Neuronal circuits within the homing pigeon hippocampal formation

The current study aimed to reveal in detail patterns of intrahippocampal connectivity in homing pigeons (Columba livia). In light of recent physiological evidence suggesting differences between dorsomedial and ventrolateral hippocampal regions and a hitherto unknown laminar organization along the transverse axis, we also aimed to gain a higher‐resolution understanding of the proposed pathway segregation. Both in vivo and high‐resolution in vitro tracing techniques were employed and revealed a complex connectivity pattern along the subdivisions of the avian hippocampus. We uncovered connectivity pathways along the transverse axis that started in the dorsolateral hippocampus and continued to the dorsomedial subdivision, from where information was relayed to the triangular region either directly or indirectly via the V‐shaped layers. The often‐reciprocal connectivity along these subdivisions displayed an intriguing topographical arrangement such that two parallel pathways could be discerned along the ventrolateral (deep) and dorsomedial (superficial) aspects of the avian hippocampus. The segregation along the transverse axis was further supported by expression patterns of the glial fibrillary acidic protein and calbindin. Moreover, we found strong expression of Ca2+/calmodulin‐dependent kinase IIα and doublecortin in the lateral but not medial V‐shape layer, indicating a difference between the two V‐shaped layers. Overall, our findings provide an unprecedented, detailed description of avian intrahippocampal pathway connectivity, and confirm the recently proposed segregation of the avian hippocampus along the transverse axis. We also provide further support for the hypothesized homology of the lateral V‐shape layer and the dorsomedial hippocampus with the dentate gyrus and Ammon's horn of mammals, respectively.


INTRODUCTION
Many bird species have outstanding spatial abilities as visible during migration, homing, or food caching (Clayton & Dickinson, 1998;Mouritsen et al., 2016). These skills rely heavily on the integrity of the hippocampus (HC) as damage to this structure is associated with an impairment in numerous spatial functions (Gagliardo et al., 1999(Gagliardo et al., , 2014Pravosudov et al., 2006;Sherry & Vaccarino, 1989). In addition, it has been found that caching bird species have a larger HC compared with those that do not cache (Garamszegi & Eens, 2004;Garamszegi & Lucas, 2005;Lucas et al., 2004). Although the anatomy and physiology of the avian HC have been investigated in several studies (e.g., Atoji & Wild, 2004;Atoji et al., 2016;Ben-Yishay et al., 2021;Erichsen et al., 1991;Herold et al., 2014;Kahn et al., 2003;Payne et al., 2021;Siegel et al., 2002;Szkely & Krebs, 1996), much is still unknown about the intrahippocampal neuronal networks that support these spatial abilities. Neurons in the homing pigeon HC display robust spatial response properties (Hough & Bingman, 2004;Siegel et al., 2005Siegel et al., , 2006Kahn et al., 2008), which interestingly do not completely align with so-called place cells in rats. More recent studies, however, indicate that the avian HC does contain place cells, at least in food-caching titmice and nonhoarding zebra finches (Payne et al., 2021), as well as head direction cells (Apostel & Rose, 2021;Ben-Yishay et al., 2021) and sharp wave ripples (SWRs) (Payne et al., 2021), and thus physiological properties that are hallmarks of hippocampal function in mammals (Hartley et al., 2014;Moser et al., 2008). Moreover, the existence of theta oscillations was reported for awake pigeons (Siegel et al., 2000), but could, however, not be replicated in subsequent studies investigating sleeping pigeons and other avian species (Martinez-Gonzalez et al., 2008;Payne et al., 2021).
Similar to the avian HC, the mammalian HC is also strongly involved in spatial tasks such as navigation and the retention of spatial information (Broadbent et al., 2004;Geva-Sagiv et al., 2015;Morandi-Raikova & Mayer, 2022). However, despite these striking functional and physiological similarities, the avian and mammalian HC show vastly different cytoarchitectures. While the mammalian HC is a three-layered cortex, consisting of a molecular, a cellular, and a polymorphic layer, this clear anatomical organization is not present in birds (Amaral & Witter, 1989;Striedter, 2016). The HC is a phylogenetically old structure that probably was already present in the last common ancestor of birds and mammals. Thus, the difference in anatomical organization at the macroscopic level could be the result of the more than 300 million years of independent evolution (Striedter, 2016). In light of the functional and physiological similarities of the avian and mammalian HC, the question arises whether some network connectivity or cellular features may be evolutionarily conserved, despite the profound macroscopic differences.
In mammals, the HC is typically subdivided along the transverse axis into the dentate gyrus (DG), the CA fields of the cornu ammonis (Ammon's horn), and the subiculum. Both the Ammon's horn and the DG show a three-layered organization where a cellular layer is flanked between two plexiform layers, which in turn have different sublayers (Nieuwenhuys et al., 2007). The Ammon's horn is typically subdivided into the areas CA1, CA2, CA3, and CA4 that can be differentiated by the density and size of pyramidal neurons in the cellular layer. The subiculum is the area where the three-layered HC transitions into the six-layered entorhinal cortex (EC) in the parahippocampal region. The EC can be distinguished from other neocortical structures as it exhibits a fiber layer with low cell numbers (lamina dissecans) instead of the typical layer 4 (Canto et al., 2008). The EC is also the main input structure into the HC and forms a link between the HC and the remaining neocortex (Kerr et al., 2007). Within the HC, the main information flow follows the trisynaptic pathway, in which the DG receives input from the EC through the perforant path (synapse 1), from where information is forwarded to the CA3 area via mossy fibers (synapse 2). Finally, CA3 provides input to CA1 via the Schaffer collaterals (synapse 3) (Amaral & Witter, 1995).
In a similar manner, the avian HC can be subdivided into subregions along the transverse axis based on connectivity as well as cytoand chemoarchitecture (Atoji & Wild, 2004Atoji et al., 2016;Casini et al., 1986;Erichsen et al., 1991;Herold et al., 2014;Kahn et al., 2003). To date, various studies have proposed different subdivisions, but most agree that the avian hippocampal formation (HF) can be subdivided into a dorsolateral (DL), a dorsomedial (DM), and a ventral V-shaped area ( Figure 1) (Atoji & Wild, 2006;Atoji et al., 2016;Erichsen et al., 1991;Gupta et al., 2012;Herold et al., 2014;Kahn et al., 2003;Siegel et al., 2002). DL can be distinguished from the laterally adjacent area corticoidea dorsolateralis (CDL) based on decreasing thickness and from the medially adjacent DM based on a sulcus on the ventricular side of the HF, which extends along the anterior posterior axis of the HF (Atoji & Wild, 2004). DL can be furthermore subdivided into a dorsal and ventral component based on connectivity and receptor architecture (Atoji & Wild, 2004;Herold et al., 2014).
For DM, different subdivisions have been proposed. While DM has been subdivided into a dorsal (DMd) and ventral (DMv) component based on hippocampal receptor densities (Herold et al., 2014), other studies further distinguished three small regions within the dorsal DM based on cell morphology and cytoarchitecture (Atoji & Wild, 2004).
Moreover, the V-shaped area consists of a triangular region (Tr) that is surrounded by a thin V-shaped layer that is clearly visible due to its higher neuron density (Atoji & Wild, 2004;Erichsen et al., 1991;Krebs et al., 1991). This V-shaped layer can be further subdivided into a lateral (Vl) and a medial (Vm) component that both are coated by cell poor fiber layers. These fiber layers have been referred to as the paraventricular (pfz) and medial fiber zone (mfz), respectively (Atoji & Wild, 2004. It needs to be noted that these V-shaped layers differ between birds and are not clearly seen, for example, in chickens, ostriches, and parrots (Striedter, 2016). So far, the consensus is that DL, DM, and the V-shaped area are reciprocally connected with each other. The main extrahippocampal inputs seem to arrive from the hyperpallium, CDL, and the thalamus, providing the avian HF with sensory information (Atoji & Wild, 2004Casini et al., 1986).
Moreover, the main extrahippocampal output of the HF is sent to the F I G U R E 1 Nissl-stained section indicating the subdivisional organization of the homing pigeon hippocampal formation (HF). Nissl-stained coronal section of the HF showing subdivisional boundaries according to Herold et al. (2014). The HF of pigeons can be subdivided into seven regions: the V-complex, which contains a lateral (Vl) and medial (Vm) cell layer, and an inner triangular area (Tr). The V-shaped cell layers are flanked by two fiber zones (pfz and mfz). Moreover, the dorsomedial region can be subdivided into ventral (DMv) and dorsal (DMd) subdivisions. Likewise, the dorsolateral region contains a ventral (DLv) and dorsal (DLd) component. Scale bar represents 500 µm.
Although several studies have already investigated the intrinsic and extrahippocampal connectivity of the three major subdivisions (Atoji & Wild, 2004;Hough et al., 2002;Kahn et al., 2003), many open questions remain. Up to now, the local connectivity pattern of the avian HF was only studied using in vivo tracer injections. Given the small size of the HF and its complex three-dimensional structure, it is extremely difficult to obtain a comprehensive description of the projections between the HF subcomponents. We therefore employed both in vitro and in vivo tracing procedures to obtain a more complete understanding of the intrahippocampal connections of the pigeon hippocampal complex.
These techniques are complementary as in vivo tracing is better suited to study long-range connectivity, while in vitro tracing offers greater spatial resolution for focal injections into small subareas. Moreover, we investigated the connectivity along the transverse axis focusing especially on a possible anatomical segregation between dorsomedial (superficial) and ventrolateral (deep) HF, which has recently been suggested based on physiological (Payne et al., 2021) and neurochemical (Medina et al., 2017;Redies et al., 2001) evidence. As this has not been identified in previous tracing experiments, our study aimed to gain a higher resolution understanding of HF intrinsic connectivity with respect to the potential segregation of subdivisions along the transverse axis.

Surgical procedure and in vivo tracer injection
In vivo tracing experiments required a surgical procedure for tracer injection. Therefore, the pigeons were anesthetized using a mixture of ketamine (Ketavet 100 mg/mL, Zoetis GmbH, 52.5 mg per kg body weight) and xylazine (Rompun 20 mg/mL, Bayer Vital GmbH, 4.5 mg per kg body weight) together with isoflurane (1.5-2.5% in pure oxygen, Forane 100%, Abbott GmbH & Co. KG, Wiesbaden, Germany).
Once the depth of anesthesia was sufficient, the birds were placed in a stereotactic apparatus and their heads were fixed. Feathers were removed from the scalp and an incision was made to expose the skull. Craniotomies above the injection targets were performed to gain access to the brain. After the meninges were opened, a glass micropipette with an inner diameter of 15-20 µm filled with Choleratoxin subunit B (CTB, Sigma, Germany) was lowered into the brain to the intended injection targets. CTB is a tracer with strong retrograde and weak anterograde transport (Köbbert et al., 2000). Stereotactic coordinates were determined using the stereotactic atlas of the pigeon brain (Karten & Hodos, 1967). A volume of 200-400 nL of tracer was pressure injected using a Nanoliterinjector 2000 (WPI, Sarasota, USA).
Injections were only performed in one hemisphere per animal. After 2 days, the pigeons were perfused.

In vitro tracing
For the in vitro tracings, the pigeons were anesthetized with equithesin (0.45 mL/100 g body weight) and subsequently decapitated.
After a rapid dissection of the brains, they were placed in ice-cooled

Perfusion and brain sectioning
The animals of the in vivo tracing experiment were anesthetized using equithesin (0.45 mL/100 g body weight) prior to the perfusion. They were then transcardially perfused with 0.9% NaCl, after which 4 • C cold paraformaldehyde (4% in 0.12 M PB, pH 7.4) was applied for tissue fixation. After sufficient fixation, the brains were extracted from the skull and further fixated in a postfix solution (4% paraformaldehyde and 30% sucrose in 0.12 M PB) at 4 • C for 2 h. Afterwards, brains were transferred to a sucrose solution for cryoprotection (30% sucrose in PBS, pH 7.4) until sinking to the bottom. Subsequently, they were embedded in gelatin (15% gelatin and 30% sucrose in PBS), which was again fixated in postfix solution for 24 h. Using a freezing microtome (Leica, Germany), the brains were sectioned coronally with 30-40 µm thickness before further staining procedures were performed on every fifth slice.  were also counterstained with Nissl.
Primary antibody specificity was ensured by comparing the resulting staining patterns of our study with previous reports in pigeons using the same DCX (sc8066, Santa Cruz) and calbindin (CB-38a, Swant) antibodies (Mehlhorn et al., 2022), as well as GFAP antibody (AB_2532994, Invitrogen) (Rook et al., 2021). All fluorescence stainings were performed as follows: brain sections were first washed in PBS (3 × 10 min), after which they were incubated in 10% normal horse serum in PBST for half an hour. Following that, the slices were transferred to the primary antibody solution (1:1000 in PBST), where they stayed for 72 h at 4 • C. Next, they were rinsed again (3 × 10 min in PBS) before being incu-

Microscopy and analysis
The slices were analyzed with a ZEISS Imager.M1 AXIO microscope equipped with an AxioCam MRm ZEISS 60N-C 2/3ʺ 0.63× camera. The

In vivo tracing experiments in the HF
We first performed in vivo tracings to investigate longer distance connectivity within the HF. For our in vivo injections and the analysis of the data, we relied on the hippocampal subdivisional scheme described Besides that, we saw long-range fibers leaving from both DMvd and DMvv and running along the dorsal and ventral aspect of the CDL, respectively. Furthermore, we took a closer look at GFAP expression, which is a marker for astrocytes and radial glia within the brain. The GFAP staining labeled a thin cell layer along the pial surface of the HF as well as cells located in the periventricular area of DMvv and DL characterized by dense radial processes ( Figure 6). Furthermore, a band of astrocytes was usually observed reaching from DMv to Tr (Figure 6a  our connectivity analysis. In addition, some of the markers (e.g., DCX, GFAP) indicate an even more subtle, possibly layered, subdifferentiation of the HF.

In vitro injections
Our in vivo tracing data and marker expression patterns demonstrated differences between dorsomedial and ventrolateral hippocampal subdivisions and suggested that DMv needs to be further subdivided into a dorsal and ventral component. In order to investigate this segregation more thoroughly, in vitro tracings were performed that enable to target small subdivisions or small areas of the brain with greater precision. Therefore, brain slices were cut using a vibratome in continually oxygenated, ice-cooled sucrose-substituted Krebs solution. After that, biocytin crystals were applied to all hippocampal subdivisions in separate slices, which then were incubated in ACSF for 4 h.

In vitro injections into DL
We first targeted DLd and DLv with biocytin crystals (Figure 9). DLd injections resulted in many fibers that ran through DMvd and DMd and continued along the mfz as well as the medial Tr (Figure 9a

In vitro injections into DM
In vitro experiments with injections into DM showed a topographical pattern of projections similar to that found in the in vivo experiments.
When crystals were placed in DMv, the resulting projection patterns also suggested a separation into a dorsal (DMvd) and a ventral (

In vitro injections into the V-shaped area
In the next step, we analyzed in vitro injections within the V-shaped area. Tr injections resulted in strong projections to DMv, especially DMvv, although not to DMd (Figure 11a,b). Interestingly, Tr neurons gave rise to long dendritic trees expanding across Vm and extending up to the mfz where they crossed with orthogonally running fibers of this fiber zone (Figure 11c). Furthermore, many fibers from Tr injection site left the HF in a ventral direction (Figure 11a).
Vm injections revealed projections to DMd, Vl, and Tr ( Figure 11d-g). On the one hand, labeled fibers were observed to run dorsally along the mfz to innervate the DMd (Figure 11d,e). On the other hand, some fibers were also found to traverse through the Tr, many reaching the Vl and the pfz (Figure 11d,f,g). A lot of these fibers contained varicosities, indicating that they might contact neurons in Tr on their way to Vl. Furthermore, within the medial Tr, some retrogradely labeled neurons could be seen close to Vm (Figure 11f).
It is possible that these were labeled due to their dendrites extending into Vm or due to tracer spread. In one of these cases, the axon of such a neuron could be followed toward the lateral Tr; however, the termination area of this axon could not be determined precisely.
Nevertheless, the observed labeling pattern opens the possibility that there might be a form of intrinsic circuitry that connects the lateral and medial parts of Tr.
In a similar fashion, Tr was the main target of Vl projections as a large number of fibers could be found in this subregion after biocytin placement into Vl (Figure 11h-j). Some fibers were also observed to be headed toward DMv. A fiber bunch was also seen in the pfz that left the HF ventral to the injection site.

DISCUSSION
The aim of the present study was to gain a deeper understanding of the connectivity pattern within the HF of homing pigeons. To this end, tracing that offers a high spatial specificity, such as the use of small biocytin crystals that can be applied to brain slices that are artificially kept alive. By using biocytin, we were therefore able to investigate, for example, the connections of the small V-shaped layers, whose connec-tivity has only been previously inferred from retrograde labeling (Kahn et al., 2003).

Intrahippocampal connectivity
Overall, we found that the avian HF comprises several reciprocal connections among subdivisions and could confirm connections that have already been established in previous tracing experiments (Atoji & Wild, 2004;Kahn et al., 2003) and with resting-state fMRI (Behroozi et al., 2017). For example, we found that DM is reciprocally connected with Tr, the V-shaped layer, and DL. Likewise, we saw that DL has reciprocal connections with DM and Tr.
Using in vitro tracings, we could reveal in much more detail the patterns of connectivity among smaller subdivisions and found that at least two parallel pathways can be distinguished within the avian HF: a more dorsomedial (superficial) and a more ventrolateral (deep) pathway. In the superficial pathway, DLd sends efferents to DMd and medial Tr, and to a lesser extent to Vm. DMd in turn projects to Vm and F I G U R E 1 0 (Continued) injections. Fibers could also be seen throughout Vm and mfz. (l) DMd injections led to anterograde labeling in DMvd. Scale bars represent 500 µm in panels (a, d, g, j); 100 µm in panels (e, i, k, l); 50 µm in panels (b, f, h); and 20 µm in panel (c).
medial Tr, as well as back to DLd. DMd also has reciprocal connections to DMvd. Finally, Vm sends projections to Vl and Tr.
In the deep pathway, DLv mainly projects to DMvv, but it is also weakly connected to Vl. From DMvv, efferents are sent to Vl and lateral Tr. Vl sends projections back to DMvv and Tr. Lastly, DMd and DMvv receive projections from the medial and lateral portions of Tr, respectively. The superficial and the deep pathways are interconnected in DL and DM, as well as in the V-shaped area through connections between lateral and medial Tr ( Figure 12).
Overall, our findings are consistent with a feedforward circuit of the avian HF that has been proposed in earlier studies (Hough et al., 2002;Kahn et al., 2003). These studies suggested that information flow starts in DL, from where information is relayed to DM. DM then projects to Vm, from where information reaches the Vl, which finally projects back to DM, which in turn projects out of the HF (Hough et al., 2002;Kahn et al., 2003). In the current study, we found an overall similar projection pattern. However, importantly, our higher spatial-anatomical resolu-  (Atoji & Wild, 2004;Kahn et al., 2003). For example, Kahn et al. (2003) report that CTB injections into the part of DM that we consider DMvv stained neurons within Vl, while injections into the medial DM, which corresponds to our DMd, led to anterogradely labeled fibers as well as retrogradely labeled neurons in Vm. Moreover, Wild (2004, 2005) found retrogradely labeled neurons within Vl after lateral DM injections, and labeled Vm neurons following medial DM injections. It is possible that the mediolateral topography described by Atoji and Wild (2004) is partly comparable to our parallel dorsomedialventrolateral topography because of the curvature of the hippocampal transverse axis. However, it is also possible that a mediolateral topography as described by Atoji and Wild (2004) is a distinctive property of DMv as some studies found medial/lateral differences in DMv with respect to its connectivity with CDL (Atoji & Wild, 2005), hyperpallium densocellulare, and cortex piriformis (Kahn et al., 2003). In this regard, it is interesting to point out that cadherins and calcium-binding proteins revealed radial hippocampal subdivisions in chickens (Redies et al., 2001;Suárez et al., 2006) that might fit well with the idea of mediolateral differences within DM. The intermediate subdivision of the parahippocampal area (APHi), which might approximately correspond to the lateral DM in Atoji and Wild (2004), is distinguishable from that of the medial parahippocampal area (APHm) that could, with some caveats, roughly correspond to the medial DM. Moreover, the APHi displays further lateral-medial subdifferentiation (Suárez et al., 2006).
The DMvd in our study might include the superficial part of the APHi, or at least its medial aspect (APHim). By contrast, the DMvv might correspond to the deep part of the (medial) APHi together with aspects of the deep part of APHm. Thus, further studies are needed to determine if and how these different topographies relate to each other, preferably leading to a unified terminology.
Moreover, we found that information from both V-shaped layers is further conveyed to Tr and that focal injections into Vm gave rise to fibers targeting Vl. The Vm to Vl connection has previously been suggested based on electrophysiological (Hough et al., 2002) and anatomical data (Kahn et al., 2003). However, previous in vivo retrograde and anterograde tracer injections into Vl and Vm, respectively, contained tracer spread in Tr and DM; therefore, the conclusions about this connection remained putative (Kahn et al., 2003). Nevertheless, we now confirmed this projection with focal in vitro anterograde tracings.
It is noteworthy that the intrahippocampal circuitry also includes connectivity between the left and right HF. Particularly interesting is that contralateral connections display a topographic pattern along the rostrocaudal axis (Atoji et al., 2002). Contralaterally projecting zones possibly include parts of DL and DM, as well as Vl and Vm (Atoji et al., 2002;Casini et al., 1986;Kahn et al., 2003). Additionally, it seems that some of the contralateral connections resemble the ipsilateral patterns of connectivity; for example, tracer injection into ventral regions including parts of DMvv and DLv resulted in retrogradely labeled neurons in the contralateral Vl (Kahn et al., 2003). On the other hand, tracer injection involving mainly Vl, lateral Tr, and possibly parts of DMvv revealed a retrogradely labeled group of cells within the contralateral DMd (Kahn et al., 2003), and anterograde and retrograde tracings suggest that Vm projects to the contralateral Vl (Atoji et al., 2002;Kahn et al., 2003). In vitro and in vivo data presented on a section of the hippocampal formation. The projections within the HF are topographically organized along the transverse axis, and the dorsomedial and ventrolateral pathways are depicted in blue and green, respectively. Denser projections that appeared to be stronger, suggesting a core hippocampal circuit, are depicted with thicker lines. In brief, in the superficial pathway (blue), DLd sends efferents to DMd and medial Tr (Trm), and to a lesser extent Vm. DMd in turn projects to Vm and Trm, as well as sends a reciprocal projection back to DLd. DMd also has reciprocal connections to DMvd. Finally, Vm sends information to Vl and Tr. In the deep pathway (green), DLv mainly projects to DMvv but also weakly to Vl. From DMvv, efferents are sent to Vl and lateral Tr (Trl). Vl sends projections back to DMvv and Tr. Lastly, DMd and DMvv receive projections from the medial and lateral portions of Tr, respectively. The dorsal and the ventral pathways are interconnected in DL, in DM, and in the V-shaped area through connections between lateral and medial Tr.

The distinction between the dorsomedial and ventrolateral pathway
Our findings indicate that the avian HF can be subdivided into two streams along its transverse axis. This finding is in line with recent electrophysiological, neurochemical, and functional data of the avian HF.
For example, Payne et al. (2021) investigated SWRs across the hippocampal transverse plane in tufted titmice (Baeolophus bicolor). They found that the sharp wave component inverted from a positive to a negative polarity between dorsal and ventral recording sites. Moreover, they reported that the current source density was structured along the radial axis as they found a current source dorsal to a sink, concluding that SWRs exhibit a laminar organization in the HF of the titmouse.
In the mammalian HC, SWRs also show a laminar distribution of sinks and sources (Ylinen et al., 1995) and are typically thought to be the result of a dense pyramidal cell layer with parallel dendrites, which give rise to large local field potential fluctuations through the summation of smaller currents (Buzsáki, 2015). More specifically, SWRs occur when clusters of pyramidal cells are spontaneously active leading to a summation of activity within the CA3 network, driving reciprocally connected interneurons to create the phase-locked, ripple-frequency spiking (Schlingloff et al., 2014). Payne et al. (2021) concluded that avian SWR might result from a more subtle laminar organization of cells, as no clear layering is visible in avian HF, or from other characteristics of hippocampal organization along the radial axis, such as differ-ences in synaptic input (Valero et al., 2015), morphology (Montagnese et al., 1993), or intrinsic cell properties (Montagnese et al., 1996).
The pattern of HF connections found in our study could provide a network that is able to produce similar SWRs. We found that avian HF contains several reciprocal circuits as the two pathways along the transverse axis are characterized by strong feedback projections that occur at many regional subdivisions. Moreover, the two pathways are connected to each other at many subdivisional areas via direct projections, as well as via neurons that are aligned orthogonally to the transverse axis and extend their dendrites across the HF. These dendrites reach the fiber bundles that course through the HF dorsally and ventrally, potentially enabling the integration of their inputs. Our tracing data are in line with early studies characterizing hippocampal neurons in chicks and pigeons  showing that hippocampal projection neurons give rise to recurrent collaterals that extend in the dorsoventral and mediolateral plane, possibly contacting other excitatory and inhibitory hippocampal neurons. Overall, the data from the current study and other data suggest that the avian HF contains a complex intrinsic network suitable to sustain and modulate incoming neural activity by a recurrent neural network. This proposed recurrent network could be a potential oscillation-producing system capable of generating SWRs.
The overall dorsomedial/ventrolateral separation observed in our tracings stands seemingly in contrast to the nonlaminar appearance of the avian HF. More recent evidence suggests, however, that the avian HF displays a laminar organization (Abellán et al., 2014;Fujita et al., 2022;Redies et al., 2001). For instance, a subset of serotonergic receptors is expressed in a laminar fashion in 1-day-old chickens (Fujita et al., 2022). Similarly, laminar expression patterns of cadherins were observed in embryonic chick brain tissue (Redies et al., 2001).
These layers are organized orthogonal to the orientation of radial glia (Abellán et al., 2014) and can be visualized with staining against calcium-binding proteins, neuronal nitric oxide synthase, and GABA in adult chicken (Suárez et al., 2006). Based on this finding, it was suggested that the ventral and dorsal subdivisions of the avian DL might correspond to deep versus superficial hippocampal layers in mammals, respectively (Medina et al., 2017). Our data confirm the presence of a laminar arrangement, which is not restricted to DL but continues along the whole transverse axis as we find topographic projections between dorsomedial (superficial) and ventrolateral (deep) subdivisions of the avian HF.
The distinction between dorsomedial and ventrolateral HF observed in our study might also reflect functional differences. For example, some studies suggest that the dorsomedial HF might play a special role in spatial memory as remembering locations of cached food, based on spatial cues, preferentially activated the superficial, but not the deep region of the HF (Mayer & Bischof, 2012;Mayer et al., 2010). In contrast, the deep HF of female zebra finches was preferentially activated when birds listened to the directed song of males (Bailey et al., 2002). Overall, these studies support the hypothesis of a functional segregation along the transverse axis that would be consistent with our anatomical findings. Future, functional studies of the avian HF should focus on investigating superficial/deep differences in hippocampal function.

The segregation of DMv into DMvd and DMvv
Another main finding of our study that is consistent with the overall segregation of HF into numerous subdivisions aligned along two parallel processing streams is that DMv can be further subdivided into a dorsal (DMvd) and a ventral (DMvv) region. This distinction was not only seen in projection patterns of DMvv and DMvd, but also in the expression of some of our immunohistochemical markers.  (Montagnese et al., 1993). Montagnese et al. (1993) found larger neurons in the dorsomedial HF compared to other hippocampal subdivisions in all four species. Moreover, the calbindin-positive neurons within the dorsomedial HF were significantly larger in the food-storing species compared to the nonstorers, suggesting a specialized role of the dorsomedial HF in spatial tasks. Moreover, the distinction between DMvv and DMvd could be observed in our GFAP staining. In general, GFAP is widely used as a marker for astrocytes, but it is also used to stain radial glia (Sievers et al., 1992). In our case, we found that the GFAP staining labeled a thin cell layer along the pial surface of the HF as well as cells located in the periventricular area of DMvv and DL, which were characterized by dense radial processes that seemed weaker in DMvd and DMd. Furthermore, a band of astrocytes was usually observed reaching from DMv to Tr. Thus, the GFAP staining might support the segregation of DMv into dorsal and ventral subdivisions. It is generally assumed that the main function of radial glia is to act as scaffolding during neuronal migration. However, in the DG of adult mammals, radial glia also acts as a precursor for differentiating neurons and glia cells (Casper & McCarthy, 2006;Garcia et al., 2004). Astrocytes within the mammalian HC are associated with memory processes (Adamsky et al., 2018), encode reward location (Doron et al., 2022), and modulate hippocampal-cortical communication during learning (Kol et al., 2020).
More specifically, astrocytes in CA1 encode spatial information that complements place cell spatial encoding (Curreli et al., 2021). If astrocytes were to serve similar functions in birds, our observed pattern could indicate a functional segregation within the avian HF.

Comparison to the mammalian HF
The question of homology between the avian and mammalian hippocampal subdivisions continues to be intensely debated (Atoji & Wild, 2004;Atoji et al., 2016;Erichsen et al., 1991;Kahn et al., 2003;Siegel et al., 2002;Szkely & Krebs, 1996). Some authors share the view that, while DM shares characteristics with the Ammon's horn, the V-shaped region could correspond to the DG (Atoji & Wild, 2004;Atoji et al., 2016;Herold et al., 2014). This comparison is based on the regions of intrinsic and extrinsic connectivity, as well as the receptor architecture of these areas (Atoji & Wild, 2004;Atoji et al., 2016;Herold et al., 2014). In the mammalian DG, granule cells project to CA3 via mossy fibers, but do not project outside the HF. In contrast, pyramidal cells in Ammon's horn project to regions outside of the HF such as to the septum (Spruston, 2008;Swanson & Cowan, 1977). In the avian HF, DM projects to the lateral septum and this projection is mainly glutamatergic, comparable to the projection neurons of the Ammon's horn (Atoji et al., 2016). In contrast, projections in birds arising from the V-shaped area are rather intrinsic to the HF and only very few neurons project to the septum (Atoji & Wild, 2004. Although we did not discuss the septum in our tracing experiments described above, we did observe a projection from DMd to the septum in our anterograde in vitro tracings.
Moreover, in vitro injections into Vl as well as in vivo tracings that predominantly targeted the V-shaped area did not result in anterograde labeling within the septum, but rather in other hippocampal subdivisions. This confirms that projections arising from the Vl are intrinsic to the HF.
Because of their intrinsic connectivity, the V-shaped layers have already previously been compared to the mammalian DG (Atoji & Wild, 2004Atoji et al., 2016). To further investigate a possible similarity, and by inference potential homology, we looked at adult neurogenesis, a prominent feature of the mammalian DG. Neurogenesis has also already been described in several bird species including pigeons (Herold et al., 2019;Melleu et al., 2013), and can be investigated using stainings against DCX (Balthazart & Ball, 2014;Kremer et al., 2013). Our qualitative analysis of DCX expression revealed that the expression was strongest in the ventral areas such as DMvv and Vl and that the dorsomedial regions including DMvd, DMd, and Vm displayed hardly any DCX. This is in line with previous studies that quantified DCX in the avian HF and found the expression to be strongest in Vl followed by DLv and DLd, while the expression was weakest in DMd and Vm (Herold et al., 2019;Melleu et al., 2013).
Similar to the DCX expression, we found GFAP-positive radial fibers mainly in DMvv and Vl. The overlapping occurrence of DCX expression and radial glia is in line with the knowledge that radial glia, in guiding neuronal migration, is involved in adult neurogenesis (Mori et al., 2005;Zupanc & Clint, 2003). These results suggest similarities of Vl and the DG, as in mammals DCX expression is strongest in DG. It is noteworthy that DCX expression in the avian brain can also be found outside of the HF (Melleu et al., 2013), which is regarded as a difference between avian and mammalian brains (Herold et al., 2014;Melleu et al., 2013). The extent of this difference, however, needs further investigation as adult neurogenesis in mammals has also been documented for other brain structures including the hypothalamus, striatum, substantia nigra, and amygdala (Jurkowski et al., 2020). Although the presence of adult neurogenesis has been established in avian HF (Barnea & Pravosudov, 2011;Hall et al., 2014;Herold et al., 2019;Hoshooley et al., 2007;Meskenaite et al., 2016;Nikolakopoulou et al., 2006), we did not further characterize the DCX-positive neurons in our study and thus cannot exclude the possibility that they represent DCX-positive neurons that were not recently generated (Bonfanti & Nacher, 2012;La Rosa et al., 2020;Luzzati et al., 2009).
Similar to DCX expression, we found that CaMKIIα expression was strongest in DMvv and Vl, although it was generally seen throughout the whole HF, albeit to a lesser extent. The CaMKIIα staining could clearly differentiate between the V-shaped layers as Vl expressed more CaMKIIα than Vm. In the mammalian HC, CaMKIIα expression is generally strong and heterogeneous but highest levels can be found within DG (Wang et al., 2013). Functionally, CaMKIIα is involved in LTP induction and in the efficiency of synaptic transmission (Wang et al., 2013).
Moreover, the depletion of CaMKIIα interferes with the integration of adult-born granule cells into the circuitry of the DG (Arruda-Carvalho et al., 2014). Thus, also our CaMKIIα data underline the distinction of Vl and Vm and provide support for the similarity of Vl, rather than Vm, to DG. This comparison has also been drawn by other studies based on mRNA expression and receptor density profiles (Atoji et al., 2016;Herold et al., 2014).
For the CDL and DL, the picture is less clear as both have already been compared to the EC (Atoji & Wild, 2004Abellán et al., 2014;Atoji et al., 2016;Herold et al., 2014Herold et al., , 2015. However, more detailed studies on CDL connectivity, which did not only investigate its connectivity to HF but also connections with other structures, came to the conclusion that CDL might be more similar to the cingulate cortex (Atoji & Wild, 2005). Considering these findings and the results from a receptor density study (Herold et al., 2014), we consider it more likely that the avian DL is comparable to the mammalian EC. Nevertheless, it is important to note that these subdivision-to-subdivision comparisons are confounded by the limited and mainly anatomical data in birds. More neurophysiological and functional studies of avian HF subdivisional contributions are needed. By employing newer methods and techniques that are now also available for birds, such as optogenetics (Roberts et al., 2012;Rook et al., 2021), fMRI in actively behaving animals (Behroozi et al., 2020), calcium imaging (Roberts et al., 2017), and wireless electrophysiology (Agarwal et al., 2021), the functional and neurophysiological properties of the avian HF can be further disentangled.
In summary, our study investigated the intrinsic connectivity of the avian HF at a higher spatial resolution than previous studies. We discovered a large number of previously undescribed intrahippocampal subdivisional connections as well as anatomically dissociable dorsomedial/ventrolateral processing streams along the transverse axis of the avian HF. This latter finding is consistent with reported dorsomedial/ventrolateral neurophysiological and functional differences in tufted titmice (Payne et al., 2021). We hope that our study will motivate future research, which will offer a more detailed understanding of the neurophysiological and behavioral functional contributions of the different HF anatomical subdivisions. Open access funding enabled and organized by Projekt DEAL.

CONFLICT OF INTEREST STATEMENT
The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.

PEER REVIEW
The peer review history for this article is available at https://publons. com/publon/10.1002/cne.25462.