Abstract
Receptor-type protein tyrosine phosphatases (RPTPs) are required for appropriate growth of axons during nervous system development in Drosophila. In the vertebrate, type IIa RPTPs [protein tyrosine phosphatase (PTP)-δ, PTP-σ, and LAR (leukocyte common-antigen-related)] and the type III RPTP, PTP receptor type O (PTPRO), have been implicated in the regulation of axon growth, but their roles in developmental axon guidance are unclear. PTPRO, PTP-δ, and PTP-σ are each expressed in chick motor neurons during the period of axonogenesis. To examine potential roles of RPTPs in axon growth and guidance in vivo, we used double-stranded RNA (dsRNA) interference combined with in ovo electroporation to knock down RPTP expression levels in the embryonic chick lumbar spinal cord. Although most branches of the developing limb nerves appeared grossly normal, a dorsal nerve identified as the anterior iliotibialis was clearly affected by dsRNA knock-down of RPTPs. In experimental embryos treated with dsRNA targeting PTP-δ, PTP-σ, or PTPRO, this nerve showed abnormal fasciculation, was reduced in size, or was missing entirely; interference with PTPRO produced the most severe phenotypes. Control embryos electroporated with vehicle, or with dsRNA targeting choline acetyltransferase or axonin-1, did not exhibit this phenotype. Surprisingly, embryos electroporated with dsRNA targeting PTP-δ together with PTPRO, or all three RPTPs combined, had less severe phenotypes than embryos treated with PTPRO alone. This result suggests that competition between type IIa and type III RPTPs can regulate motor axon outgrowth, consistent with findings in Drosophila. Our results indicate that RPTPs, and especially PTPRO, are required for axon growth and guidance in the developing vertebrate limb.
Introduction
Receptor protein tyrosine phosphatases (RPTPs) are transmembrane proteins implicated in the regulation of axon growth and guidance (Bixby, 2000; Johnson and Van Vactor, 2003). In particular, genetic evidence implicates type IIa and type III RPTPs in pathway selection by photoreceptors, CNS axons, and motor neurons in Drosophila (Van Vactor et al., 1998; Newsome et al., 2000; Clandinin et al., 2001; Maurel-Zaffran et al., 2001; Schindelholz et al., 2001; Sun et al., 2001). The low penetrance of defects in RPTP null mutants, and similarities in phenotypes between null mutants of different RPTPs, suggest redundancy in RPTP function during axonogenesis. However, examination of multiple mutants indicates both cooperation and competition among RPTPs in guiding axons (Desai et al., 1997; Schindelholz et al., 2001).
Evidence of RPTP function in vertebrate axon pathfinding is less clear. In vitro evidence links the vertebrate type IIa RPTPs leukocyte common-antigen-related (LAR), protein tyrosine phosphatase (PTP)-δ, and PTP-σ to axon growth regulation (Ledig et al., 1999b; Wang and Bixby, 1999; Q. L. Sun et al., 2000; Yang et al., 2003), and similar evidence has been obtained for the type III RPTP, PTP receptor type O (PTPRO) (Stepanek et al., 2001). Experiments using putative dominant-negative mutants of PTP-σ and PTP-δ suggest involvement of these RPTPs in pathfinding in vivo (Johnson et al., 2001; Rashid-Doubell et al., 2002). LAR, PTP-δ, and PTP-σ mutant mice have nervous system phenotypes, some of which may be related to axon growth (Yeo et al., 1997; Elchebly et al., 1999; Wallace et al., 1999; Uetani et al., 2000). Finally, LAR and PTP-σ null mutants exhibit aberrant peripheral nerve regeneration (Xie et al., 2001; McLean et al., 2002; Thompson et al., 2003; Van der Zee et al., 2003). Direct loss-of-function experiments linking these RPTPs to developmental axon guidance have not been reported.
The routes taken by embryonic motor axons to their targets are relatively well understood. In the chick spinal cord, axons from different motor pools are intermingled, because they grow toward a “choice point” at the base of the limb; there, axons segregate and extend dorsally or ventrally (Landmesser, 1978b; Lance-Jones and Landmesser, 1981). In the limb, axons diverge again to grow to their proper target muscles (Lance-Jones and Landmesser, 1981; Tosney and Landmesser, 1985). Type IIa and type III RPTPs are involved in decision making at motor neuron choice points in Drosophila, and these functions may be conserved. Because several such RPTPs are expressed in embryonic chick motor neurons, these neurons should be a useful system for examining the roles of RPTPs in vertebrate axon guidance.
In the chick, RNA interference (RNAi) can be combined with in ovo electroporation to knock down expression of guidance proteins in the developing spinal cord (Pekarik et al., 2003). We used this methodology to perform loss-of-function experiments for PTP-δ, PTP-σ, and PTPRO in developing spinal motor neurons. Our studies provide the first evidence for the involvement of a vertebrate type III RPTP in axon growth in vivo, and support the idea of cooperation and competition among vertebrate RPTPs in regulation of motor axon pathfinding.
Materials and Methods
Materials. Fertilized White Leghorn chicken eggs were purchased from SPAFAS (Norwich, CT) and incubated in a humidified incubator at 38°C until the desired stage of development. All of the chemicals were purchased from Sigma (St. Louis, MO) unless stated otherwise.
In situ hybridization. Plasmids containing cDNA for base pairs 382-1004 of chick (c)PTPRO (accession number U65891), base pairs 1093-1489 of cPTP-σ (accession number L32780), and a portion of cPTP-δ that aligns to base pairs 877-1478 of human PTP-δ (accession number L38929) were linearized and used as templates for transcription of sense and antisense digoxygenin-labeled riboprobes using the recommended procedure (Roche, Mannheim, Germany). Embryos were dissected at embryonic day 5.5 (E5.5), fixed in 4% paraformaldehyde overnight, cryoprotected in 30% sucrose overnight, and frozen in OCT (Tissue Tek; Fisher Scientific, Pittsburgh, PA) for cryosectioning. Ten-micrometer-thick sections were allowed to dry and processed for in situ hybridization following a standard protocol (Schaeren-Wiemers and Gerfin-Moser, 1993). Sections were hybridized overnight with 600 ng/ml riboprobe at 56°C.
Production of double-stranded RNA. Sense and antisense RNAs were produced from plasmids encoding portions of cPTP-σ (base pairs 1093-1489), cPTP-δ (aligns to base pairs 877-1478 of human PTP-δ), cPTPRO (base pairs 1458-2334), and cChAT (base pairs 1344-1665; accession number AY044155). After linearization with the appropriate restriction endonucleases, RNA was transcribed using the SP6 and T7 MegaScript kits from Ambion (Austin, TX). RNA was cleaned, analyzed, and annealed as described previously (Pekarik et al., 2003).
In ovo RNA interference. Double-stranded RNA (dsRNA) injection and electroporation were performed as described previously (Pekarik et al., 2003). Embryos were injected at E3 to E3.5 [Hamburger and Hamilton stages (HH) 18-20] and dissected 48-60 h later (HH 25-27). dsRNA for combinatorial RNA interference (RNAi) was injected at a final concentration of 500 ng/μl for each RNA. This gave molarities of 1.7 μm for PTP-σ, 1.1 μm for PTP-δ, 0.8 μm for PTPRO, and 2.0 μm for choline acetyltransferase (ChAT).
Western blotting. Embryos were dissected 48 h postelectroporation in cold PBS-glucose. The lumbosacral portions of 10-15 spinal cords were split into right and left sides and lysed in 1% Igepal CA-630, 5 mm EDTA, 100 μg/ml PMSF, 1 mg/ml leupeptin, 1 mg/ml aprotinin, and 1 mg/ml pepstatin A in TBS (25 mm Tris, 150 mm NaCl, pH 7.4) for 30 min. Undissolved pellets were cleared by centrifugation (12,000 × g; 30 min), and 50 μg of supernatant protein was separated on SDS-PAGE and transferred to nitrocellulose as described previously (Bixby and Jhabvala, 1990). Primary antibodies were rabbit anti-cPTPRO (Ledig et al., 1999a) at 1:2500, or rabbit anti-cPTP-σ (Stoker et al., 1995) at 1:5000. Secondary antibody was Alexa Fluor 680 goat anti-rabbit IgG (Molecular Probes, Eugene, OR). After rinsing, blots were incubated with anti-chick L1 cell adhesion molecule (L1CAM) (Lemmon and McLoon, 1986), followed by infrared dye 800-conjugated anti-mouse IgG (Rockland, Gilbertsville, PA). Bound antibody was visualized, and band intensity was quantified using the LI-COR (Lincoln, NE) Odyssey Infrared Imaging System.
Whole-mount embryo staining. Embryos were dissected and fixed in 4% paraformaldehyde overnight. After rinsing in PBS, embryos were incubated in 1% Triton X-100 for 1 h, followed by 20 mm lysine in 0.1 m sodium phosphate for 1 h, followed by blocking in 10% goat serum in PBS for 4 h. Embryos were incubated in anti-neurofilament-M antibody (RMO270; Zymed, San Francisco, CA) diluted 1:1500 in blocking solution for 48 h at 4°C. Embryos were rinsed overnight in PBS, and then incubated in cyanine 3 (Jackson ImmunoResearch, West Grove, PA) or Alexa Fluor 594 goat anti-mouse IgG (Molecular Probes) diluted 1:2000, for 16 h at 4°C.
Phenotype quantification. Embryos were categorized as unaffected or affected by experimenters blinded to treatment group. An initial subgroup, containing embryos from all single-treatment, PBS, and control groups, were analyzed separately by three experimenters and categorized by each based on comparison of the treated anterior iliotibialis (AITIB) with the untreated AITIB. Embryos were categorized as affected if there was a gross difference in morphology or nerve size between the electroporated and nonelectroporated sides, including missing nerves, severely truncated nerves, or gross changes in branching/fasciculation.
For calculation of nerve area, images of neurofilament-stained nerves were captured with a Nikon (Melville, NY) Coolpix 990 camera mounted on a Nikon Eclipse TE300 microscope. An algorithm designed in Igor Pro 4.01 (A. R. Rosendahl, University of Miami, Miami, FL) was used to analyze the images. Briefly, the algorithm used an adaptive thresholding process to extract nerve pixels from background. The same images were used for Neurolucida analysis. An experimenter traced each branch of the nerves in Neurolucida 5.04.3 (MicroBrightField, Williston, VT). This tracing was then used by Neuorexplorer 3.60.3 (MicroBrightField) to calculate branch length and the number of primary branches (defined as branching of axon bundles from the main axon fascicle) for each nerve. For all of the quantitative measures, we calculated the ratio of the measurements between treated and untreated nerve. These values were then used to compare all of the treatment groups using nonparametric ANOVA (Kruskal-Wallis test) followed by Dunn's multiple comparisons test to compare individual treatment groups.
1,1′-Dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate tracing. After dissection, embryos were stored in 4% paraformaldehyde. For retrograde tracing, nitrocellulose membrane was soaked with a 1 mg/μg dimethylformamide solution of 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI) (D-282; Molecular Probes) and allowed to dry. Small pieces of nitrocellulose were inserted into the limb bud in the termination zone of the AITIB. Embryos were stored at 37°C for 5 d. After tracing was complete, embryos were embedded in 6.25% agar type IX, and 250 μm transverse sections were cut on an oscillating tissue slicer (Electron Microscopy Sciences, Ft. Washington, PA). Anterograde tracing was performed on 250 μm sections cut in the same way. The untreated side of the embryo was marked with india ink. Glass capillaries with a tip diameter of ∼5 μm were dipped into 5 mg/ml DiI dissolved in ethanol and placed into the ventrolateral spinal cord. After 1 h, capillaries were removed, and the sections were stored in the dark at room temperature for 7 d. Images were scanned using a Zeiss (Oberkochen, Germany) LSM510 confocal microscope (Miami Project to Cure Paralysis), and manipulated in Adobe Photoshop.
Islet-2 staining and design-based stereology. Embryos were dissected and fixed at 4°C for 1.5 h. Every fifth section of 20 μm cryostat sections from the rostral lumbosacral region was processed for Islet-2 (Isl-2) immunohistochemistry. The Islet-2 primary antibody (51.4H9) developed by T. M. Jessell (Columbia University, New York, NY) was obtained from the Developmental Studies Hybridoma Bank (Iowa City, IA) developed under the auspices of the National Institute of Child Health and Human Development and maintained by the University of Iowa (Iowa City, IA). Sections were incubated in a 1:100 dilution of 51.4H9 at 4°C for 48 h. After rinsing, they were incubated in a 1:1000 dilution of biotinylated anti-mouse IgG (Vector Laboratories, Burlingame, CA). Bound antibody was visualized with the Vectastain Elite ABC kit (Vector Laboratories) followed by a nickel-intensified DAB reaction. Stereo Investigator 5.05.04 (MicroBrightField) was used to estimate the number of Islet-2-positive cells through the volume of the sectioned area. DAB-positive cells were counted in the right and left ventral horns of each section using a counting frame of 25 × 25 μm and a scan grid size of 45 × 45 μm. At least 300 cells on each side were counted per embryo. From this, the software estimated upward of 5000 Isl-2-positive cells per side of each embryo in the sectioned region.
PTPRO-GFP (green fluorescent protein) electroporation. PCR was used to add a unique SalI site at base pair 3100 just 3′ of the sequence encoding the 28-aa juxtamembrane region of cPTPRO. The entire extracellular domain, transmembrane region, and juxtamembrane region were cloned into the pEGFP-N2 vector (Clontech, Palo Alto, CA) in-frame with the GFP-coding sequence. This cPTPRO-GFP construct was released from the pEGFP vector and cloned into a modified pIRES plasmid with a β-actin promoter (Pekarik et al., 2003). PTPRO-GFP was electroporated with dsChAT RNA or dsPTPRO RNA at a final concentration of 1.5 μg/μl. After 48 h, embryos were dissected and 100 μm sections of the lumbosacral segments were cut on an oscillating tissue slicer as described above. All of the in-focus green cells were counted in each section, and at least 10 sections were examined per embryo.
Statistics. Statistics were done using InStat (version 2.03).
Results
RPTPs are expressed in the lateral motor column of the embryonic chick spinal cord
The pathways taken by chick motor axons to their targets in the hindlimb are well understood, and the motor neurons are accessible for embryonic manipulation. We therefore chose to examine regulation of motor axon growth in the chick lumbosacral spinal cord. Two type IIa RPTPs, PTP-δ and PTP-σ, and one type III RPTP, PTPRO, are expressed during the period of axon outgrowth in motor columns of the rodent spinal cord and the chick brachial spinal cord (Yan et al., 1993; Sommer et al., 1997; Chilton and Stoker, 2000; Beltran et al., 2003; Thompson et al., 2003). To determine whether these RPTPs are also expressed in chick lumbosacral motor neurons, we performed in situ hybridization using digoxygenin-labeled riboprobes. PTPRO mRNA was expressed in the ventral horn, including large neurons, and in the dorsolateral commissural neurons (Fig. 1) as well as in the dorsal root ganglia (data not shown) of E5 chick embryos. This is fully consistent with the pattern seen previously in the brachial spinal cord (Chilton and Stoker, 2000). PTP-δ mRNA was expressed diffusely throughout the lumbar spinal cord, including the ventral horn, and PTP-σ mRNA was found in the ventral horn and in the ventricular zone (data not shown). This pattern of mRNA expression is consistent with our hypothesis that these three RPTPs are involved in the growth of hindlimb motor axons from the embryonic chick spinal cord.
Electroporation of dsRNA specifically “knocks down” protein levels of target RPTP
To assess the roles of RPTPs in motor axon guidance, we used in ovo electroporation combined with RNAi. Electroporation of long dsRNAs has been shown to produce specific and reproducible decreases in expression of target proteins (Pekarik et al., 2003). The phenotypes resulting from gene silencing by RNAi are specific, in the sense that dsRNA targeting of known axonal guidance proteins produces the same axonal pathfinding defects as injection of cognate function-blocking antibodies, and expression levels of nontargeted proteins from the same family as the targeted protein are not changed (Stoeckli and Landmesser, 1995; Pekarik et al., 2003). Recently, in ovo electroporation of short interfering RNA (siRNA) hairpins was shown to successfully knock down neuropilin-1 in the chick spinal cord (Bron et al., 2004). Although this methodology is advantageous in that interfering RNA is continuously produced, using one long (400-800 bp) dsRNA allowed us to sidestep issues of sequence selection, which is important because siRNAs differ in their effectiveness (Elbashir et al., 2001; Holen et al., 2002). We produced dsRNAs targeting the type III RPTP, PTPRO, and the type IIa RPTPs, PTP-δ and PTP-σ. These dsRNAs were injected into the central canal of E3.5 chick spinal cords (HH 18-20) and electroporated into the lumbosacral region. Because the electroporation confines transfection to one side of the spinal cord (Pekarik et al., 2003) (data not shown), this side will be referred to as the “treated” side; the other side acts as a control. After electroporation, the chicks were allowed to develop for 48 h. During this time period, motor neurons extend axons into the periphery (Hamburger, 1975; Oppenheim and Heaton, 1975).
We evaluated the efficacy of gene silencing by RNAi in two ways, using PTPRO as our test protein. In an initial series of experiments, we combined in ovo electroporation with coinjection of (1) a plasmid encoding a GFP fusion protein containing the extracellular, transmembrane, and juxtamembrane domains of PTPRO, and (2) dsRNA targeting either PTPRO or ChAT as a control. This allowed us to use GFP fluorescence to evaluate any change in transfected PTPRO expression attributable to dsRNAi. In embryos transfected with PTPRO-GFP and dsRNA targeting ChAT, a large number of GFP-expressing cells was always observed (Fig. 2B). In contrast, many fewer GFP-positive cells were observed after cotransfection of dsRNA targeting PTPRO (Fig. 2D). This effect was quantified by counting the number of GFP-positive cells per section in the two conditions. On average, we found 18.6 ± 1.8 (mean ± SEM) green fluorescent cells per section in the embryos cotransfected with ChAT dsRNA, compared with 1.4 ± 0.3 cells per section with PTPRO dsRNA (n = 3 embryos in each group; p < 0.0001). This drastic decrease in PTPRO-GFP expression was apparent throughout the electroporated region; there was no evidence that a subset of cells was more sensitive to electroporation or RNAi.
These experiments demonstrate that our dsRNA can effectively target PTPRO, but do not resolve the extent to which endogenous PTPRO is affected. Because our cPTPRO antibody is ineffective in immunohistochemistry, we examined this issue by Western blotting. The lumbosacral spinal cords from 10-15 embryos were dissected and split into left and right halves, lysed, separated on SDS-PAGE, and probed with antibodies to RPTPs on Western blots. When we examined PTPRO expression in embryos treated with dsRNA targeting PTPRO, we found the ratio of PTPRO expression [left (or treated) side/right (or untreated) side] to be decreased by 29% relative to this ratio in untreated control embryos (p < 0.05) (Fig. 3). Thus, dsRNA treatment reduced PTPRO expression levels significantly. To evaluate the specificity of this result, we examined the relative levels of PTP-σ on the two sides of spinal cords in control embryos and embryos treated with PTPRO dsRNA. In this case, the expression ratio was unchanged (5.5% increase; NS) (Fig. 3). Thus, there is a specific decrease of the target RPTP (PTPRO) in embryos that are electroporated with PTPRO dsRNA.
We also tested knock-down of a second RPTP using dsRNA targeting PTP-σ; knock-down of PTP-σ protein was clearly seen in Western blots (Fig. 3). Analysis of blots from three different experiments revealed that knock-down of PTP-σ protein averaged 25%, similar to what was seen with PTPRO dsRNA. Our results are consistent with previous experiments demonstrating specific knock-down of axonin-1 expression by the same protocol (Pekarik et al., 2003). In this study, estimated knock-down averaged 28%, quantitatively similar to the present findings with PTPRO and PTP-σ. Together, the data suggest that our RNAi protocol specifically and effectively knocks down expression of the target RPTP. Because our evaluation of PTPRO and PTP-σ knock-down is quantitatively consistent with axonin-1 knock-down, we expect that knock-down efficiency for PTP-δ is similar; we could not test this directly because we lack a useful antibody to cPTP-δ.
It is important to note that knock-down of PTPRO in transfected neurons is underestimated by our Western blot data. First, only a fraction (≤60%) of the cells on the treated side are transfected (Pekarik et al., 2003). Thus, knock-down in transfected cells must be at least 50%. Second, by using the entire hemicord over the entire lumbosacral region, we are including tissue not likely to be efficiently electroporated. Finally, PTPRO is present in axons and growth cones (Bodden and Bixby, 1996; Stepanek et al., 2001), and our samples will therefore contain axonal protein from neurons outside the targeted area. It is difficult to estimate the magnitude of the latter two effects, but our results with PTPRO-GFP suggest that knock-down is likely to be substantially more than 50%.
Knock-down of RPTPs produces a deficit in motor axon pathfinding
PTP-δ, PTP-σ, and PTPRO have been shown to regulate neurite outgrowth in vitro (Ledig et al., 1999b; Wang and Bixby, 1999; Q. L. Sun et al., 2000; Stepanek et al., 2001), and expression of putative dominant-negative constructs suggests that PTP-δ and PTP-σ are involved in axon growth and guidance in the visual system (Johnson et al., 2001; Rashid-Doubell et al., 2002). However, the involvement of these RPTPs in developmental axon guidance has not been tested by direct loss-of-function experiments in vivo. We used in ovo RNAi to examine loss-of-function phenotypes for these RPTPs in motor axon growth.
Embryos were electroporated with dsRNA targeting PTP-δ, PTP-σ, or PTPRO, and allowed to develop for 48-60 h (to HH 26-27). At HH 24-26, motor axons diverge from the main nerve trunks and grow toward specific muscles (Hamburger, 1975; Oppenheim and Heaton, 1975). To evaluate the growth of peripheral nerves, we examined neurofilament staining in whole mounts of the lumbosacral region of the embryo. RNAi of RPTPs did not grossly affect the major nerve branches in the hindlimb (data not shown). However, knock-down of RPTPs resulted in a conspicuous phenotype in a superficial dorsal nerve (Fig. 4A). In control embryos, the untreated and treated nerves had three to four major bundles, with substantial lateral branching. In embryos treated with dsRNA targeting RPTPs, we saw changes in the morphology of this nerve on the treated side. In some cases, there were defasciculation errors such that the nerve had only one or two major bundles, and limited subbranching. In other cases, the nerve was much smaller, seeming to be delayed in its outgrowth, or was missing entirely.
To quantify these effects, we characterized embryos as “affected” (gross differences in the dorsal nerve between treated and untreated sides) or “unaffected” (only minor differences between the two sides) (see Materials and Methods). Approximately 70% of embryos treated with dsRNA targeting either the PTP-δ or PTP-σ were categorized as affected (Fig. 4B). Strikingly, >85% of embryos treated with dsRNA targeting the type III RPTP PTPRO were affected (Fig. 4B). These results suggest that the function of these RPTPs is essential for proper guidance of the axons in this dorsal nerve.
We used several types of controls to determine the specificity of these effects. As controls for the experimental procedures, we examined embryos in which the egg was windowed but the embryo was left untouched, and embryos injected with PBS and electroporated. We also used controls to determine the specificity of the RNAi itself. First, embryos were electroporated with dsRNA targeting ChAT. ChAT is expressed in motor neurons at this time in development but does not appear to be necessary for initial axon outgrowth to muscle targets (Misgeld et al., 2002). To control for the possibility that interference with any axon guidance cue would produce the phenotype we observed, we electroporated embryos with dsRNA targeting axonin-1. Axonin-1 is expressed in the axons of this dorsal nerve (Fig. 5A). The dsRNA we used has been shown to knock down axonin expression and to cause clear and predictable defects in commissural axon pathfinding (Pekarik et al., 2003). In none of the control situations did we observe more than ∼20% affected embryos (Figs. 4B, 5B). Indeed, RNAi for ChAT or axonin-1 produced no increase in affected embryos above that seen for PBS electroporation. Thus, knock-down of RPTPs specifically affects axonal morphology in this dorsal nerve.
To supplement our qualitative categorizations of embryos, we used several quantitative measures of phenotype. A computer algorithm designed to delineate nerve from background (see Materials and Methods) counted pixels in images of each nerve. This gave us an estimate of nerve area, which we used to calculate a ratio of treated nerve area to untreated nerve area (Fig. 6). In control embryos, the mean ratio was slightly larger than 1; in PBS and ChAT embryos, the mean ratio was ∼0.9. Embryos electroporated with dsRNA targeting PTP-δ or PTP-σ had a ratio of 0.7. Remarkably, embryos electroporated with dsRNA targeting PTPRO had nerves that were on average only one-half as large as those of controls; this difference was highly statistically significant. Similar results were obtained using the Neurolucida program to calculate total nerve length from traced images. Embryos treated with dsRNA targeting PTPRO had significantly shorter treated nerves than ChAT control embryos (p < 0.001); PTP-δ and PTP-σ embryos also showed a decrease in nerve length on the treated side but were not significantly different from control.
Analysis of nerve branching (number of primary branches on control and treated sides) demonstrated a significant effect of knocking down each of the RPTPs we tested. Embryos treated with dsRNA targeting PTP-σ had an average of 3.2 ± 0.2 primary branches on the untreated side, but only an average of 2.0 ± 0.2 primary branches on the treated side (p = 0.0001; Wilcoxon matched-pairs t test). Embryos treated with dsRNA targeting PTP-δ or with PTPRO had an average of 2.8 ± 0.1 primary branches on the untreated sides compared with 2.1 ± 0.2 and 2.0 ± 0.2 primary branches on the treated sides, respectively (p < 0.01 and p < 0.001). In contrast, control embryos treated with dsRNA targeting ChAT had an average of 2.9 ± 0.2 primary branches on both the untreated and treated side. Overall, our quantitative description of the phenotype followed the same pattern as that of our qualitative data; knock-down of each of the three RPTPs changed the size and morphology of the treated nerve, although interference with PTPRO had the greatest effect in our experiments. These data suggest that type IIa and type III RPTPs are involved in axon pathfinding by chick motor neurons. The effect we saw with PTPRO knock-down represents the first evidence for an in vivo role of a type III RPTP in vertebrate axon growth.
The affected dorsal nerve is the anterior iliotibialis
To identify the dorsal nerve affected by RPTP knock-down, we followed the projection pattern of this nerve during development by neurofilament staining of embryos at HH 29-32 (E6-E7.5) (data not shown). Over time, the nerve spreads as a thin sheet covering the lateral surface of the thigh. This pattern corresponds to the position of the anterior iliotibialis, a thin, sheet-like muscle that contains both fast- and slow-twitch muscle fibers (Landmesser, 1978a; Milner et al., 1998). To confirm this identification, we used retrograde DiI labeling in E5.5 embryos to identify the neurons of origin of the dorsal nerve. We found that the motor portion of the nerve arises from neurons in the lateral portion of the ventral horn in lumbosacral segments (LS) 1-3 (Fig. 7). This is fully consistent with the position of the anterior iliotibialis motor neuron pool (Landmesser, 1978a). Based on the position and projection pattern of the dorsal nerve, and the position of the motor neurons giving rise to this nerve, we will refer to it as the AITIB. PTPRO mRNA is found in, but is not limited to, neurons of the E5 ventral motor column in which the AITIB motor pool is found (Figs. 1, 7B) (data not shown). Thus, PTPRO appears to be expressed in the motor neuron pool that gives rise to the AITIB during the time of AITIB outgrowth. Sensory neurons were also labeled by the retrograde tracing (Fig. 7B, inset), showing that the affected AITIB is a mixed nerve.
To examine the effects of RNAi on individual motor axons, we performed anterograde DiI tracing. DiI was placed in the lateral portion of the ventral horn in vibratome sections from LS 1-3 of embryos electroporated with dsRNA targeting PTPRO, and tracing was examined using a confocal microscope. Because of limitations on the thickness of sections suitable for this type of microscopy, the entire length of the AITIB was rarely identified. However, we consistently observed that labeled motor axons on the treated side of the embryo (Fig. 8A) were shorter than those on the untreated side of the embryo (B). We did not observe differences in motor axon trajectories within the spinal cord or the plexus region. Thus, these experiments indicate that AITIB motor axon growth is retarded by knock-down of PTPRO.
The AITIB phenotype is not caused by increased death of motor neurons
If in ovo RNAi of RPTPs led to increased motor neuron death, this could contribute to a reduction in the size of the AITIB. To investigate this possibility, we counted the number of motor neurons in embryos electroporated with PTPRO dsRNA. Serial sections from PTPRO or control embryos were stained with an antibody to the transcription factor Isl-2, an early marker for motor neurons (Tsuchida et al., 1994). Isl-2-positive cells were counted using design-based stereology (see Materials and Methods) in treated and untreated sides of spinal cord lumbosacral segments 1-5. Similar numbers of Isl-2-positive cells in individual sections were seen in both PTPRO RNAi embryos and in controls (Fig. 9). Additionally, the estimates of total motor neurons in the entire volume examined were similar between the two sides of the embryos. The ratio of Isl-2-positive cells in treated compared with untreated spinal cord halves was 0.98 ± 0.02 for control embryos, and 0.99 ± 0.01 for PTPRO RNAi embryos (n = 3 for both groups). Thus, cell death of motor neurons does not contribute substantially to the mean 50% decrease in AITIB size in PTPRO-targeted embryos.
Combinatorial RNAi: evidence for competition between type IIa and type III RPTPs
Experiments in Drosophila have shown that type IIa and type III RPTPs can cooperate or compete in the guidance of both central axon fascicles and peripheral motor axons (Desai et al., 1997; Q. Sun et al., 2000; Schindelholz et al., 2001). Do vertebrate RPTPs also interact in the regulation of motor axon pathfinding? To answer this question, we used combinatorial RNAi. It has been demonstrated that several genes can be silenced together by injecting a combination of dsRNAs targeting the genes of interest, if the RNAs are injected at reasonably high concentrations (Schmid et al., 2002). For our experiments, we first determined that 500 ng/μl was the dsRNA concentration that produced the most reliable PTPRO phenotype (data not shown). In the combination experiments, we therefore used all of the dsRNAs at a final concentration of 500 ng/μl.
Embryos in which both type IIa RPTPs (PTP-δ and PTP-σ) were targeted had approximately the same severity of phenotypes as embryos in which either PTP-δ or PTP-σ alone were targeted, although the percentage affected was slightly higher in the combination experiments than for either RPTP alone (Fig. 10A). These data provide no evidence for a strong interaction between PTP-δ and PTP-σ in the growth of the AITIB.
Embryos in which either PTP-δ or PTP-σ was targeted concomitantly with the type III RPTP, PTPRO, were on average less severely affected than embryos in which PTPRO alone was targeted (Fig. 10B). In the case of PTPRO/PTP-δ, only 50% of embryos were classified as affected, compared with 87% for PTPRO alone, a significant reduction (p < 0.01). To ensure that this reduction in phenotype severity was not simply attributable to addition of a second dsRNA, we used a combination of PTPRO and ChAT dsRNAs. A slight and nonsignificant reduction in the percentage of affected embryos was seen in this case, indicating that the reduction in phenotype severity seen with PTPRO/PTP-δ was specific to this combination. These data suggest that PTPRO and PTP-δ compete functionally during the growth of the AITIB nerve. A similar trend was seen for PTP-σ/PTPRO, but the data do not allow a conclusion in this case.
Finally, we tested a combination of dsRNAs targeting all three RPTPs (Fig. 10C). In this case, the reversion of the PTPRO phenotype was even more dramatic, with only 41% of the embryos categorized as affected (p < 0.01). As a control for the use of large amounts of dsRNA, we tested PTPRO dsRNA in conjunction with 1000 ng/μl ChAT dsRNA; no significant difference was seen compared with PTPRO alone. These data suggest that PTP-δ, together with PTP-σ, counteracts the role of PTPRO in controlling the outgrowth of the AITIB. The need to balance the functional activities of type IIa and type III RPTPs in motor axon guidance is similar to the situation in Drosophila. The data may indicate a conserved function for RPTPs in motor axon outgrowth from flies to vertebrates.
Discussion
RNAi targeting of PTP-δ, PTP-σ, or PTPRO in spinal motor neurons led to dramatic changes in the morphology of the developing AITIB nerve. Although RNAi produced similar changes for all of the RPTPs examined, knock-down of the type III RPTP PTPRO produced the strongest effects. Knock-down of the type IIa RPTPs in combination with PTPRO suggests competition among these RPTPs in axon guidance. Notably, our results provide the first evidence implicating a type III RPTP in vertebrate axon pathfinding.
RPTPs could be acting at several places to regulate outgrowth of AITIB axons. Embryos lacking AITIB nerves suggest involvement of RPTPs in axon initiation or spinal nerve exiting. Alternatively, RPTPs could function at choice points along axonal pathways; loss of RPTP function could cause axons to stall or make incorrect decisions. In Drosophila, loss-of-function mutants of type IIa RPTPs exhibit “stall” or “stop short” phenotypes in motor axon fascicles (Desai et al., 1996; Desai et al., 1997). As in these mutants, phenotypes in our RPTP RNAi embryos varied considerably. We cannot determine whether this phenotypic variation is attributable to incomplete “penetrance” or variation in the efficacy of the RNAi, but incomplete penetrance has been observed in many cases for Drosophila RPTPs.
Retrograde tracing from the AITIB also labeled sensory neurons. Thus, the affected nerve contains sensory fibers. It is unlikely that our electroporation targeted sensory neurons, because we found transfection of sensory ganglia only when we electroporated before HH 18. Our phenotypic analysis suggests that the whole nerve is affected by RPTP dsRNA electroporation, including the sensory component. This finding is consistent with a body of evidence linking sensory axon pathfinding to decisions made by earlier-extending motor axons (Honig et al., 1986; Landmesser and Honig, 1986; Scott, 1986; Wang and Scott, 1997; Honig et al., 1998).
The RPTP knock-down phenotypes also included gross changes in fasciculation pattern. Some embryos exhibited phenotypes in which primary or secondary branches of the nerves appeared to be fused together in the treated side. In some cases, secondary and tertiary branches developed more distally in the treated AITIB than in the control AITIB. These phenotypes implicate RPTPs, and particularly PTPRO, in defasciculation. The Drosophila PTPRO relative DPTP10D is involved in defasciculation of the intersegmental nerve b, along with other RPTPs (Sun et al., 2001).
RPTPs likely act both as ligands and as receptors, and the effects we observe may be attributable to loss of either or both functions. Proper guidance of photoreceptor cell axons in Drosophila, for example, is dependent on both cell-autonomous (presumably receptor) functions of RPTPs, as well as non-cell-autonomous (presumably ligand) functions (Garrity et al., 1999; Clandinin et al., 2001; Maurel-Zaffran et al., 2001). Additionally, RPTPs might act in cis to regulate other guidance molecule-receptor interactions within the axon or on growth cones. The Caenorhabditis elegans RPTP Clr-1, for example, inhibits signaling in cis through the netrin receptor UNC40 (Chang et al., 2004). In the spinal cord, effects of RNAi could be attributable to loss of RPTPs as receptors on AITIB motor neurons, making the neurons unresponsive to growth-promoting factors. Effects could also be attributable to loss of RPTPs from other cells, so that receptors on the AITIB are not activated by their growth-promoting ligands. In the spinal nerves and crural plexus, RPTPs could also be acting autonomously within the AITIB axons, or could be signaling the AITIB axons from neighboring nerve fibers. Once in the limb, changes in AITIB morphology would presumably be attributable to cell-autonomous RPTP action.
Interpretation of these knock-down experiments is complicated by our limited understanding of the binding partners of RPTPs and their effects on RPTP function. PTP-δ binds homophilically (Wang and Bixby, 1999), although it may have other partners. PTP-σ binds to heparan sulfate proteoglycans and to other uncharacterized ligands in retinal basement membranes (Ledig et al., 1999b; Aricescu et al., 2002). The binding partners of PTPRO are unknown, although it is unlikely to bind homophilically (Stepanek et al., 2001).
It is difficult to predict effects of RPTP knock-down on motor axon growth from available data. Because PTP-δ promotes neurite outgrowth in vitro, and putative dominant-negative mutants of PTP-δ cause decreased growth of retinotectal axons (Johnson et al., 2001), one might expect the removal of PTP-δ to cause a decrease in size of the nerve. Knock-down of PTP-δ indeed decreased some aspects of AITIB growth. In contrast, available evidence suggests that unliganded PTP-σ is inhibitory to axon growth, and that ligand binding may promote growth by inactivating PTP-σ (Ledig et al., 1999b; Johnson et al., 2001). Consistent with this idea, loss of PTP-σ function accelerates regeneration of adult peripheral nerves (McLean et al., 2002). In our experiments, however, knock-down of PTP-σ alone did not appear to increase nerve growth. It is possible that PTP-σ has distinct functions in different axonal populations. Alternatively, ligand rather than receptor functions may be important. The situation for PTPRO is even less clear. Although the extracellular domain of PTPRO is neurite inhibitory in vitro (Stepanek et al., 2001), effects of PTPRO as a receptor in PTPRO-expressing neurons have not been studied.
Although the three RPTPs we tested appear to have distinct activities in vitro, the gross phenotypes caused by single knock-downs in motor neurons were similar. Because there is no exacerbation of the single-knock-down phenotypes in embryos treated with dsRNA targeting PTP-δ and PTP-σ, these RPTPs might have partially redundant functions in regulating the outgrowth of the AITIB. However, double- and triple-combination experiments indicate a functional competition between PTPRO and the type IIa RPTPs in AITIB growth. This unexpected complexity is likely to be partially attributable to our level of analysis. It is clear that functional interactions among RPTPs cannot be reliably predicted on the basis of observations of gross phenotypes of single null mutants (Krueger et al., 1996; Desai et al., 1997; Schindelholz et al., 2001; Sun et al., 2001).
At what level are these RPTPs interacting to regulate axon growth? There is no indication thus far that any RPTPs share the same ligands or receptors, although LAR, PTP-δ, and PTP-σ do share potential heparan sulfate proteoglycan binding motifs (Aricescu et al., 2002; Johnson and Van Vactor, 2003). It seems more likely that the RPTPs are interacting downstream of ligand-receptor interactions. Little is known about neuronal substrates for these RPTPs. PTP-δ, PTP-σ, and LAR bind to the liprin protein family, which is implicated in several aspects of synapse formation (Serra-Pages et al., 1995; Zhen and Jin, 1999; Kaufmann et al., 2002; Wyszynski et al., 2002). Substrates for PTPRO may include Trk receptors (Beltran et al., 2003), in addition to the novel neural protein NPCD (neuronal pentraxin with chromo domain) (Chen and Bixby, 2005a,b). More progress has been made in the discovery of potential substrates for LAR; it can act together with Abl to coordinate phosphorylation/dephosphorylation of the profilin-interacting protein Ena. Additionally, LAR interacts with the guanine nucleotide exchange factor Trio, which can activate Rho family GTPases to regulate actin filaments (Bixby, 2001; Johnson and Van Vactor, 2003). A recent paper shows that PTP-δ interacts with an actin-binding protein called MIM in fibroblasts, suggesting that cytoskeletal interactions are common to type IIa RPTPs (Gonzalez-Quevedo et al., 2005). The type IIa and type III RPTPs could have opposing effects on the same signaling pathway, or affect opposing signaling pathways, to control cytoskeletal rearrangements in the growth cone.
It is surprising that the axon pathfinding defects we saw with RPTP knock-down were confined to the AITIB. PTP-δ, PTP-σ, and PTPRO are expressed in many ventrolateral neurons throughout the length of the lumbosacral spinal cord. Based on GFP transfection experiments, we produce efficient electroporation in 7-10 segments of the cord, in both medial and lateral cells. Thus, it does not seem likely that the dsRNA is selectively entering the AITIB motor pool. Our PTPRO-GFP experiments do not provide evidence for selective efficacy of RNAi in different cell populations. Possible explanations for the apparent selectivity include the following: (1) lower baseline levels of RPTP expression in AITIB neurons, (2) lack of neighboring axon bundles in the AITIB pathway to provide guidance cues, and (3) the timing of growth relative to knock-down—the AITIB emerges precociously compared with other nerves from the crural plexus. The isolation of the AITIB from other nerves may also play another role. Perhaps pathfinding is disrupted in other axon fascicles, but cannot be observed against a background of normal axons with the assessment techniques we used.
The accessibility of chick spinal motor neurons combined with the ability to perform RNAi with long dsRNAs has allowed us to perform direct loss-of-function experiments with several RPTPs in developing projection neurons. The strongest pathfinding defects were seen by targeting the type III RPTP, PTPRO. Our data also support the idea of competition and cooperation between type IIa and type III RPTPs in regulating vertebrate axon guidance, suggesting an evolutionary conservation of RPTP function. Among the major challenges remaining is the identification of extracellular and intracellular binding partners of these RPTPs, providing insight into the mechanisms underlying RPTP function in axonal pathfinding.
Footnotes
This work was supported by National Institutes of Health Grant NS38920 (J.L.B.). L.S. was a Lois Pope Leaders in Furthering Education Fellow. We thank Dr. Vance Lemmon for the L1 antibody and for helpful criticism of the manuscript, Andrew Rosendahl for the nerve branch area analysis program, and Dr. Lynn Landmesser for assistance with chick neuroanatomy. We also thank Beata Frydel and the Miami Project Core Imaging Facility for imaging help. Drs. Vlad Pekarik, Dan Liebl, and Ken Muller provided excellent technical advice. We thank Katy Reinhard for help with tissue sectioning.
Correspondence should be addressed to Dr. John L. Bixby, Miami Project to Cure Paralysis, University of Miami School of Medicine, 1095 Northwest 14th Terrace, Miami, FL 33136. E-mail: jbixby{at}miami.edu.
Copyright © 2005 Society for Neuroscience 0270-6474/05/253813-11$15.00/0