The hippocampus of birds in a view of evolutionary connectomics

The avian brain displays a different brain architecture compared to mammals. This has led the first pioneers of comparative neuroanatomy to wrong conclusions about bird brain evolution by assuming that the avian telencephalon is a hypertrophied striatum. Based on growing evidence from divers analysis demonstrating that most of the avian forebrain is pallial in nature, this view has substantially changed during the past decades. Further, birds show cognitive abilities comparable to or even exceeding those of some mammals, even without a "six-layered" cortex. Beside higher associative regions, most of these cognitive functions include the processing of information in the hippocampal formation (HF) that shares a homologue structure in birds and mammals. Here we show with 3D polarized light imaging (3D-PLI) that the HF of pigeons like the mammalian HF shows regional specializations along the anterior-posterior axis in connectivity. In addition, different levels of adult neurogenesis were observed in the subdivisions of the HF per se and in the most caudal regions pointing towards a functional specialization along the anterior-posterior axis. Taken together our results point to species specific morphologies but still conserved hippocampal characteristics of connectivity, cells and adult neurogenesis if compared to the mammalian situation. Here our data provides new aspects for the ongoing discussion on hippocampal evolution and mind.

The avian brain displays a different brain architecture compared to mammals. This has led the first pioneers of comparative neuroanatomy to wrong conclusions about bird brain evolution by assuming that the avian telencephalon is a hypertrophied striatum. Based on growing evidence from divers analysis demonstrating that most of the avian forebrain is pallial in nature, this view has substantially changed during the past decades. Further, birds show cognitive abilities comparable to or even exceeding those of some mammals, even without a "six-layered" cortex. Beside higher associative regions, most of these cognitive functions include the processing of information in the hippocampal formation (HF) that shares a homologue structure in birds and mammals. Here we show with 3D polarized light imaging (3D-PLI) that the HF of pigeons like the mammalian HF shows regional specializations along the anterioreposterior axis in connectivity. In addition, different levels of adult neurogenesis were observed in the subdivisions of the HF per se and in the most caudal regions pointing towards a functional specialization along the anterioreposterior axis.

Introduction
Understanding the routes of brain evolution in the context of cognition seems to be a challenging task for comparative neuroscientists following the fact that there are still many gaps to fill with a limited amount of species living in the 21st century. Nevertheless, in the past 50 decades researchers have presented data reporting that birds, cephalopods and even other taxa without a "six-layered" neocortex, show complex and intelligent behavior, tool use, social altruism and far more other skills comparable to or even exceeding those of some mammals (Gü ntü rkü n & Bugnyar, 2016; Karten, 2015). Together all of these cognitive functions and skills involve a functional corticalebasal ganglia circuit, which seemed to be highly conserved between species, like for example, the song system of birds and speech production in humans (Chakraborty & Jarvis, 2015;Karten, 2015). These findings however, have increased comparative studies and ongoing discussions dealing with the evolution of the "six-layered" neocortex (Bolhuis & Wynne, 2009;Butler, Reiner, & Karten, 2011;Gabi et al.., 2016;Gü ntü rkü n, 2005; Gü ntü rkü n, Str€ ockens, Scarf, & Colombo, 2017;Herculano-Houzel, Catania, Manger, & Kaas, 2015;Hevner, 2016;Kaas & Stepniewska, 2015;Montiel, Vasistha, Garcia-Moreno, & Molnar, 2016;Olkowicz et al., 2016;Shepherd & Rowe, 2017). In contrast, debates about hippocampal evolution are relatively tame (Striedter, 2016), although hippocampal (archicortical) structures evolved sometime in-between the paleo-and the neocortex, and many cognitive functions involve hippocampal circuits that are further integrated in the corticalebasal ganglia network. In addition, recent evidence emerged that it is likely that the "six-layered" neocortex has evolved from a common ancestor with a "three-layered" cortex, a morphology that is still present in hippocampal regions of different species (Gü ntü rkü n et al., 2017; Rowe & Shepherd, 2016;Shepherd & Rowe, 2017;Striedter, 2016;Ulinski, 1983). In our view, this observation makes the hippocampus an excellent candidate to study cortical evolution because the hippocampus shares a common origin in the amniote species (Medina & Abellan, 2009;Puelles, 2001Puelles, , 2011 and many features of hippocampal functions are conserved, particularly spatial memory (Bingman & Muzio, 2017;Gagliardo, Ioale, & Bingman, 1999;Herold, Coppola, & Bingman, 2015;Witter, Kleven, & Kobro Flatmoen, 2017). Even more interesting, the hippocampal formation (HF) of birds displays a special morphology along its extension through the anterioreposterior axis. Most of it exhibits no clear three-layered laminar organization like the mammalian hippocampus or its homologue in the amniote species (Herold et al., 2014;Nomura & Hirata, 2017;Striedter, 2005Striedter, , 2016. This variation however, raises the question what other properties of neurons or neuronal networks are required to serve conserved functions and what differences may exist and have evolved over more than 350 Million years of separate evolution in adaption to an ecologic niche (Fig. 1). The HF in birds can be divided into several subdivisions, the V-complex, with its ventromedial (Vm), triangular (Tr) and ventrolateral (Vl) region, the dorsomedial region with its dorso-dorsomedial (DMd) and ventro-dorsomedial (DMv) subdivision and the dorsolateral region with its dorsodorsolateral (DLd) and ventro-dorsolateral (DLv) subdivision (Fig. 2;Herold et al., 2014). Despite the differences in the overall architecture, cell types and connections of the HF in birds show many similarities to those of mammals (Atoji, Wild, Yamamoto, & Suzuki, 2002;Casini, Bingman, & Bagnoli, 1986;Colombo & Broadbent, 2000;Hough, Pang, & Bingman, 2002;Krebs, Erichsen, & Bingman, 1991;MacPhail, 2002; Sherry, Grella, Guigueno, White, & Marrone, 2017). However, there is an ongoing discussion, which subdivisions of the HF in birds correspond to their mammalian counterparts, i.e., the dentate gyrus (DG), the cornu ammonis (CA) fields, the subiculum and the entorhinal cortex (Abellan, Desfilis, & Medina, 2014;Atoji, Sarkar, & Wild, 2016;Herold et al., 2015;Medina, Abell an, & Desfilis, 2017). New theories of the evolution of DG as a late evolving structure and an exclusive addition to the hippocampus of mammals have further questioned whether a DG as strict as in mammals even exists in birds or not (Atoji et al., 2016;Bingman & Muzio, 2017;Bingman, Rodriguez, & Salas, 2017;Kempermann, 2012;Striedter, 2016). However, recent findings by analysis of single-cell transcriptomics of the reptilian medial most cortex, particularly the dorsomedial cortex, support the hypothesis that the hippocampal regions, including DG, CA1 and CA3 were already present in the ancestor of all amniotes (Tosches et al., 2018).
Yet, another clue comes from new insights of functionally different domains in the mammalian hippocampus. Based on genetic and connectional data the hippocampus of mammals can be compartmentalized into multiple domains along the longitudinal/dorsal-ventral axis (Dong, Swanson  Analysis of single cell transcriptomics support the idea that the anlage of the hippocampal formation, i.e., dentate gyrus and cornu ammonis was already present in the ancestor of all amniotes (Tosches et al., 2018). Since then, several million years of separate evolution occurred between birds and mammals. c o r t e x 1 1 8 ( 2 0 1 9 ) 1 6 5 e1 8 7 2017). Further, a large body of connectional and functional data, including adult neurogenesis in rodents and primates, revealed that the dorsal hippocampus (septal pole or posterior in primates) is involved in spatial memory and navigation, and showed higher levels of adult neurogenesis, while the ventral hippocampus (temporal pole or anterior in primates) mediates emotion, anxiety, stress-related behavior and cognitive flexibility, and exhibits lower levels of adult neurogenesis (Anacker & Hen, 2017;Colombo, Fernandez, Nakamura, & Gross, 1998;Fanselow & Dong, 2010;Strange, Witter, Lein, & Moser, 2014;Witter et al., 2017). This functional differentiation in case of stressrelated behavior might also exist in the avian HF (Robertson et al., 2017;Smulders, 2017).
Up to now, only a few researchers have explored the rostro-caudal axis in the avian HF (Fig. 2) and less is known about the basis of functional differences along the axis. Therefore, we decided to analyze two different components of hippocampal functional circuity. First, we will present data for the homing pigeon (Columba livia f.d.) with a focus on rostro-caudal differences in hippocampal connectivity at the microscale level by using the newly developed method of 3D-polarized light imaging (3D-PLI). 3D-PLI has been shown to be a powerful tool to analyze fiber architecture, and the course of fibers and fiber tracts in mammals (Axer, Amunts, et al., 2011;Zeineh et al., 2017;Zilles et al., 2016). Second, we will present data of neurogenic markers for adult neurogenesis in the pigeon rostro-caudal axis of the HF.

2.
Material and methods

Animals
In total, we examined eleven adult homing pigeon brains (C. livia f.d.) of the same age originating from our local breeding colony, which were raised in the same loft (250 cm Â 190 cm Â 190 cm) and lived there under identical conditions. The pigeons were allowed to fly freely and food, grit and water were provided ad libitum. All experimental procedures were approved by the national authority (LANUV NRW, Germany) and were carried out in accordance with the National Institute of Health Guide for Care and Use of Laboratory Animals.

Tissue-processing
Animals for the Polarized Light Imaging (n ¼ 2) were euthanized with Pentobarbital (70 mg/kg), decapitated, brains were removed from the skull, and immediately fixed in 4% buffered formalin pH 7 and stored at 4 C. Two weeks later brains were transferred into a solution of 10% Glycerin, 2% DMSO and 4% formalin pH 7 for five days, and subsequently transferred into a solution of 20% Glycerin, 2% DMSO and 4% formalin pH 7 for additional two weeks. Brains were frozen and stored at À80 C. They were cut in serial coronal or sagittal 60 mm sections with a cryostat microtome (Leica Biosystems, Nussloch Germany). Slices were thaw mounted on un-gelatinized slides, freeze dried and coated in 20% Glycerin three days before the 3D-PLI measurements started. After the measurements, slices were further processed for Nissl staining with Cresyl violet.

Imaging
In order to make the fiber architecture of brain tissue visible, linearly polarized light has been applied to the histological sections and the light transmission has been sampled at 9 or 18 vertical polarization planes covering 180 by means of a circular analyzer (Axer, Amunts, et al., 2011;. The interaction of polarized light with the birefringent components of the nerve tissue, i.e., the myelin, allows for the extraction of predominating fiber orientations in tiny volumes of 1.3 mm Â 1.3 mm Â 60 mm. The polarimetric setup is based on a standard bright field microscope (with K€ ohler illumination) with two polarizing filters and a movable specimen stage (M€ arzh€ auser Wetzlar, Germany) introduced into the beam path (Fig. 3). The wavelength spectrum (550 ± 5 nm) has been adapted to the polarizing filter specifications by means of a white light LED in combination with a narrowband pass filter next in line. The size of the square field of view of the monochrome CCD camera (QImaging Retiga 4000R) is 2.7 Â 2.7 mm 2 providing a pixel resolution of 1.3 microns. The transmission intensity of the whole specimen is imaged with a tile overlap of 1.0 mm at 18 vertical polarization plane angles (r) for the frontal sections and with a tile overlap of .75 mm at 9 vertical polarization angles for the sagittal sections respectively, in order to determine accurately the predominant fiber orientation in each of 1.3 mm Â 1.3 mm Â 60 mm voxel (Fig. 4f). Schematic outlines along the anterioreposterior axis according to Herold et al. (2014). Atlas levels referring to Karten and Hodos (1967, p. 195). The major subdivisions are the V-region comprising the lateral (Vl) and medial (Vm) layers and the triangular area in between (Tr), the dorsomedial region with its dorsal (DMd) and ventral (DMv) subdifferentiation, and the dorsolateral region (DL) including the further subdivisions dorsal (DLd) and ventral (DLv). c o r t e x 1 1 8 ( 2 0 1 9 ) 1 6 5 e1 8 7 Using the Jones calculus for flat polarization optics (Jones, 1941) three modalities can be derived from the sinusoidal dependence ( Fig. 4e) of the transmission intensity I on the polarization plane angle (r): Transmittance (arbitrary units) is the mean intensity of the light transmission through the tissue (Fig. 4a). Retardation (0e1) is the vertical projection of the cumulative tissue birefringence normalized by the transmittance at a given light wavelength (Fig. 4b). Direction (0 e180 ) is the predominant inplane nerve fiber orientation (Fig. 4c).
The out-of-plane elevation angle of the nerve fiber is called Inclination (Fig. 4d) and can be estimated from retardation and transmittance by means of four parameters determined on the basis of their gray value distributions over the whole section: 1) retmaxwm: maximum retardation of the white matter 2) retmaxgm: maximum retardation of the gray matter 3) tmaxgm: maximum transmittance of the gray matter 4) tmeanwm: mean transmittance of the white matter (Reckfort, 2015). The parameter retmaxgm has been introduced in addition to account for the weak influence of the gray matter retardation. These four parameters approximately determine the relation between the retardation and the inclination by means of the transmittance: Due to the presence of artificial inhomogeneity in the transmittance induced by glycerol at variable time delays between tissue embedding and 3D-PLI-measurement, these parameters had to be slightly adapted subsequently to avoid the saturation of the inclination at 0 or 90 degrees in areas prone to this artifact like deep white matter or scarcely myelinated nuclei. The inclination values are compromised by this modification to the point of about þ5 up to þ10 .

Fiber orientation maps
Direction, inclination and mask provide a full set of polar coordinates for the calculation of a 3D-vector array of fiber orientations. The vector data of a single section provided by 3D-PLI are represented by color images called fiber orientation maps (FOMs). A point on the colored surface of a hemisphere represents the color of a corresponding 3D fiber orientation (Fig. 4f). Yet a quarter sphere is sufficient because flat 3D-PLI without tilting the light beam does neither differentiate forward and backward orientation nor the sign of the fiber inclination (downward or upward). The hue corresponds to the direction of the in-plane projection of the fiber orientation.
The fiber inclination on the other hand is coded by saturation and brightness (Fig. 4g). In case of the RGB-color-code ( Fig. 5A) hemisphere fiber orientations are reflected by colors only. The principal directions (lefteright, upedown, fronterear) are corresponding to the fundamental colors (red, green, blue). In the HSV-color-code ( Fig. 5B) hemisphere fiber orientations are reflected by hue, saturation and brightness. In the HSV-black version the brightness decreases with increasing inclination staining the poles at 90 black.
The reference for a single vector representation is generally not a single fiber but all birefringent tissue compartments inside a volume element (voxel) contributing to an image pixel. There are two reasons for signal loss in the white matter: fiber crossings and steep fibers with inclinations nearby 90 . Therefore, in areas of massive fiber intermingling at scales below the section thickness of 60 microns the direct representation of the fiber orientation by color and saturation is replaced by an extinction texture. Hence fibers stay visible in the FOM, however, orientation values are getting lost (Fig. 5C).
1) Briefly, free-floating sections were rinsed 2 times in phosphate buffered saline (PBS). Subsequently slices were incubated in 2 N HCl at 45 C for 30 min and .1 M borate buffer pH 8.5 for 10 min. Sections were than washed in PBS for 10 min that was followed by a blocking step with 3% goat serum (Vector, USA) in PBS þ Triton-X .3% (PBS-T) for 60 min. This was followed by incubation with a primary antibody mix anti-BrdU (1:200) and anti-DCX (1:500) overnight at 4 C. After that sections were washed 3 times in PBS that was followed by incubation with the secondary antibody mix containing goat anti-rat Alexa 488 (1:200) and goat antirabbit CY3 (1:500; both Jackson-Immuno Research, USA) for 2 h at darkness. At all further steps, sections were handled at a minimum of light or in complete darkness. After incubation sections were washed 3 times with PBS and 2 times with PB, mounted on gelatin-coated glass slides, airdried for 24 h and then dipped into distilled water and directly coverslipped with Fluoromount-G (Southern Biotec, USA). The slides were than taken to a fluorescent microscope system AxioScan.Z1 (Zeiss, Germany) and the whole sections were scanned for digitalization of photomicrographs at 20Â magnification. All sections were analyzed with the software Zen2 (Zeiss, Germany), by delineation of ROIs (i.e., subdivisions of the HF) and counting immunereactive cells with the counting tool so that a value for each region of the number of immunoreactive cells (either BrdU, or DCX)/area (mm 2 ) was determined. 2) To detect BrdUþ cells, sections were treated like in the labeling procedure of 1). After the pre-processing, sections were incubated with the primary antibody mix anti-BrdU   All antibodies were validated by Western-blotting procedures or showed specific binding in previous studies ( Fig. 5E; Melleu, Santos, Lino-de-Oliveira, & Marino-Neto, 2013;Robertson et al., 2017). Tissue samples were lysed at 4 C with 50 mM TriseHCL buffer (pH 7.4) containing 1% IGEPAL, 150 mM NaCl, 1 mM ethylenediaminetetraacetic acid, .1% SDS, 1 mM dithiothreitol and 2% glycerin. After homogenization, the lysates were centrifuged at 20.000 g at 4 C. For sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-Page) and Western Blot analysis, the supernatant was added 1:4 to a 4Â gel loading buffer containing 1.25% mercaptoethanol (pH 6.8) and heated to 95 C for five minutes. The probes were subject to gel electrophoresis (12%) using .3 mg protein per lane. Proteins were transferred to nitrocellulose membranes using a dry transfer apparatus (Invitrogen, Carlsbad, USA). Blots were blocked for 1 h with a blocking solution containing bovine serum albumin solubilized in 20 mM TriseHCL pH 7.6, 150 mM NaCl and .1% Tween 20, and incubated overnight with the primary antibody (see list above) at 4 C. After several rinsing steps and incubation with horseradish peroxidasecoupled immunoglobulin G, anti-rabbit IgG or anti-mouse IgG (both Jackson-Immuno Research, USA) at room temperature for 1 h, blots were washed and developed by using enhanced chemiluminescent detection (BioRad).

Data analysis
At a first step, the atlas levels (Karten & Hodos, 1967, p. 195) of the brain sections from pigeons for 3D-PLI and immunohistochemistry measurements were determined. For analysis of the 3D-PLI data, visual analyses of frontal and sagittal sections enabled us to take advantage of the extremely high in-plane resolution of 3D-PLI to optimally detect the dorsaleventral and medialelateral trajectories in the frontal images, and dorsaleventral and anterioreposterior trajectories in the sagittal images. We identified the orientation of fibers by examining the color-coded fiber orientation maps, examining the color as well as the image texture (e.g., numerous lines pointing in one direction). Dark regions on transmittance images that were not cell bodies were generally considered to represent myelin. 3D-PLI is maximally sensitive for fibers oriented in-plane; on frontal section images, fibers oriented through-plane (i.e., along the anterioreposterior axis of the hippocampus formation) would appear dark on the transmittance images, but show reduced or no color orientation. Therefore, the sagittal images were used to confirm through-plane orientation of pathways. Further anterioreposterior long-range connections were first analyzed and identified in the sagittal sections, and then combined with the analysis of the frontal sections.
Combining the results of 3D-PLI with our subregion segmentation, we identified the major pathways according to known connectivity patterns from earlier avian studies (see ; Shanahan, Bingman, Shimizu, Wild, & Gü ntü rkü n, 2013 for review), and compared it with the PLI data. A known limitation of 3D-PLI fiber tracking, however, is that we cannot determine, whether a fiber, or a fiber bundle, is entering or leaving the HF, so we used earlier results of tracing studies to guide the interpretation.
For immunohistochemistry, data of the hippocampal subregions defined by Herold et al. (2014) from atlas levels A 10.25 (anterior) to A 3.75 (posterior) were analyzed. The number of immunoreactive cells (BrdUþ, DCXþ, BrdUþ/NeuNþ, BrdUþ/GFAPþ)/area (mm 2 ) was averaged for the following coordinates ±500 mm: A 4.25, A 5.50, A 6.75, A 8.25 and A 9.50. To determine differences in BrdUþ, DCXþ, BrdUþ/NeuNþ, BrdUþ/GFAPþ cells among the HF at different anterioreposterior levels or across subdivisions, we first applied a Friedman ANOVA for each tested neurogenic marker or marker pair. If significant, pair-wise comparisons were run with the Wilcoxon-rank test. For the general statistical analyses, Statistica 13 (StatSoft, Tulsa, USA) was used. The significance level was set at .05.

Connectivity of the pigeon hippocampal formation along the anterioreposterior axis
Along the anterioreposterior axis, the pigeon HF is moving from a rostro-medial to a caudo-lateral position in the dorsal forebrain ( Fig. 2). We generally confirmed earlier tracing studies showing that the pigeon HF is connected to the lateral and medial septum (SL and SM), the nucleus taeniae of the amygdala (TnA), the lateral part of the bed nucleus of the stria terminalis lateralis (BSTL), the nucleus of the diagonal band (NDB), the hypothalamus, i.e., the nucleus periventricularis magnocellularis (PVM), the dorsolateral corticoid area (CDL), the hyperpallium apicale (HA) and laterale (HL), the cortex piriformis (CPi), the arcopallium ventrale (AV) and intermediale (AI) and inter-hippocampally via commissural fibers through the commissura pallii (CPa; Atoji, Saito, & Wild, 2006;Atoji & Wild, 2004;Atoji et al., 2002;Benowitz & Karten, 1976;Casini et al., 1986;Krayniak & Siegel, 1978a, 1987bShanahan et al., 2013;Szekely & Krebs, 1996). Because the focus of this study was to examine the differences in anterioreposterior trajectories we analyzed the above mentioned main neural pathways/long-range connections and mostly skipped the c o r t e x 1 1 8 ( 2 0 1 9 ) 1 6 5 e1 8 7 local circuitry. To do so, we combined our observations with results of former tracing studies and included whether a connection was afferent or efferent or reciprocal (for details see Atoji and Wild, 2004;Shanahan et al., 2013 for review).
Beginning very close to the zero point of the medial level, the interhemispheric fissure (medial telencephalic wall), we observed a massive amount of fibers connecting the hippocampus with the septal region in the dorso-ventral axis that were topographically organized along the anterioreposterior axis ( Fig. 7AeD). At this level, direct hypothalamic connections of the PVM and the caudal DMd region of the hippocampus were visible and hippocampaleseptal projections to SL and SM were prominent (Fig. 7AeC). More dorsally fibers of the septal region joined the tractus cortico-habenularis et cortico-septalis (CHCS). Fibers that popped out from the dorso-rostral HF to the most medial portion of the tractus septopallio-mesencephalicus (TSM) were also present at this level. The commissural fibers leaving the CPa, curved to the hippocampaleseptal junction, ran upwards to the medial wall before they spread into the subdivisions of the HF (caudal route). Most of them originate in the V-complex and DM of the contralateral HF. Visualization of CPa in frontal sections around A 7.50 confirmed this and showed fibers of the V-complex and DM running downwards to the contralateral side, crossing between the two hemispheres indicated by the color-switch, as soon as they reach the contralateral side along the rostral route (Fig. 7D). In the dorsal regions of the HF many thin fibers travel along the anterioreposterior axis connecting anterior and posterior regions ( Fig. 7B and D).
Moving more laterally four major pathways were recognized where efferents and afferents pass through the medial wall (Fig. 8AeE). The details of the different paths are presented in Fig. 8BeE. At the lateral level around Lat .50, all subdivisions of the HF along the anterioreposterior axis could be differentiated by the fiber architecture (Fig. 8B) and an additional superficial pathway was recognized at the caudal pole of the HF along Vm and DM where fibers inclines to a 90 direction, i.e., travel in a medialelateral direction or vice versa (Figs. 7B and 8C). Inspection of the most lateral sagittal slides showed that fibers from caudal CPi, PoA and Fig. 7 e The spatial course of massive long-range connections of the hippocampal formation along the anterioreposterior axis. A: 3D-PLI image at the sagittal level around Lat .00 close to the interhemispheric fissure/medial wall. The hue of the color wheel indicates the direction of the in-plane fiber orientation (for explanation see 2.2.2), and the brightness/darkness of the color (e.g., more peripheral/central in the color wheel) indicates a primarily in-plane/through-plane orientation. Red outlines mark the hippocampal formation. White lines show direct connections of the hypothalamus, the nucleus periventricularis magnocellularis (PVM) and the caudal dorsomedial region. Hippocampaleseptal connections are indicated by orange lines (lateral septum, SL) and light pink lines (medial septum, SM). The course of commissural fibers running in the commissura pallii (CPa) is marked with yellow. Fibers popping out from the most medial position of the anterior tractus septopallio-mesencephalicus (TSM) are in light blue revealing connections to rostral HF, the hyperpallium apicale (HA) and the mesopallium (M). More dorsally fibers of the septal region belong to the tractus cortico-habenularis et cortico-septalis (CHCS). B: Along the anterioreposterior axis, the V-region and the ventral dorsomedial region (DMv) show a high density of thick fibers or fiber bundles traveling in the dorso-ventral axis that are topographically organized along the anterioreposterior axis, while in the dorsal regions many thin fibers traveling along the anterioreposterior axis connecting anterior and posterior regions (white line). C: Enlargement of the connection between PVM and the dorsomedial region (DM), as well as fiber organization of the CPa and septal regions. D: Visualization of the CPa in a frontal section at A.7.50. As indicated by the color-switch (white arrow), fibers of the V-region and DM cross between the two hemispheres and run upwards to the contralateral side (yellow line). Comparisons between frontal and sagittal sections as well as inclination images show that these fibers stay in plane, i.e., running towards the same level at the anterioreposterior axis they originated. E: Additional legend showing the sinusoidal dependence of the 3D-PLI images. Hemisphere fiber orientations are reflected by hue, saturation and brightness. A change in color corresponds to a change in direction of the fiber. With increased inclination fibers appear in black.
AV as well as intra-hippocampal fibers (reciprocal) connect along this route, while at more medial sagittal levels fibers from SPC join this pathway. The four major pathways included fibers in frontal as well as sagittal sections that connect BSTL and NDB with the HF traveling together from the basal tip of the ventricle along the medial side of the ventricle close to latero-dorsal SL. Further, a few olfactory fibers also join this pathway. Fibers and fiber bundles of the hippocampal connection between SL and SM were additionally separated into two pathways at the level of the nuclei in the septal region (Fig. 8D). The fourth path comprised the crossing fibers of the HF through CPa that get intermingled with the other pathways by passing through the medial wall and then curve separately slightly rostroventrally to reach the commissure or vice versa (Fig. 8E).
The visualization of the fiber courses in the tractus CHCS and in the dorsal regions of the HF showed in frontal sections around atlas level A6.50 fibers traveling along the tractus CHCS started to run outwards of DM and the V-region from dorsal to ventral joining the CHCS with increased inclination in caudal direction (Fig. 9A, named post-commissural in Krayniak & Siegel, 1978b). At frontal sections around level A 7.25, at the most caudal point of the septal nuclei SL and SM, some fibers from DL and DM still travel to the medial wall along the periphery of the HF or along the ventricle in a dorso-ventral direction with increased inclination, while a few fibers already started to run further ventrally to SL (Fig. 9B). At both frontal levels, the fiber orientation maps of the DMd and DLd region mostly appeared in dark color, indicating intense fiber intermingling and/or fibers coursing in the anterioreposterior axis ( Fig. 9A and B). As verified in the sagittal sections (Fig. 9C), indeed many fibers travel through/from DLd and DMd along the anterioreposterior axis. Those fibers mainly represent the connections between the CPi, the hyperpallium and the HF, as well as intrinsic connections of DM and DL. Fibers from CPi travel dorsally from anterior to posterior along the outer surface of HA reaching the HF at the border of DL (asterisk). In addition, radial fibers from HD, HL and HA cross this fiber tract (arrows), join along its path, and get indistinguishable from olfactory fibers (Fig. 9D). At lateral level Lat 1.00, the CHCS parallels the stria medullaris (SMe) as shown in a sagittal section of the right hemisphere and was clearly separated from the SMe (Fig. 9E). At this level fibers from the habenula, as well as fibers from nucleus taeniae of the amygdala (TnA) have joined the tract.
Around sagittal level Lat 1.80, the fiber courses of TnA that emerged at the ventro(basal)-medial site of the ventricle add to the medial route of afferents from TnA to HF (Fig. 10A). After reaching this point, fibers travel to the caudal end of the hippocampus and then turn into a dorso-medial direction providing the input to the caudal half of the HF. This was verified by inspection of frontal sections around atlas level A 6.25. Fibers travel from TnA medially along the bottom of the ventricle and incline to the caudal pole. Further, we observed that the fiber bundle of TnA and the fiber bundle from the habenula joined the hippocampaleseptal bundle at this coordinate (Fig. 10B).
Moving more laterally to level Lat 3.50 the area corticoidea dorsolateralis (CDL) was visible ( Fig. 10C and D). The analysis of the connectivity between the HF and the CDL revealed many connections between CDL and DMv, CDL and the V- region and to a lesser extent between CDL and DMd, and CDL and DL.

3.2.
Adult neurogenesis in the pigeon hippocampal formation along the rostro-caudal axis DCXþ cells were differentially expressed in the HF [c 2 (n ¼ 9, df ¼ 6) ¼ 51.05, p < .001; Fig. 11A]. The highest numbers were detected in Vl (to all subdivisions of the HF, p < .01, except compared to Dlv, p < .05) and the lowest in DMd (to all, p < .01). Additionally, Tr and Vm showed relatively low levels of DCX expressing cells (to all, p < .01, except Tr compared to Vm, p < .05). In DLd, DLv and DMv ovoid cells were higher expressed compared to Tr cells (to all p < .05), while in Vl more often Tr cells expressed DCX (p < .01; Fig. 11A). Along the anterioreposterior axis overall DCXþ cells showed a different distribution in the HF [c 2 (n ¼ 9, df ¼ 4) ¼ 17.78, p < .01; Fig. 11B]. The number of DCXþ cells specifically dropped at the most . Fibers begin to run outwards of DM and the V-region running from dorsal to ventral joining the CHCS with increased inclination. B: Frontal section around atlas level A 7.25 at the most caudal point of the septal nuclei SL and SM. Fibers from DL and DM travel dorso-medially along the periphery of the HF or along the ventricle to a ventral direction with increased inclination. In both, (A) and (B), further inspection of the sagittal slices showed that fibers run to the septal nuclei and then travel rostral. Additionally, most of the DMd and DLd region appears dark, indicating intense fiber intermingling and/or fibers coursing in the anterioreposterior axis. As verified in (C) in an example of a sagittal section, indeed many fibers travel through/from DLd and DMd along the anterioreposterior axis. C: Fibers in DMd running in the anterioreposterior axis. Fibers from the cortex piriformis (CPi) travel dorsally from anterior to posterior along the outer surface of HA reaching the hippocampal formation at the border of DL (asterisk). In addition, radial fibers from HA/HL cross this fiber tract (arrows), join along its way particularly from more anterior portions and get indistinguishable from olfactory fibers (D). E: The CHCS parallels the stria medullaris (SMe) in a sagittal section of the right hemisphere around atlas level Lat 1.00 and is clearly separated (see white arrows; CHCS: upper arrow, SMe lower arrow).
posterior parts of the HF around atlas level A 4.25 ± .5 (to all other atlas levels, p < .01). This effect mainly results from a decrease of Tr DCXþ cells at the most posterior level [c 2 (n ¼ 9, df ¼ 4) ¼ 12.18, p < .05], while ovoid DCXþ cells were almost equally distributed along the anterioreposterior axis (Fig. 11B). Additionally, very few DCXþ/BrdUþ positive cells in the HF were detected, indicating that even after three month of injections of BrdU immature neurons in the pigeon HF are still in a "waiting phase" on their way to mature neurons.

Discussion
The present study disclosed the axonal architecture of the pigeon along the anterioreposterior axis including long range connection using 3D-PLI as a high-resolution imaging approach to study the connectivity structure in the whole brain (Axer, Amunts, et al., 2011;Schubert et al., 2016;Zilles et al., 2016). It verified that subdivisions of the pigeon HF are partly topographically connected along the anterioreposterior axis with regard to the long-range connectivity and replicated earlier studies of the connectome of the HF in pigeons (Atoji & Wild, 2005;Atoji et al., 2002Wild, 2004, Casini et al., 1986;Krayniak & Siegel, 1978a, 1987b. To visualize the long-range connections, we used 3D-PLI in two different cutting directions, i.e., sagittal and frontal. The data presented here show for the first time the useful application of 3D-PLI in a bird brain. We furthermore demonstrate regional differences in adult neurogenesis/gliogenesis in individual subdivisions of the pigeon HF (as defined by Herold et al., 2014) by quantitative measures of different markers that have been demonstrated to reliably map neurogenesis in a wide range of species. Here, detailed analysis of five different atlas levels along the anterioreposterior axis indicated higher rates of immature neurons along the axis compared to the most caudal pole of the HF, while globally, matured newborn neurons/glia only showed a trend towards different levels. However, analysis of subdivisions with high or moderate neurogenesis levels showed that newborn glial cells decreased from anterior to posterior levels in the lateral blade of the V-region (Vl). Our  data supports the idea that there is indeed a specialization to regional subdivisions of the HF along the rostro-medial to the caudo-lateral axis of the HF in birds (Smulders, 2017).

Connectivity of the hippocampal formation along the anterioreposterior axis in the avian brain
Numerous studies have investigated the connectome of the avian HF (see Shanahan et al., 2013;Szekely, 1999 for review), but, none of them did explicitly studied the anterioreposterior axis, although some functional differences were reported in divers bird species. Nevertheless, a few studies mentioned or showed to some extent data of connections along the anterioreposterior axis of the HF (Atoji & Wild, 2004;Atoji et al., 2002;Krayniak & Siegel, 1978a, 1978bMontagnese, Zachar, B alint, & Csillag, 2008). In general, the HF projects massively to the septal nuclei and these projections are topographically organized (our data and Atoji & Wild, 2004;Atoji et al., 2002;Krayniak & Siegel, 1978a, 1978bMontagnese et al., 2008). Caudal HF projects to the post-commissural septum and more rostral HF projects to the rostral septum and the NDB. However, the rostral reciprocal connections to NDB may be more widespread along the anterioreposterior axis according to Atoji and Wild (2002), but are in line with observations from Krayniak and Siegel (1978a). At least, more rostro-medial NDB connects along the medial wall with the HF, while connections with more caudo-lateral NDB were observed with fibers that travel from/to the tip of the ventricle on the same route that connect the HF and the BSTL. According to , projections from HF to BSTL originate all along the rostro-caudal axis. However, we could not confirm a lateral route from DM to BSTL via the CDL. The interhippocampal commissural projections are also topographically organized, i.e., rostral, middle and caudal HF projects to corresponding levels of the contralateral HF. Commissural fibers were organized in a caudal (levels posterior to A 7.5) and a rostral route (levels anterior to A 7.5). These findings are in line with observations by Atoji et al. (2002) and Krayniak and Siegel (1978a). Input from the nucleus taenia (TnA) that belongs to a group of nuclei in the arcopallium/amygdala complex (Herold, Paulitschek, Palomero-Gallagher, Gü ntü rkü n, & Zilles, 2018) is limited to the middle/caudal HF to approximately level A 6.25. Fibers between the nucleus posterioris amygdalopallii (PoA) and caudal hippocampal subdivisions course mostly transversally (dorso-laterally) along the outer surface of the telencephalon, while fibers between the ventral arcopallium (AV) or intermediate arcopallium (AI) and the HF use both, the "outer" dorsolateral and the "inner" Vm path along the ventralemedial edge of the ventricle. Hereby, more rostral levels, i.e., more laterally AV fibers travel through the dorsolateral pathway and at more caudal levels, i.e., more medially AV/AIv fibers join the Vm pathway. Whether these two pathways reflect afferent and efferent pathways, or different subtle pathways or only different rostro-caudal connectivity have to be further explored, as well as reinterpretation of data according to the different arcopallial subdivisions (Herold et al., 2018) that receive projections from the HF or vice versa. Both pathways have been described earlier, but with unclear demarcation of the arcopallium/amygdala complex nuclei (Atoji et al., 2002;Casini et al., 1986;Kr€ oner et al., 1999). We further observed a direct connection to the nucleus PVM that was limited to a small portion in caudal DM. This connection was described earlier (Bons, Bouill e, Bayl e, & Assenmacher, 1976;Bouill e & Bayl e, 1973) and represents afferents from the dorsomedial HF that modulate corticosterone plasma levels. Fibers that join the TSM were limited to the anterior HF, thereby possibly connecting to brain stem sites (Reiner et al., 2004). Further connections with other extrahippocampal structures like hyperpallial subdivisions, olfactory regions or CDL did not differ in their connectivity along the anterioreposterior axis. However, from the border of HA and DL, an enlarging number of thin fibers travel along anterioreposterior axis along the dorsal regions through/over/in DL and DM to the caudal pole of the HF. Before reaching the caudal pole, some of them turn into the V-region, and then turn ventro-laterally to Vl. To our knowledge, none of the earlier studies mentioned these numerous fibers traveling in the anterioreposterior axis before. To sum up, our results support the idea of a specialization of subdivisions of the HF along the anterioreposterior axis and may also imply functionally different contributions of dorsal and ventral DM and DL. Thereby our results may stimulate future studies to investigate these differences, for example, in relation to spatial learning and emotional behavior.
A number of markers that are expressed at specific stages during neuronal development or that indicate actively dividing cells have been tested in addition to the "gold standard" BrdU, a thymidine analogue that is inserted into the DNA in actively dividing cells after administration. Doublecortin (DCX), for example, is expressed in many advanced precursor or at an early post-mitotic stage of immature neurons and has been widely accepted as an indicator for adult neurogenesis and neuronal plasticity although some challenges still exist (Balthazart & Ball, 2014;Kremer et al., 2013). Other markers, like Ki67 or PCNA are strictly associated with cell proliferation (Moldovan, Pfander, & Jentsch, 2007;Scholzen & Gerdes, 2000) and Hu expression marks neuronal progenitor cells at an early stage (Barami, Iversen, Furneaux, & Goldman, 1995). NeuN and GFAP on the other hand serve as markers for mature neurons and glia cells. Although most of the antibodies used in this study were used in different species before, including different bird species, we validated the antibodies not only with negative controls but also with Western blot procedures by using fresh brain tissue, and confirmed the correct binding site through protein weights. Immature neurons were verified with DCX, and newborn neurons were proved with BrdU/NeuN double labeling, while newborn glial cells were determined with BrdU/GFAP double labeling.
In pigeons, qualitative and quantitative analysis showed lower amounts of DCXþ cells in Tr and Vm compared to Vl, while DM and DL were only sparsely labeled (Melleu et al., 2013) and lower amounts of DCXþ/PCNAþ cells in Vm compared to moderate levels in Tr and Vl, and higher amounts in DL compared to DM (Mazengenya, Bhagwandin, Nkomozepi, Manger, & Ihunwo, 2017). Our quantitative approach confirmed the highest number of DCXþ cells in Vl followed by DLv and DLd, while the lowest numbers of cells/mm 2 were detected in DMd. Further, ovoid migrating immature neurons were more abundant in regions of high to moderate levels of DCX expressing cells, except in Vl, where Tr immature neurons that have already reached their destination were more frequent. We underpinned the DCX results by the finding that mature, adult-born neurons also showed the highest amounts in Vl and DLv, which was determined with BrdUþ/ NeuNþ labeling after three month of BrdU injections. Functionally, newborn cells in the avian HF have been related to age (Meskenaite et al., 2016), food-deprivation i.e., stress (Robertson et al., 2017;Smulders, 2017), housing, i.e., environmental enrichment (Melleu et al., 2016), food storing (Sherry & Hoshooley, 2010), seasonal changes (Sherry & MacDougall-Shackleton, 2015) and spatial memory . To our knowledge, none of the studies mentioned earlier in this section studied the genesis of glial cells in the HF. Here we found higher numbers of BrdUþ/GFAPþ labeled cells in all estimated subdivisions compared to newborn neurons. In contrast to newborn neurons, newborn glial cells were higher in Vm and DMd compared to all other subdivisions. In general, GFAPþ cells of different size were recognized all over the pigeon HF, with some of them showing long radial branches from the outer surface of Vm and Vl into the Tr region.
Anterioreposterior differences in adult neurogenesis were reported in a study that used radioactive labeled thymidine to mark newborn cells in the HF of black-capped chickadees (Barnea & Nottebohm, 1994). In the study by Barnea and Nottebohm (1994), higher levels of adult neurogenesis in the anterior HF, compared to middle and posterior HF in wild, but not in captive hold birds was demonstrated, which was interpreted in relation to spatial memory acquisition in wild birds visiting different places to collect food. Further, newborn cell numbers were found to decline from anterior to middle to caudal levels, and in relation to the survival time of birds after the injections. In addition, higher levels in the caudal HF of birds that socially interact compared to single caged birds were described (Barnea et al., 2006). The levels of adult neurogenesis changed during season of these food-hoarding species, showing that seasonal changes of behavior affect this rostrocaudal gradient and can result in similar levels of adult neurogenesis in the HF along the anterioreposterior axis (Hoshooley et al., 2007). Under stress conditions, Robertson et al. (2017) reported increased levels of corticosterone plasma levels and reduced levels of newborn cells in the neuronal lineage (BrdUþ/Huþ) in food-restricted chicken in rostral HF. However, levels of BrdUþ cells alone decreased in both, caudal and rostral HF after one week of the injection with BrdU. No general differences in BrdUþ/Huþ or BrdUþ cells between rostral and caudal HF were reported, but BrdUþ cells in the ventricular zone showed higher amounts at rostral levels (Robertson et al., 2017). In line with this we found BrdUþ/ GFAPþ cells gradually decrease from anterior to posterior positions in Vl, showing that a higher number of dividing cells in the more anterior subdivision most likely find their fate in providing an optimal environment for neuronal function and survival. Taken together, our results fit in with former studies, reporting lower amounts of immature neurons at the caudal c o r t e x 1 1 8 ( 2 0 1 9 ) 1 6 5 e1 8 7 pole, and no gradient in overall BrdUþ cells, although none of the studies investigated values for individual subdivisions. In other species, like zebra finches, brown-headed cowbirds and red-winged blackbirds variable neurogenesis levels along the anterioreposterior axis dependent on different housing conditions (social interaction), food storing behavior (spatial memory) and breeding conditions were reported (Barkan et al., 2016;Barnea et al., 2006;Barnea & Nottebohm, 1994;Guigueno, MacDougall-Shackleton, & Sherry, 2016). In our view, more precise studies with respect to the different subdivisions of the HF have to be conducted to relate our findings to functional specializations along the anterioreposterior axis that depend on or impact neurogenesis. Here, our findings provide a foundation upon which future studies can be conducted. Based on previous information on connectivity, neurogenesis, volume change and neurochemical data, Smulders (2017) introduced the idea of a functional specialization of the avian HF along the anterioreposterior axis, and compared the avian anterior pole of the HF to the dorsal pole in rodents (i.e., posterior HF in humans), and the avian posterior pole to the ventral pole in rodents (i.e., anterior HF in humans). Therefore, in the following sections we will now complement this idea with our results by comparing our findings to those in mammals.

4.3.
Comparison to the mammalian hippocampal formation e connectivity To date, there is no general consent which subdivisions of the avian HF correspond to their mammalian counterparts, i.e., the DG, CA fields, subiculum and entorhinal cortex, although many similarities between birds and mammals exist at all levels, i.e., anatomical, neurochemical, electrophysiological and functional, many similarities between birds and mammals exist (Abellan et al., 2014;Atoji et al., 2016;Bingman et al., 2017;Herold et al., 2014Herold et al., , 2015Medina et al., 2017). Some authors argued therefore that thinking about the avian HF in terms of the strict organization such as seen in the mammalian hippocampus is likely insufficient to understand the HF of birds (Bingman & Muzio, 2017;Bingman et al., 2017;Herold et al., 2014Herold et al., , 2015. At the time, most researchers agree that the V-region/DM corresponds to DG/CA-fields and DL corresponds to the entorhinal cortex. Given the remarkable cognitive abilities of birds and the conserved role of the hippocampus in spatial memory, this however, makes the hippocampus of birds an exciting structure to explore the relationship between structure and function against the background of divergent evolutionary paths between mammals and birds of more than 300 Million years. Beside the morphological differences of avian and mammalian hippocampal structures, both structure share many similarities in connectivity (our data; . But what about rostro-caudal differences in connectivity? Strong evidence for a functional specialization along the dorsaleventral axis (anterioreposterior in primates) of the hippocampus in mammals originate from lesion or inactivation studies (Anacker & Hen, 2017;Colombo et al., 1998;Fanselow & Dong, 2010;O'Leary and Cryan, 2014;Strange et al., 2014;Thompson et al., 2008;Witter et al., 2017). Consequently, the dorsal (septal) hippocampus, which corresponds to the posterior hippocampus in primates is involved in spatial memory and contextual memory encoding and performs primarily cognitive functions, while the ventral (temporal) hippocampus (anterior in primates), processes emotional (stress) and social behavior. Based on the small literature of functional differences in the avian HF along the anterioreposterior axis, this would fit with the idea that the avian anterior pole of the HF corresponds to the septal pole and the posterior HF to the temporal pole of the mammalian HF (Smulders, 2017). Besides, genetic analysis of several marker genes in mice has shown that the hippocampus exhibits different regional and laminar patterns of molecular domains along the dorso-ventral axis and that it can be further compartmentalized in dorsal, intermediate and ventral regions (Fanselow & Dong, 2010;Thompson et al., 2008) that partially overlap with the classification of these domains originally illustrated by Swanson and Cowan (1977). The molecular differences in hippocampal domains are underpinned by anatomic differences in connectivity, which are more prominent between the dorsal and ventral pole, while the intermediate region exhibits a mixture of connections of both, dorsal and ventral parts that discerns on the one hand but have yet not been fully functionally elucidated.
The dorsal (septal, posterior in primates) hippocampus sends massive projections to the dorsal subiculum, while both hippocampus and dorsal subiculum further project to the retrosplenial and anterior cingulate cortices, which are involved in cognitive processing of visuospatial information and memory. The dorsal subicular complex sends further massive parallel projections through the postcommisural fornix to the mammillary nuclei and anterior thalamic regions that in turn project back to the dorsal hippocampus and retrosplenial cortex. Additionally, dorsal CA1 and CA3 project to the caudal and lesser to the dorsal part of the medial zone of the lateral septal complex, which in turn project to the medial septal complex. The lack of a subdivision that is comparable to the subiculum and a more precise classification of the lateral septum (SL) in birds, makes it hard to compare this to the avian situation, but at least we can say that more anterior SL is connected to the anterior DM and further fibers that travel through the TSM connect to anterior DM and DL. Whether these fibers connect the hippocampus to brain stem nuclei and/or diencephalic nuclei or branch to other septal nuclei was not investigated in this study, but will be of interest for future analysis. Based on prior studies, in birds, HF-septal connectivity is less reciprocal, with a significant higher number of afferents from the HF to the septum that is different compared to mammals (Atoji and Wild, 2004;Casini et al., 1986;Montagnese et al., 2004). SL in birds is further receiving a high quantity of visual inputs from diencephalic nuclei, and has prominent connections with hypothalamic nuclei, as well as dopaminergic, serotonergic, noradrenergic and cholinergic inputs (Atoji and Wild, 2004). In addition, the dorsal subiculum as well as the lateral and medial entorhinal cortex of mammals sent further projections to the rostrolateral nucleus accumbens (NAc) and rostral caudateputamen, which are in turn connected in the so called functional "caudal behavior control column" underlying expression of exploratory behavior, including locomotion, spatial direction and orientation of movements (Swanson & Kalivas, 2000). Direct fiber connections between rostral hippocampal subfields and the NAc were not observed in this study but c o r t e x 1 1 8 ( 2 0 1 9 ) 1 6 5 e1 8 7 between NAc and anterior SL. This finding is in contrast to the tracing study of Atoji et al. (2002) and Atoji and Wild (2004), reporting rostral projections from DL and DM to the medial striatum. However, it is possible that these small number of projections travel under the massive fiber bundles of TSM along the medial wall at the anterior position around A 10.50 (beginning of NAc) and get indistinguishable in our study. Although, no rostro-caudal differences were seen, connectivity to visual areas in the hyperpallium (HA, HL, HD) and the area CDL were confirmed, which in turn connect to the nidopallium caudolaterale (NCL), the analogue of the prefrontal cortex in the pigeon brain and other nidopallial areas (Gü ntü rkü n, 2005). This opens up the possibility for the avian HF to connect to higher order areas like the mammalian HF connects to the prefrontal and cingulate areas. Together, our data supports the idea that the anterior hippocampus in birds is stronger involved in visuo-spatial processing but the specialization is less clear compared to the mammalian situation.
The mammalian ventral CA1 on the other hand, projects to several olfactory areas, and shares, together with the ventral subiculum, a bidirectional connectivity to the amygdala nuclei, preferably those that receive massive olfactory inputs (Anderson, Morris, Amaral, Bliss, & O'Keefe, 2007;Cenquizca and Swanson, 2007;Pitk€ annen et al., 2000;Roberts et al., 2007;Witter & Amaral, 2004). In the pigeon, we confirmed connectivity at medial to caudal levels along the anterioreposterior axis between the HF and the nucleus taeniae of the amygdala (TnA), which is the only amygdala nucleus that receives direct olfactory input from the olfactory bulb according to Patzke, Manns, and Gü ntü rkü n (2011). Connectivity to the posterior nucleus of the amygdala (PoA) did not show any specialization along the anterioreposterior axis. In mammals, ventral CA1/subiculum and the amygdala nuclei also share a bidirectional connectivity with pre-and infralimbic and agranular insular cortices (Roberts et al., 2007;Thierry et al., 2000). It was speculated that the hyperpallium densocellulare (HD) in birds shares similarities with the insular cortex of mammals, despite a completely different connectivity (Medina & Reiner, 2000). However, we observed connections between HD and HF, but like the connectivity with the CPi no anterioreposterior specialization was revealed. In mammals, the ventral CA1 and subiculum are further connected either directly or indirectly to the central and medial amygdala, massively to rostral and ventral parts of the lateral septum and the bed nucleus of the stria terminalis to innervate the periventricular and medial zones of the hypothalamus. Thereby, the ventral hippocampus is integrated in a network controlling neuroendocrine, autonomic and somatic motor activities in motivated behaviors, which explains its involvement in fear or aversive learning tasks or stress related responses. Here, direct connections with BSTL were detected but seem to be widespread, while septal connections are topological to the medial and lateral caudal septum. In contrast to SL, the medial septum (SM) is more specifically connected to individual nuclei of the hypothalamus (Atoji and Wild, 2004). In mammals, additionally the ventral CA1 and subiculum, as well as the medial band of the lateral and medial entorhinal cortex give rise to projections to the caudomedial NAc, which plays an important role in reward processing and feeding behavior (Bagot et al., 2015;Walker, Miles, & Davis, 2009). Nothing comparable was observed in the pigeon brain. However, we detected a direct connection of caudal DM with the nucleus PVM, which is the homologue of a portion of the nucleus paraventricularis of the hypothalamus (PVN) in mammals (Berk, 1987). This connection was described earlier in pigeons as projections from the caudal dorsomedial portion of the HF to the PVM (Bons et al., 1976;Bouill e & Bayl e, 1973). In mammals, the PVN receives afferents from the subiculum, primary the ventral subiculum (Silverman, Hoffmann, & Zimmermann, 1981) and CA1 (Pakhomova, 1981), which would fit with the theory that caudal HF is comparable to the temporal pole of the HF in mammals, i.e., ventral HF in rodents, and anterior HF in primates.
In rodents, commissural fibers of the hippocampus are organized topographically along the dorso-ventral axis, which is also true for the pigeon anterioreposterior axis. In monkeys, commissural fibers were detected only in the rostral part of both, the DG and the CA fields, whereas the subicular subdivisions and the entorhinal cortex show many commissural connections and most of them are topographically organized (Anderson et al., 2007), while in humans, commissural fibers seem to be absent (Wilson et al., 1991).
Further, we detected numerous thin fibers traveling along anterioreposterior axis along the dorsal regions DLd and DMd. In rodents and monkeys, longitudinally fibers were mostly found in CA3 and DG to propagate information along the septo-temporal axis. CA3 longitudinal fibers originate from the pyramidal cells of the pyramidal layer of CA3, while in DG cells of the hilar region are the source of these fibers that are called associative fibers and are very important for building associative spatial relations between cues and environment (Anderson et al., 2007). In addition, the subiculum gives rise to a longitudinal associative projection that extends from the level of origin to the subiculum at the temporal pole. Recently, in humans, a large number of longitudinally traveling fibers have been observed with 3D-PLI in the alveus of CA1 that are perforant path fibers in the stratum lacunosum-moleculare that cross the hippocampal fissure en route to the dentate molecular layer as well as between entorhinal fibers and the subiculum (Zeineh et al., 2017). However, at the moment, any comparison to the mammalian situation would be purely speculative and our findings of the local circuitry need to be further evaluated by precise tracing studies controlling for the origin of these fibers, although they do not seem to complement the unmyelinated fibers of mossy cells, which would be rarely seen in 3D-PLI due to the maximal sensitivity of PLI to the birefringent properties of myelin.
To sum up, our data and data from tracing studies clearly support a specialization along the anterioreposterior HF in birds, with anterior HF resembling the septal pole and posterior HF resembling the temporal pole of mammals. Further, the finding of a direct connection to PVM with caudal DM as well as the emergence of longitudinal fibers in DM makes it likely that DM exhibits subicular/CA1/CA3 characteristics that need to be revisited and presumably leads to an additional parceling of DM into more subdivisions instead of two (DMd, DMv). c o r t e x 1 1 8 ( 2 0 1 9 ) 1 6 5 e1 8 7

4.4.
Comparison to the mammalian hippocampal formation e adult neurogenesis Although deriving from a different neurogenic niche compared to birds, the mammalian hippocampus of diverse species including humans is capable of adult neurogenesis during lifespan, i.e., generating new neurons from neural stem cells in the subgranular zone (but see Bingman et al., 2017;Gage, 2000;Kempermann, 2012;Kempermann et al., 2018;Ming and Song, 2005 for review). In mammals, the highest levels of newborn neurons marked with BrdU and additional markers were detected in DG, while we detected the highest numbers of newborn neurons in the Vl-region. However, in birds, the occurrence of immature neurons appeared not to be exclusively to the V-region but is also prominent in other subdivisions of the HF, including the proposed homologues of CA subfields, subicular and entorhinal cortex, which are relatively sparse labeled with DCX in mammals, although DCX expression is not completely absent (Kremer et al., 2013). This is a major but known species difference, reflecting generally higher levels of plasticity in the avian brain (as well as in fish, reptiles and amphibians) compared to mammals. The same is also true for DCXþ cells that have been detected outside the HF (Ernst & Fris en, 2015;Melleu et al., 2013). In rodents, it has been shown that DCXþ cells have different electrophysiological firing properties compared to mature neurons. This seems to be independent of their brain localization and presumably indicates particular plasticity that is necessary to adapt to environmental changes and to process specific information (Klempin, Kronenberg, Cheung, Kettenmann, & Kempermann, 2011;Spamanato et al., 2012). To our knowledge, no comparable studies in birds exist and thus our findings and those from others should stimulate future electrophysiological studies in birds. However, assuming that the response properties of immature neurons in the avian HF are comparable to those of mammals, birds may have additional mechanisms and faster cellular responses to adopt to environmental changes because they express higher levels of immature neurons.
Equally to the avian HF, adult neurogenesis in mammals varies according to different tasks that have been used to study the role of newborn neurons in relation to function (activity), physiological and pathological stimuli (Christian, Song, & Ming, 2014;Ming & Song, 2011;Snyder, Radik, Wojtowicz, & Cameron, 2009). In mammals, adult neurogenesis in the DG supports memory pattern separation based on space to contextualize (disambiguate) multiple memories (Clelland et al., 2009;Yassa & Stark, 2011), a function that is also supported by the avian HF (Herold et al., 2015). Further, in mammals, adult neurogenesis is involved in inflammatory pain (Zheng et al., 2017), cognitive flexibility (Anacker & Hen, 2017), stress (Tanti et al., 2013) and anxiety/depression like behaviors (O'Leary and Cryan, 2014). In line with anatomical specializations along the dorso-ventral axis, general differences and functional specializations of mammalian ventral and dorsal domains in the context of adult neurogenesis were observed. Generally, DXCþ cells were found at higher levels in the septal (dorsal) pole of the DG of marmosets and mice compared to the temporal (ventral) pole, while prolif-erating Ki67 þ or Ki67þ/PCNAþ cells did not differ (Amrein et al., 2015;Anacker & Hen, 2017). Additionally, DCXþ cells were generally higher in the suprapyramidal blade compared to the infrapyramidal blade and radial glia/progenitor cells were detected at lower levels in the infrapyramidal blade of ventral DG compared to all DG subdivisions (Jinno, 2011). Higher levels of DCXþ cells were also observed in canine dorsal hippocampus (Lowe et al., 2015). In the dorsal hippocampus of rats, DCXþ, BrdUþ, Ki67þ and BrdUþ/DCXþ cells showed higher levels compared to ventral 9 days after BrdU injections (Ho & Wang, 2010). In line with these observations our data complement those findings and show further similarities between the avian and mammalian HF. Interestingly, we also found differences in the number of DCXþ cells and BrdU/NeuNþ newborn neurons between Vl and Vm, with higher levels of adult neurogenesis in Vl compared to Vm. In our study, pigeons generally lived in an enriched environment in their loft and were allowed to fly freely. Further, in rodents that performed complex spatial and contextual stimulation in an enriched environment neurogenesis was increased in dorsal DG (Kempermann, Kuhn, & Gage, 1997;Snyder et al., 2009;Tanti et al., 2012Tanti et al., , 2013, while exposure to chronic stress severely reduces levels of proliferation, differentiation and survival of neurons in the ventral DG but also affects dorsal portions to some levels if simultaneously contextual memory learning is acquired (Anacker & Hen, 2017;Hawley, Morch, Christie, & Leasure, 2012;O'Leary and Cryan, 2014;Snyder et al., 2009;Tanti et al., 2013). Together, data from neurogenesis experiments further point to similarities of septal HC in mammals and anterior HF in birds, and temporal HC and posterior HF. However, one recent study in chickens showed that food deprivation, which was accompanied by increased corticosterone levels results in decreased levels of neuronal progenitors in the rostral pole, while new neurons decreased in both, caudal and rostral poles (Robertson et al., 2017). However, in rodents it was shown that sometimes chronic stress also affects neurogenesis in the septal pole, and seems to be highly dependent on the context in which animals were tested (Ho & Wang, 2010;O'Leary and Cryan, 2014;Snyder et al., 2009). Clearly more functional studies in birds, as well as in mammals are needed that explore the relationship of adult neurogenesis and functional specialization along the septo-temporal axis to a better understanding of these interactions.

Evolutionary connectomics
The hippocampus in all vertebrates emerges during development from the same dorsomedial region of the procencephalic alar plate (Bingman, Salas, & Rodriguez, 2009;Butler & Hodos, 2005;Hevner, 2016;Nomura & Hirata, 2017;Striedter, 2005), except ray-finned fish, where the homologue is at a dorsolateral position, which adopts because the embryonic telencephalon does not evaginate but, instead, everts (Striedter & Northcutt, 2006). In non-mammalian species, the homologue of the hippocampus is called medial or dorsomedial pallium (Hevner, 2016). During more than 300 Million years of separate evolution from a basal hippocampus of amphibians or nonsauropsids, compared to both, avian and mammalian hippocampi have increased in size and complexity and evolved different cytoarchitectonical organizations Striedter, 2016). Further positional changes occurred. While in reptiles and birds, the HF is largely located periventricular in a dorsal region of the medial wall of the telencephalon, in mammals, only the primordium of the hippocampal complex (fimbria, DG, Ammon's horn, subiculum, and entorhinal cortex) is located dorso-medially but comes to lie ventrally within the development of the neocortices. In addition, the DG becomes variably convoluted among mammals (Hevner, 2016;Witter et al., 2017). However, a thin remnant of the HF surrounds the splenium of the corpus callosum that ascends dorsally over the corpus callosum to form the induseum griseum, which comprises remnants of the subiculum and other fields of the HF (Anderson et al., 2007).
In the hippocampus of reptiles, tri-lamination emerged that was accompanied by an enlargement of other pallial regions, while in birds, the HF displays a more cellular structure reflecting a presumptive reduction of lamination that was accompanied by a substantial increase in hippocampus size Hevner, 2016;Striedter, 2016). In mammals, lamination occurs inside-out in the HF, accept in DG were migration occurs outside-in. This process strongly depends on radial glia scaffold and reelin secretion of Cajal-Retzius cells (Frotscher & Seress, 2007). Cajal-Retzius cells are also prominent during development of the avian medial pallium (Nomura, Takahashi, Hara, & Osumi, 2008) but in contrast to the mammalian medial pallium do not migrate to the dorsal preplate, nor do other progenitor cells from the cortical hem or the medial pallium give rise to migrating cells to the dorsal pallium (García-Moreno et al., 2018). However, this seems to be different from lizards (Cabrera-Soccoro et al., 2007), and needs to be further explored along the axis of the cortical hem/medial pallium, but may be one reason of the different structure of mammalian, reptilian and avian hippocampal as well as neocortical regions. In case of the avian hippocampus, which derives from a more dorsomedial position and has largely expanded, this might also explain the different morphology of the most caudaleventral parts of the V-region, where a clear tri-laminar structure is visible (Fig. 14) compared to more rostral parts of the V-region, where the cellular line splits up into two blades or becomes convoluted. Analysis of single-cell transcriptomics of the reptilian medial most pallium, particularly the dorsomedial cortex, has been shown that preliminary precursors of the hippocampal regions, including the DG, cornu ammonis 1 (CA1) and cornu ammonis (CA3) were already present in the ancestor of all amniotes (Tosches et al., 2018), which is also supported by a recently developmental analysis of chick, lizard and mouse hippocampus by Medina et al. (2017). These findings and our recent data, however, make it unlikely that the DG has recently evolved as an add-on structure of the mammalian species suggested earlier by Kempermann (2012). Instead an anlage of the DG in the least common ancestor must have been already existed that was highly morphologically transformed during million years of separate evolution with still conserved functionality.

Conclusion
Birds have extraordinary cognitive abilities that even exceed those of some primates, and they achieve this even without a layered structure of their pallial derivatives but with conserved functional loops between brain structures, including hippocampal connectivity and some similar properties of cells contributing to these circuits. In addition, in many cases the organization principles of hippocampal network connectivity, i.e., anterioreposterior differences seem to be conserved. On the other hand, over more than 300 Million years of separate evolution in adaption to an ecologic niche may have led to further neuronal specializations resembling an alternative route to functionally adopt, reaching the apex of avian evolution in a highly competitive world among other animals including humans. Moreover, high forms of plasticity in the avian brain, validated by DCX expression in various brain regions, may indicate the promotion of further development and evolution of brain structures, and perhaps the ability for rapid adaptation to environmental changes, which may be a selective advantage at some point.

Funding
The 3D-PLI related part of the presented study was supported by the European Union's Horizon 2020 Research and Innovation Program under Grant Agreement No. 720270 (Human Brain Project, SGA1) and No. 785907 (Human Brain Project, SGA2).