stimulation threshold (in vivo) or lack of monosynaptic responses during afferent stimulation (Sessle, 1977) are both not feasible and definitive methods to use in the neonatal in vitro brain slice preparation. Although we have previously demonstrated stimulation techniques in brain slices to study the physiology and pharmacology of synaptic inputs from the sensory mesencephalic nucleus to TMNs (Trueblood et al., 1996; Del Negro and Chandler, 1998), many of these connections are lost during brain-slice preparation and cannot be used as a reliable approach to test the lack of afferent connections to segregate gamma TMNs. Furthermore, gamma TMNs are spontaneously active in vivo (Sessle, 1977) and we did not encounter such MNs in our in vitro dataset.
Assumptions and limitations of motor pool simulation model
First, we assume that our TMN dataset predominantly represents the JC motor pool(s) neurons because the TMNs filled with the red dye (AlexaFluor 568) during patch-clamp recording were identified post hoc in the dorsolateral Mot V where JC MNs are localized (Mizuno et al., 1975). Second, we use membrane properties including rheobase and input resistance that are suggested to predict MU types with >90% accuracy (Zengel et al., 1985). We further directly extrapolate the MN recruitment threshold and f–I gains found in this dataset into a single JC motor pool model. Nonetheless, due to the simplified representation of each model MN based on experimentally derived properties (see Materials and Methods, and Results), the insight from the model simulations that rheobase change would have a profound effect on steady-state motor pool input–output relationship should be reasonable. However, more detailed MN models are essential to examine the effect of alterations in MN discharge patterns on motor pool behavior. Last, we assume no aberrations in the muscle fiber properties in the mutant compared with WT that requires empirical validation.
Discharge patterns of TMNs and OMNs
Similar to spinal MNs (Bennett et al., 2001a; Button et al., 2006; Turkin et al., 2010), we found that the brainstem TMNs and OMNs also displayed heterogeneous discharge patterns in response to triangular ramp current injections. However, the OMNs were somewhat less heterogeneous. Specifically, all four discharge patterns described for spinal MNs were present among TMNs, whereas only three types of patterns were present among OMNs. In comparison with discharge patterns observed in TMNs, Type I OMNs were qualitatively similar. However, Type II OMNs showed lesser degree of rate adaptation than TMNs. For example, for the Type II OMN shown in Figure 7B, the firing frequencies during descending ramp gradually merge with the ascending frequencies, thus abolishing the rate adaptation near recruitment frequency. Among Type III OMNs often an initial phase of rate adaptation was observed during the descending current ramp followed by a crossover of frequencies resulting in sustained firing. Such frequency crossover was not observed in TMNs. Type IV cells with a sustained hysteretic pattern were not observed among OMNs. Interestingly there was a significant change in the proportion of cells displaying the different discharge patterns in TMNs compared with OMNs in the SOD1G93A animals. This suggests an altered homeostasis in the excitability of the vulnerable Mot V.
Effect of development on MN membrane properties
Based on in vitro studies, membrane properties of motoneurons change during the postnatal period (for review, see Carrascal et al., 2005). Moreover, among brainstem motoneurons, some membrane properties show opposite changes between different brainstem motor nuclei. For example, rheobase is suggested to decrease with postnatal age in oculomotor neurons (Carrascal et al., 2006), whereas it increases with postnatal age in hypoglossal neurons (Núñez-Abades et al., 1993). We show that the two types of postnatal age-matched motoneurons (TMNs and OMNs) present fundamental differences in the membrane properties in the mutant compared with WT. The cells were proportionately distributed across ages considered (P8–P12) in both genotypes for TMNs (Fig. 2D). For OMNs, we noted a slight bias toward P8 among WT and toward P12 in SOD1G93A. Although this was a random occurrence, we further note that this did not lead to manifestation of age-related differences between WT and SOD1G93A groups because we did not observe significant changes in any of the measured membrane properties between the two groups among either of the predicted MU types. Moreover, developmental changes in OMN membrane properties are more dramatic at P0–P5, whereas such changes are not significant between P6–P10 and P11–P15 (Carrascal et al., 2006). Together, our results on ALS-vulnerable TMNs clearly demonstrate motor unit specific modifications in the mutant motoneurons, whereas the ALS-resistant mutant OMNs do not present such changes in the postnatal age range tested.
Possible mechanisms underlying MN excitability changes
Spinal MNs with sustained discharge patterns showed large persistent inward sodium and calcium currents (PICs) activated by supraspinal monoaminergic inputs. These PICs are modulated following spinal trauma and in ALS (Li and Bennett, 2003; Quinlan et al., 2011). Alternatively, MNs with linear discharge patterns are likely to have negligible PICs (Bennett et al., 2001a; Li et al., 2004). The fact that these two types of discharge patterns showed opposite changes in proportion and excitability in TMNs suggests that PICs play an integral role in the homeostatic regulation of MN and motor pool excitability (Elbasiouny et al., 2010; Martin et al., 2013). Recently we reported evidence for tonic GABAergic currents that can modulate TMN excitability and their downregulation in SOD1G93A TMNs (Venugopal and Chandler, 2012). Such reductions in tonic inhibition may underlie the reduced firing threshold in large TMNs. Alterations in potassium (Wainger et al., 2014) and persistent sodium conductance (Kuo et al., 2005; Bellingham, 2013) may also play a role. Concomitant opposite changes in large versus small TMN excitability highlights the existence of mechanisms coupling the excitability of different MNs within a motor pool. Noncell autonomous changes in astrocytes could play a role in such coupling (Bataveljić et al., 2012).
Role for intrinsic MN excitability changes in ALS pathogenesis
Motoneuron hyperexcitability has been widely reported in ALS patients (Vucic and Kiernan, 2006) and in animal and cell culture models (Pieri et al., 2003; Kuo et al., 2004; Durand et al., 2006; van Zundert et al., 2008; Quinlan et al., 2011; Wainger et al., 2014). Presymptomatically, hyperexcitability could make MNs hyper-responsive to glutamatergic inputs in turn leading to excessive intracellular calcium causing ER stress and triggering apoptotic cascade through mitochondrial overload (Grosskreutz et al., 2010; van Zundert et al., 2012). Therefore, factors reducing hyperexcitability are considered potential therapeutic targets. However, if we consider our results, it is clear that hypoexcitability concomitantly occurs in a distinct population of low-threshold TMNs with distinct discharge properties compared with the high-threshold cells that show hyperexcitable shifts. Within a homogeneous motor pool, hypoexcitability in low-threshold motoneurons can restrict/suppress muscle activation as shown by the model. Indeed, inhibiting MN excitability presymptomatically has been shown to induce pathological cellular responses and enhanced compensatory responses, whereas enhancing excitability was neuroprotective (Saxena et al., 2013).Together with the additional findings that hyperexcitability is not observed in adult mutant mice (Delestrée et al., 2014), it appears that a change toward hypoexcitability may indeed represent the disease process, whereas early hyperexcitability may represent a compensation. Therefore, identifying the distinct cellular and molecular properties of cells displaying reduced excitability in addition to hyperexcitability would serve as opportunistic targets for investigation into the underlying mechanisms contributing to disease progression.
This work was supported by Grant NS071348 to S.H.C. We thank Dr Dwayne Simmons' laboratory for the confocal imaging facility, and Kelly Lucchesi for assisting with mouse genotyping.
The authors declare no competing financial interests.
- Correspondence should be addressed to Dr Scott H. Chandler, UCLA, 2859 Slichter Hall, 603 Charles East Young Drive East, Los Angeles, CA 90095-7246.