Ephrin-B2 expression in the proprioceptive sensory system

Abstract

Recent studies have shown that ephrin-B2 on sensory afferent fibers from the dorsal root ganglia (DRG) controls transmission of pain sensation to the spinal cord. We examined ephrin-B2 expression in mouse DRG and spinal cord using an ephrin-B2/beta-galactosidase chimeric allele. We found that ephrin-B2 is expressed exclusively on proprioceptive neurons and fibers in neonates, while expression in lamina III and IV of the adult spinal cord was observed in addition to that in the deeper laminae. We confirmed that ephrin-B2 protein causes co-clustering of EphB2 receptor and glutamate receptors in spinal cord neurons. Our data are consistent with a role for ephrin-B2 in transmission of positional information to the CNS, and thus suggest a role in the synaptic plasticity of spinal cord locomotor circuits that are known to be sensitive to proprioceptive sensory input after spinal cord injury.

Introduction

Spinal cord injury (SCI) severs the spinal cord locomotor circuitry from its supraspinal input and forms an environment that is refractory to axonal regrowth [10, 31, 38]. Much of SCI research focuses on ways to promote regeneration of severed axons so as to restore voluntary control. Recent studies using genetically modified mouse lines have demonstrated that significant long tract regeneration can be accomplished by targeting specific genes [28]. However, accomplishing such regeneration in a therapeutic setting is difficult [36].

Locomotion in mammals is coordinated by a complex set of neuronal circuits called the central pattern generator (CPG) that remain intact below the level of SCI. The CPG can produce stepping even when isolated from the brain by SCI. This stepping is exquisitely sensitive to sensory input from proprioceptive afferents, and plasticity of synapses from these afferents onto the CPG is associated with training-induced recovery of locomotion after SCI. Because of this, spinalized cats and rodents and humans with complete injuries can be trained to perform complicated weight bearing and stepping tasks independent of voluntary control [5, 14, 29, 33]. Understanding the connectivity of the sensory system to the spinal cord is therefore necessary to exploit any voluntary control that remains after incomplete SCI or that may be regained through regenerative therapies.

Synapses in the CNS form from highly mobile filopodia that extend from growing dendrites. Once they make connections with their target axons, they grow mature spines, from which boutons form with post-synaptic densities (PSD) to match the pre-synaptic ones across the synaptic cleft. The neurotransmitter of choice in the CNS is glutamate, and is detected in the PSD by N-methyl-D-aspartate receptors (NMDAR) or α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPAR) [24]. Part of the stabilization process of the synapse is the clustering of these receptors, and Eph/ephrin signaling is an integral part of this process [7, 23].

Eph receptor tyrosine kinases are classified as A or B based on phylogenetic similarity and their affinities for the glycosylphosphatidyl inositol-linked ephrin-As or the transmembrane ephrin-Bs, although some members bind promiscuously across classes. Binding to Ephs can also initiate “reverse signaling” into ephrin-bearing cells. Both forward and reverse signaling are intimately involved with axon guidance and synaptic plasticity in the brain[24, 27].

There is mounting evidence that forward signaling also plays a critical role in synaptic plasticity between the sensory system and its spinal cord targets. Battaglia, et al. reported that treatment of rat spinal cords with ephrin/Fc soluble protein induced thermal hyperalgesia and mechanical allodynia in rats and that administration of an EphB1/Fc protein to block ephrin signaling in the spinal cord alleviated this hyper-sensitivity [1]. These and other authors’ data indicate that ephrin-Bs on presynaptic sensory afferents, especially ephrin-B2 (EB2), interact with EphBs on spinal cord neurons in the dorsal horns to cause clustering and activation of NMDARs and thus hypersensitivity to painful stimuli. A recent study by Han, et al. indicates that EphB1 may be the post-synaptic mediator of this effect, as EphB1 knockout mice in their hands demonstrated a reduced response to neuropathic pain compared to wild type [18]. And Zhao, et al. documented deficits in inflammatory and neuropathic (but not acute) pain responses in a nociceptor-specific EB2 conditional knockout [40].

We examined EB2 expression in the spinal cord using a β-galactosidase (β-gal) indicator mouse line, and found expression in fibers that innervated deeper laminae corresponding to mechanoreceptive and proprioceptive input. Our findings suggest that ephrin-B2 may play a role in sensory input to the CPG, and thus may modulate locomotion in a previously undescribed manner.

Materials and Methods

Mouse strains

Ephrin-B2/LacZ (EB2LZ) mice encoding a chimeric protein with a β-gal moiety replacing the intracellular segment of ephrin-B2 (EB2), are described in [13]. EphB2/LacZ mice, carrying a LacZ cDNA in place of the intracellular domain of EphB2 are described in [22].

Mouse surgeries

Adult EB2LZ mice were anesthetized and their left dorsal roots at lumbar levels four and five were cut and ligated as described in [32]. The mice were sacrificed 15 days later, and perfused with 4% formaldehyde.

DiI tracing

Lumbar DRGs still attached to the spinal cords of newborn EB2LZ homozygotes were fixed and injected with 1,1′-dioctadecyl-3,3,3′3′-tetramethylindocarbocyanine perchlorate (DiI). After one week in 4C, the tissues were embedded in agarose and sectioned serially at 50 μm on a vibratome. Sections were examined immediately under fluorescence microscopy with a Texas Red filter. Digital images were recorded with a Cool Pix digital camera. Wild type littermate controls were processed alongside homozygotes and their afferent patterns compared.

Immunofluorescence and β-galactosidase stains

Tissues perfused with 4% formaldehyde were equilibrated in 30% sucrose and cryosectioned at 12 mm. Sections were stained with X-gal to detect β-gal activity [2] or immunolabeled with antibodies against β-gal (Millipore AB986) and parvalbumin (Millipore MAB1572). Myelinated and unmyelinated afferent fibers to the spinal cord dorsal horn were detected with antibodies against NF200 (Sigma N5389) and CGRP (Sigma C7113), respectively. Primary spinal cord neurons were incubated with anti-EphB2 (R&D Systems) and anti-NR1 (Cell Signaling). Primary antibodies were detected with donkey anti-goat, goat anti-mouse, or goat anti-rabbit Cy2 or Cy3 conjugated secondary antibodies from Jackson ImmunoResearch.

In situ hybridization

EB2 cDNA from RT-PCR of mouse brain RNA was subcloned into pBluescriptSK(−) (Stratagene) and used to create DIG-labeled cRNA probes by in vitro transcription [2].

Primary neuron cell cultures

DRGs from e14.5 EB2LZ +/− mouse embryos were isolated as previously described [25] and grown in Neurobasal medium supplemented with B27 and Nerve Growth Factor (NGF, 10 ng/ml), Brain-Derived Neurotrophic Factor (BDNF, 2 ng/ml), or Neurotrophin-3 (NT-3, 2 ng/ml). After 48h in culture, cells were fixed and stained with X-gal. Adult spinal cord neurons were isolated as previously described [2] and grown for 48h in the presence of 4 μg/ml EB2/Fc protein preclustered with anti-Fc antibody. Area and number analyses were done with ImageJ software.

Results

Expression of EB2 in proprioceptive afferents

We examined EB2LZ mice in which a β-gal moiety replaces the intracellular domain of the ephrin. β-gal expression in this mouse gives a faithful representation of the EB2 expression pattern [6, 11, 12, 34]. We used this mouse rather than anti-EB2 immunocytochemistry because the available EB2 antibodies are known to be non-specific in that they cross-react heavily with EB1 and may also cross-react with other ephrins depending on experimental conditions (manufacturer’s literature and author’s personal observations). X-gal staining of adult EB2LZ spinal cord revealed a sweeping pattern from the dorsal horns down to the inner laminae on each side of the central canal, and to a lesser extant into the ventral horns (Fig. 1). The same pattern was observed at cervical, thoracic, and lumbar levels. Although cell bodies in the deeper laminae stained positive, three pieces of evidence indicated that this staining was primarily in the afferent fibers emanating from the DRG rather than from cells of the spinal cord. First, EB2LZ staining was evident in neurons and axons in the DRG (Fig 1D). Second, in situ hybridization of spinal cord sections showed little EB2 mRNA signal except in a few cells around the central canal, consistent with the large blue cells seen in the same area in X-gal stains (Fig 1A, C, F). Third, we performed unilateral dorsal rhizotomies on EB2LZ mice, and examined X-gal staining in the spinal cords seven days later. Staining was evident on the uncut side, but much reduced on the cut side (Fig. 2). These data are in agreement with published findings of EB2 on sensory afferents, but identify different targets within the spinal cord [1, 26, 37, 40].

Figure 1. EB2LZ is expression in spinal cord and afferent fibers from the DRG.

(A–C) transverse sections of EB2LZ/+ spinal cords stained with X-gal. A is 40x. B and C are 200x magnification of the boxes in A. (D) DRG at 100x. (E) histogram of the sizes of 206 DRG neurons from three EB2LZ/+ mice. X-gal-positive (blue) cells are closed bars; X-gal-negative cells are open bars. 18% of cells were blue. Avg. size=1056 ± 77 μm² for blue vs. 433 ± 20 μm² for red; p<0.0001 by Student’s t-test). (F) EB2 in situ hybridization of spinal cord at 40x.

Figure 2. Loss of EB2LZ fibers after dorsal rhizotomy.

X-gal stain of EB2LZ spinal cord with L4 dorsal root transection (right side) 7 days before sacrifice.

Four pieces of evidence in our hands strongly suggest that EB2 is on those fibers that carry proprioceptive input. First, the innervation pattern in adult X-gal stained EB2LZ spinal cords penetrates past the superficial laminae that receive nociceptive input and deep into the ventral horns, where the motor neurons that control locomotion are found. Many of the stained fibers appear to terminate at the base of the dorsal horn in the area of Clark’s Column, a spinal cord nucleus that runs from lumbar to cervical levels and carries proprioceptive information to the cerebellum (Fig. 1A, C) [19]. Second, a small subset (18%) of the largest neurons in the DRG of adult EB2LZ mice are β-gal positive, consistent with expression in the medium-sized mechanoreceptive and large proprioceptive neurons, rather than the more numerous and smaller nociceptors [32] (Fig. 1D, E). Third, We cultured embryonic day 13.5 mouse DRG neurons from EB2LZ mice and grew them in the presence of NGF, BDNF, or NT-3. At this stage in development, DRG neurons that send out nociceptive, mechanoreceptive, or proprioceptive afferents are dependent for survival on these neurotrophins, respectively [25]. We found β-gal positive neurons only in the NT-3-dependent population that gives rise to proprioceptive neurons (data not shown). Fourth, the pattern of β-gal in EB2LZ neonatal spinal cord and DRG overlaps that of parvalbumin, a marker for proprioceptive neurons [4] (Fig. 3A).

Figure 3. EB2 expression in proprioceptive sensory neurons and afferents at birth and in adult dorsal horn.

(A) Neonatal EB2LZ spinal cords and DRGs immunolabeled with anti-βgal (red) and anti-parvalbumin (green) antibodies. (B) EB2LZ/+ spinal cord stained with X-gal (panel i) and processed for immunofluorescence to detect NF200 (panel ii, red) and CGRP (panel iii, green). Panels i-iii are merged in iv. Rexed’s laminae are indicated by dotted lines. Arrowheads show boundaries of CGRP expression. Arrows indicate a bundle of NF200 and LacZ positive fibers passing through lamina II.

Strong X-gal stain was seen in the adult dorsal horn, indicating a broader pattern of expression than in the neonate. Nevertheless, the EB2/β-gal signal was still restricted to myelinated fibers (identified by NF200 immunoreactivity) that innervate laminae III and IV. These are primarily from Aβ mechanoreceptors and A∂ nociceptors. EB2 was specifically excluded from the CGRP-positive, unmyelinated C fibers that innervate lamina II (Fig. 3B).

EB2 does not mediate axon guidance from sensory neurons

We examined the anatomy of EB2LZ homozygous neonates, in which reverse signaling is absent due to the lack of an EB2 intracellular domain, using both anti-parvalbumin immunofluorescence (Fig. 3) and DiI filling of L4 DRG to label sensory afferents in the cord (Suppl. fig. 1). These mice showed no gross axon guidance defects. Thus, we conclude that EB2 is not likely to control developmental guidance of sensory axons.

Eph expression in the spinal cord

Our examination of the EphB2-LacZ knock-in mouse [22] revealed EphB2/β-gal expression primarily in the dorsal horns of the spinal cord in a pattern that mirrors the path of incoming sensory fibers, and also in some cells of the ventral horn. We did not observe β-gal in the DRG or in the dorsal root (Suppl. fig. 2). EphA4 is also expressed in spinal cord motor neurons [21]. These expression patterns agree with the in situ hybridization patterns for these two Ephs documented in the Allen Brain Atlas. Both Ephs are physiological binding partners for EB2 in other situations, and have been described in synapse formation elsewhere in the CNS [16, 39]. Thus, both are candidate receptors for the EB2 on sensory afferents. We isolated spinal cord neurons and treated them in culture with clustered recombinant EB2/Fc protein and observed EphB2 in these neurons co-clustering with NMDA receptors. EB2 treatment increased the size of these clusters, consistent with a role in synapse strengthening (Suppl. fig. 3). This data agrees with previous reports and underscores the likelihood that EB2 controls sensory plasticity in the spinal cord [1, 3].

Discussion

Whereas previous studies described immunodetection of EB2 in the dorsal horns, we discovered EB2/β-gal in fibers innervating the deeper laminae assigned to proprioceptive afferents. Of particular note was the concentration of β-gal positive fibers and cells in the region corresponding to Clarke’s Column. This nucleus stretches from cervical to lumbar levels in the cord and relays proprioceptive input to the brain. Given recent work that described a role for EB2 in chronic pain [1, 9, 26, 37, 40], we were surprised to find EB2LZ in the developing sensory system confined to proprioceptive DRG neurons and their fibers, with no expression in nociceptors. In adult spinal cords, we observed EB2LZ fibers in laminae III and IV in the dorsal horn, but not in lamina II. Our data contradict the findings of Battaglia, et al., which described EB2 immunoreactivity in this superficial lamina and in small diameter DRG neurons [1]. This disagreement may be explained by cross-reactivity of the antibodies used by these investigators for B ephrins. Nevertheless, the literature firmly implicates EB2 signaling in pain sensation. Zhao and colleagues generated an EB2 conditional knockout mouse using a nociceptor-selective Nav1.8 cre and reported that the mutants exhibited deficits in chronic pain perception [40]. Nav1.8 is expressed in A∂ and Aβ as well as C fibers [8], and our data suggest that knockout of EB2 in A∂ nociceptors may explain the observed phenotype, as well as in low-threshold Aβ mechanoreceptors that are sensitized by chronic inflammation [15]. Our observation that EB2 is not expressed in nociceptors at birth might explain why Zhao’s mutant showed no acute pain phenotype resulting from developmental defects, but rather a chronic pain phenotype that involves post-developmental plasticity.

Our findings have practical implications for translational manipulation of sensory circuits in the spinal cord. First, as noted above, CPG locomotor circuits are highly plastic and susceptible to positional imput once seperated from the brain. This principle is being used to increase weight-bearing and stepping in quadruplegics through Lokomat therapy [20, 35]. If activation of ephrin-induced plasticity can be activated, perhaps by injection of ephrin/Fc proteins, it might greatly increase the gains earned with training. At the same time, others have reported that injecting the rodent spinal cord with unclustered ephrin blocking proteins supresses chronic pain [9, 37], thus potentially causing a therapeutic “conflict of interest”. Activating Eph signaling in the spinal cord to improve locomotion might increase pain in the patient, while blocking pain by blocking Eph activation might inhibit the plasticity necessary for motor gains. Additionally, blocking of Eph signaling post-SCI has been reported to reduce the glial scar around the injury and increase regeneration and recovery [17, 30]. However, the same blocking strategy may impair the plasticity necessary for the spinal circuits that would supposedly be targeted by regenerating supraspinal axons. Thus, our work highlights the necessity of learning the exact role of each ephrin and Eph in the spinal cord and finding a balance between each of these systems.

Supplementary Material

01. Supplemental figure 1. DiI tracing of neonatal sensory afferents.

DiI stain of DRG afferent fibers in the spinal cord of EB2LZ homozygous neonates. The pattern observed was indistinguishable from that of wild type controls.

02. Supplemental figure 2. Expression of EphB2-LacZ in adult spinal cord.

Spinal cords of EphB2LZ mice stained with X-gal. A is 40x magnification. B is the area within the box in A shown at 100x.

03. Supplemental figure 3. Co-clustering of EphB2 and NMDAR1 in spinal cord neurons by EB2/Fc treatment.

Neonatal mouse neurons with clustered EB2/Fc or control IgFc. Fixed cells were labeled with anti-NMDAR1 (red) and anti-EphB2 (green). EphB2 puncta frequently colocalized in the same optical section with NMDAR1 puncta (arrowheads). Analysis of mean number of NMDAR clusters per cell showed a trend toward more clusters in ephrin-treated cells, but without statistical significance. Mean size of NMDAR clusters (pixel area) was doubled in ephrin-treated cells. Error bars are +/− SEM. N=5 cells per group.

Highlights.

• Ephrins are membrane-bound ligands that control synaptic plasticity.

• We described expression of ephrin-B2 in proprioceptive DRG neurons and their afferent fibers to the spinal cord.

• These data have substantial implications for training of spinal locomotor circuits.