The branching of vertebrate neuronal processes, axons or dendrites, is predominantly binary, where a proximal neuronal process bifurcates into two distal progeny. Here, we illustrate counterexamples where neuronal processes branched into more than 2, up to 6, progeny branches, in one step. These were branch points of myelinated dendrites in the peripheral terminals of ALLn cranial nerve afferents innervating the Lorenzinian-type ampullary electroreceptors on the rostrum of paddlefish, Polyodon spathula. We imaged afferent terminals fluorescently (in widefield stacks, with deconvolution) after immunolabling of afferent terminals for neuronal cytoplasmic neurofilament-H (NEFH), myelin markers (MBP, P0), or nodal ion channels (Nav1.x, Kv1.1), or after migration of DiI in dendrite membranes. Branched afferent terminals formed a laminar radial radiation beneath a receptive field, parallel to the skin surface, with two serial stages: (1) Starting at each afferent's centric first branchpoint, each of a small group of 2–4 (most often 3) afferents branched radially into 2–4 (usually 3) generations of myelinated dendrites, whose internodes were covered by sheaths immunoreactive for antigenic markers of myelin (MBP+/P0+), ending distally at ~15 heminode presumed spike initiation zones. (2) From the latter, bundles of unmyelinated (MBP-/P0-) sensory neuron (NEFH+) processes projected distally to innervate electrosensory neuroepithelia of adjacent ampullary organs. The nonbinary branch points that we imaged were in stage-1, on myelinated dendrites. Branch points always coincided with composite systems of nodes, in which each progeny dendrite started at a narrowed nodal segment expressing voltage gated sodium ion channels at high density. These narrowed nodal segments of multiple progeny branched in parallel from a parent’s blunt distal end. Hence the branch point nodal complexes formed multisite excitable systems, likely strongly coupled due to proximity.

Branchings of vertebrate dendrites and axons, the elongated processes of neurons, are considered predominantly binary^[1][2], whereby a proximal parent process gives rise to only two distal progeny processes, at branch points resembling ‘Y’ or ‘T’, forming binary trees. Branch points seeming to have >2 progeny typically prove to be composed of a series of bifurcations. In dendrites of vertebrate CNS neurons, binary branching can range from progeny having equal size (symmetrical), to highly asymmetrical dendritic spines. Progeny diameters (d1, d2) are typically smaller than the parent’s (D) as: $D^{N}=d_{1}^{N} + d_{2}^{N}$ where N varies from 1.5 to 2 corresponding to symmetrical d/D ratios of 0.63–0.71^[2].

The branching patterns of vertebrate myelinated axons have been studied less. The peripheral terminations of large myelinated sensory neuron axons (afferents) at sensory receptor organs typically include short myelinated branches^[3][4]. Their branch points coincide with nodes of Ranvier^[5]. Myelinated afferent branches at vertebrate sensory organs are termed ‘dendrites’ because they receive signalling input from presynaptic receptor cells or distal transduction sites, the branches perform like dendrites in collecting input from different sites, and because they transmit information (in spike trains) toward the soma of a sensory neuron (near the vertebrate CNS)^[6].

This report illustrates clear counterexamples to the ‘rule’ of binary vertebrate neuronal branching. We analyzed nonbinary branchpoints in the myelinated terminal arbors of sensory neurons innervating Lorenzinian-type ampullary electroreceptors (ERs) in North American paddlefish, Polyodon spathula. These sensory receptor organs respond to weak external electrical signals from prey. They are embedded in specialized skin on the rostrum of paddlefish, a flattened electrosensory appendage in front of the head, and adjacently. Their Lorenzinian-type receptor cells synaptically excite continuous firing in the terminals of a small group of parallel primary afferents, each with a myelinated projection axon in the ipsilateral Anterior Lateral Line sensory cranial nerve (ALLn)^{[7][8][9][10][11]}.

Abbreviations: BP: Branch point. DiI: diI-C18-(3); orange-red fluorescent lipophilic dye that diffuses planarly in cell membranes; 1,1'-dioctadecyl-3,3,3',3'-tetramethylindo-carbocyanine perchlorate; CAS 41085-99-8. ER: Electroreceptor. +ir: Positive immunoreactivity. µ: Micron ($1x10^{-6}$ meter).

By immunolabeling or DiI tracing, and fluorescence imaging, we demonstrated nonbinary branch points on the myelinated dendrites of afferent terminals innervating paddlefish ERs, and we analyzed components of nodal complexes that always coincided with branch points of myelinated afferent dendrites. Our broader aim was to describe and identify motifs in the afferent innervation of electroreceptors on the rostrum of adult paddlefish.

Figures 1A and 1B portray the global organization of afferent innervations of ampullary electroreceptors (ERs), in specialized skin covering the rostrum. We studied tissue from the rostrum base in 1–2 year paddlefish, representative of adult structure. The ~0.5–2 mm-wide receptive field of an ER consisted of ~15–30 adjacent ampullary organs (bright oval profiles, Fig. 1A inset and 1B), each leading to a skin pore, in a few rosettes. Different ampullary organs were 0.25–0.45 mm deep. The interior basal face of each was covered by an electrosensory neuroepithelium containing receptor cells^{[8][12][13][14]}. In thick frozen sections or in vibratome wholemounts, imaged parallel to the skin surface, we counted 3.08 ± 0.51 (range 2–4, usually 3) parallel afferents innervating the receptive field of 12 ERs from 4 paddlefish. For example, the ER in figure 1A was innervated by 3 different afferents. The terminals of the individual afferents branched from a small compact electrosensory afferent nerve (N, Fig. 1A), which entered an ER from deeper tissue. Such nerves derived from ALLn subdivisions^[7]. The nerve to an ER routinely entered centrically, near the center of the receptive field (imaged as a group of innervated ampullary organs). Branching of each afferent radiated away from its first branch point (1º, primary, or initial BP), and formed a laminar radial radiation parallel to the skin, at ~550–750 µ subsurface depth, that innervated all of the ampullary organs in a receptive field. This laminar stratified organization of ERs on the rostrum base, with afferent terminals in a layer deep to ampullary organs, differed from the organization of other specialized ERs in the thin flexible hypobranchial skin^[10][15]. In cross sections of electrosensory nerves near ERs (like N in Fig. 1A), the sensory neuron axons had large 8–16 µ diameters, of homogeneous size in a given nerve.

We immunolabeled afferent terminals at ERs with commercial 1º Abs to key marker proteins of peripheral myelinated axons (Methods). Afferent branches were myelinated proximally near the nerve entry, then unmyelinated distally near ampullary organs, in two distinct radial stages in series (Fig. 1A and 1B). Proximal myelinated stage 1: From the entry site of an electrosensory nerve, afferents underwent 2–4 generations of radial branching in which the parent and progeny afferent processes appeared myelinated, with thick‑wall MBP+ or P0+ sheaths surrounding NEFH+ sensory neuron cores interrupted by NaV1.x+ and KV1.1+ nodes. These proximal MBP+/P0+ sheathed afferent branches were termed ‘myelinated dendrites’ (Introduction). Elongated 1st‑generation (gen1) myelinated dendrites, 2 to 4 in number, radiated away from the central 1º branch point (BP) of each afferent, directed along approximately straight paths in different planar directions of a laminar radiation, ~200 µ thick parallel to skin. Gen1 dendrites had few (0–2) inline nodes. The gen1 dendrites of different afferents were grouped in parallel in loose bundles (fascicles; #1,#2,#3 in Fig. 1A). Each gen1 dendrite ended distally at a 2º BP. The bundles of gen1 afferent dendrites projected to distal zones of 2º and 3º BPs and in some cases 4º BPs, which matched up with subgroups of ampullary organs (Fig. 1A inset). Distal myelinated branches were shorter and had multiple short segments of myelin (arrowhead, Fig. 1A) interrupted by gaps. We confirmed by triple labeling of NaVpan, NEFH, and MBP that such gaps corresponded to NaV+ focal rings on afferents, and hence were inline nodes. Thus inline nodes were more frequent distally. Some distal nodes were less- or not narrowed (not shown). Myelination terminated at a final NaV1.x+ and KV1.1+ (not shown) node of Ranvier; such heminodes (h, Fig. 1A,1B) were likely spike initiation zones as in other myelinated afferents. Thus there were likely multiple (~12 in Fig. 1A) presumed spike initiation sites at the distal extremities of each afferent’s myelinated branches^[16], as at other vertebrate sensory organs having myelinated afferent branches^{[17][18][19][20][21]}. Distal unmyelinated stage 2: Distal to each heminode was a local unmyelinated (MBP– and P0–) continuation of each afferent, projecting as bundles of fine parallel NEFH+ neurites (arrowheads in Fig. 1B) that contacted and branched profusely on the basal faces of electrosensory neuroepithelia of 1–3 adjacent ampullary organs, ending at synapses with receptor cells^[8]. The unmyelinated afferent projection bundles were ensheathed by terminal glia (not shown). These unmyelinated afferent projections were akin to neurite endings at mammalian cutaneous mechanoreceptors^[22][23][24]. Variants: The myelinated branching patterns of the different afferents innervating a given ER were similar and parallel but usually not identical. The count of myelinated branches in any dendritic bundle was similar to the number of nerve afferents, but could differ. For example, in figure 1A, bundle #3 had 3 dendrites (like the nerve) but #2 had 4. A minority of myelinated progeny continued from a 2º BP to a heminode without further branching (*, Fig. 1B), or crossed a zone of other BPs without branching. Some BPs gave rise to supernumerary progeny branches. Also, some 1º or 2º branchings were compound, comprising a pair of adjacent BPs in series (*, Fig. 1A). As a result of these variations, the count/bundle of myelinated dendrites tended to increase at more distal stage‑1 branching.

Nonbinary branch points, with ≥3 progeny neuronal processes, were frequently observed on myelinated dendrites (Fig. 1B–1E). While binary BPs were most frequent, BPs with 3 progeny were comparable in abundance, and BPs with 4–6 progeny (Fig. 1D) were routine. Multiple progeny branches typically arose in one step, not as serial bifurcations (Fig. 1D). The diameters of sensory neuron progeny (d) branches were smaller than the parent diameter (D). For example, in figure C1 at 50 µ radius from the BP center, the NEFH+ parent dendrite was 17.6 µ wide, compared to 10.7, 16.9, 11.9 µ widths of the 3 NEFH+ progeny branches, for d/D ratios of 0.61, 0.96, 0.68, mean 0.76. Thus the summed cross sectional area of the progeny NEFH+ dendrites was 1.75‑fold greater than the parent’s area (calculated from widths). In a sample of 30 NEFH-labeled BPs having 2–6 progeny dendrites (from 7 ERs, 3 paddlefish), the d/D ratio was 0.66 ± 0.13 (82 values). The d/D ratio tended to vary inversely with progeny count (correlation coefficient r = –0.52). The summed cross sectional area of a BP's progeny (at 50–100 μ from a BP’s center) was 1.24 ± 0.37 (30 values) normalized to the respective parental area; the inequality posed questions about cytoskeletal transport^[25] at these myelinated afferent BPs.

The branch points of myelinated dendrites always coincided with compound nodes of Ranvier. These networks of nodes, the intersection of 3–7 neuronal processes, immunolabeled positively for neurofilament‑H (Fig. 1B, 1C1, 1C3, 1D) in neuronal cytoplasm, myelin basic protein (Fig. 1A, 1B, 1C2, 1C3) or P0 (not shown) in dendrite sheaths, and voltage gated ion channels including NaV1.x sodium channels (Fig. 1C3, 1C4) and KV1.1 potassium channels (not shown), as at classic inline axonal nodes^[6][26][27]. Each of a BP’s progeny neuronal processes started at a nodal segment of narrowed diameter. These 2–6 narrowed progeny segments connected to the blunt slightly dilated end of a parent neuronal dendrite. For example, in figure 1C1, the 3 narrowed segments were 4.3, 7.0, 5.0 µ wide, or 40, 42, 39 % of the full inline widths of progeny (at 50 μ distance, as above). For a sample of BPs (as above), the widths of the narrowed nodal segments of progeny were 41 ± 8% (62 values) that of their respective full width progeny dendrite more distally. Also, stripped ALLn myelinated axons had narrowed inline nodes (not shown). The structural motif of narrowed neuronal segments bearing focal NaV+ rings, like classic axonal nodes, confirmed that these initial segments of afferent progeny at BPs were true nodes of Ranvier.

The NEFH+ immunofluorescent labeling of dendrite neuronal cytoplasm was less bright at branched nodes, compared to dendritic shafts. Such reduction of NEFH+ ir was greatest at the narrow-diameter segments of nodes, and at the distal end of parent neuronal processes. For example, in figure 1C1, a profile plot (inset) across this Ψ‑shaped node (white line, starting at ►), showed ~75% maximal reduction of NEFH +ir. It fell sharply on the parent side of the node, but on the progeny side the NEFH+ ir recovered only gradually over a ~50 µ space distal of the BP center. Low NEFH ir was also visible at BP nodes in figure 1D, and at narrowed node-like inline sites on afferent dendrite branches (*, Fig. 1B). The nodes of stripped ALLn axons also showed reduced NEFH ir, colocalized with NaV1.x ir on narrowed nodal segments (not shown). Thus, low NEFH ir was characteristic of nodes on myelinated ALLn afferents innervating paddlefish ERs.

MBP+ sheaths extended into nodes, and covered most of the narrowed progeny nodal segments, except at narrow slit-like gaps between adjacent ‘myelin’ sheaths at BPs, 3.3 µ wide in figure 1C2. Hence the NaV1.x foci were unsheathed on these BP nodes, as at inline axonal nodes. The MBP+ sheaths had 2.5–4 μ thick walls, and in figure 1C3 the sheaths had ~45% greater outer diameter than the NEFH+ cytoplasm of respective neuronal branches. Reduced diameters and altered morphologies of MBP+ sheaths were often observed on short branches from 3º or 4º BPs, but on other dendrites (*, Fig. 1B) the MBP+ sheath remained thick all the way to a heminode (h, Fig. 1B).

The spatial distribution of NaV1.x ion channels at BP nodes was analyzed by immunolabeling (green, Fig. 1C3, 1C4), and by profile plots of NaV ir along different paths across BPs (Fig. 1C4, 1C5). The enlarged view of NaV +ir in figure 1C4 was a single raw optical section with ~1 μ depth of field, in the plane of a large 1º BP from a >2 year fish (parent afferent axon diameter >17 µ). It suggested a single high-density ring of NaV1.x +ir around each narrowed progeny nodal segment, near its mid-length, as at axonal nodes^[6][22][27]. Hence the red line path in figure 1C4 yielded a profile plot with sharp peaks on either side of the narrowed progeny segment, consistent with a focal NaV ring around it. The double peak (*, Fig. 1C5) was consistent with side-by-side focal NaV rings (*, Fig. 1C4) around the narrowed segments of two adjacent progeny branches, near where they joined to the parent’s distal end. Also, diffuse green fluorescence suggested possible nonfocal NaV +ir distributed at lower density within the BP (see Alternative Explanations, below).

We also imaged nonbinary BPs by an alternate approach, DiI tracing (Fig. 1E). DiI migrated in the membranes of sensory neuron segments and also in ‘myelin’ sheaths, retrogradely and transcellularly^[28] from small DiI crystals placed on adjacent electrosensory neuroepithelia (not shown). Figure 1E shows 2 BPs of individual afferent dendrites having 3 progeny branches (‘3’). The frequent dark striations on dendritic sheaths were due to surface melanin pigment; melanin was abundant in paddlefish ERs. Such striations provided visual confirmation that parent and progeny branches were covered by ‘myelin’ sheaths. A profile plot (Fig. 1E, inset) of DiI emission along a segmented trans‑BP path (white line in Fig. 1E) indicated that DiI emission fell by 60–70% at nodes, presumably on account of ‘myelin’ sheath (carrying DiI) being absent from nodal gaps (as Fig. 1C2). Low DiI emission at nodes was confirmed for a nonbinary BP node (green circle in Fig. 1E and its inset plot). DiI emission was similarly low at other plausible inline nodal sites marked by corresponding colored circles. Thus, local sites of abruptly lower DiI emission were candidate markers for nodes along myelinated afferent branches.

The NaV1.x ion channels at BP nodes formed spatially distributed excitable systems, in which the NaV+ nodal segments of progeny dendrites were parallel elements, whose proximal ends were interconnected through the interior of the parent’s blunt distal end. Terms to describe this glove-like spatial configuration might include ‘network’, ‘composite’, or ‘multifingered’, but ‘branched node’ would seem inaccurate because each branch point included multiple nodal components, and because the narrowed nodal segments per se were never branched. Close proximity likely ensured that the ensemble of NaV+ loci at a BP was strongly coupled via intracellular current flows. In an alternate configuration of nodes at branch points, symmetrical bifurcation of a mouse motorneuron axon had 3 narrowed ‘nodes’ joined end-to-end by a brief 3‑port full width segment (Fig. 3 in^[29]).

Branching of myelinated afferents occurs commonly at other vertebrate cutaneous and proprioceptive sensory organs^[4][17], and muscle spindle “pre-terminal” branch points with 3 or 4 myelinated progeny were reported^[3], as were nodes at branch points of mammalian muscle afferents^[5]. Hence our results on myelinated afferent nonbinary branching and branch point nodal complexes may apply to other species and sensory modalities.

We showed that myelinated afferents innervating the ampullary organs of ‘adult’ paddlefish electroreceptors may branch in a nonbinary manner, with up to 6 myelinated progeny branches. Each started at an initial narrowed nodal process, expressing expected molecular components of nodes including focal voltage gated sodium ion channels. The progeny nodal segments branched in parallel from a parent’s blunt distal end. We documented that sites of reduced NEFH ir or reduced DiI emission can serve as markers for node loci. Myelinated afferent branches with nodes and nodal branch points can be considered within the contexts of active neuronal dendrites or strongly coupled oscillators^[30][31].

The thick MBP+ or P0+ sheaths of paddlefish ER afferent dendrites are likely myelin. Although the “pre-terminal” branches of other vertebrate afferents have been assumed to be myelinated^[3][4][5], proof of this would require electron microscope confirmation of compact myelin wrappings there.

Scattered or out-of-focus fluorescence may be an alternative explanation of what appeared as low density diffuse NaV +ir inside a BP area (Fig. 1C4). For example, a broad peak as the black profile plot of figure 1C5 could be explained as out-of-focus light across a NaV focal ring around a progeny branch.

Specific molecular components related to branching may be present at myelinated afferent branch points^[32].

Possibly the afferent branching illustrated here in 1–2 year fish might have occurred symmetrically at early developmental stages, then been modified to assume the >2 branchings observed.

Simulations may clarify whether nodal networks with geometries as observed here at branch points may have directionality or filtering properties appropriate for centripetal transmission of spikes in myelinated afferents.

We analyzed soft skin tissue on the base of the paddlefish rostrum, from n=6 paddlefish, 1–2 years old and 25–42 cm in eye-to-fork length. For immunolabeling, paddlefish were minimally fixed by brief vascular perfusion of 4% paraformaldehyde and picrate, and immersion. Tissue blocks were then dissected from the rostrum base, washed, cryoprotected, and stored at -30°C. Frozen sections were cut 10–50 µ–thick on a cryostat, and air dried onto amine-adhesive slides.

Indirect immunofluorescent labeling was conventional, using commercial primary antibodies (1º Abs) against marker proteins of myelinated axons, and goat secondary (2º) Abs, along with blocking by goat serum, permeabilization by 0.3% v/v Triton X100 detergent, and mounting/clearing in VectaShield. The 1º Abs to NaV1.x voltage gated sodium ion channels included polyclonal Alomone ASC-003 (rabbit, AB_2040204), and monoclonal Sigma-Aldrich S8809 (mouse, AB_477552). Voltage gated potassium KV1.1 ion channels were labeled using Alomone APC-009 (rabbit, AB_2040144). The 1º Abs to myelin markers included anti-myelin basic protein (MBP; GenScript A01407, rabbit, AB_1720890) and anti-protein zero (P0; Neuromics CH23009, chicken, AB_1619444). The cytoplasm of neuronal processes was labeled using anti-neurofilament‑H (NEFH; Aves Labs NFH, chicken serum, AB_2313552). Rabbit Abs were affinity-purified.

Afferent processes were also traced by fluorescent DiI migration in membranes^[28]. Tissue blocks from the rostrum base were hard-fixed by vascular perfusion with 4% paraformaldehyde (without picrate or alcohol) then continuously immersed in it. Blocks were coated with agarose to restrict contamination. Small DiI crystals (Molecular Probes D282) were placed manually.

A widefield Nikon epifluorescence microscope had stepper-motorized focus, an electrical shutter for a Hg arc, and 3 single-fluor filter sets (Chroma) with ~50 nm wide blue, green, or orange/red emission bands. Its CCD camera was a monochrome 2048×2048 pixel Diagnostic Instruments camera like model IN1410, or a model IN421 color camera with 1600×1200 pixels, both with 7.4×7.4 μ pixels. The colors in illustrations match the emission bands of the 2º Abs used. Images were processed or measured using Adobe Photoshop CS6, Fiji-ImageJ, or AutoQuant 9.3 software. Most illustrations show flattened projections from image z‑stacks, but panels C1–4 show individual images. Images were routinely 2D-deconvolved. Values show mean ± SD.

Supported by National Institutes of Health (USA) grant 5R21GM103494, and by research funds and Biomimetic Nanoscience and NanoTechnology infrastructure funds from Ohio University (USA), and by a pilot grant from the Ohio University Research Council.

A.B. Neiman participated in valuable discussion.

The authors declare no conflicts of interest.

A protocol for these experiments (16-L-020) was approved by the institutional animal care and use committee at Ohio University (USA), conforming to National Institutes of Health (USA) guidelines.

No fraudulence is committed in performing these experiments or during processing of the data. We understand that in the case of fraudulence, the study can be retracted by ScienceMatters.