Elephant facial motor control

Abstract

We studied facial motor control in elephants, animals with muscular dexterous trunks. Facial nucleus neurons (~54,000 in Asian elephants, ~63,000 in African elephants) outnumbered those of other land-living mammals. The large-eared African elephants had more medial facial subnucleus neurons than Asian elephants, reflecting a numerically more extensive ear-motor control. Elephant dorsal and lateral facial subnuclei were unusual in elongation, neuron numerosity, and a proximal-to-distal neuron size increase. We suggest that this subnucleus organization is related to trunk representation, with the huge distal neurons innervating the trunk tip with long axons. African elephants pinch objects with two trunk tip fingers, whereas Asian elephants grasp/wrap objects with larger parts of their trunk. Finger “motor foveae” and a positional bias of neurons toward the trunk tip representation in African elephant facial nuclei reflect their motor strategy. Thus, elephant brains reveal neural adaptations to facial morphology, body size, and dexterity.

Elephant trunk control involves huge numbers of facial motor neurons, cell size gradients, and motor foveae.

INTRODUCTION

Elephant brains (1, 2) are functionally largely uncharted because of a lack of access to brains and the nonfeasibility of invasive mapping. We sought to break this barrier by a several decade collection effort, quantitative nerve tracing, comparative analysis, simple/robust staining methods, and detailed analysis of cell number, size, and position. We investigated four African savanna elephants, Loxodonta africana, further referred to as African elephants, and four Asian elephants, Elephas maximus (3). Our material had a variety of limitations (see Materials and Methods), and not all measurements could be taken on all specimens; a detailed list of specimens and how they were processed is given in Table 1.

Table 1. Overview elephants, treatment of corresponding specimen, and derived data.

Name (species)	Sex	Location at death	Age (years), died on (dd/mm/yyyy)	Specimen treatment	Data derived
Aruba (L. africana)	F	Opel-Zoo Kronberg, Germany	41, 11/03/2020	Brain removed, put in fixative	Cell count
Zimba (L. africana)	F	Opel-Zoo Kronberg, Germany	39, 10/04/2021	Brain removed, briefly frozen, put in fixative	Cell count, nerve tracing, fiber count
Linda (L. africana)	F	Zoo Poznan, Poland	35, 02/02/2021	Brain removed, briefly frozen, put in fixative	Cell count, nerve tracing, fiber count
AM 1 (L. africana)	M	Colchester Zoo, Great Britain	0, 2011	Frozen for several years	Bilateral nerve tracing, fiber count
Dumba (E. maximus)	F	Elefantenhof Platschow, Germany	44, 18/02/2022	Brain removed, put in fixative	Cell count
Burma (E. maximus)	F	Zoo Augsburg, Germany	52, 16/06/2021	Brain removed, put in fixative	Cell count, nerve tracing, fiber count
Asian Baby (E. maximus)	?	Tierpark Hagenbeck, Germany	0, 2012	Frozen for several years	Cell count
Brainless Baby (E. maximus)	M	Parc Pairi Daiza, Belgium	0, 2013	Frozen for several years	Bilateral nerve tracing, fiber count

RESULTS

The elephant facial nucleus differs from other mammals

A Nissl-stained coronal section through the brain stem of adult Asian elephant cow Burma is shown in Fig. 1A. The section is taken at midlevel of the right elephant facial nucleus (white outline). A somata drawing reveals a large number of neurons (Fig. 1B; micrograph, Fig. 1C). The elephant facial nucleus was subdivided in several subnuclei (color-coded in Fig. 1D, top). A musclelotopy suggestion (Fig. 1D, bottom) is discussed below. Somata drawings of other mammalian facial nuclei are shown in Fig. 1E. The elephant facial nucleus differs from other mammals in its vast size (4) and low cell density. We counted neurons in the elephant facial nucleus either by a classical optical fractionator approach (Table 2), which is based on the randomized sampling of facial nucleus neurons, or by simply sampling all neurons in every 10th section of the facial nucleus (Table 3); both approaches led to similar counts and conclusions. We compared cell counts of individual elephant facial nuclei (Fig. 1F, red filled dots), with the species averages reported by Furutani and Sugita’s (5) and Sheerwood’s (6) seminal comparative studies. Elephant facial nucleus neurons are more numerous than in all land-living mammals (54,103 ± 5804 in Asian elephants; 62,961 ± 4645 in African elephants; mean ± SD, averages over neuron counts). Only dolphins have higher facial nucleus neuron counts than elephants. Elephant facial nucleus counts were significantly different between Asian and African elephants and come to roughly five times the gross mammalian average. We also counted facial nerve axons in two nerves of one elephant and came to numbers (Table 4 and Fig. 1F, red empty squares) similar to our cell counts. Elephant facial nuclei are very large and differ between Asian and African elephants.

Fig. 1. Extraordinary size and neuron number of the elephant facial nucleus.

(A) Micrograph of a coronal Nissl-stained 60-μm brainstem section of the adult Asian elephant cow Burma. The section stems from midlevel of the facial nucleus (white outline). D, dorsal; L, lateral. (B) Somata drawing and facial nucleus outline. (C) Micrograph [dashed area in (B)]. (D) Top: Facial subnuclei of the section in (B) highlighted in color. Bottom: Color-coded musclelotopy suggestion. (E) Somata drawings of facial nuclei of various mammals, scale as (B). (F) Facial nucleus neuron number in mammals. Species averages are given in black [dots data (5); triangles data (6)]. In red, data are given for individual African and Asian elephant facial nuclei; filled symbols (cell counts), empty symbols (facial nerve fiber counts), squares (stillborn elephants), dots (adult elephants) (see fig. S1). Photo credit (D): Michael Brecht, Humboldt-Universität zu Berlin; Zoologischer Garten Berlin, Berlin, Germany.

Table 2. Optical fractionator counts.

Cell count estimates were derived using the optical fractionator; every 10th section was sampled. See text for details. FN, facial nucleus.

Specimen	Cells sampled	Gundersen Coefficient of error	Facial nucleus neuron estimate
Aruba left FN (African)	281	0.06	57,550
Aruba right FN (African)	302	0.06	60,731
Linda left FN (African)	305	0.07	70,585
Linda right FN (African)	289	0.06	67,868
Zimba left FN (African)	312	0.06	59,785
Zimba right FN (African)	318	0.06	61,252
Average African elephants (±SD)	301	0.062	62,961 ± 4,645
Burma left FN (Asian)	207	0.08	52,574
Burma right FN (Asian)	218	0.08	53,599
Dumba left FN (Asian)	259	0.06	59,376
Dumba right FN (Asian)	285	0.06	63,609
Asian Baby left FN (Asian)	239	0.06	47,457
Asian Baby right FN (Asian)	238	0.07	48,004
Average Asian elephants (±SD)	54,103 ± 5,804	0.068	54,103 ± 5,804

Table 3. Full counts of every 10th section corrected for double counts.

Cell count estimates were derived by counting all cells in every 10th section and correcting for double counted cells.

Specimen	Cells sampled	Double count correction	Facial nucleus neuron estimate
Aruba left FN (African)	6,966	0.191	56,421
Aruba right FN (African)	6,519	0.191	52,804
Linda left FN (African)	7,754	0.14	68,387
Linda right FN (African)	7,952	0.14	66,684
Zimba left FN (African)	7,855	0.14	67,863
Zimba right FN (African)	7,761	0.14	67,051
Average African elephants (±SD)	7,468		63,202 ± 6,187
Burma left FN (Asian)	6,205	0.1705	51,469
Burma right FN (Asian)	6,318	0.1705	52,405
Dumba left FN (Asian)	7,076	0.205	56,112
Dumba right FN (Asian)	7,434	0.205	58,952
Asian Baby left FN (Asian)	5,548	0.134	47,036
Asian Baby right FN (Asian)	5,532	0.134	46,900
Average Asian elephants (±SD)	6,352	0.068	52,146 ± 4,401

Table 4. Facial nerve axon counts.

Specimen	Axons sampled	Total bundles/ bundles counted	Facial nerve axon estimate
Brainless Baby Left facial nerve (Asian)	8,973	236/33	48,846
Brainless Baby Right facial nerve (Asian)	1,580	207/10	45,896

Allometric analysis of elephant brain size and facial nucleus neuron number

We wondered whether the large number of elephant facial nucleus neurons is simply a reflection of the large brain size of elephants with their brain weight of 4.7 kg (2). To assess this possibility, we plotted facial nucleus neuron number of a variety of mammalian species against brain weight in a log-log plot (Fig. 2). To better understand the relationship of facial nucleus neuron number and brain weight, we fitted power functions to data from all mammals and to the data from primates [a mammalian group for which Sherwood assembled a large dataset (6)]. This analysis indicated two conclusions. First, it suggested that elephants have a large facial nucleus compared to primates (i.e., the red dots of the two elephant species are well above the black line fitted to the primate data). To assess how different elephant facial nucleus neuron numbers were from primates, we fitted a linear model using a combination of the fitlm and predict Matlab functions (The MathWorks Inc.) to the primate data. We computed 95% confidence intervals for the regression of the primate data (Fig. 2, black dots). We found that facial nucleus neuron number for both elephant species is about three times 95% confidence intervals away from the primate data prediction (when plotted on the log scale), suggesting that elephant facial nucleus neuron number is significantly different from what would be expected for primates. Second, the data suggest that elephants are not the only mammals with excess numbers of facial nucleus neurons; dolphins also have unexpectedly numerous facial nucleus neurons. We note that our allometric analysis is based on across-study comparisons, which comes with a grain of salt. Specifically, one needs to take into account that the data come from a range of different studies over the course of decades that have all used different methods for estimating neuron numbers, i.e., the data might not be strictly comparable. We tentatively conclude that elephants have comparatively many facial nucleus neurons, but they are not the only mammals with large numbers of such neurons.

Fig. 2. Allometry of brain weight and facial nucleus neuron number in mammals.

Facial nucleus neuron number (y axis) versus brain weight (x axis). All data points are species averages; data as specified in Fig. 1F. Blue dashed line, power function (exponent 0.2988) fitted to the data of all mammals. Black line power function (exponent 0.1641) fitted to the data of primates. Both dolphins and elephants have many more facial neurons than predicted by the power function for primates.

Subnuclei of African and Asian elephant facial nucleus

The same subnuclei could be recognized in all Asian and African elephants (fig. S1). The medial, ventral, and interposed subnuclei resembled subnuclei of other mammals, and we suggest that they represent ear, lower lip, and upper lip muscles, respectively. This suggestion is supported by a match of axon numbers of facial nerve branches and cell numbers in the respective subnuclei. The dorsolateral region of the facial nucleus innervates muscles above the mouth in mammals. We suggest that dorsal and lateral subnuclei represent dorsal and ventral elephant trunk muscles (fig. S1). This idea rests on the following reasoning: (i) According to the general mammalian musclelotopy (6), the dorsal and lateral subnuclei are found in the nose region of the facial nucleus. (ii) The dorsal and lateral subnuclei are extraordinarily large, as expected of the trunk representation. (iii) The dorsal and lateral subnuclei are very elongated, consistent with an expected trunk representation. (iv) Both the dorsal and lateral subnuclei broaden toward the distal end and have very large cells in the distal part; this—as we discuss in detail below—might represent the dexterous trunk tip. (v) The dorsal and lateral subnuclei extend into the surrounding neuropil, an unusual feature consistent with enormous expansion of elephant trunk musculature in evolution. (vi) Topographically, we expect the dorsal subnucleus to represent the dorsal trunk muscles and the (more ventrally situated) lateral subnucleus to represent the ventral trunk muscles (6). (vii) Other subnuclei of elephant facial nucleus (besides the dorsal and lateral subnuclei) look similar to the facial nuclei of other mammals, and we putatively assigned these subnuclei other homologous facial representations.

Differences between African and Asian elephant ears, ear motor (auricularis) innervation, and medial facial subnuclei

Why do African elephants have more facial nucleus neurons than Asian elephants? African elephants have much larger ears than the Asian elephants (Fig. 3A). The auricularis branch of the facial nerve (providing ear motor innervation) is thicker in African compared to Asian elephants (Fig. 3B). The medial facial subnucleus was much larger in African than in Asian elephants (Fig. 3C). The medial subnucleus represents ear muscles in diverse mammals including mice (7, 8), rats (9), rabbits (10), bats (9), cats (11), and Cebus monkeys (12). Medial subnucleus neurons were more numerous in African compared to Asian elephants (Fig. 3D, left). We also counted axons in the auricularis branches of an African and an Asian elephant and observed very similar counts and species differences as for the medial subnucleus (Fig. 3D, square symbols, left), suggesting that the medial nucleus represents ear motor innervation. Averaging cell counts, we arrive at 11,965 ± 291 (mean ± SEM) neurons for African elephants and 7498 ± 568 neurons for Asian elephants. African elephants devote a larger relative share of facial neurons to the medial subnucleus than Asian elephants (Fig. 3D, right). Given the comparative evidence and auricularis nerve branch axon counts, it appears that African elephants devote more facial nucleus neurons to ear motor control than Asian elephants.

Fig. 3. Asian and African elephant ears, ear (auricularis) motor innervation, and the medial facial subnuclei.

(A) Ears of Asian (left) and African (right) elephants differ. (B) Auricularis and main branch of the facial nerve of an Asian (left) and African (right) baby elephant. The auricularis branch (innervating the ear) is thicker in the African elephant. (C) Drawings of left facial nucleus and somata of an Asian elephant (left) and an African elephant (right). Medial subnucleus cells are highlighted in red. Sections are taken at the maximal extent of medial subnuclei. The medial subnucleus innervates ear muscles in mammals. (D) Absolute neuron (circles) and axon (squares) number (left) and percentage (of total facial nucleus neurons, right) of medial subnucleus in Asian (empty symbols) and African (solid symbols) elephant facial nuclei. P values refer to unpaired t tests of cell counts only and would be lower if fiber counts were included. Photo credit (A and B): Lena V. Kaufmann, Humboldt-Universität zu Berlin; (A, left) Zoologischer Garten Berlin, Berlin, Germany; (A, right) Zoo Schönbrunn, Vienna, Austria.

Dorsal and lateral subnuclei show cell-size gradients possibly related to trunk innervation

The dorsal and lateral facial subnuclei of elephants are situated where we expect the trunk representation (6, 13) and share several features as described above (elongated shape, large neuron number, protrusion into the surrounding neuropil, and proximal-to-distal neuron size increase). We suggest that these features are explained by a representation of dorsal and ventral trunk muscles along dorsal and lateral subnuclei, respectively. The elephant trunk is enormous (Fig. 4A). A somata drawing of a coronal section through the facial nucleus of the elephant cow Aruba is shown in Fig. 4B, and the dorsal subnucleus is enlarged in Fig. 4C. The elongated dorsal subnucleus shape is atypical for mammals (5), and the proximal-to-distal cell size increase is conspicuous. While the proximal cells are also large (Fig. 4D, top), the distally situated neurons are extremely large with soma diameters of <100 μm (Fig. 4D, bottom). Such cell size heterogeneity is unusual in mammalian facial subnuclei, which typically encompass neurons of similar size (5, 6). Position-dependent cell size is unexpected in motor neurons, where size typically determines function and recruitment [e.g., Henneman’s size principle (14)]. What factors account for the proximal-to-distal neuron size increase? We reason that extreme axonal volumes associated with trunk innervation may drive this cell-size gradient. In Fig. 4E (left), we plotted estimated soma volume against the longitudinal axis of the dorsal subnucleus. In Fig. 4E (right), we plotted estimated axonal volume (see Materials and Methods) of cells innervating the trunk. Estimated axonal volumes are huge, reaching more than 200 million μm³, and increase along the trunk similarly as soma volumes increase along the dorsal subnucleus. Huge axon volumes resulting from trunk innervation might explain the subnuclear cell-size gradient.

Fig. 4. Cell size gradients and extremely large neurons in the dorsal subnucleus of the elephant facial nucleus might be related to trunk innervation.

(A) Trunk of African elephant cow Aruba. (B) Somata drawing of a coronal Nissl-stained 60-μm section through the left facial nucleus of Aruba (dorsal subnucleus outlined). (C) Drawing of the elongated dorsal subnucleus with its soma size gradient. (D) Top: Proximal neuron micrograph. Bottom: Large distal neuron micrograph (soma diameter, 104 μm). (E) Left: Plot of estimated soma volume in the dorsal subnucleus as shown in (B) and (C) against subnucleus position. The correlation is significant (P < 0.05). Right: Plot of estimated axon volume for trunk innervating facial nucleus neurons. See fig. S2. Photo credit (A): Petra Prager.

We observed cell size gradients in the dorsal and lateral subnuclei of all adult elephants, but not in a newborn elephant (fig. S2). Cross sections of axons in the newborn elephant were smaller than in adults (fig. S2), and the axonal volume gradient along the trunk was shallower in the newborn elephant. These data suggest that large axonal volumes drive cell size gradients in adult elephants.

Cell size gradients in the dorsal and lateral facial subnuclei of the elephants differ from dorsolateral facial subnuclei of other mammals

We also considered the possibility that cell size gradients are simply a feature of the dorsolateral facial subnucleus and the representation of nose muscles in mammals. To this end, we studied cell size in the mammalian dorsolateral facial subnucleus, which typically represents nose/anterior rostrum muscles, i.e., putatively homologous muscles as represented by the elephant dorsal and lateral facial subnuclei. We studied the sections where the dorsal subnucleus has its maximal extent and found cell size gradients in African (Fig. 5A) and Asian (Fig. 5B) elephants, i.e., a substantial correlation of cell position and soma area in elephant dorsal subnucleus. We then studied the dorsolateral facial subnucleus in other mammals. As in elephants, we studied the sections (from our own data or the online accessible BrainMaps database) where the dorsolateral subnucleus has its maximal extent. In other mammals, however, we found either weak (for the macaque monkey, Fig. 5C) or no correlation of cell position and soma area in the dorsolateral facial subnuclei of cats, mice, and opossums (Fig. 5, D to F). It also appeared that cell size was more heterogeneous in the elephant dorsal facial subnucleus than in the dorsolateral facial subnuclei of other mammals. We conclude that cell size gradients are a special feature of the elephant facial nucleus.

Fig. 5. Cell size gradients in the dorsal facial subnucleus of the elephants are not seen in dorsolateral facial subnuclei of other mammals.

(A) Left: Somata drawing of a coronal Nissl-stained 60-μm section through the left facial nucleus of African elephant cow Aruba (dorsal subnucleus outlined). Right: Plot of soma area (normalized to average soma size in the subnucleus) in the dorsal subnucleus against cell position in the dorsal subnucleus. Cell position is plotted in percentage of the longitudinal axis of the subnucleus; the counting direction and orientation of the longitudinal axis is indicated by the arrow (left). A linear regression line is superimposed (black); the correlation was significant [Spearman’s Rho; P (two-tailed) = 0.0]. (B) Left: Same as (A) but for the Asian elephant cow Burma. Right: Same as (A) but for the Asian elephant cow Burma. [Spearman’s Rho; P (two-tailed) = 0.0]. (C to F) Same analysis as in (A) and (B) but for the dorsolateral facial subnucleus of various mammals, which typically represents nose/anterior rostrum muscles. The data stem from the online accessible BrainMaps database. (C) Left: Same as (A) but for the macaque monkey; cells in the dorsolateral nucleus are shown in black; other facial nucleus cells are shown in gray [Spearman’s Rho; P (two-tailed) = 0.046]. (D) Same as (C) but for the cat. A linear regression line is superimposed (dashed); the correlation was not significant [Spearman’s Rho; P (two-tailed) = 0.28]. (E) Same as (D) but for the mouse [Spearman’s Rho; P (two-tailed) = 0.20]. (F) Same as (D) but for the short-tailed opossum [Spearman’s Rho; P (two-tailed) = 0.65].

Motor foveae and a differential neural trunk tip bias reflect African-Asian elephant differences in finger morphology and motor strategy

African elephants tend to pinch objects (15) (Fig. 6A) with their triangular dorsal and ventral finger (Fig. 6B), thus engaging the trunk tip. To assess how trunk morphology is reflected in the facial nucleus, we overlaid somata drawings from the African elephant cow Aruba (Fig. 6C). The overlay reveals two triangular high-density motor foveae at the expected position of the dorsal finger (distally in the dorsal subnucleus) and the expected position of the ventral finger (distally in the lateral subnucleus). The overall resemblance of the two triangular cell-dense regions (Fig. 6C) and trunk tip (Fig. 6B) is very suggestive. In African elephants, cell density at central regions of the lateral and dorsal subnucleus was severalfold higher than across the facial nucleus area (lateral, 35 ± 5; dorsal, 27 ± 4; overall area, 14 ± 1 (mean ± SEM) cells/mm² per 60-μm section; paired t tests, P < 0.05).

Fig. 6. Asian-African elephant differences in putative trunk finger motor foveae and motor strategies.

(A) African elephants tend to pinch objects with their two fingers (15), thus engaging the trunk tip. (B) African elephant trunks have two fingers, a dorsal (DF) and ventral (VF) one. (C) Overlay of somata from five coronal 60-μm sections from around the anterior third of the facial nucleus of African elephant Aruba. DF? and VF? denote the putative dorsal and ventral finger representation. Note the high cell density motor foveae and the shape similarity to (B). (D) Asian elephants tend to grasp/wrap objects with their trunk (15), thus engaging larger trunk parts. (E) Asian elephant trunks have one dorsal finger (DF). (F) Somata overlay from the Asian elephant Burma, conventions as in (C). High cell density foveae are less evident than in (C). (G) Neurons are preferentially distributed distally (toward the putative trunk tip) in the dorsal subnucleus of African elephants compared to Asian elephants. Bottom: Somata overlay of dorsal subnucleus from African elephant Aruba (C) and Asian elephant Burma (F) arranged horizontally (distal to the right). Note the rightward/distal cell bias in the African elephant. Top: Cell distribution along the proximal-distal axis of dorsal subnuclei in six African (solid line) and six Asian (dashed line) elephant facial nuclei. Cell numbers were similar proximally but very different distally (t tests, *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001). (H) Distal distribution bias in the lateral subnucleus (toward the putative tip representation) in African elephants. Conventions as in (G). Photo credit (A), (B), and (E): Lena V. Kaufmann, Humboldt-Universität zu Berlin; (A) Zoo Schönbrunn, Vienna, Austria. Photo credit (D): Michael Brecht, Humboldt-Universität zu Berlin; Zoologischer Garten Berlin, Berlin, Germany.

Asian elephants tend to grasp/wrap objects with their trunk (15) (Fig. 6D) as they have only one dorsal trunk finger (Fig. 6E), thus engaging larger parts of the trunk. We observed cell-dense regions at the putative position of the dorsal finger in overlays (Fig. 6F), but the cell density differences in Asian elephants were less marked than in Africans. The pinching motor strategy of African elephants puts more emphasis on the trunk tip than the Asian wrapping strategy. We therefore wondered how neurons are distributed along the longitudinal axis of the dorsal and lateral subnucleus, the putative trunk-axis representation. In African and Asian elephants, cell numbers were similar in the proximal but were very different in the distal part of the dorsal subnucleus (Fig. 6G), and we made similar observations for lateral subnuclei (Fig. 6H). We suggest that African elephant trunk fingers are represented by motor foveae and that African elephants devote more neurons to the trunk tip than do Asian elephants.

DISCUSSION

Our approach

We present a multifaceted approach to the elephant facial nucleus. Our techniques (Nissl staining, axonal osmium tetroxide stains, somata drawings, cell fiber counting, nerve tracing, and comparative considerations) were matched to the elephant material available. The elephant facial nucleus is large and highly differentiated. Our findings reflect the muscularity (16) and dexterity of the elephant trunk and are a testament to Krogh’s principle, which states that for every problem there will be some animal of choice, on which it can be most conveniently studied (17). Specifically, we think that elephants are a model for highly differentiated facial motor control. Our comparative analysis suggests that the relation between facial nucleus neuron number and facial motor control is complex. We note that facial motor neuron number is not the sole predictor of facial motor ability. In particular, the facial nucleus appears to be involved in functions other than the control of facial expressions, as animals such as dolphins—which are not known for elaborate gesturing—have numerous facial nucleus neurons. The comparative analysis of subnuclei combined with quantitative nerve tracing provides a suggestion for elephant facial nucleus musclelotopy. We admit, however, that such methods cannot provide the same certainty as invasive back-labeling experiments, and our conclusions must remain tentative at this point. Our methods combination could elucidate other motor systems, such as the enigmatic facial motor control of cetaceans.

Neural specializations of Asian and African elephants

A novel insight is that brains of Asian and African elephants differ. Such differences concern both the overall number of cells and the relative partitioning of the facial nucleus. The medial facial subnucleus of African elephants is larger than the Asian medial facial subnucleus or the entire human facial nucleus. Comparative considerations, nerve tracing, and fiber counting strongly suggest that the medial subnucleus relates to ear motor control. Pictures of an attacking African elephant bull with laterally erected ears (3) point to the behavioral relevance of ear motor control in facial gesturing. In terms of the relative fraction of medial subnucleus neurons of the total facial nucleus, Friauf and Herbert (9) showed that flying foxes (Rousettus aegyptiacus) seem to have an even larger medial (ventromedial in their terminology) subnucleus. In these echolocating fruit bats, the medial subnucleus accounts for about half of the facial nucleus and innervates ear muscles. It will be worthwhile to investigate the role of ear movements in elephant auditory and infrasound perception.

Body size, cell size, and motor neuron size gradients

We observed very large motor neurons and cell size gradients in the putative trunk representation. We suggest that cell size in these neurons is driven by large axonal volumes resulting from distal trunk innervation and quantitative considerations, and the absence of such cell size gradients in smaller mammals makes this suggestion plausible. Plausibility is not enough, however, and our hypothesis needs further testing. The trunk length–cell size gradient hypothesis could be tested by dissecting an adult African elephant bull facial nucleus, an elephant substantially larger than the cows we studied. We predict that cell size in the proximal lateral/dorsal nucleus of bulls will be only slightly larger than in cows but will be markedly larger in the distal lateral/dorsal subnucleus of bulls. The issue of body size and neural control emerged early in neuroscience, when von Helmholtz’s (18) measurements revealed the finite (and actually quite moderate) nerve conduction velocity. He argued that the delays imposed might be manageable for humans but will pose serious problems for larger animals such as whales. We wondered whether elephants might compensate conduction delays across the trunk with extra-thick axons, but nerve bundles innervating different parts of the elephant face were homogeneous in axon thickness. Elephant trunks will therefore impose notable conduction delays. Elephants might need to rely on predictive trunk motor control, a function thought to be supported by the cerebellum (19). Elephants have the relatively largest mammalian cerebellum (20), and the cerebellum is also huge in other large mammals (21, 22).

Motor foveae and elephant trunk motor strategies

We observed motor foveae that putatively represent African elephant trunk fingers. The visual fovea is a pit of high cone density and is constantly kept on visual targets. The two motor foveae of African elephants are regions of markedly increased cell density (Fig. 6C) resembling trunk fingers (Fig. 6B). African elephants constantly bring trunk fingers on haptic targets. Structurally and behaviorally, these motor regions qualify as foveae. In the auditory system of cf-bats, an auditory fovea has been described (23), and by Doppler shift compensation (24), bats keep echoes on the frequency fovea. Similar foveation is observed in the star-nosed mole somatosensory system (25). The idea that African elephants have a trunk tip motor fovea aligns with their behavior (15), specializations of the trunk tip (26), and the tactile trunk innervation (27).

We identified Asian versus African elephant neural specializations, neural adaptations to body size and motor foveae, and putative mechanisms of dexterity. We conclude that elephant brains offer insights for neuroscience.

MATERIALS AND METHODS

Elephant specimens

We worked with a variety of elephant specimens in our study. All specimens came from zoo elephants and were collected by the IZW (Leibniz Institute for Zoo and Wildlife Research, Berlin) over the past 3 decades in agreement with International Trade in Endangered Species of Wild Fauna and Flora (CITES) regulations. Specifically, specimen reports and CITES documentation for all animals included are held at the IZW. All animals included in the study died of natural causes or were euthanized by experienced zoo veterinarians for humanitarian reasons, because of insurmountable health complications.

Asian elephants, E. maximus

Data from two Asian elephant babies, which died around birth, were included. In one of them, a male from Parc Pairi Daiza in Belgium, we had access only to the animal’s head but not the brain. The brain of this animal was described in a study by Rasenberger (28). We also included data from the adult Asian elephant cow Burma (52 years old) from the Augsburg Zoo and from Asian elephant cow Dumba (44 years old) from the elephant farm Platschow. Different data were derived from the various Asian elephant specimen. In the other Asian baby elephant, which originated from Tierpark Hagenbeck in Germany, we collected the brain with two facial nuclei. In the brainless Asian baby elephant, we performed extensive bilateral facial nerve tracing and nerve branch fiber counting. In two adult Asian elephants, we collected the two brains with two facial nuclei and performed a limited amount of facial nerve tracing and fiber counting.

African savanna elephants, L. africana

Data from one stillborn African elephant baby, a male from Colchester Zoo in Great Britain, were included. We also included data from three adult African elephant cows: Aruba (41 years old) and Zimba (39 years old), both from Opel-Zoo Kronberg (Germany), and Linda (35 years old) from the Zoo Poznan (Poland). Different data were derived from the various African elephant specimen. In the African baby elephant, we collected the brain with two facial nuclei and performed extensive bilateral nerve tracing and nerve branch fiber counting. While the nerves of this animal were well preserved, the facial nuclei of this animal showed a not well-preserved cellular structure and could not be analyzed. In the three adult African elephant cows, we collected the brains with two facial nuclei each and performed a limited amount of nerve tracing and fiber counting.

Specimen condition

Specimen condition varied widely in our study, and information about all specimens is summarized in Table 1. Most heads or other material reached us frozen, and none of the elephant heads/brains were perfused. Although many of the animals included were dissected by professional veterinarians, the preservation of material varied across specimens. A variety of reasons contributes to a suboptimal preservation of elephant material. Specifically, it often takes days to dissect elephants and the animals’ carcasses cool down only very slowly. Furthermore, the freezing leads to freezing artifacts, and even in extracted brains, fixative action is slow because of elephant brain size. Some of these problems are discussed and have been partially solved in the following references (29, 30).

Elephant preparation, facial nucleus and facial nerve collection, and nerve measurements

Elephant preparation

Heads of deeply frozen baby elephants were removed at the IZW. In adult elephants, heads and trunks were removed at the respective zoos, and the remaining skull was trimmed with motorized saws and axes at the IZW Berlin. Brains of baby elephants (after thawing) and brains from trimmed skulls of adult elephants were extracted by F. Egelhofer and A. Sebastiampillai at the Neuropathology of the Charité, Berlin.

Facial nucleus extraction

We proceeded with facial nucleus collection after extraction of the brain and dura removal followed by several weeks of fixation in 4% paraformaldehyde solution. To remove facial nuclei, we positioned entire elephant brains with their ventral side up in a dissection tray. We then dissected away blood vessels and the pia arachnoidea from the elephant brain stem. To dissect out facial nuclei, we oriented ourselves at the division of medulla and pons, which is prominently visible in elephants. A vertical cut was made approximately 3.5 mm anterior to and parallel to the medulla-pons border. Another second vertical cut was made approximately 12 mm posterior to the first cut. A third vertical cut was made approximately 12 mm posterior to the second cut. Vertical cuts were also placed laterally from the brain stem to free the sections from the cerebellum. After these cuts, two roughly 12-mm-thick coronal brainstem sections were carefully removed from cerebellum, each of which were sectioned in roughly 200 successive coronal brain stem sections after paraffin embedding (see below). In some cases, the entire facial nucleus with its anterior posterior extent of ~11 mm was contained in the more anterior of the two thick sections; in other cases, small parts of the facial nucleus were also contained in the more posterior of the two thick sections. Directly before and after removal of the two thick sections, we took photographs of the elephant brainstem and the isolated sections.

Nerve tracing and nerve collection

Facial nerves were collected from elephant skulls, where the nerve leaves the skull through the stylomastoid foramen. To find the foramen, we first identified the occipital condyles and dissected soft tissue from the adjacent bone in the rostrolateral direction until the styloid process became palpable through the soft tissue. The process was exposed, and soft tissue was removed carefully in the lateral direction until the base of the process where the facial nerve leaves the skull through the stylomastoid foramen and runs in the rostral direction. We carefully dissected out the facial nerve and its major branches in one African baby elephant and in one Asian baby elephant. Facial nerve branches were only considered for further analysis if we were certain that we were able to collect the entire branch or set of branches. The auricularis branch in particular branched off early after the facial nerve emerged from the skull and consisted of two or three branches. We traced the facial nerve until it entered the trunk and started to mingle with the (sensory) infraorbital nerve, which also innervated the trunk. Ideally, we could also identify major nerve branches innervating the upper and lower lip (see fig. S1). Comprehensive and complete tracing of the facial nerve into the trunk was done for four facial nerves from baby elephants (see above) and took approximately 1 day per nerve. Facial nerve segments were also removed from the heads of both Asian and African elephant cows, but in these cases, nerve collection was less complete and was done en passant, while elephant heads were trimmed (see above) for further processing of elephant brains. In these cases, we could only occasionally identify the auricularis branch and other nerve branches were not accessible.

Nerve measurements

We visually identified intact nerve parts and determined the minimal and maximal thicknesses of the nerve. Both minimal and maximal thicknesses were measured with a caliper independently by two investigators, and measurements were averaged.

Nerve sectioning, preparation, and staining

Preparation and staining procedures of the facial nerve were performed analogous to earlier work (27, 31, 32). In brief, we dissected out the facial nerves, performed an osmium tetroxide staining (see below) on them, embedded them in paraffin, and did 4- to 8-μm cross sections.

Osmium tetroxide stain

To assess the axon diameter and to verify counts of myelinated fibers, osmium tetroxide myelin stains were performed on tissue segments directly before paraffin embedding. For this purpose, sections of the facial nerve of approximately 5 mm width were placed in 2% osmium tetroxide solution (osmium tetroxide 4%, SERVA, catalog no. 31253.02) for about 1 hour under constant shaking. After several washing steps with double-distilled water (ddH₂O) to remove residual staining solution, the nerve segments were embedded in paraffin, cut into 4- to 8-μm-thick cross sections, and mounted on Carl Roth Adhesion slides Superfrost Plus Gold. The slides were stored overnight in a furnace at 45°C. The following day, sections were deparaffinized in xylol, isopropanol, and 100% ethanol and covered with Eukitt mounting medium (catalog no. 03989-100ML, Sigma-Aldrich).

Facial nucleus sectioning, preparation, and staining

Facial nuclei were treated similar to nerve tissue with the exception that there was no osmium tetroxide staining, thicker sections were cut (20 to 60 μm thickness), and these were stained for Nissl substance. Most facial nuclei were sectioned in 60 μm thickness, and instead of every 60-μm 10th section, we did three 20-μm sections that we used for various antibody stains, the results of which we do not discuss in the current study. Series of sections were processed, alternating with Nissl and antibody staining.

Cellular and axon size measurements

Thin Nissl or osmium tetroxide–stained sections were viewed using Stereo Investigator software (MBF Bioscience, Williston, USA) using an Olympus BX51 microscope (Olympus, Japan) with a MBFCX9000 camera (MBF Bioscience, Williston, USA) mounted on the microscope. The microscope was equipped with a motorized stage (LUDL Electronics, Hawthorne, USA) and a z-encoder (Heidenhain, Schaumburg, USA). Stereo Investigator software was used for stereological procedures, cell size, and axon diameter measurement and for acquiring images. Digitized images were adjusted for brightness and contrast using Adobe Photoshop (Adobe Systems Inc., San Jose, CA, USA), but they were not otherwise altered.

Stereology based on the optical fractionator

We used three approaches to quantify elephant facial motor innervation, an optical fractionator approach (described in Table 2), full cell counts of every 10th section (described in Table 3), and facial nerve axon counts (described in Table 4). As seen in Tables 2 to 4, the results of these approaches aligned well. For optical fractionator approach, we estimated the total number of neurons in the elephant facial nucleus using stereology methods. We estimated the total number with Stereo Investigator software (MBF Bioscience, Williston, USA) using a sampling scheme called the optical fractionator method. Our region of interest was identified and outlined at low (2× objective) magnifications. The neurons were identified by their shape and staining intensity and large size at high magnification (20×) and counted individually. Without exception, facial nucleus neurons were larger than neurons in surrounding brain structures. The standard stereological sampling scheme is independent of volume, measurements, and shrinkage because the number of neurons is estimated directly without referring to neuron densities. Using the optical fractionator technique, we counted the nucleoli that come into focus and fall within the acceptance lines of the dissector, which were randomly placed on the series of sections (33, 34). We counted neurons in 12 facial nuclei of six elephants. Elephant brain stems were stained with Nissl substance and sliced at 60 μm. A detailed overview of count results is given in Table 2. We used the following parameters. The dissector laid a grid of squares over our region of interest with a size of 2000 × 1000 μm, where we counted the neurons at each dissector in the counting frame area of 350 μm × 350 μm. At each counting frame, we counted between 0 and 5 neurons. A total of 200 to 300 neurons were counted in each facial nucleus for assessing the total number of neurons (see Table 2). The entire elephant facial nucleus spanned 60-μm sections in the baby elephant and 200 to 220 60-μm sections in adult animals; every 10th section was counted. Guard zones were set to zero. The mean thickness measured at every counting site was measured to be around 15.5 μm and used to estimate the total number of neurons.

Somata drawings and subnucleus assignment

An initial analysis indicated a very intricate organization of cell size and position in the elephant facial nucleus. To obtain a full picture of this organization, we decided to prepare drawings of all facial nucleus neuronal somata in every 10th section of the elephant facial nucleus with the Stereo Investigator/Neurolucida software (MBF Bioscience, Williston, USA). In case there was damage to the facial nucleus in a specific section, we would deviate from the 10-section interval. We found that the subnuclei described in fig. S1 could be easily recognized in almost all of the elephant facial nucleus sections in somata drawings, because different subnuclei were separated by cell-poor regions. A small fraction of cells could not be assigned to these subnuclei either because the cells were situated between subnucleus borders or because they came from sections where the subnucleus organization was not clear to us (i.e., the most anterior or posterior section of the facial nucleus; fig. S1).

Cell number estimates based on full counts of every 10th section

A second approach to estimate facial nucleus neuron number was the full count of neurons in every 10th section. The results of this approach are shown in Table 3. The elephant facial nucleus is characterized (compared to other species’ facial nuclei) by a low cell density and large neurons. Given these characteristics, we found it easy to detect double-counted neurons in carefully aligned adjacent sections, and we decided to correct the raw cell counts by an estimate of double counts; for each specimen, we determined a double-count correction factor. Our double-count analysis revealed that the same somata were never counted in three adjacent sections and that double counts of cells in adjacent sections accounted for between 13.8 and 20.5% of entries and final counts were adjusted accordingly (Table 3).

Axonal fiber counts

Axonal fiber counts were made as described previously (27, 31, 32), and the results of the counting effort are shown in Table 4. In brief, our counting strategy was the following. We only counted sections orthogonal to nerve fiber trajectories. We then drew outlines of all nerve bundle in the respective nerve branch and thus determined the cross-sectional nerve bundle area. We counted all fibers in a subset (5 to 20% of bundles) of nerve bundles and determined the cross section of these bundles. From these measurements, we determined a fiber/cross-section area estimate and extrapolated the total fiber count. For the subset of counted fibers, we always observed a strictly linear relationship between fiber count and bundle cross section. We also visually inspected all nerve bundles for outliers of particularly low- or high-fiber packing density, but we did not observe such outliers. As far as we can tell, all or close to all facial nerve fibers are myelinated. Counts of axon stains were based on osmium tetroxide myelin stains.

Somata drawing overlays and cell density analysis

We found overlays of somata drawings over multiple sections helpful in visualizing and analyzing the elephant facial nucleus. In particular, we prepared overlays of five sections centered around a section in the anterior third of the facial nucleus for all elephant facial nuclei. At this level, the lateral and dorsal subnuclei have their maximal extent. To prepare overlays, we started with the central section and overlaid adjacent sections by transversal movement and image rotation of the added section. In a few cases, we also applied slight isometric expansion/contraction of an added section. We aligned sections such that we obtained an optimal fit across the entire facial nucleus according to visual inspection.

Cell density analysis was also performed on these section overlays from the anterior third of the facial nucleus. We determined overall cell density (i.e., cells/mm² per 60-μm section) in the entire facial nucleus area and at the center of the lateral and dorsal subnucleus in each elephant. The center of these subnuclei was observed, where these elongated subnuclei had their greatest width. Specifically, we defined the center area of the lateral nucleus as the midline area of half the width of the subnucleus and at an elongation between 45 and 50% of length of the subnucleus (with 0% being proximal and 100% being distal; see Fig. 6, G and H). The center area of the dorsal nucleus was defined as the midline area of half the width of the subnucleus and at an elongation between 70 and 75% of length of the subnucleus. We made measurements at these coordinates in each lateral and dorsal elephant subnucleus, irrespective of where the maximal cell density was observed in the individual specimen.

Soma and axonal volume estimates

Soma volume was approximated as a sphere with the radius of the mean of the minimal and maximal Feret diameter. Axonal volume estimates for trunk innervating facial nucleus neurons were made on the basis of cross-section and axon length estimates. Estimates of cross sections came from outlines drawn of osmium tetroxide–stained myelin sheets of the elephant facial nerve. Such measurements include the myelin sheet and thus overestimate axon diameter. We did not correct estimates for a ~20% volume shrinkage due to fixation, and we reckon that errors of axon volume overestimation (because of myelin sheet inclusion) and underestimation (because of shrinkage) will roughly cancel out. Our axonal volume estimates were based on the average cross section of drawn axon outlines in two newborn elephants and one adult elephant. Specifically, we found an average cross-section area of 43.8 ± 13.8 μm² (mean ± SD) in 11,533 facial nerve axons measured in newborn elephants and an average cross-section area of 109 ± 45 μm² in 3492 axons measured in one adult African elephant (Zimba). Axon length estimations were performed based on measurements of newborn and adult skulls, as well as facial nerve tracing and photographs of trunks of adult and newborn elephants. Specifically, we estimated the axon length to trunk base to be 33 and 66 cm to the trunk tip in newborn elephants and to be 50 cm to the trunk base and 210 cm to the trunk tip in adult elephant cows.

Statistical analysis

All statistical tests are specified in the respective figures or legends or in the text. All tests were two-tailed. We used t tests and Spearman’s Rho to test the significance of correlations between set of values. For assessment of allometry (Fig. 2), we used the linear model fitted to the primate data to predicted values and confidence intervals of the number of facial nucleus neurons for different values of brain weight. We used a combination of the fitlm and predict Matlab functions (The MathWorks Inc.), with confidence intervals representing α equal to 0.05.

Supplementary Materials

This PDF file includes:

Figs. S1 and S2

View/request a protocol for this paper from Bio-protocol.