In contrast to many other mammals, cetaceans have relatively small hippocampi that appear to lack adult neurogenesis

Immunohistochemistry

The brains of all animals used for immunohistochemistry were, following euthanasia, perfusion fixed with 4 % paraformaldehyde in 0.1 M phosphate buffer and then stored in an antifreeze solution until processed for immunohistochemistry (Manger et al. 2009). To investigate the presence of adult hippocampal neurogenesis, we used standard immunohistochemical procedures with an antibody directed against doublecortin (goat-anti DCX C-18 primary antibody, Santa Cruz Biotechnology) (Patzke et al. 2013a, b; Chawana et al. 2013). Using DCX immunohistochemistry, we examined the hippocampus and adjacent piriform cortex of 71 mammalian species (Table 2) from 13 mammalian orders covering a range of brain sizes (from less than 1 g through to 5 kg). This study was provided with ethical clearance by the University of the Witwatersrand Animal Ethics Committee, which uses guidelines similar to those of the NIH regarding the use of animals in scientific research. The animals used in the current study were all collected under appropriate governmental permissions.

From each animal used in the current study, blocks of hippocampal tissue were dissected in a plane orthogonal to the ventricular surface of the hippocampus at approximately the middle portion of the hippocampus. Each tissue block was cryosectioned into 50-μm-thick sections on a freezing microtome. Consecutive sections were stained for Nissl substance and reacted immunohistochemically for DCX, with a minimum of 12 sections per stain from hippocampi from two individuals of each species. The sections used for Nissl staining were mounted on 0.5 % gelatine-coated slides, dried overnight, cleared in a 1:1 mixture of 100 % ethanol and 100 % chloroform and stained with 1 % cresyl violet.

The sections used for free-floating immunohistochemical staining were treated for 30 min in an endogenous peroxidise inhibitor (49.2 % methanol:49.2 % 0.1 M PB:1.6 % of 30 % hydrogen peroxide) followed by three 10 min rinses in 0.1 M PB. To block unspecific binding sites, the sections were then pre-incubated for 2 h, at room temperature, in blocking buffer (3 % normal rabbit serum, 2 % bovine serum albumin, BSA and 0.25 % Triton X-100 in 0.1 M PB). Thereafter, sections were incubated in the primary antibody solution, made up of the appropriate dilution of the primary antibody in blocking buffer for 48 h at 4 °C under gentle agitation. In the current study we used immunolabelling of DCX, an endogenous marker of immature neurons, to ascertain the presence or absence of adult hippocampal neurogenesis. While the presence of DCX in neurons outside of the hippocampus may not relate to adult neurogenesis in these regions, such as the piriform cortex (Klempin et al. 2011), it has been established that DCX immunolabelling of granule cells of the dentate gyrus is a good proxy for the presence of adult hippocampal neurogenesis (Rao and Shetty 2004; Couillard-Despres et al. 2005). The presence of DCX also reflects cumulative adult hippocampal neurogenesis over a period of 2 weeks to 6 months, although this period is species specific (Rao and Shetty 2004; Kohler et al. 2011). In this sense, lack of DCX staining should be a reliable indicator of the absence of adult hippocampal neurogenesis. DCX immunolabelling is therefore particularly useful when studying a wide variety of field-caught mammalian species, as no specific intervention is required to reveal adult hippocampal neurogenesis.

To visualize DCX, we used the goat-anti DCX C-18 primary antibody from Santa Cruz (catalogue number sc-8066) at a dilution of 1:300. This antibody is an affinity-purified goat polyclonal antibody raised against a peptide mapping at the C-terminus of doublecortin of human origin. The amino acid sequences of the C-terminus of the doublecortin protein are highly conserved across mammalian species based on the Protein database provided by the National Center for Biotechnology Information. The primary antibody incubation was followed by three 10 min rinses in 0.1 M PB and the sections were then incubated in a secondary antibody solution (1:1,000 dilution of biotinylated anti-goat IgG, BA 5000, Vector Labs) for 2 h at room temperature. This was followed by three 10 min rinses in 0.1 M PB, after which the sections were incubated for 1 h in an avidin–biotin solution (1:125; Vector Labs), followed by three 10 min rinses in 0.1 M PB. The sections were then placed in a solution containing 0.05 % 3,3’-diaminobenzidine (DAB) in 0.1 M PB for 5 min, followed by the addition of 3.3 μl of 30 % hydrogen peroxide per 1 ml of DAB solution. Chromatic precipitation was visually monitored under a low-power stereomicroscope. Staining continued until such time as the background stain was at a level that would allow for accurate architectonic matching to the Nissl sections without obscuring the immunoreactive structures. Development was arrested by placing sections in 0.1 M PB for 10 min, followed by two more rinses in this solution. Sections were then mounted on 0.5 % gelatine-coated glass slides, dried overnight, dehydrated in a graded series of alcohols, cleared in xylene and covers-lipped with Depex. To ensure non-specific staining of the immunohistochemical protocol, we ran tests on sections where we omitted the primary antibody, and sections where we omitted the secondary antibody. In both cases, no staining was observed. In 11 species (African elephant, four-toed sengi, Hammer-headed fruit bat, ring-tailed lemur, Beecroft’s flying squirrel, Arabian spiny mouse, greater kudu, river hippopotamus, West Indian manatee, harbour porpoise and minke whale), an absorption control in sections encompassing the dentate gyrus and piriform cortex was also run using the blocking peptide sc-8066 P (Santa Cruz) as recommended by the supplier. In all cases, no staining was evident. Digital photomicrographs were captured using a Zeiss Axioskop and Axiovision software. No pixelation adjustments or manipulation of the captured images were undertaken, except for the adjustment of contrast, brightness and levels using Adobe Photoshop 7.

Hippocampal volume increases as an exponential function across mammalian species

Previous studies investigating the relationship of how the hippocampus scales relative to brain size in adult mammals have used standard linear regression models (Finlay and Darlington 1995; Reep et al. 2007). In the current study, we analysed a larger database (375 species belonging to 17 orders; Table 1) and found that the relationship between brain and hippocampal volume in mature mammals was best described by an exponential function that approximated a growth curve (an exponential decay increasing form model) (Fig. 1). The exponential function depicted (Fig. 1) is based on values for chiropterans, insectivores, primates, artiodactyls, carnivores and other species for which data were available apart from the cetaceans, elephants, hippopotami and manatee (Table 1). On the basis of these tests, an exponential curve [y = a × (b - exp(−c × x)); where a = 9.26; b = 0.72 and c = 0.097] was fit across the groups as this model performed the best of all models tested(r² = 0.97;DOF = 359; AICC =−894.55; sum of squares = 30.25; nruns = 182, P = 50.71 %). Using Akaike’s information criteria, the exponential model was shown to have a 100 % likelihood of being a better fit than the linear model (Delta = 122.50; P = 2.5 × 10⁻²⁷). From this exponential function, we calculated 95 % confidence and prediction intervals, which demonstrated that the vast majority of the mammalian species fell within these statistically derived boundaries of the relationship of hippocampus to brain volume. Onto the plot, we superimposed data on hippocampal volume from African elephant, river hippopotamus, West Indian manatee and four species of cetaceans (Fig. 1; Table 1). The hippocampal volumes for the African elephant, hippopotamus and manatee all lie within the 95 % prediction intervals and close to or within the 95 % confidence intervals.

In Fig. 3a, we provide a graphical representation of the two best-performing regression models, i.e. the non-linear exponential model and the least squares linear model. The exponential model ranked best in terms of goodness of fit criteria displaying the following regression statistics (r² = 0.97; AICC =−894.55; sum of squares = 30.25; DOF = 359) in comparison to that of the weaker-performing linear model (r² = 0.98; AICC =−772.05; sum of squares = 42.66; DOF = 360). An F test comparing the sum of squares of the exponential model with that of the linear model indicated a 1.18 9 10⁻²⁶ % (F = 147.39) probability that the exponential model was a better fit to the data than the linear model. Furthermore, both visual and statistical comparison of the accompanying residuals confirmed that a linear model was not suitable for describing these data (Fig. 4). The residuals as based on the linear model are not randomly scattered about zero as is confirmed by a runs test, while both visual and statistical comparison of the non-linear model confirms its appropriateness for this data (nruns = 182, P = 50.71 %).

While independent contrast analysis is commonly used when exploring cross-taxonomic relationships, this technique is known to perform poorly when the underlying relationship between characters is non-linear. In accordance with suggestions pertaining to non-linearity (Garland et al. 1992; Quader et al. 2004), we log transformed our data and performed independent contrast analysis to evaluate the scaling of hippocampal volume with brain volume if a linear model were valid. In Fig. 5, we present a plot of the phylogenetic correct least square regression and associated confidence intervals and prediction intervals, mapped onto the original tip data space (Garland and Ives 2000). The resultant coefficient of determination for this model is r² = 0.85/0.83 with a slope of 0.77/0.75. This plot indicates that even after phylogenetic correction, the cetaceans lie well below the confidence and prediction intervals of the mammalian line and are characterized by a markedly different scaling of the hippocampus relative to brain volume compared to all other mammals. In addition, the non-line-arity of the mammalian data is also evident in these plots.

Thus, in contrast to all other mammalian species examined to date, the data for the four species of cetaceans examined (harbour porpoise, bottlenose dolphin, Atlantic white-sided dolphin and minke whale) fall well below the 95 % prediction intervals (Figs. 1, 2, 3, 4, 5). Our data indicate that the cetaceans have hippocampal volumes that range between 8 and 20 % of the volume that would be predicted based on their brain size. Across all mammals analysed, the cetaceans were the only species that were different with regard to hippocampal size, and even their closest relative, the semi-aquatic river hippopotamus and the West Indian manatee, a species within the only other obligatorily aquatic order of mammals, did not show a trend towards a reduction of hippocampal volume.

Amygdala volume increases as a linear function across mammalian species

Previous studies investigating the relationship of how the amygdala scales relative to overall brain size in adult mammals used standard linear regression models (Finlay and Darlington 1995; Reep et al. 2007). In the current study, we analysed a larger database (373 species belonging to 17 orders) and found that the relationship between brain and amygdala volume in mature mammals was best described by a linear function (Fig. 2) as previously demonstrated (Finlay and Darlington 1995; Reep et al. 2007). On the basis of the tests undertaken, a linear function was fit across the groups as this model performed the best of all models tested (r² = 0.98; DOF = 360; AICC = −1,120.68; nruns = 190, P = 79.63 %). From this linear function, we calculated 95 % confidence and prediction intervals, which demonstrated that the vast majority of the mammalian species fell within these statistically derived boundaries of the relationship of amygdala to brain volume. Onto the plot, we superimposed data on amygdala volume from African elephant, river hippopotamus, West Indian manatee and three species of cetaceans (Fig. 2). The amygdala volumes for the African elephant, hippopotamus and manatee all lie within the 95 % prediction intervals and close to or within the 95 % confidence intervals, but those of the three cetacean species fell below the 95 % prediction intervals.

In Fig. 3b, we provide a graphical representation of the two best-performing regression models, i.e. the least squares linear model and the non-linear exponential model. The linear model ranked best in terms of goodness of fit criteria displaying the following regression statistics (r² = 0.98; AICC = −1,106.35; sum of squares = 16.94; DOF = 360) in comparison to that of the slightly weaker performing exponential model (r² = 0.98; AICC = −1,106.28; sum of squares = 16.85; DOF = 359). Using Akaike’s information criteria, the linear model was shown to have a 51 % likelihood of being a better fit than the exponential model (Delta = 0.07; P = 0.49). An F test comparing the sum of squares of the exponential model with that of the linear model indicated a 16.44 % (F = 1.94) probability that the linear model was a better fit to the data. Furthermore, both visual and statistical comparison of the accompanying residuals confirmed that a linear model was more suitable for describing these data (Fig. 4).

Thus, in contrast to all other mammalian species examined, the data for the three species of cetaceans examined (harbour porpoise, bottlenose dolphin and minke whale) fall well below the 95 % prediction intervals (Figs. 2, 5). Our data indicate that the cetaceans have amygdala volumes that range between 37 and 42 % that would be predicted based on their brain size. Across all mammals analysed, the cetaceans were the only species that were different with regard to amygdala size, and even their closest relative, the semi-aquatic river hippopotamus, did not show a trend towards reduction in amygdala size; however, the West Indian manatee, a species within the only other obligatorily aquatic order of mammals, did show a trend towards a reduction of amygdala volume. For both cetaceans and the manatee, these reductions in relative amygdala volumes are likely related to the reduction/absence of the olfactory system in these species.

Adult hippocampal neurogenesis is apparent in all mammals except cetaceans

Our investigation of adult hippocampal neurogenesis across 71 species of mammals using immunohistochemistry to visualize DCX (Kempermann 2012) revealed robust staining of immature neurons across all species examined, except for the two cetacean species (Figs. 6, 7, 8). In addition, a survey of the literature (Table 2) indicates that all 93 mammalian species (from 16 different mammalian orders) studied to date, except the two cetaceans studied herein, possess robust adult hippocampal neurogenesis. Internal controls for antibody staining revealed positive staining of immature neurons in the piriform cortex of the minke whale and in the remnants of piriform cortex in the harbour porpoise (we use the term remnants as the odontocete cetaceans lack an olfactory bulb, and thus the size of the piriform/olfactory cortex is greatly reduced) (Fig. 8). The piriform cortex is known to contain neurons immunoreactive to DCX in mammals (Klempin et al. 2011). Thus, we can conclude that there are no specific problems with the cetacean tissue used or the immunohistochemical methodology. Moreover, we obtained robust staining in the only other obligatorily aquatic marine mammal investigated, the West Indian manatee, and in several species of semi-aquatic mammals, including the river hippopotamus (Fig. 7), seals from both phocid and otariid lineages (Fig. 7), Asian small-clawed otters and giant otter shrews (Patzke et al. 2013b). Given the success of DCX immunohistochemistry acting as a proxy marker for adult hippocampal neurogenesis across such a diverse array of species, we feel confident in reporting its apparent absence in the cetaceans.

In addition to the apparent lack of adult hippocampal neurogenesis and the small relative and absolute size of the cetacean hippocampus, the architecture of the cetacean hippocampus contrasts with that seen in all other mammals examined. In most mammals, the granule layer of the dentate gyrus is observed to be a tightly packed layer of cells within a distinctly organized three-layered cortical region (Fig. 7); however, in the minke whale, a mysticete cetacean, while evident, the packing of the neurons in the granule cell layer of the dentate gyrus is not as dense as that seen in other mammals. In the harbour porpoise, an odontocete cetacean, the granule cell layer is so loosely organized as to be difficult to discern in normal histological preparations (Fig. 8). Thus, in contrast to all other mammals, the cetaceans have three distinct aspects of hippocampal anatomy that indicate they are neuroanatomically different to all other mammals—a small hippocampus, an apparent lack of adult hippocampal neurogenesis and a loosely organized dentate gyrus.

Hippocampal scaling and adult hippocampal neurogenesis

The current study, using a larger database than previous studies (Finlay and Darlington 1995; Reep et al. 2007), including large-brained mammals such as elephants, indicates that the manner in which the volume of the hippocampus scales with the volume of the brain is best described as an exponential function, rather than as a linear function. To date, this is the only demonstration that a component of the brain scales in a non-linear manner with overall brain volume and indeed our own calculations of the scaling of another limbic structure, the amygdala that lies in close apposition to the hippocampus, provide support for this distinction of the hippocampus. Interestingly, the only species that do not adhere to this non-linear scaling of the hippocampus are the cetaceans, which have small hippocampi and seem to lack adult hippocampal neurogenesis. Thus, it would appear that as adult hippocampal neurogenesis is a feature common to most mammals, this persistent growth phenomenon may have some bearing on the manner in which the hippocampus scales with the brain across mammalian species; however, to postulate a direct link between the two and what a potential mechanism might be is difficult at this stage. The hippocampal volume scaling relationship is likely to be affected not only by the addition of new neurons in the dentate gyrus, but also by their constituent parts (dendrites and mossy fibres), differing rates of neurogenesis across the life span, rates of apoptosis, brain size of each species and the associated neuronal density, and epigenetic and phylogenetic factors. Thus, at this stage we cannot propose any direct link between neurogenesis and hippocampal scaling, although our results indicate that this would be a potentially interesting avenue for future study.

As adult hippocampal neurogenesis appears to be a common mammalian trait (apart from cetaceans), this has important implications for the understanding of this neural phenomenon. While many factors influence the rate of proliferation and survival of newly born neurons in the adult hippocampus (e.g. Kempermann 2012), the fact that the vast majority of mammals are likely to have this trait indicates that adult hippocampal neurogenesis probably subserves an invariant function across mammalian species. Our broad survey of species examined questions concepts related to the environment and adult hippocampal neurogenesis, as the species investigated inhabit most of the environments in which mammals are found, from rainforests to deserts and terrestrial to aquatic. As mentioned earlier, newly formed neurons in the hippocampus appear to play a role in pattern separation, thus enhancing the circuits involved in learning and memory and has led to the memory resolution hypothesis for adult hippocampal neurogenesis (Sahay et al. 2011; Aimone et al. 2011). All mammalian species are likely to benefit from this circuitry enhancement, or increased memory resolution, no matter what environment they inhabit.

Why are cetaceans different from all other mammals?

Our findings raise the question of how cetaceans came to have small hippocampi that seem to lack adult neurogenesis and are loosely organized. In terms of general neuroanatomical structure (Manger 2006; Manger et al. 2004, 2012) and sleep physiology (Lyamin et al. 2008), cetaceans are different from all other mammals and the current study adds further support to this interpretation of cetacean neurobiology. Adult cetaceans lack, or have minimal, REM sleep (Lyamin et al. 2008) and appear to lack a clear sleep state for the first month of life, likely having less than 30 s of sleep during the first postnatal month (Lyamin et al. 2005), both aspects appearing to be features of their evolutionary adaptation to the thermally challenging aquatic environment. Studies of the effect of sleep deprivation on adult hippocampal neurogenesis in laboratory mammals (Meerlo et al. 2009) have shown that prolonged REM deprivation decreases cell proliferation rates and that prolonged deprivation of both NREM and REM sleep inhibits cell maturation and integration. This can occur independently of the release of adrenal stress hormones (Meerlo et al. 2009), as seen in mother cetaceans prior to and after birth (Lyamin et al. 2005). The lack of postnatal sleep and the continued lack of REM sleep throughout life may lead to a cessation of hippocampal neurogenesis immediately after birth in cetaceans, despite this not appearing to be a stress-related reduction in hippocampal neurogenesis. This cessation, sustained by a lack of REM sleep in older cetaceans, may prevent any postnatal enlargement of the hippocampus as seen in other mammals (Bayer 1980; Thompson 2012), leading to the observed small size of this structure in adult cetaceans. Moreover, if hippocampal development were arrested immediately postnatally in cetaceans as proposed, the loosely organized cetacean dentate gyrus is likely to be one result of this premature cessation of hippocampal development. It has been postulated that the risk of hypothermia in neonatal cetaceans underlies the lack of postnatal sleep (Lyamin et al. 2005, 2008), thus the current observations lend support to the thermogenesis hypothesis of cetacean brain evolution (Manger 2006).

What do these findings mean regarding cetacean cognitive capacities?

That cetaceans have small, loosely organized hippocampi that apparently lack adult hippocampal neurogenesis poses a serious problem for the hypothesis that these animals are, in comparison to most other mammals except great apes, highly cognitively complex (Marino et al. 2008). Here, we provide three examples of cognitive studies in which the hippocampus plays a central role that are instructive in understanding the results of behavioural experiments on cetaceans. In an object permanence task (invisible displacement/transposition task), cetaceans have been shown to possibly only reach Piaget stage 4 (visible displacement), whereas other mammals and birds tested readily reach stage 5 and apes achieve stage 6 (Mitchell and Hoban 2010; Jaakkola et al. 2010). Object permanence tasks are strongly hippocampus dependent, as they rely on spatial memory. Thus, the failure of cetaceans to clearly achieve higher than stage 4 on these tasks is in agreement with the lack of hippocampal development and adult neurogenesis demonstrated here. Additionally, as Piaget stage 6 of object permanence is thought to be a necessary requirement for mirror-self recognition (Mitchell and Hoban 2010), the lack of achievement of this level of object permanence by cetaceans questions the results of a previous study suggesting that dolphins have this cognitive ability (Reiss and Marino 2001). As a second example, the much lauded language comprehension studies of dolphins (Herman et al. 1984) can be appropriately contextualized. It should be noted that for a dolphin to begin to participate in the trials that probe semantic understanding requires at least 4 years of training (Herman et al. 1984). In a comparable experimental situation, sea lions were shown to reach similar levels of performance to dolphins on these tasks in 2 years or less (Schusterman and Kreiger 1984). In the current study, we have observed that pinnipeds have normal-sized hippocampi that possess adult hippocampal neurogenesis. Thus, success in these types of cognitive experiments that requires the formation and recall of hippocampus-dependent explicit memories and cognitive flexibility was clearly achieved more rapidly in sea lions than dolphins. As a third example, it has been shown that bottlenose dolphins fail to complete a spatial maze task associated with an “if and only if, then” construct on their own volition, whereas several other mammalian and vertebrate species tested readily achieved this combined maze and rule task (Nikolskaya 2005). To complete this task successfully, the animals were required to form memories of places and events, which, given the structure of the cetacean hippocampus, appears to be a cognitive task beyond their neural means.

It may be argued that the functions associated with the hippocampus with regard to complex cognition, learning, memory and spatial orientation have been subsumed into the circuitry in other parts of the cetacean brain, thus facilitating the expression of normally hippocampus dependent cognitive functions. This situation has been observed in rats, where the prefrontal cortex of rats with lesions of the dorsal hippocampus assumes hippocampal functions (Zelikowsky et al. 2013). Despite this, given the known neuroanatomy of the cetacean brain, where the prefrontal cortex appears almost absent (Manger 2006) and the entorhinal and subicular regions of the hippocampal formation appear to be proportionally smaller in the cetaceans mirroring the decrease in hippocampal size (Jacobs et al. 1971, 1979), it is difficult to speculate where this alternative circuitry might lie, how this may facilitate hippocampus-dependent functions or even if it would be as effective as the typical mammalian hippocampal circuitry in undertaking hippocampus-related tasks. Given the fact that it is far more difficult to misinterpret neuroanatomical structure than behavioural studies, the current and previous findings (Manger 2006; Manger et al. 2012) regarding cetacean brain structure appear to necessitate a reappraisal of our notions regarding the cognitive capabilities and behavioural studies of cetaceans (Manger 2013) and the evolution of relatively and absolutely large brain size in this mammalian order (Manger 2006).