DiI tracing and immunofluorescent staining of cryostat sections (16 μm) from paraformaldehyde-fixed tissue were performed as described previously (Schmidt et al., 2007; Schmidt and Rathjen, 2011).

The following primary and secondary antibodies using the indicated concentrations or dilutions were applied to tissue sections or Western blotting: Guinea pig antiserum to the extracellular domain of Npr2 (dilution: 1:5,000) (Ter-Avetisyan et al., 2014), guinea pig antiserum to amino acids 2-89 of cGKIα (1:25,000 for immunofluorescence staining), affinity-purified rabbit antibody to amino acids 2-89 of cGKIα (0.25 μg/ml for Western blot analysis), rabbit antibody to full length cGKIα expressed in Sf9 cells (1 μg/ml of the IgG fraction for Western blot analysis), chicken anti-β-galactosidase (1:5,000; Abcam, ab 9361; RRID:AB_307210), mouse anti-clathrin heavy chain (0.05 μg/ml; BD Biosciences, 610499; RRID:AB_397865), mouse anti-GAPDH (1:7,500; Novus Biologicals, NB300-221; RRID:AB_10077627), rabbit anti-PGP9.5 (1:1,000, Dako, Z5116, RRID:AB_2622233), rabbit anti-trkA (1:1,000; Millipore, 06-574; RRID:AB_310180), rabbit anti-parvalbumin (1:1,000; Swant, PV28; RRID:AB_2315235), rabbit anti-CGRP (1:1,000; Chemicon, AB1971; RRID:AB_2313629), rabbit anti-red fluorescent protein (1:2,500, ABIN129578; RRID:AB_10781500), isolectin GS-IB4-Alexa-488 (1:50; Life Technologies; RRID:AB_2314662), donkey anti-chick-IgY-Cy3 (1:1,000; Dianova), goat-anti-rabbit-Alexa488 (1:1,000; Dianova), goat anti-guinea pig-Alexa488 (1:1,000; Dianova), donkey anti-guinea pig-HRP (Dianova), goat anti-rabbit-HRP, and goat anti-mouse-HRP (all 1:20,000; Dianova,). 0.1 mg per g body weight of tamoxifen was applied by oral gavaging as described (Ter-Avetisyan et al., 2014).

Microscopic images were obtained at room temperature by confocal imaging using a Carl Zeiss LSM 710 NLO Laser Scanning Microscope equipped with ZEN 2010 software and the following lenses: a Plan-Neofluar 10x/0.30 NA objective, a Plan-Achromat 40x/1.40 NA oil objective, or a Plan-Achromat 63x/1.40 NA oil objective (all from Carl Zeiss MicroImaging, GmbH). Images were imported into Photoshop CS5 (Adobe) for uniform adjustment of contrast and brightness. Figures were assembled using Illustrator CS5 (Adobe).

Five to six-week old mice of either sex were anesthetized by an intraperitoneal injection of ketamine (100 mg/kg) and xylazine (10 mg/kg). 0.2 μl of 1.5% cholera toxin subunit B conjugated with Alexa Fluor 594 (CTB-AL594) (Wan et al., 1982; Robertson and Arvidsson, 1985; Conte et al., 2009) in 0.1 M phosphate buffered saline (PBS) were injected subcutaneously into the plantar surface of the left second hindpaw digit using a pulled glass capillary attached to a Hamilton microliter syringe. The glass capillary was inserted into the most distal interphalangeal crease and advanced under the skin toward the next proximal crease where the tracer was slowly injected. Five days post-injection, allowing for transganglionic transport of the tracer, mice were transcardially perfused with PBS and ice-cold 4% paraformaldehyde (PFA). Subsequently, the lumbar spinal cord was dissected out and post-fixed overnight in 4% PFA at 4°C. Fixed spinal cords were washed three times with 0.1 M PBS for 10 min each at RT. Subsequently, the tissue was immersed in ascending concentration series of 2,2'-thiodiethanol (TDE) (Staudt et al., 2007; Kloepper et al., 2010; Aoyagi et al., 2015; Costantini et al., 2015) for 24 h each at RT. The applied concentrations were 10, 25, 50, and 97% TDE diluted with 0.1 M PBS. During all incubation steps, the samples were kept on a vibrating table in the dark. For two-photon imaging, the cleared spinal cords were mounted on glass slides in 97% TDE which has a refractive index of 1.52 (Staudt et al., 2007) using press-to-seal silicone isolators with the dorsal surface facing up.

Two-photon imaging was performed using a laser scanning microscope equipped with a tunable Ti:sapphire laser. Two channels were recorded sequentially to collect Alexa Fluor 594 fluorescence (excitation wavelength: 810 nm; emission range: 600–735 nm) and tissue autofluorescence (excitation wavelength: 810 nm; emission range: 505–575 nm). A 25x multi-immersion objective (0.8 numerical aperture) was used with immersion oil for cleared sample imaging. Tiled stacks were taken through the spinal cord dorsal horn (pixel size: 0.5 × 0.5 μm; z step size: 1.5 μm). All images were processed using ImageJ (Schneider et al., 2012). Tiled stacks were stitched using either the imaging software ZEN 2010 or the ImageJ plugin “Stitching 2D/3D” (Preibisch et al., 2009). Subsequently, images were cropped to the same size and reduced to the same slice number. Next, background fluorescence was reduced by subtracting the autofluorescence channel from the CTB-AL594 channel. Using stack histogram-based thresholding, the image stacks were binarized. The threshold was set as the mean gray value plus three times the standard deviation. Finally, single pixels were removed to reduce noise, e.g., hot pixels. In order to enable comparative analyses of spinal terminal fields of cutaneous myelinated afferents, the three-dimensional centers of mass of the voxel clouds representing CTB-AL594 labeled fiber terminals were determined using the ImageJ plugin “3D ImageJ Suite” (Ollion et al., 2013). All images were aligned to the center of mass of the voxel cloud, i.e., images were cropped to the same size and reduced to the same slice number around the respective centers of mass. Summed dorsoventral, rostrocaudal, and/or mediolateral projections of the binary image stacks were constructed to enable two-dimensional visualization of terminal fields.

Ex vivo skin nerve preparations were performed as described (Moshourab et al., 2013). Briefly, mice were sacrificed and the glabrous skin on the left hindlimb was removed. The saphenous nerve was exposed dissected free along the lower leg up. Subsequently, the skin was carefully removed from the musculoskeletal and the connective tissue of the paw. The skin-nerve preparation was placed in an organ bath filled with oxygenated 32°C warm synthetic interstitial fluid (SIF) consisting of 123 mM NaCl, 3.5 mM KCl, 0.7 mM MgSO₄, 1.7 mM NaH₂PO₄, 2.0 mM CaCl₂, 9.5 mM sodium gluconate, 5.5 mM glucose, 7.5 mM sucrose and 10 mM HEPES, at a pH of 7.4. Using insect needles, the skin was mounted in the organ bath with its epidermis facing the bottom of the chamber, exposing the dermis to the solution. The nerve was pulled through a hole into the adjacent recording chamber which was filled with mineral oil. Finally, using fine forceps the nerve was desheathed by removing its epineurium and small filaments were teased of the nerve. Throughout the whole experiment the skin was superfused with oxygenated SIF at a flow rate of 15 ml/min.

Littermate mice of either sex were used in all behavioral tests (8 to 10-weeks old). The Npr2^fl/fl;Wnt1^Cre mice were on a mixed genetic background (Bl6/SV129) and cGKI^KO/fl;Wnt1^Cre mice were on a Bl6 background. Animals were habituated to the experimental room and were investigated by observers blinded for the genotype.

Sample size for behavioral experiments and for analysis of axon branching was deduced from previously published studies (Schmidt et al., 2007) or in publications listed above in the sub-section on behavioral testing. No further statistical methods were used to predetermine sample size. Sample size are given directly in the figures or in the legends. Experiments were done blind with respect to genotype. Littermates were used as control. Statistical analysis was performed with SPSS software (RRID:SCR_002865) using the Student's t-test for paired comparisons or Mann-Whitney U-test. Power analysis of statistical data was performed with the software “nQuery + nTerim 4.0.” Depending on the equality or inequality of variances and on the equality and inequality of the number of measurements t-test or Satterthwaite t-test were applied.

To generate a conditional allele of Npr2, two loxP sites were introduced into the introns flanking exons 17 and 18 of the mouse Npr2 gene which encode the guanylyl cyclase domain of Npr2. The neo selection cassette flanked by two FRT sites was removed by crossing this mouse line with a FLPE-deleter strain. The resulting floxed Npr2 allele contains two loxP sites surrounding exon 17 and 18 (Figure 1A). The correct integration of the loxP sites was verified by Southern blotting using (Figure 1B) a probe as indicated in Figure 1A and PCR was applied for genotyping using two distinct primer pairs that amplify genomic regions including loxP1 or loxP2, respectively (Figure 1C). Conditional inactivation of Npr2 in DRG neurons at early embryonic stages when sensory axons enter the spinal cord was obtained by crossbreeding with the Wnt1-Cre transgenic mouse line. At early embryonic stages Wnt1 is restricted to the midbrain, dorsal spinal cord and DRGs as detected by the Wnt1-lacZ reporter mouse (Echelard et al., 1994). Cre-mediated excision of the exon 17 and exon 18 generated a frameshift in exon 19 which leads to a premature stop codon. This resulted in nearly undetectable expression of Npr2 protein in DRGs as indicated by Western blotting using antibodies to the extracellular domain of Npr2 (Figure 1D).

About 50% of these Wnt1-Cre inactivated Npr2 mutants developed normally as demonstrated by a nasal to anal body length and a body weight that were indistinguishable from wild type or the homozygous floxed strain (Figures 1E–J). Furthermore, grip strength in these mutant mice appeared normal (Figure 1N). However, about 50% of Npr2^fl/fl;Wnt1^Cre mice were retarded in their development which revealed also a reduced survival rate at post-weaning stages (Figure 1F). Premature death of these mutants was accompanied by a severe malocclusion of the upper and lower incisors and intestinal distension which both affected normal nutrition (Figure S1). These mice were excluded from further testing. A malocclusion in conditional mutants of the Npr2 ligand CNP (Nakao et al., 2015) and milk retention in the stomach and intestinal distension have also been observed in the Npr2 mutant slw/slw (Sogawa et al., 2010).

Analysis of sensory axon bifurcation by DiI tracing at embryonic day 12.5 or by the reporter line Thy1-YFP-H at postnatal day P15 indicated a complete absence of axon bifurcation at the dorsal root entry zone of the spinal cord in the conditional mutant but not in the homozygous floxed allele (Figures 1K–M). DRG axons turn either in a rostral or caudal direction at the lateral margin of the cord. In contrast, the formation of collaterals from stem sensory axons in the dorsal funiculus is not impaired in the absence of Npr2 (data not shown) indicating that this type of branch formation is not dependent on Npr2 signaling as described previously (Schmidt et al., 2007). In contrast to conditional inactivation of Npr2 by Wnt1-Cre the use of other Cre-lines such as trkA-Cre or trkC-Cre did not induce bifurcation errors most likely due to later initiation of Cre expression with respect to Npr2 in embryonic DRG neurons (Figure S2). Overall, our data indicate that the Npr2^fl/fl;Wnt1^Cre mutant is a suitable model to study the consequences of impaired axon bifurcation in adult mice.

Sensory information from a large number of afferent axons converges within the spinal cord in nociceptive, mechanoreceptive or proprioceptive fields by the generation of collaterals that terminate in specific layers. Bifurcation of sensory axons at the DREZ contributes to this overall representational organization (Brown, 1981). The normal body length of the conditional Npr2 mutant allowed us to clarify whether the absence of bifurcation affects the size or shape of termination fields of afferents from the skin. A lack of bifurcation at the DREZ predicts altered receptive fields and less overlap between different sensory modalities. Therefore, we mapped the spinal terminations of cutaneous myelinated afferents from the hind paw in wild type and Npr2^fl/fl;Wnt1^Cre mice. The topography of primary afferent termination was visualized by transganglionic transport of fluorescently labeled cholera toxin B (CTB) subcutaneously injected into the second digit of the left hind paw of 5-week old mice (Figure 2A). CTB binds to the ganglioside GM1 which is highly expressed on axons (Kobbert et al., 2000). Injection in digit two resulted in a characteristic fork-like afferent fiber termination field in control Npr2^fl/fl spinal cords (Figure 2B). Total number of positive voxels representing CTB-labeled fibers as well as spans and area densities in dorso-ventral, medio-lateral, and rostro-caudal planes were calculated (Figures 2C–I). The total number of voxels was reduced to 64% for the 2nd digit (Figure 2C) indicating that the number of incoming sensory axons and collaterals were strongly reduced in comparison to wild type - surprisingly not to 50% as one might have been expected from the total lack of bifurcation at the DREZ. This observation might be explained by an increase in the density of collaterals protruding from the stem axons or by an increase in the terminal branching of collaterals in the mutant. Interestingly, the dorsoventral span of fluorescence intensity was increased to about 131% of control for the 2nd digit in the mutant whereas the medial-to-lateral distribution of terminal field was decreased by 25%. No difference was observed in the rostral-to-caudal span of the projection (Figures 2D–F). The density, the voxels per area, decreased to 66 and 58% in the rostrocaudal and in the mediolateral direction, respectively, whereas the dorsoventral direction decreased only to 76% (Figures 2G–I) for digit two. In summary, the absence of bifurcation caused decreased projection intensities together with a reduction in the mediolateral span as well as an increase in the dorsoventral extension of sensory afferent termination fields in the spinal cord.

Since axon bifurcation is compromised in all DRG neurons of Npr2-deficient mice we asked whether proprioception might be impaired. Therefore, balance and motor coordination in Npr2^fl/fl;Wnt1^Cre mice was examined using the balance beam test, rotarod, staircase assay, food grasping and reaching assay, and walking track analysis (MouseWalker). Balance beam and rotarod measure the ability to perform complex and coordinated movements. Npr2-deficient mice did not show a reduced latency to fall off or to slip off the beam (Figures 3A,B) or to fall off the rotating rod (Figure 3C). Similarly, no deficits were observed in the staircase and reaching and grasping test (Figures 3E–I), indicating that motor coordination was not altered in the absence of sensory axon bifurcation. To assess sensory and motor abilities related to the forepaw and mouth coordination the adhesion removal test was performed. Adhesive tape was fixed gently to the hairless parts of the forepaws and the time to first contact and the removal of both adhesive strips was recorded. Control mice and Npr2 mutants made regular attempts to remove the tape (Figure 3D). To ask whether the absence of sensory axon bifurcation interferes with the control of locomotion we applied the MouseWalker system which allows a comprehensive and quantitative description of parameters of freely walking Npr2-deficient mice (Figures 3J–L).

In summary, no deficits were observed in these behavioral assays in conditional Npr2 mutants (Figure 3). Npr2-deficient mutants possess considerable coordination capabilities.

To assess the consequences of impaired axonal bifurcation of sensory neurons for pain processing in vivo, we analyzed the nociceptive behavior of Npr2^fl/fl;Wnt1^Cre and respective control mice in various animal models of pain. We tested both male and female mice, but no significant effects of sex were detected in any assay.

We first determined the acute pain thresholds for noxious thermal stimuli in Npr2^fl/fl;Wnt1^Cre and littermate control mice using the hot plate test (50–54°C). Npr2^fl/fl;Wnt1^Cre mice displayed typical nocifensive behaviors (shaking or licking of the hindpaws or jumping off the plate), but at all tested temperatures they did so at consistently longer latencies than control mice (Figure 4A). Paw withdrawal responses to radiant heat (Hargreaves method) were also significantly increased in Npr2^fl/fl;Wnt1^Cre mice as compared to control mice (Figure 4B). In contrast to the impaired heat sensitivity, responses to acetone-evoked evaporative cooling were normal in Npr2^fl/fl;Wnt1^Cre mice in the group tested (Figure 4C; p = 0.295). Moreover, mechanical sensitivity of the hindpaw assessed using a dynamic plantar aesthesiometer was similar in Npr2^fl/fl;Wnt1^Cre and control mice (Figure 4D). Together, these data suggest that the loss of axonal bifurcation impairs the rapid response to avoid noxious heat, whereas behavioral thresholds and response latencies to cold or mechanical stimuli were not affected.

We next examined the tonic pain behavior in Npr2^fl/fl;Wnt1^Cre and control mice after injection of capsaicin or formalin into a hindpaw. The capsaicin injection (5 μg) resulted in licking and shaking the injected paw in both genotypes. Whereas, the licking behavior was not statistically different between groups (p = 0.104), the number of paw shakes was significantly reduced in Npr2^fl/fl;Wnt1^Cre mice compared to control mice in the test group (Figure 4E). Injection of 0.5% formalin into a hindpaw resulted in the typical biphasic paw licking. In the first phase (1–10 min), which results from acute activation of nociceptors, the licking behavior of Npr2^fl/fl;Wnt1^Cre and control littermates was comparable (Figure 4F). However, the formalin-induced paw licking behavior was significantly reduced in Npr2^fl/fl;Wnt1^Cre mice in the second phase (11–60 min), which involves a combination of ongoing afferent firing and central sensitization (Shields et al., 2010). These data indicate that nociception induced by the chemical irritants capsaicin or formalin are impaired by the loss of axonal bifurcation. Consistent with these results are our previously published electrophysiological recordings on global Npr2 and CNP knockouts which revealed a reduced number of neurons in the dorsal spinal cord responding to capsaicin (Schmidt et al., 2007, 2009).

In addition to its role in sensory axon bifurcation Npr2 could potentially have an acute function in the processing of sensory information for example in noxious heat perception or nociception induced by chemical irritants. Furthermore, the second messenger cGMP and cGKI contribute to the processing of inflammatory and neuropathic pain in the spinal cord (Tegeder et al., 2004; Schmidtko et al., 2008, 2009; Luo et al., 2012). To further explore whether the measured deficits in pain behaviors in Npr2 mutants are caused by the absence of sensory axon bifurcation or by changes in primary acute pain perception, we studied the developmental expression profile of Npr2 in DRGs by histochemical determination of β-galactosidase activity in cryostat sections from the Npr2^wt/lacZ reporter mouse and by Western blot analysis of tissue extracts of DRGs (Ter-Avetisyan et al., 2014). Consistent with our previous in situ hybridization data we observed that all DRG neurons strongly express Npr2 at early embryonic stages when sensory axons enter the spinal cord (Figures 5A, C–E). At postnatal stages, however, the level and the number of Npr2 expressing cells gradually decreases in DRGs and at post-natal day 75 only 10% were positive for Npr2. Similar results were obtained by analyzing red fluorescent protein expression in DRGs at P75 which was induced by crossing the Rosa26-tdTomato reporter mouse line with Npr2-CreERT2 and tamoxifen application at 3 consecutive days (P71, P72, and P73) (data not shown). Further analysis revealed that Npr2 is not confined to a specific subset of DRG neurons. It is expressed in a small group of trkA-, parvalbumin-, CGRP-, or IB4-expressing DRG neurons. The percentage of Npr2-positive cells in these subpopulations - except for parvalbumin - decreased during maturation (Figures 5F–I).

Co-stainings indicated that a small percentage of Npr2-positive DRG neurons at P75 express parvalbumin or trkA; therefore, it is possible that changes in peripheral nociceptor function may underlie altered heat sensitivity in Npr2^fl/fl;Wnt1^Cre mice. To test this, extracellular recordings from single C-fiber units in the hindpaw skin were performed using the ex vivo skin nerve preparation. C-fiber responses to heat, cold and mechanical stimulation of receptive fields were characterized in a total of 24 thermal-responsive C-fibers in Npr2^fl/fl;Wnt1^Cre mice, and 33 units in littermate control mice (Table S1). Example heat-evoked firing activity of single C-fibers in Npr2^fl/fl;Wnt1^Cre mice and littermate controls are shown in Figure 6A.

C-fiber heat thresholds, defined as the temperature required to evoke the first action potential, did not differ between Npr2^fl/fl;Wnt1^Cre mice and littermate controls (Figure 6B). Similarly, there was no significant difference in the number of action potentials in C-fibers during heat stimulation between Npr2^fl/fl;Wnt1^Cre mice and littermate controls in response to suprathreshold heat stimulation (Figure 6C, exemplified in Figure 6A). Cold thresholds of C-fibers were also not different between Npr2^fl/fl;Wnt1^Cre mice and littermate controls (Figure 6D). Mechanical thresholds and suprathreshold mechanical stimulus-evoked firing activity of C-fibers also did not differ between Npr2^fl/fl;Wnt1^Cre mice and littermate controls (Figures 6E,F). Example mechanical-evoked firing activity of a C-fibers recorded from a Npr2^fl/fl;Wnt1^Cre mouse is shown in Figure 6G. These data indicate that thermal and mechanical coding properties of cutaneous C-fibers are not functionally impaired in Npr2 mutant mice.

cGKI is an essential signaling component downstream of Npr2 to induce sensory axon bifurcation as shown previously (Schmidt et al., 2002, 2007; Zhao et al., 2009). However, the developmental expression profile of cGKI is distinct from Npr2. At early embryonic stages when sensory axons are bifurcating in the spinal cord a complete overlap between Npr2 and cGKIα in all sensory neurons was observed (Figures 5B,D,E). In contrast to Npr2, cGKIα remains expressed in all sensory neurons also at mature stages suggesting that its function might be regulated by other cGMP synthesizing enzymes than by Npr2 in adult mice.

The absence of cGKI in a constitutive knockout does not interfere with bone growth but shows deficits in the regulation of smooth muscle contraction with vascular and intestinal dysfunctions which cause pre-mature death (Hofmann et al., 2006). Behavioral testing therefore also required conditional inactivation of cGKI in DRG neurons, which was obtained by a floxed allele of the cGKI gene (Wegener et al., 2002) and using Wnt1-Cre as described for Npr2. This resulted in a complete loss of cGKI protein in DRGs as indicated by Western blotting using antibodies to cGKI (Figure 7A). Consequently, sensory axons failed to form bifurcational branches at the DREZ (Figures 7B,C). In contrast to conditional Npr2 mutants, no decrease in the survival rate or any problems on the health status of 23 inspected cGKI^KO/fl;Wnt1^Cre animals in which cGKI was conditionally inactivated was observed. In some tissues Npr2-induced cGMP signals are mediated by the cGKII instead of cGKI which might explain these phenotypic differences on the survival rate between cGKI and Npr2.

To estimate whether the observed Npr2-dependent changes in pain processing are mediated by a Npr2/cGMP/cGKI signaling pathway, we analyzed the nociceptive behavior of cGKI^KO/fl;Wnt1^Cre and littermate control mice (cGKI^KO/fl). In the hot plate test, cGKI^KO/fl;Wnt1^Cre mice demonstrated reduced noxious heat sensation (Figure 8A), albeit with significant differences compared to control mice only at 50 and 54°C. In the acetone test (Figure 8B) and dynamic plantar test (Figure 8C), the responses to cold and mechanical stimuli were normal in cGKI^KO/fl;Wnt1^Cre mice. After capsaicin injection into a hindpaw, the paw licking behavior of cGKI^KO/fl;Wnt1^Cre mice was reduced as compared to cGKI^KO/fl control mice (Figure 8D). Altogether, both Npr2^fl/fl;Wnt1^Cre and cGKI^KO/fl;Wnt1^Cre mice demonstrated impaired responses to noxious heat and chemical stimuli but normal responses to cold and mechanical stimuli, suggesting that the loss of bifurcation causes specific deficits in nociceptive processing.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.