Functionally Reduced Sensorimotor Connections Form with Normal Specificity Despite Abnormal Muscle Spindle Development: The Role of Spindle-Derived Neurotrophin 3

Inhibition of muscle spindle induction by the conditional elimination of ErbB2

We first sought to investigate the dependence of the central connectivity of muscle spindle afferents on the muscle spindle. Because neuronal induction of muscle spindle development depends on Nrg1 signaling (Hippenmeyer et al., 2002) through the activation of ErbB2 receptors on primary myocytes (Andrechek et al., 2002; Leu et al., 2003), we generated mice in which ErbB2 expression was eliminated from intrafusal muscle precursors using two targeted alleles of the ErbB2 gene [a null (ErbB2^NULL) (Lee et al., 1995) and a conditional (ErbB2^FLOX) (Crone et al., 2003) allele] and the deleter line (myf5^CRE) (Tallquist et al., 2000) in which the Cre recombinase is expressed in myoblast precursors of intrafusal muscle before the arrival of sensory axons at embryonic day 15 (E15). At approximately postnatal day 15, the ErbB2^NULL/FLOX/myf5^CRE mutant (referred to as ErbB2Δ in the subsequent text and figures) develops a motor phenotype characterized by a flexed posture when inverted and a wide-based, ataxic gait, similar to previously generated mutants that lack muscle ErbB2 (Andrechek et al., 2002; Leu et al., 2003). The control animals ErbB2^FLOX/FLOX and ErbB2^NULL/FLOX (no Cre; referred to collectively as ErbB2 control) demonstrated no motor abnormality.

When ErbB2 expression is selectively eliminated from muscle, mutant spindles form that are characterized by a single immature intrafusal fiber and that these rudimentary spindles fail to progress beyond E18 and never develop into mature muscle spindles (Leu et al., 2003). To characterize the molecular development of muscle spindles in the ErbB2Δ mutant, we used in situ hybridization analysis of P1, P3, and P7 animals with probes specific for the transcription factors Egr3 (Tourtellotte and Milbrandt, 1998) and Er81 (Arber et al., 2000), two molecular markers of muscle spindle differentiation that are first expressed at E15.5 and E18.5, respectively (Hippenmeyer et al., 2002)

Muscle spindle counts were done on TA muscles isolated from ErbB2Δ mutant animals, as well as ErbB2^FLOX/FLOX and ErbB2 ^NULL/FLOX controls. At postnatal day 3 in the ErbB2 control animals, there were 31.0 ± 1.9 (n = 5) muscle spindles per TA muscle, and these spindles were characterized by the strong expression of Egr3 in multiple intrafusal muscle fibers (Fig. 1A). In contrast, TA muscles in the ErbB2Δ mutant expressed Egr3 weakly in only 9.6 ± 3.8 (n = 5) rudimentary muscle spindles that comprised only a single Egr3⁺ myofiber (Fig. 1D). Conditional elimination of ErbB2 by myf5^CRE therefore results in a 70% loss in the total number of nascent muscle spindles as determined by the expression of Egr3. Moreover, because normal muscle spindles each contain up to four intrafusal fibers (Tourtellotte et al., 2001), there is a loss of >90% of the total number of intrafusal fibers in the ErbB2Δ mutant.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Late muscle spindle marker expression is decreased in ErbB2Δ mutants. In situ hybridization analysis on serial sections of P3 tibialis anterior muscle from an ErbB2Δ mutant (D–F) ErbB2^NULL/FLOX (no CRE) control (A–C) animal with probes for the transcription factors Egr3 (A, D), Er81 (B, E), and neurotrophic factor NT3 (C, F) reveals low levels of Egr3 expression in a single myofiber of the mutant muscle spindle (D); Egr3 is robustly expressed in multiple intrafusal fibers in the control (A). In situ analysis also demonstrates the complete absence of late spindle markers Er81 and NT3 in mutant muscle spindles (E, F) compared with the control (B, C). Immunohistochemical analysis by whole-mount staining of P4 gluteus maximus muscle with antibodies against PV and VGluT1 in ErbB2^FLOX/FLOX (no CRE) control (G) and ErbB2Δ mutant (H) animals demonstrates the relatively simple morphology of the muscle spindle afferent terminal around a single, rudimentary intrafusal fiber in the mutant. I, Quantitative in situ hybridization studies of the TA muscle of the ErbB2Δ mutant demonstrate induction of Egr3 expression (black) in ∼30% (9.6 ± 3.8 SEM; n = 5) of the number of spindles observed in ErbB2 (ErbB2^NULL/FLOX or ErbB2^FLOX/FLOX) controls (31.0 ± 1.9; n = 5). In contrast, no significant Er81 expression (gray) was observed in P3 mutant muscle. Statistical significance: *p < 0.001, ANOVA (Fisher's least significant difference method); there was no statistically significant difference between the number of muscle spindles expressing Egr3 and Er81 in the ErbB2 controls.

Additional in situ hybridization analysis with probes for Egr3 and Er81 on contiguous sections of TA muscle found no evidence of Er81 expression in any of the Egr3⁺ myofibers (Fig. 1E), in contrast to the ErbB2 controls in which Er81 is coexpressed in Egr3⁺ spindles (Fig. 1B). In the TA muscle at P3, there is a near absence of Er81⁺ muscle spindles in the ErbB2Δ mutant (0.6 ± 0.2; n = 5) compared with the ErbB2 (27.5 ± 1.9; n = 4) controls (Fig. 1I). Similar results were obtained at postnatal days 1 and 7 (data not shown).

We also examined the expression of another molecular marker of muscle spindles, the neurotrophic factor NT3 that is initially expressed diffusely in embryonic muscle but becomes restricted to the intrafusal fibers late in embryogenesis (Copray and Brouwer, 1994; Wright et al., 1997). In situ analysis of NT3 expression revealed no detectable NT3 message at P3 in the ErbB2Δ mutant (Fig. 1F) in contrast to the ErbB2 control (Fig. 1C).

These data demonstrate that, in the ErbB2Δ mutants, there is widespread failure of muscle spindle induction, although the inhibition is not complete. In approximately one-third of the normal number of muscle spindles, induction initiates at E15, as demonstrated by the limited expression of Egr3, but is aborted before the expression of Er81 and NT3 at E18.

We next assessed the development of proprioceptive afferent terminals on muscle spindles in the ErbB2Δ mutant by immunostaining with antibodies for the calcium binding protein PV, a selective marker for proprioceptive sensory neurons (Honda, 1995; Arber et al., 2000), and VGluT1 that labels the proprioceptive afferent terminals on the muscle spindle (Wu et al., 2004). Whole-mount immunostaining of the P4 gluteus maximus muscle was used to visualize the morphology of afferent terminals on the muscle spindle. Few of the rudimentary spindles remaining in the ErbB2Δ mutant (Fig. 1H) showed the characteristic, complex annulospiral afferent terminal normally found around the equatorial region of each intrafusal fiber (Fig. 1G). Moreover, the VGluT1 staining normally associated with the spindle afferent terminal was diminished around the ErbB2Δ mutant muscle spindle compared with the ErbB2^FLOX/FLOX and wild-type controls (Fig. 1G,H) (WT data not shown). In comparison, the morphology of the Golgi tendon organ, a muscle mechanoreceptor innervated by class I_B afferents, was unaffected by the elimination of muscle ErbB2 (data not shown), consistent with the selective elimination of muscle spindles in animals lacking neuronal Nrg1 (Hippenmeyer et al., 2002).

The widespread failure of muscle spindle differentiation in the ErbB2Δ mutant had no obvious effect on muscle spindle afferent survival, as expected from the analysis of the neuronal Nrg1 mutant (Hippenmeyer et al., 2002), and similar mutants in which ErbB2 expression is eliminated from developing muscle (Leu et al., 2003). Anterograde labeling of sensory projections to the spinal cord from the L5 dorsal root (Fig. 2A,D) together with parvalbumin immunostaining (Fig. 2B,E) demonstrated that muscle spindle afferents projected to the ventral spinal cord in their characteristic pattern (P4 through P7).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

The gross pattern of central projections is normal in ErbB2Δ mutant mice. A–F, The pattern of central projections of proprioceptive afferents in the L5 spinal cord of the ErbB2Δ mutant is similar to that of the ErbB2 ^FLOX/FLOX control at postnatal day 6. The images are z-stack projections of ∼10 scans of 2 μm thickness. A, B, D, E, The L5 dorsal root was filled with Texas Red Dextran (red) to reveal all afferent projections; motor neurons (MN) were retrogradely labeled from the L5 ventral root with Cascade Blue Dextran (blue). Proprioceptive afferents are selectively labeled by PV immunostaining (green). C, F, The spatial distribution of immunostaining for the VGluT1 is normal in the ErbB2Δ mutants. G–L, Proprioceptive afferent terminals, triple-labeled with Texas Red Dextran, PV, and VGlut1on L5 motor neurons in the ErbB2^FLOX/FLOX control (G–I) and ErbB2Δ (J–L) mutant. The insets in the top corners are enlargements of the yellow boxes; synaptic contacts are indicated by the yellow arrows.

Immunostaining for VGluT1, which is enriched in the central axonal terminals of postnatal DRG sensory neurons (Oliveira et al., 2003), was also observed throughout the mutant spinal cord in a pattern indistinguishable from the control animals (Fig. 2C,F). To determine whether proprioceptive afferents make direct contacts on motor neurons in the mutant animals, we visualized presynaptic afferent terminals by the colocalization of Texas Red Dextran (from anterograde L5 dorsal root fills), PV, and VGluT1 (Fig. 2G–I, J–L, respectively) on motor neurons labeled retrogradely from the ventral root. By this method, we observed triple-labeled L5 afferent terminals in contact with L5 motor neurons in the ErbB2Δ mutants throughout the neonatal period, to as late as P8.

We extended our histochemical analysis of these mutants to P21 to establish whether the afferent synaptic contacts we detected at P6 persisted as the animal matured. For this purpose, we counted the number of VGluT1⁺ contacts on the cell bodies of large-diameter motor neurons in the L4 lateral motor column in the mutant and wild-type controls (Fig. 3). In this analysis, three-dimensional reconstructions of motor neuron somata were first obtained from z-series of confocal optical images, from which estimates of individual cell surface area were calculated to determine approximate VGluT1⁺ contact density per soma. Small-diameter motor neurons <10,000 μm² were excluded from analysis, and there was no significant difference between the average surface area of mutant (13,263 ± 372 μm²; n = 20) and control (12,188 ± 433 μm²; n = 9) motor neurons (p = 0.1, t test). On average, there were 18.5 ± 1.15 (n = 20) VGluT1⁺ contacts on motor neurons in the ErbB2Δ mutant compared with 26 ± 2.35 (n = 9) in the control, corresponding to a contact density of 0.14 ± 0.008 and 0.20 ± 0.025 contacts per 100 μm² surface area on mutant and control motor neurons, respectively. Although the reduction in the number of VGluT1 did not reach significance, the density of VGluT1 contacts was reduced by ∼30% in the ErbB2Δ mutant (Fig. 3C,E). These data demonstrate that, in the absence of normal spindle development, there is only a modest reduction in the normal density of VGluT1⁺ contacts on motor neurons, which led us to examine whether these anatomical contacts between muscle afferents and motor neurons are functional in the ErbB2Δ muscle spindle mutants.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

At P21, VGluT1⁺ contacts on motor neurons are reduced in the ErbB2Δ mice. A–D, Single optical plane (step, 0.6 μm) (A, C) and projections of ∼10 optical planes (single optical thickness, 0.6 μm) (B, D) of two L4 motor neurons from a P21 wild-type control (A,B) and P21 mutant (C,D). Motor neurons were identified by ChAT staining (blue); VGluT1 contacts are shown in red. E, Quantitative analysis of VGluT1⁺ contacts per motor neuron soma and corresponding densities in the mutant and wild-type control demonstrate no significant (NS) loss of VGluT1 contacts on motor neuron somata in the mutant and a modest but significant (p < 0.05) decrement of ∼30% in the density of VGluT1⁺ terminals. No significant difference was observed in the size of motor neurons analyzed in the mutant versus control.

Electrophysiological characterization of monosynaptic connections between muscle-identified primary afferents and motor neurons

To record afferent-evoked responses from muscle-identified motor neurons, we visually targeted motor neurons prelabeled 2–3 d earlier by retrograde transport of a tracer injected into the TA or the LG muscles. During the physiological experiment, the recorded motor neuron was injected with a different dye. After processing the tissue for immunohistochemistry, we confirmed that the recorded cells were labeled with both the intracellular and the retrogradely transported marker (Fig. 4(also see Fig. 6A,B, bottom left insets). By prelabeling TA and LG motor pools in vivo, we were able to cut the ventral roots and thereby avoid using antidromic stimulation of a muscle nerve to determine the identity of motor neurons (Seebach and Ziskind-Conhaim, 1994; Mears and Frank 1997). Muscle–nerve stimulation results in an antidromic action potential in the homonymous motor neuron and the disynaptic recurrent IPSP (Mentis et al., 2005), which would have complicated detection of the afferent-evoked EPSPs. Moreover, monosynaptic inputs from other motor neurons (Mentis et al., 2005; Nishimaru et al., 2005) would have been impossible to distinguish from monosynaptic EPSPs generated by stimulation of muscle afferents. Cutting the ventral roots made it possible to restrict the effects of muscle nerve stimulation to those attributable solely to the activation of afferents.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

The isolated spinal cord preparation and identification of motor neurons. A, Shows a drawing of the lumbar spinal cord dissected together with the sciatic nerve and its major peripheral nerves. Recording and stimulating electrodes were applied to the L4 and L5 lumbar ventral roots (vr-L4 and vr-L5), en passant on the L4 and L5 dorsal root ganglia (DRG-L4 and DRG-L5), and to the following nerves: the deep peroneal nerve supplying the tibialis anterior (TA nerve) and extensor digitorum longus muscles, the superficial peroneal nerve supplying the peroneus longus and brevis (SP nerve), the first muscular branch of the tibial nerve supplying the lateral head of gastrocnemius (LG nerve), soleus and plantaris muscle and the third muscular branch of the tibial nerve supplying the flexor hallucis longus, tibialis posterior, and flexor digitorum longus and the lateral and medial plantar nerves (tibial nerve). In some experiments, an extracellular glass electrode was used to record afferent-evoked field potentials. B, C, Images of TA motor neurons (green) retrogradely labeled in vivo at P0 using cholera toxin B subunit conjugated with Alexa 488 together with a microelectrode filled with Alexa 543 hydrazide (red) cell attached to a TA motor neuron (B) and after rupture of the membrane and intracellular dye-filling (C). D, A two-photon image of lateral gastrocnemius motor neurons retrogradely labeled with cholera toxin B subunit-conjugated Alexa 647 (blue) and showing one cell filled with the intracellular dye after 15 min of whole-cell recording (pink because of red and blue colocalization). E, an antidromically evoked action potential recorded from a wild-type LG motor neuron and evoked by stimulation of the L4 ventral root.

We examined the synaptic potentials evoked by afferent stimulation in nine motor neurons in nine wild-type animals (P4 and P5) and in eight motor neurons in eight ErbB2Δ mutant animals (P4 and P5). To establish that the afferent-evoked synaptic potentials recorded in wild-type and mutant animals were monosynaptic, we compared the timing of the afferent-evoked extracellular field potentials recorded within the motor nucleus with the onset of the homonymous ventral root potential and the intracellularly recorded afferent-evoked synaptic potential (Fig. 5). As has been shown previously in the mouse (Li and Burke, 2001), the afferent-evoked field potential comprises two main components. The first is the terminal potential representing orthodromic invasion of muscle afferent fibers within the motor nucleus (Fig. 5A, red trace, field potential) (Lev-Tov and Pinco, 1992; Li and Burke, 2001), and the second component is the synaptic field potential. The synaptic field potential was identified by comparing the field potentials in the presence and absence of low calcium to block synaptic transmission (Fig. 5A). These recordings revealed that the onset of the synaptic field potential coincided with the onset of the electrotonically recorded ventral root potential (Fig. 5A, right panel, double-headed arrow). In three experiments, this allowed us to approximate the synaptic delay as the time between the onsets of the terminal and synaptic field potentials (mean of 0.85 ms). However, this is probably an overestimate of the delay because extracellular recordings may not be sensitive enough to detect the earliest synaptic events in motor neurons. In accord with this interpretation, we found that the latency of the intracellularly recorded EPSP generated by stimulation of the homonymous muscle nerve was 0.9 ± 0.4 ms (range, −1.9 to 0.6 ms) less than that of the onset of the ventral root potential (Fig. 5C,D), consistent with the monosynaptic origin of the earliest part of the EPSP.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Temporal relationships between afferent-evoked field potentials, ventral root potentials, and intracellularly recorded EPSPs from muscle-identified motor neurons. A, Simultaneous recording of the extracellular field potentials (Field potential; top) recorded within the motor nucleus at the L4/5 border and the L5 ventral root potential (vr-L5; bottom) in control aCSF (blue traces) and in a low (0.1 mm) calcium aCSF (red traces) to block chemical synaptic transmission. The recordings were made in a wild-type P5 mouse. The traces on the right are shown at an expanded time base indicated by the dotted box on the left panel. The field potential recorded in low-calcium aCSF is the terminal potential (labeled as “Terminal field potential onset” and is marked by red arrowed dotted line) and is generated by action potentials invading primary afferent arbors within the motor nucleus. The synaptic field potential starts where the field potential recorded in normal aCSF deviates from the low calcium trace (labeled as “Synaptic field potential onset” and marked by black arrow). Note that the onset of the synaptic field potential coincides with the onset of the electrotonically recorded ventral root potential (labeled as “Ventral root potential onset” and marked by dotted line with arrowhead). The field potential traces are averages from 250 individual responses (10 Hz at supramaximal stimulus intensity) so that the synaptic field potential amplitude is depressed. The ventral root potential is not averaged. The response to the first stimulus is shown for each condition. The synaptic delay is approximated as the time from the onset of the terminal field (red arrow) to the onset of the synaptic field (black arrow). B, Field potential and ventral root potentials recorded in a P5 ErbB2Δ preparation. The color code is the same as in A. Note that there was no difference in the synaptic delay between wild-type and ErbB2Δ mice. C, Comparison of the timing of the synaptic potential recorded in a tibialis anterior motor neuron (TA mn) evoked by stimulation of the TA muscle nerve (5× T at 0.1 Hz) with the simultaneously recorded L4 ventral root potential (vr-L4) and the afferent volley recorded from the L4 dorsal root ganglion (DRG-L4) in a P4 wild-type mouse. Each recording has been averaged from five traces. Insets above the top (TA mn) and middle (vr-L4) traces show the responses at a slower timescale. Note that, in this cell, the onset of the homonymous synaptic potential precedes the onset of the ventral root potential consistent with its origin monosynaptically. The red arrow under the DRG-L4 trace identifies the timing of the earliest part of the volley that was used to calculate the conduction velocity of the primary afferent axons. D, Graph showing the difference in time between the onset of intracellularly recorded EPSPs from wild type TA and LG motor neurons and the ventral potentials generated by stimulation of the homonymous muscle nerve or the DRG (DRG-L4 for TA and DRG-L5 for LG). The stimulation intensity was 5× T at 0.1 Hz. The onset times were calculated from averages of three to five sweeps. The black symbols show the average and SEM for the data. E, Graph showing the conduction velocities of the primary afferents calculated from the latency from stimulation of the peripheral nerve to the volley recorded in the DRG (Nerve-DRG) or the latencies from stimulation of the peripheral nerve to the onset of the EPSP recorded in the motor neuron (Nerve-MN). There were no statistical differences in the conduction velocities between the wild type (n = 5; blue) and ErbB2Δ (n = 5; red) mice (t test; p = 0.47 for Nerve-DRG; p = 0.38 for Nerve-MN).

As an additional control, because ErbB2 is also involved in the differentiation and growth of Schwann cells that could be affected in the ErbB2Δ mutant (for review, see Britsch, 2007), we measured the conduction velocity of the primary afferents activated from the peripheral nerves and found that, in wild-type mice (n = 4), the average conduction velocity from the nerve to the DRG was 1.45 ± 0.06 m/s and from the nerve to the recorded motor neuron was 1.26 ± 0.04 m/s (n = 5). Similarly for the ErbB2Δ mice, the average conduction velocity from the muscle nerve to the DRG was 1.82 ± 0.14 m/s (n = 4) and from the muscle nerve to motor neuron was 1.44 ± 0.19 m/s (n = 5) (Fig. 5E). There was no statistical difference between wild-type and ErbB2Δ mice (t test, p > 0.50).

Monosynaptic inputs from muscle spindle afferents to motor neurons are functional but reduced in amplitude in the ErbB2Δ mutant

Figure 6 illustrates the method used to determine the pool identity of two LG motor neurons from a wild-type (Fig. 6A) and an ErbB2Δ (Fig. 6B) mouse. In each case, we used visual targeting to record from cells labeled with the retrograde tracer CTb–647 and then injected the cell with Alexa 543 hydrazide (Fig. 6A,B, red cells in main panel). After fixing the spinal cords and processing the tissue for immunohistochemistry, we confirmed that both motor neurons were double labeled with the intracellular and the retrograde dyes (Fig. 6A,B, yellow arrows in the small insets). This approach also allowed us to determine whether VGluT1 terminals were also present on the dendrites of the recorded motor neurons. Although we demonstrated the presence of VGluT1⁺ terminals on motor neuron somata in both wild-type and ErbB2Δ animals, in adult animals it is known that most of the muscle afferent input onto motor neurons is on the dendrites (Burke and Glenn, 1996). As shown in the large insets in Figure 6, we confirmed the presence of VGluT1⁺ terminals on motor neuron dendrites for both wild-type and ErbB2Δ animals.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Monosynaptic connections between primary afferents and motor neurons are preserved but reduced in strength in ErbB2Δ mice. A, B, Images of a wild-type (A) and an ErbB2Δ (B) LG motor neuron recorded from and subsequently filled with Alexa 543 hydrazide (shown in red). The images are z-stack projections of either 40 (A) or 44 (B) scans of 1.9 μm optical thickness. The identification of the cell as an LG motor neuron was established by double labeling with the retrograde tracer CTb–647 (blue) as illustrated in the two insets on the bottom left of each image. Note that the double-labeled cells identified by the arrows are pink (blue and red), indicating colocalization, whereas the dye-filled cells shown in the main part of the image are red. Immunohistochemistry against VGluT1 revealed several VGluT1⁺ terminals apposed to the motor neuron dendrites. These can be seen in the large insets in both figures (white arrows), which show an expanded region off the dendrites indicated by the dotted yellow box. Identification of VGluT1⁺ terminal appositions was made on individual optical sections. Each image is a z-stack projection of 18 sections each 0.92 μm thick. D, Dorsal; V, ventral; L, lateral; M, medial. C, Intracellularly recorded EPSPs together with simultaneously recorded ventral root potential in response to stimulation of the LG nerve (5× T, average of 5 stimuli at 0.1 Hz) in wild-type and ErbB2Δ mice (P5). The amplitude of the monosynaptic response was measured for all records at 3 ms (indicated) after the onset of the EPSP or the ventral root potential (see Results). The black arrowheads indicate the stimulus artifact. Note that both the EPSP and the monosynaptic component of the ventral root potential are smaller in the ErbB2Δ motor neurons compared with WT (indicated by the red arrows). D, Bar graphs showing the average amplitude of the ventral root potentials (left bars) and the intracellularly recorded EPSPS (right bars) in wild-type (blue) and ErbB2Δ (red) animals in response to stimulation of the sites indicated below the plot. DR, Dorsal root; TA, tibialis anterior nerve; LG, lateral gastrocnemius nerve. The average intracellular response of the monosynaptic EPSP is shown for five wild-type (blue) and six ErbB2Δ (red) LG and TA motor neurons (per, peripheral nerve). Statistical significance: *p < 0.05, **p < 0.025, t test.

We measured the amplitude of the monosynaptic response in both wild-type and ErbB2Δ mice 3 ms after the onset of the EPSP. We chose this time based on two observations. First, the difference in the latency of the homonymous and antagonist EPSPs was ∼5–6 ms for wild-type and ErbB2Δ animals (see below). Second, we sometimes observed an inflection on the rising phase of the EPSP (supplemental Fig. 1A, available at www.jneurosci.org as supplemental material) that occurred at 4.3 ± 0.4 ms after the EPSP onset for three wild-type motor neurons (one LG and two TA) and 4.5 ± 0.4 ms for three ErbB2Δ motor neurons (two LG and one TA; 2× or 5× T stimulation). It seems reasonable to suppose that the inflection coincides with activation of a disynaptic component to the afferent-evoked EPSP. Consistent with this interpretation, we found that the delayed component exhibited greater synaptic depression during a stimulus train of 10 Hz than did the initial component of the EPSP. Because of the rapid depression that occurs during such a stimulus train (supplemental Fig. 4, available at www.jneurosci.org as supplemental material), we quantified the depression by expressing the amplitude of the second response in the stimulus train as a percentage of the first response. In two wild-type motor neurons, the initial component of the response (measured at 3 ms after the EPSP onset) fell to 70.5% of the amplitude of the first response in the stimulus train. The later peak (measured at 6–6.5 ms after the EPSP onset) fell to 54% of the amplitude of the first response in the stimulus train. Similarly for three ErbB2Δ motor neurons, the early response (3 ms) fell to 63 ± 3.6% of the amplitude of the first response in the train, whereas the later response (6–6.5 ms) fell to 48.3 ± 2.4% (t test, p < 0.05). The latencies of the inflections (4.3 ms wild type; 4.5 ms ErbB2Δ) are similar to those of the Ia inhibitory disynaptic IPSP in neonatal (P7) mouse motor neurons (Wang et al., 2008). These authors showed that the onset of the reciprocal Ia IPSP occurred 3.5 ms after the onset of the homonymous EPSP in P7 mouse motor neurons.

Our method for measuring the amplitude of the monosynaptic component of the afferent-evoked potential differs from that used by Frank and his coworkers (for a description of the method, see Mears and Frank, 1997) who used a “model” PSP to measure the amplitude of the monosynaptic potentials. In supplemental Figure 1B (available at www.jneurosci.org as supplemental material), we compare the two methods and show that they produce very similar estimates for the amplitude of the monosynaptic EPSP, with the latency method used here producing a slightly lower estimate of the amplitude than the “model” method.

We found that the amplitude of monosynaptic potentials was reduced by ∼80% in the ErbB2Δ animals compared with wild-type animals (Fig. 6C,D). The average amplitude of five homonymous EPSPs in wild type animals (three LG and two TA motor neurons) was 5.5 ± 2.4 mV, and for six cells in the ErbB2Δ animals (four LG and two TA motor neurons), it was 1.2 ± 0.5 mV. The difference was statistically significant (p < 0.05, Mann–Whitney test). As with the intracellular EPSPs, the measurements of extracellularly recorded ventral root potentials were made at 3 ms after the onset of the potential (or at the peak if it occurred within 3 ms of the stimulus) to ensure that only the monosynaptic component was measured. Stimulation of the dorsal root resulted in a ventral root potential whose amplitude was 5.0 ± 0.7 mV for the wild-type and 1.3 ± 0.2 mV in the ErbB2Δ mice. This difference in the amplitude of the afferent-evoked ventral root potentials was also statistically significant (p < 0.05, Mann–Whitney test). Similarly, the ventral root potentials generated in response to stimulation of the muscle nerves were also significantly smaller in the ErbB2Δ animals (p < 0.05). In the wild-type animals, stimulation of the LG or TA nerves generated average ventral root potentials of 0.89 ± 0.28 or 0.94 ± 0.19 mV, respectively. For the ErbB2Δ animals, these values were 0.19 ± 0.03 and 0.14 ± 0.07 mV, respectively, a reduction of 80–85%.

The differences in the amplitude of the EPSPs in wild-type and ErbB2Δ mice are not attributable to differences in the input resistance of the motor neurons. The input resistance measured in four LG wild-type motor neurons was 37.1 ± 7.3 MΩ, which was not statistically different (t test) from the values measured in five LG ErbB2Δ motor neurons (28.1 ± 9.5 MΩ). In addition, five TA wild-type motor neurons had a mean of input resistance of 36.7 ± 7.3 MΩ, similar to the value obtained from three TA ErbB2Δ motor neurons (38.8 ± 9.5 MΩ).

The reduction in the amplitude of the afferent-evoked synaptic potentials in the ErbB2Δ animals cannot be explained by an inability to stimulate afferent axons in the peripheral nerve, either because their electrical thresholds were too high or because they had retracted within the muscle nerve beyond the point of stimulation. This is because the monosynaptic EPSPs evoked by dorsal root stimulation (which would have stimulated all afferent fibers) were reduced by a similar amount (26% of wild-type amplitude) to those evoked by muscle nerve stimulation (22% of wild-type amplitude).

The effects of conditional elimination of spindle-derived NT3 on the amplitude of monosynaptic ventral root potentials evoked by muscle afferent stimulation

As described above, the elimination of ErbB2 expression in muscle precursors results in the arrest of muscle spindle development before E18 and a variety of molecular and morphological defects in spindle development. One of these is the loss of NT3 expression from muscle spindles (Fig. 1C,F), which previous studies suggest accounts for the loss of the monosynaptic sensorimotor responses observed in the Erg3 knock-out mouse in which muscle spindle development is abnormal, and spindles do not express NT3 (Chen et al., 2002). To determine genetically whether the attenuation of the afferent-evoked monosynaptic EPSPs in the ErbB2Δ animals was attributable to the loss of spindle-derived NT3 rather than some other abnormality of spindle development, we generated a novel Egr3^CRE allele and used it in combination with a floxed allele of NT3 (Bates et al., 1999) to selectively eliminate the expression of NT3 from the intrafusal fibers of muscle spindles. The Egr3^CRE allele was designed to preserve the expression of endogenous Egr3 by placing an IRES–CRE expression sequence downstream of the last coding exon of the Egr3 gene (supplemental Fig. 2, available at www.jneurosci.org as supplemental material). Animals homozygous for both the NT3^FLOX and Egr3^CRE alleles were generated in a predominantly C57BL/6J genetic background, and serial sections of neonatal muscle were analyzed by in situ hybridization for the expression of the early spindle markers Egr3, Er81, and NT3. At P1, spindles of the wild-type (data not shown) and NT3^FLOX/FLOX (no CRE) control animals expressed each of these late spindle markers; however, in the NT3^FLOX/FLOX/Egr3^CRE/CRE mutants (referred to as NT3Δ in the subsequent text and figures), NT3 was selectively absent (Fig. 7). Similar analysis at P3 demonstrates that NT3 expression is not simply delayed (data not shown). The NT3Δ mutants demonstrated the same ataxic gait and flexure posturing as the ErbB2Δ mutants, consistent with a deficit in proprioceptive signaling, although the motor phenotype observed was delayed in onset and less pronounced than that observed in the ErbB2Δ animals.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Selective elimination of NT3 expression in NT3Δ mutants does not result in reduced afferent-evoked monosynaptic ventral root potentials at P5. In situ hybridization analysis of serial sections of P1 (A–F) hindlimb from NT3^FLOX/FLOX/Egr3^CRE/CRE mutant (D–F), NT3^FLOX/FLOX (no CRE) control (A–C) animals with probes for the transcription factors Eg3 (A, D), Er81 (B, E), and NT3 (C, F) reveals selective elimination of NT3 from mutant muscle spindles. G, Averaged extracellular recordings from the L5 ventral root after stimulation of the L5 dorsal root at 5× T from three groups of animals (1 control and 2 mutants) at P5. The frequency of stimulation was 0.1 Hz. Note that the amplitude of the monosynaptic component the ventral root potential (arrows) did not differ significantly between the controls (NT3^FLOX/FLOX; no Cre) and either of the two mutants (NT3^FLOX/FLOX/Egr3^CRE/CRE or NT3^FLOX/FLOX/myf5^CRE). H, Graph showing the average response from three preparations from each group of animals. There was no significant difference between the groups (p = 0.10, ANOVA). I, Recordings from the same classes of animal as in G but at P14. Note that the monosynaptic reflex was greatly reduced in the mutants compared with the controls. J, Graph showing the averaged amplitude of the monosynaptic response from three animals per group. There was a significant reduction in the monosynaptic reflex for both mutants compared with that of the control group (*p < 0.01, ANOVA Fisher's test).

We compared the amplitude of ventral root potentials at P5 in NT3Δ animals and wild-type controls, and surprisingly the amplitude of the monosynaptic response was normal in NT3Δ mutants at P5, as illustrated in Figure 7G. However, at P14, by which age the NT3^FLOX/FLOX/Egr3^CRE/CRE animals develop a proprioceptive phenotype, the amplitude of the potentials was significantly reduced (Fig. 7I). At P5, the average amplitude of the potentials measured was 8.3 ± 1.7 mV for the NT3Δ animals (n = 3) and 6.4 ± 1.1 mV for the controls; none of these values differed statistically (p = 0.10, ANOVA). In contrast, at P15, the amplitude of the potentials measured was 0.35 ± 0.03 mV for the NT3Δ animals (n = 3) and 1.98 ± 0.44 mV for the controls (n = 3), which represents a reduction of >80%.

Because Egr3^CRE and NT3 are also expressed in Schwann cells and the peripheral vasculature (supplemental Fig. 2C,D, available at www.jneurosci.org as supplemental material) (Scarisbrick et al., 1993; Jessen and Mirsky, 2005), the effect on monosynaptic potentials we observe at P14 in the NT3Δ mice could be attributable to the loss of NT3 from these other, nonspindle sources. To address this possibility, we compared the amplitude of ventral root potentials at P5 and P14 in NT3Δ animals to those in which NT3 was selectively eliminated from all (intrafusal and extrafusal) muscle fibers using the same myf5^CRE allele used in the generation of the ErbB2Δ mutants. As illustrated in Figure 7, G and H, the amplitude of the monosynaptic response was also normal at P5 in NT3^FLOX/FLOX/myf5^CRE mutants despite the absence of NT3 expression in intrafusal muscle demonstrated by in situ hybridization analysis (data not shown). However, at P14, when the NT3^FLOX/FLOX/myf5^CRE animals exhibit a motor phenotype similar to the NT3Δ and ErbB2Δ animals, the amplitude of the potentials was significantly reduced (Fig. 7I,J). At P5, the amplitude of the potentials measured was 5.1 ± 0.3 mV for the NT3^FLOX/FLOX/myf5^CRE animals (n = 4) compared with 6.4 ± 1.1 mV for the controls. None of these values differed statistically (p = 0.10, ANOVA). In contrast, at P15, the amplitude of the potentials measured was 0.47 ± 0.14 mV for the NT3^FLOX/FLOX/myf5^CRE animals (n = 3) and 1.98 ± 0.44 mV for the controls (n = 3), a reduction of >75%, comparable with the ErbB2Δ and NT3Δ mutants.

These data demonstrate that spindle-derived NT3 has a late role in the maintenance of functional muscle spindle afferent connections on motor neurons and that the regulation of these monosynaptic inputs by muscle spindles in the first postnatal week does not depend on NT3 expression.

The functional specificity of connections between afferents and motor neurons is preserved in ErbB2Δ animals

In normal mice, muscle spindle afferents project monosynaptically to homonymous and some synergist motor neurons but do not make direct connections with functionally antagonist motor neurons (Mears and Frank, 1997). To examine whether this specific pattern of connectivity was preserved in the ErbB2Δ, mice we compared the sensorimotor connections between the functionally antagonist LG and TA motor neurons. Figure 8, A and B, illustrates EPSPs evoked in wild-type and ErbB2Δ LG motor neurons by stimulation of homonymous and antagonist muscle nerves. As shown in the figure, the latency of the intracellularly recorded EPSP evoked by antagonist nerve stimulation was at least 5 ms longer than that evoked by homonymous stimulation.

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

Muscle-identified motor neurons retain specific inputs from primary afferents in ErbB2Δ mutant mice. A, Comparison of the potentials evoked by homonymous (blue trace, LG) and antagonist (red trace, TA) muscle nerve stimulation at 5 T intensity averaged from five stimuli applied once every 10 s. The blue and red arrows indicate the onset of the EPSPs. B, Recordings of homonymous and antagonist evoked synaptic potentials in an ErbB2Δ LG motor neuron. The conditions and stimulation parameters were the same as for the wild-type records in A. C, D, Bar graph showing the average latency of the synaptic potential in evoked in LG (C) and TA (D) motor neurons by stimulation of different nerves for wild-type (blue bars) and ErbB2Δ (red bars) mice. The black bars show the latency of the EPSP after simulation of the DRG. Error bars are SEM. E, The left panels show five superimposed intracellular responses from a wild-type LG motor neuron after stimulation of the homonymous LG nerve. Two different frequencies of stimulation are illustrated (0.1 Hz, top; and 10 Hz, bottom). Five sequential color-coded trials are illustrated. Note that the onset of the EPSP after stimulation of the homonymous nerve results in very little variation (“jitter”) in the latency at the two different frequencies (as indicated by the dotted line). In contrast, stimulation of the antagonist nerve (right panels) results in more variation in the latency particularly at 10 Hz. The color-coded arrows denote the onset of the EPSP in each trial. F, Intracellular responses from an LG ErbB2Δ motor neuron under identical testing. Note that, at 10 Hz, both WT and ErbB2Δ motor neurons exhibited reduced amplitude in EPSPs as expected, presumably attributable to neurotransmitter depletion. G, H, Bar graphs for wild-type (G) and ErbB2Δ (H) motor neurons comparing the average latency jitter of the homonymous and antagonist nerve-evoked EPSPs. There was no significant difference in the latency-jitter EPSP after the homonymous nerve stimulation for either wild-type or ErbB2Δ motor neurons at all stimulation frequencies. In contrast, stimulation of the antagonist nerve resulted in significantly more jitter at stimulation frequencies of 1 and 10 Hz (*p < 0.05, **p < 0.01, ANOVA Fisher's test).

The latency of the homonymously evoked EPSP was 9.4 ± 0.6 ms (range, 8.4–10.4 ms; n = 3) for wild-type LG motor neurons and 11.4 ± 0.7 ms (range, 9.9–12.8 ms; n = 4) for wild type TA motor neurons (Fig. 8C,D). When the antagonist nerves were stimulated (TA nerve and recordings made in LG motor neurons; LG with recordings made in TA motor neurons), the latencies were significantly longer (p < 0.05, t test). For TA to LG stimulation, the mean latency of the EPSP was 15.9 ± 2.3 ms (range, 12.5–20.2 ms; n = 3) and for LG to TA, it was 17.2 ± 0.6 ms (range, 16.5–20.4 ms; n = 3). Thus, in LG motor neurons, the latency of the EPSP evoked by antagonist nerve stimulation was 6.5 ms longer than that evoked by homonymous nerve stimulation, and, for TA motor neurons, this difference was 5.8 ms.

It has been shown recently that stimulation of muscle afferents activates a disynaptic inhibitory pathway in neonatal mouse motor neurons (Wang et al., 2008). We did not, however, observe hyperpolarizing responses in antagonist motor neurons after stimulation of antagonist muscle nerves. The responses below the baseline in the bottom panels of Figure 8, E and F (antagonist stimulation), are misleading because the traces have all been adjusted to a common baseline. During 10 Hz stimulation, the potentials summate so that individual responses are superimposed on this tonic depolarization. We do not consider the depolarizing potentials shown in Figure 8 (A,B, and E,F) to be inverted IPSPs. The chloride equilibrium potential in neonatal motor neurons of this age has been reported to be close to the resting membrane potential or slightly hyperpolarizing (Delpy et al., 2008). Moreover, using whole-cell recording, the equilibrium potential will be more negative than normal because of cellular dialysis by the low chloride electrode solution. The depolarizing potentials activated by antagonist nerve stimulation most likely reflect activation of disynaptic excitatory pathway activated in parallel to the Ia inhibitory pathway.

The same measurements performed on ErbB2Δ animals revealed latencies of 8.6 ± 0.2 ms (range, 8.3–9.1 ms; n = 3) for homonymous stimulation of LG motor neurons and 10.2 ± 0.48 (range, 9.2–10.8 ms; n = 3) for homonymous stimulation of TA motor neurons. When the antagonist nerves were stimulated, the latencies were 14.2 ± 1.3 (range, 11.3–20.2 ms; n = 6) for LG motor neurons and 15.3 ± 0.5 (range, 11.1–23.4 ms; n = 6 for TA motor neurons). None of these latencies were statistically different from the wild-type animals. For the ErbB2Δ animals, the difference in homonymous and antagonist latency was 5.6 ms for LG motor neurons and 5.1 ms for TA motor neurons, indicating that monosynaptic inputs from muscle afferents are functional in the ErbB2Δ animals, and, for the TA and LG motor neurons, the specificity of the connections is preserved.

Stimulation of other nerves, including the SP or another muscular branch of the tibial nerve, also resulted in synaptic potentials in LG wild-type motor neurons whose latencies were significantly longer (p < 0.05) than those obtained with homonymous activation (tibial to LG, 14.9 ± 1.6, range of 12.3–17.9 ms, n = 3; SP to LG, 18.2 ± 1.5 ms, range of 15.3–20.3 ms, n = 3). Similarly, stimulation of the SP nerve in ErbB2Δ mice resulted in EPSP latencies of 15.8 ± 1.7 ms (range, 14.1–19.2 ms; n = 3). These results are consistent with the idea that homonymous nerve stimulation results in monosynaptic activation of the motor neurons and antagonist stimulation results in disynaptic or multisynaptic activation in wild-type animals.

Comparison of the latency jitter of homonymous and antagonist evoked EPSPs to different frequencies of afferent stimulation

We also examined the behavior of the homonymous and antagonist evoked potentials to stimulation at different frequencies (0.1, 1, and 10 Hz) to confirm that the homonymously evoked EPSPs were produced monosynaptically, whereas those evoked by antagonist nerve stimulation were not. Previous studies have shown that high-frequency stimulation of the monosynaptic potential results in a progressive decline in the amplitude of the response (Fig. 8E,F, bottom panels). However, multisynaptic potentials, including those generated by antagonist nerve stimulation, do not show a monotonic decrease in amplitude but tend to fail in an all-or-none manner as the intervening interneurons fail to reach threshold (Lee and O'Donovan, 1991; Lev-Tov and Pinco, 1992; Li and Burke, 2001). As a result, the latency of disynaptic and polysynaptic responses will show much more variability as the stimulus frequency is increased than monosynaptic responses. This is illustrated in Figure 8, E and F, which compares the EPSPs generated by homonymous and antagonist nerve stimulation at different frequencies of stimulation for wild-type (Fig. 8E) and ErbB2Δ (Fig. 8F) preparations. It can be seen from the figure that stimulation of the homonymous nerve results in a small latency jitter (the difference between the minimum and maximum latency) as stimulation frequency increases. However, the jitter is substantially greater for the antagonist nerve stimulation at the different frequencies. For five wild-type motor neurons, the jitter for the homonymous stimulus at the different frequencies (0.1, 1, and 10 Hz) was 0.4 ± 0.1, 0.7 ± 0.2, and 1.1 ± 0.1 ms, respectively. For antagonist stimulation, these values were 0.8 ± 0.2, 12.8 ± 8.5, and 20.4 ± 8.6 ms, respectively. The jitter was significantly different (p < 0.05) between homonymous and antagonist stimulation at 1 and 10 Hz (Fig. 8G). This clearly demonstrates a major difference in the behavior of the homonymous and antagonist EPSPs. Similar findings were made for homonymous and antagonist stimulation in ErbB2Δ mice (Fig. 8H). For homonymous stimulation, the latency jitter was 0.8 ± 0.1, 1.0 ± 0.3, and 1.1 ± 0.3 ms in six motor neurons for stimulation at the three different frequencies. For antagonist stimulation, these values were 1.9 ± 1.2, 5.6 ± 3, and 15.6 ± 5.3 ms, respectively, in six motor neurons. The difference in latency jitter between homonymous and antagonist stimulation was significant (p < 0.05) at 1 and 10 Hz. There was no difference in the jitter at the different frequencies between the wild-type and ErbB2Δ animals (ANOVA on ranks).

The anatomical specificity of the connections between muscle afferents projecting to agonist and antagonist motor neurons is preserved in ErbB2Δ animals

Our data demonstrate that the specificity of functional connections between muscle afferents and motor neurons is preserved in the ErbB2Δ mutant. However, because the monosynaptic EPSP in these animals is so dramatically diminished, it is possible that aberrant projections are formed in the mutant but are too weak to detect physiologically. To address this possibility, we examined the anatomical specificity of the projections in six P14 mice (three wild type mice; three ErbB2Δ mice) by labeling the afferents from the TA and the LG muscle using CTb conjugated to either Alexa 488 (for TA) or Alexa 647 (for LG; see Materials and Methods). To assess the specificity of the projection, we counted the number of VGluT1⁺ boutons that were labeled with either tracer on muscle identified motor neurons. As illustrated in Figure 9, the tracer labeling is punctate so that axons and dendrites cannot be followed in continuity (supplemental Fig. 3, available at www.jneurosci.org as supplemental material). Therefore, it was necessary to label motor neurons using Texas Red Dextran, which fills the cytoplasm of the soma and proximal 50–100 μm of the dendrites. This was essential to distinguish between the cytoplasmic CTb and that from afferent terminals located on the dendritic surface (Fig. 9E,F). Although we occasionally observed VGlut1⁺ boutons labeled with both CTb tracers, the overwhelming majority of VGlut1⁺ boutons were labeled with a single tracer.

• Download figure
• Open in new tab
• Download powerpoint

Figure 9.

Motor neurons receive primary afferent boutons from homonymous but not from antagonistic muscle afferents in control and ErbB2Δ mice. A–D, Confocal images from the same transverse section of a hemisected wild-type spinal cord in the L4/L5 region showing retrograde labeling of TA motor neurons with CTb488 (A, in green), LG motor neurons labeled with CTb–647 (B, in white), motor neurons (mns) retrogradely labeled with Texas Red Dextran (C, in red), and VGluT1 immunoreactivity (D, in blue). The images in the E series are from a wild-type LG motor neuron. E1, A z-stack projection (total scan depth, 9.1 μm) of the motor neuron retrogradely labeled with Texas Red Dextran. The insets show the soma of the motor neuron double labeled with Texas Red Dextran and CTb–647 (red and white) and labeled with CTb–647 alone (white), confirming the identity of the motor neuron as LG. The dotted yellow rectangle is shown at higher magnification with various combinations of fluorochromes in the E2–E4 series. Each image is a single optical scan (0.69 μm thickness). E2 shows VGluT1 immunoreactivity (blue) and the Texas Red Dextran-labeled dendrite of the LG motor neuron. The circle identifies a putative primary afferent bouton labeled with VGluT1, which is also labeled with CTb–647 (E3) as seen in the merged image (E5), indicating that it is a bouton from an afferent originating in the LG muscle. E4, No VGut1⁺ boutons were labeled with the tracer (CTb488, green) from the antagonist muscle (TA). The series of images F1–F5 are from an LG ErbB2 motor neuron and were acquired in the same way as those of the images shown in E1–E5.

To quantify the number of muscle identified synaptic (VGlut1⁺) boutons on motor neuron dendrites, we examined three TA and three LG motor neurons that were labeled with a single CTb and also with Texas Red Dextran. To quantify the number of labeled afferent terminals, we chose dendrites on which at least one VGluT1⁺ bouton also labeled with a single CTb (488 for TA or 647 for LG motor neurons) (Fig. 9E,F). We first counted the number of VGlut1⁺ boutons apposed to the motor neuron dendrite over 51–117 μm (six motor neurons in three mice). There were 5.6 ± 0.9 boutons per 50 μm of dendritic length for three LG control motor neurons and 7.2 ± 0.6 boutons per 50 μm of dendritic length for three TA motor neurons. Because there was no statistical difference between number of TA and LG motor neurons (p = 0.24, t test), we pooled the data to give 6.4 ± 0.6 boutons per 50 μm of dendritic length for the wild-type mice. Approximately 15 ± 2.3% of these boutons were also labeled with one of the CTb tracers (Fig. 9E). In every case, the tracer was the same as that labeling the motor neuron. We found no examples of inappropriate projections (e.g., TA boutons on an LG motor neuron or vice versa).

We made similar observations in the ErbB2Δ mutant motor neurons. We analyzed seven LG motor neurons with dendrites spanning 39–107 μm. The frequency of VGluT1⁺ boutons on dendrites was 5.5 ± 0.9 per 50 μm of dendritic length, and, of these, 25.8 ± 5.3% were labeled with the appropriate CTb for the motor neuron (Fig. 9F). As with the wild-type animals, we found no instances of inappropriate labeling. There was no statistical difference in the proportion of double-labeled VGluT1 boutons between the wild-type and ErbB2Δ mice (p = 0.11, t test).

In two of six wild-type motor neurons and in four of seven ErbB2Δ motor neurons, we observed VGlut1⁺ boutons labeled with the appropriate tracer on the somata. As with the dendrites, we found no inappropriately labeled boutons on the motor neuron somata.