For both ADCYAP1 analysis and astrocyte morphometry, we collected adult individuals with mist-nets during the wintering period in the mangroves of the Amazon River estuary. All individuals were collected between 2012 and 2017, in the northeast of Pará state, at the municipality of Bragança, Pará, Brazil, on Canela Isle (0° 47′33.52″ S 46° 43′ 8.55″ W), Lombo Grande Isle (0 ° 47′33.52″ S 46 ° 43′8.55″ W), Praia do Pilão (0 ° 47′46.08″ S 46 ° 40′29.64″ W), Baiacu Beach (0 ° 47′32.55″ S 46 ° 46′52.05″ W), Quatipuru Mirim Beach (0 ° 46′35.61″ S 46 ° 52′57.66″ W) and Otelina Isle (0 ° 45′42.57″ S 46 ° 55′ 51.86″ W). Captured wintering birds included C. semipalmatus, C. pusilla, and A. macularius. The non-migratory C. collaris, a resident of South America, was also captured and compared with the long-distance migratory species (Del Hoyo et al., 1992; Rodrigues, 2000, 2006). Immediately after capture, biometric data were obtained from all individuals.

Birds were captured in compliance with license No. 44551-2 of the Chico Mendes Institute for Biodiversity Conservation (ICMBio), minimizing discomfort during handling as much as possible.

Actitis macularius presents a pattern of migration on broad fronts with many stopover sites and a broad dispersion on spring and summer grounds, whereas C. pusilla has a narrow band of migration and moderate dispersion on spring and summer sites (Reed and Oring, 1993; Billerman et al., 2020). Migration timing for these species is very similar but greater migratory distance is performed by C. pusilla compared to A. macularius (Skagen et al., 1999). C. semipalmatus flies towards coastal areas in the southern United States, the Caribbean and much of South America. Like A. macularius, C. semipalmatus travels long distances with flights interrupted for resting and feeding (Campos et al., 2008; Serrano, 2010; Nóbrega et al., 2015).

Microsatellites have been used in ecological and conservation studies since the 1990s (Moura et al., 2017). The comparison between the length of the alleles and the nucleotide composition of the base of the sequence makes it possible to identify variation in the microsatellites (Ellegren, 2004). Allele size variation is identified by polymerase chain reaction (PCR) amplification (Primmer et al., 1996). In the present report we carried out DNA isolation and purification from stored blood samples of 53 individuals with distinct migratory behaviors: A. macularius (n = 12), C. pusilla (n = 14), C. semipalmatus (n = 13), and C. collaris (n = 14). After blood collection (less than 100 μL), all captured animals were released back into the wild, except for five individuals of each species used in morphometric studies. We followed the recommendations of the DNA extraction protocol of Wizard^® Genomic Purification Kit (PROMEGA).

For PCR, specific primers were used that flank microsatellite repetitions of the ADCYAP1 locus (Steinmeyer et al., 2009). To amplify the ADCYAP1 loci, the M13 tail technique proposed by Schuelke (2000) was used. The principle of the technique is to use primer pairs that flank repetitive DNA sequences to amplify samples of genomic DNA and to examine the size of the amplified alleles on a sequencing device (Boutin et al., 1997; Boutin-Ganache et al., 2001). The PCR reaction was performed using a total volume of 13 μL containing 5 ng of DNA, 10 μL Buffer, 1.5 mM MgCl₂, 1.2 mM dNTP, 8 pM M13 probe and reverse primer, 0.5 pM of primer forward and 1 unit (U) of Taq DNA polymerase. To find the best hybridization temperature for the studied species, PCRs were performed with temperature gradients between 50 and 60°C. The PCR reaction consisted of an initial denaturation of 94°C for 5min, followed by 30 cycles of 94°C for 30s, 51°C for 45s and 72°C for 45s, followed later by 8 cycles of 94°C for 30s, 53°C for 45s and 72°C for 45s, with a final extension of 72°C for 10min [for more details on PCR see Schuelke (2000)]. We used the standard microsatellite genotyping method in the ABI 3500XL fragment analyzer (GeneMapper, Applied Biosystems). Subsequently, peak patterns were analyzed by the Fragment Profiler 1.2 program (Amersham Biosciences) and organized in Microsoft Excel 2019.

To detect and identify genotyping errors resulting from null alleles, allele drop out and stuttering related to ADCYAP1 locus, we used the software MICROCHECKER (Barros et al., 2020). We used ARLEQUIN v3.5 (Excoffier et al., 2007) to measure genetic diversity in terms of number of alleles per locus (A), observed (HO), and expected (HE) heterozygosity (Nei, 1978). We tested for Hardy–Weinberg Equilibrium (Mayo, 2008; Meirmans et al., 2018) and analyzed population differentiation at gene ADCYAP1 through FST and RST statistics (Meirmans, 2020) with a significance level α < 0.05.

Five individuals of each species were used for morphometric astrocyte studies. All birds were anesthetized with Isoflurane (Olkowski and Classen, 1998), euthanized with an anesthetic overdose, and perfused transcardially with 0.1% heparinized phosphate buffered saline (PBS) for 10 min, followed by 4% paraformaldehyde pH 7.2–7.4 for another 30 min. After craniotomy, brains were removed and stored in 9% phosphate buffer (Sigma Aldrich - S3264) and then cut in the coronal plane. Eighty micrometer thick sections were obtained using a vibrating blade microtome (LeicaVibratomeVT1000S). Serial anatomical sections (1:6 interval) were subjected to immunohistochemical reactions using GFAP selective marker for astrocytes.

In previous reports we used three-dimensional microscopic reconstructions and applied hierarchical cluster analysis to classify astrocytes based on morphometric features using the largest Euclidean distance between the groups. We found two large morphological families (Type I and Type II). These families were differentially affected by contrasting migratory patterns (Carvalho-Paulo et al., 2018; Henrique et al., 2020). Here, we performed hierarchical clustering on variance-shrunk logarithmized values of multimodal morphometrical features to define the number of morphological families and compared hippocampal astrocyte morphologies of long-distance migratory birds with contrasting migratory flights (C. semipalmatus, C. pusilla, and A. macularius) with a non-migratory specie (C. collaris).

For the three-dimensional reconstruction of positive GFAP astrocytes, an optical microscope (Eclipse Ci, NIKON) with motorized stage and analog-digital converters (MAC6000 System, Ludl Electronic Products, Hawthorne, NY, United States) was used. This system was coupled to a microprocessor that controlled the movements of the microscopic stage with the aid of a specialized program (Neurolucida, MBF Bioscience, Williston, VT, United States) to store the spatial information (X, Y, Z coordinates) of each digitized point of interest. Three types of cells were identified: protoplasmic, fibrous, and radial astrocytes similar to previous descriptions (Carvalho-Paulo et al., 2017; Mendes de Lima et al., 2019; Henrique et al., 2020). In this study, we used only protoplasmic astrocytes.

The contours of the hippocampal formation (Atoji and Wild, 2004; Atoji et al., 2016) were determined on a low power objective (4× lens). To identify astrocyte morphological details and to ensure greater detail in 3D reconstructions, the low power lens was replaced by a high-power oil immersion 100× lens PLANFLUOR (NA 1.3; DF = 0.2 μm; Nikon, Japan).

A total of 264 astrocytes from the hippocampal formation of A. macularius, 251 of C. pusilla, 302 of C. semipalmatus, and 260 of C. collaris were reconstructed in three dimensions. To select astrocytes for reconstruction a random and systematic stereological sampling approach was adopted (West, 2002). For this, we used squared probes (50 μm × 50 μm) separated from each other by a 900 μm × 900 μm grid interval (Figure 1). The grid interval was estimated to obtain a minimum of 50 reconstructed astrocytes per animal and it was selected based on the total area of the hippocampal formation. The number of probes per section was proportional to the area of the hippocampal formation of each section.

Only astrocytes located within the limits of the probe box were selected for analysis. In cases where there were no cells within the probe box that met the pattern of complete immunostaining and branch integrity, the closest astrocyte outside the probe box border was chosen for reconstruction. Due to minimal shrinkage in the X/Y axes, the correction for shrinkage induced by histological processing was applied exclusively for the z axis, and corresponded to 1.75×, as previously recommended by Carlo and Stevens (2011).

Supplementary Table 1 shows the definition of all morphometric features of hippocampal astrocytes. The following equation adapted from previous neuronal dendritic reconstruction (Pillai et al., 2012) was used to estimate morphological complexity:

Complexity =[Sum of the terminal orders + Number of terminals] × [Total branch length/Number of primary branches]

Thus, 20 morphological variables were used for morphometry: total branch length, branch surface area, tortuosity, total branch volume, base diameter of primary branches, total number of segments, segments/mm, number of branching points, tree surface area (μm²), planar angle, complexity, convex hull perimeter (μm), convex hull area (μm²) 2-D, convex hull surface area (μm²), convex hull volume (μm³), and Vertex: Va, Vb, Vc, K-dim (fractal dimension) (see Supplementary Table 1 for details).

For photomicrographs, a digital camera (Microfire, Optronics, CA, United States) coupled with a Nikon microscope (Eclipse Ci, NIKON) was used, and acquired images were post-processed for brightness and contrast with Adobe Photoshop software (Adobe Inc San José, CA, United States). We selected images of the most representative astrocytes of each cell type indicated by the hierarchical cluster analysis. For the choice of the representative cell of each group (“average cell”), the distance matrix was used to obtain the sum of the distances of each cell relative to all others. It is assumed that the cell that best represents a group has the smallest sum of distances. The matrices were constructed with the combination of all cells of a given group taken pairwise, followed by the weighted calculation of a scalar Euclidean distance between cells using all morphometric variables (Henrique et al., 2020).

We identified 17 alleles for the ADCYAP1 locus, varying in size from 168 to 204 base pairs (bp). The most common alleles found were 172 and 174 bp. The ADCYAP1 locus was polymorphic in all four species with the number of alleles varying from 5 to 6, and number of exclusive alleles varying from 1 to 4 across populations (Supplementary Table 3). Analysis showed no evidence of null alleles and genotype errors such as stuttering, or allele drop out. However, two out of the four populations exhibited significant deviation (p < 0.01) from Hardy–Weinberg equilibrium and higher numbers of homozygotes compared to the other populations. Observed and expected heterozygosity ranged from 0.417 to 0.786 and from 0.712 to 0.792, respectively (Supplementary Table 4).

All pairwise F_st and R_st values based on ADCYAP1 allele frequencies presented significant differentiation across species, which was expected since all are distinct and known species (Supplementary Table 5). Higher levels of genetic differentiation were observed across species pairwise comparisons and low levels were observed between species in the Charadrius genus, corroborating the evolutionary relations of these groups.

The association of size of microsatellites with migratory distance showed a positive and strongly supported association (Spearman correlation test, Rho = 0.915; p = 0.000; Figure 3 and Supplementary Table 6).

We also performed a comparison between the size of the microsatellites in the four species (Supplementary Table 7). We found significant differences (PERMANOVA, F = 180.98, p = 0.0001) in all possible comparisons (C. semipalmatus vs C. collaris, p < 0.026; C. semipalmatus vs C. pusilla, p < 0.025; C. semipalmatus vs A. macularius, p < 0.0001; C. collaris vs C. pusilla, p < 0.0001; C. collaris vs C. pusilla, p < 0.0001; A. macularius vs C pusilla, p < 0.0127).

Stellate astrocytes in the hippocampal formation of A. macularius, C. pusilla, C. semipalmatus, and C. collaris show glial fibrillary acidic protein expression in the cell body, from which GFAP positive primary thicker branches emerge and progressively ramify, terminating as tiny branches (Figure 4).

From hierarchical cluster and discriminant function analysis emerged that in the non-migratory species, the morphological complexity, the volume of the convex hull and the number of segments were the multimodal variables that most contributed to the formation of clusters (Figures 5A,B). In the analysis of the canonical discriminant function, it was shown that the complexity and volume of the convex hull accounted for 99.4% of the variance (Figures 5C,D). Wilks’s lambda value correspondent to 1 and 2 canonical discriminant functions was the most important for group separation (Wilks’s lambda = 0.185; p = 0.000) (Figure 5E) with 95.4% of the group data classified correctly (Figure 5F).

In migratory species, the morphological complexity, volume, surface, area, and perimeter of the convex hull were the multimodal variables that contributed most to the cluster formation in the different species (Figures 5, 6). The morphological complexity and volume of the convex hull were the morphometric features that contributed most for cluster formation in all species (Figures 5I,J, 6C,D,I,J). Function 1 explained 77.1, 98.4, and 94.7% of the sample variance in the species C. semipalmatus, C. pusilla, and A. macularius, respectively (Figures 5C,D). Wilks’s lambda value correspondent to 1 and 2 canonical discriminant functions was the most important for groups separation in the species C. semipalmatus and C. pusilla (Wilks lambda: C. semipalmatus = 0.185; p = 0.000; C. pusilla = 0.211; p = 0.000) (Figures 5K, 6E) and Wilks’s lambda value correspondent to 1 and 3 canonical discriminant functions was the most important for groups separation in A. macularius (Wilks lambda = 0.141; p = 0.000) (Figure 6K) with 95.4% of the group data correctly classified in the three species (Figures 5H, 6F,H).

After evaluation of the data to select multimodal variables for hierarchical cluster analysis using variance-shrunk (logarithmized) parameters, we identified the groups in the dendrogram with statistically significant differences, followed by canonical discriminant analysis (Wiltschko W. and Wiltschko, 2012) (see, Figures 5–7). The canonical discriminant function analysis performed very well demonstrating its ability to predict astrocyte morphology, with high classification accuracy of individuals in the three different groups in C. collaris (97.4% for type I, 98.3% for type II and 87.5% for type III), and C. semipalmatus (95% for type I, 90% for type II and 96.7% for type III) and C. pusilla (95.5% for type I, 99.4% for type II and 96% for type III) and in four different groups in A. macularius (91.4% for type I, 80% for type 2, 89.4% for type III and 78.7% for type IV), confirming the existence of three distinct groups in C. collaris, C. semipalmatus, and C. pusilla and four groups in A. macularius. See Figures 5, 6 for details.

The comparative analysis between species, with all morphotypes revealed significant differences in all comparisons for morphological complexity [PERMANOVA: C. collaris—F (3;260) = 190.33, p = 0.0001; C. semipalmatus—F (3;302) = 226.65, p = 0.0001; C. pusilla—F (3;251) = 164.06, p = 0.0001; A. macularius—F (4;264) = 216.8, p = 0.0001] and significant differences in pairwise comparisons within all species (p = 0.0001) (Supplementary Tables 8–11).

Figure 7 is a graphic representation for these findings of morphological complexity and convex hull volume. As mentioned before these are the variables that contributed most to cluster formation in all species. It is important to highlight that there was no linear correspondence between morphotype mean values of convex hull volume in C. pusilla (with three morphotypes) and A. macularius (with four morphotypes) in the graphic representation of this variable in Figure 7.

The percent distribution of each morphotype in each species is shown in Figure 8. The semipalmated sandpiper C. pusilla had a higher percentage of type II (62.55%) than type I (17.53%) or type III (19.92%), and these values contrast with the percentual distributions of type II, type I, and type III astrocytes of C. semipalmatus (36.42, vs 33.47 vs 29.8%), A. macularius (24.62 vs 21.97 vs 35.61 vs 18% of type IV), or C. collaris (45.98 vs 29.12 vs 24.52%) (Figure 8). Thus, except for A. macularius where type III showed higher frequency, astrocytes of intermediate morphological complexity (Type II morphotype) are more frequent in all other species. Please remember that Type I designates the morphotype with greater morphological complexity mean value in all species.

Permutational analysis of multivariate dispersions of 17 morphometric features showed significant interaction between species and astrocyte type [PERMIDISP (7,1069) F: 5,277; p = 0.0001). In the paired tests for species there were differences in dispersion for almost all species (p < 0.05), except between non-migratory C. collaris and the species with the longest migratory distance A. macularius (t = 0.711; p = 0.477). Considering only the factor “type,” however, there was no significant dispersion (type I × type II; t = 0.973; p = 0.338) (see Supplementary Table 12).

Regarding the differences in Euclidean space, we found significant differences in the interaction of the factors [PERMANOVA; F (3.1069) = 5,479; p = 0.0001], “species” [F (3.1069) = 106.82; p = 0.0001] and “type” [F (3.1069) = 186.69; p = 0.0001]. All pairwise comparisons showed significant differences, including comparisons between species, between types and interactions between the two factors (p < 0.001). Thus, when we analyze the results of PERMIDISP and PERMANOVA, we can say that only the factor “type” showed differences in the location of the samples in the Euclidean space. Regarding the factors “species and “species × type” the differences found were due to the dispersion of the samples.

Figure 9 exhibits a Kernel density plot of the distribution of morphological complexity values of distinct morphotypes of hippocampal astrocytes from the dataset of C. collaris, C. semipalmatus, C. pusilla, and A. macularius. As expected, except for A. macularius where we identified four peaks, all other species showed three peaks, confirming that morphological complexity can be used in isolation to distinguish the morphotypes of astrocytes of the hippocampal formation.

Another way of visualizing the influence of migratory behavior on the morphology of astrocytes is to observe the representative 3D mean cells of each group (Figure 10). While in A. macularius we named type IV the hippocampal astrocytes with the lowest complexity, in species with three morphotypes (C. collaris, C. semipalmatus, and C. pusilla) the lowest complexity astrocytes corresponded to type III. Notice the greater morphological complexity mean values of hippocampal astrocytes in migratory species as compared to the non-migratory C. collaris.

Two raw astrocytes images overlayed with branch reconstructions used to illustrate contrasting morphological values are exhibited on Figure 11. They are examples of higher and lower morphological complexity astrocytes from the hippocampal formation of shorebirds which were microscopically reconstructed in 3D.

Before and during migration, birds undergo significant changes in physiology and behavior that are adapted and motivated differently in spring and autumn: spring migration takes place in search of reproductive sites while autumn migration takes place in search of wintering sites with milder temperatures and greater food availability (Sharma et al., 2018a,b). As compared with autumn migration, buntings show greater body mass, higher levels of triglycerides and free fatty acids, accumulate more subcutaneous fat and liver lipids, and more intense Zugunruhe in spring migration (Samojłowicz et al., 2019). A relação entre o tamanho do microsatellite e diferentes funções fisiológicas foi previamente descrita (Ralston et al., 2019). In the present report all individuals were captured between August 15 and April 8, during the wintering period in the mangroves of Bragança (Brazil) in the Amazon River estuary, where they experienced temperatures between 20.4 and 32.8°C with the minimum values coincident with the increase in pluviometry mean values in March (Ribeiro, 2001).

In all comparisons of mean allele length between migrant species (A. macularius, C. pusilla, and C. semipalmatus), and the non-migratory species (C. collaris) we found significant differences, with the lower mean value for Euclidean distance found in the non-migratory species. In addition, the mean values of allele length in migratory birds increased significantly as a function of migratory distance.

Thus, based on earlier studies of ADCYAP1 variation (Mueller et al., 2011), our prediction that migratory species A. macularius, C. pusilla, and C. semipalmatus would possess longer microsatellite repeat-length alleles of ADCYAP1 compared with those of non-migratory C. collaris, was confirmed. In addition, we found proportional direct association between the size of microsatellite repeat-length alleles at the ADCYAP1 locus and autumnal migratory distances. These findings are in line with previous studies that demonstrated that ADCYAP1 polymorphism covaries with breeding latitude (Bazzi et al., 2016b).

Differences in migratory behavior affect not only the number of astrocytes in the hippocampal formation of migrating birds, but also their morphology (Henrique et al., 2020). Indeed, Henrique and collaborators compared the number and 3-D morphology of hippocampal astrocytes of C. semipalmatus before and after autumnal migration with those of C. pusilla to test the hypothesis that the contrasting migratory flights of these species could differentially shape hippocampal astrocyte number and morphology. C. pusilla migration to the southern hemisphere includes a 5-day non-stop flight over the Atlantic Ocean (Brown, 2014), whereas C. semipalmatus migration, to the same area, is largely over land with stopovers for feeding and rest. After hierarchical cluster analysis of astrocyte morphological features, two families of morphological phenotypes in C. pusilla and C. semipalmatus hippocampal formations named Type I and Type II were distinguished, which were differentially affected after autumnal migratory flights. Stereological counts of hippocampal astrocytes demonstrated that the number of astrocytes decreased significantly in C. pusilla but did not change in C. semipalmatus.

Thus, after hierarchical cluster analysis with non-normalized data, we previously found at the greater Euclidian distances two morphotypes of hippocampal astrocytes (Carvalho-Paulo et al., 2018; Henrique et al., 2020). Here, we performed hierarchical clustering on variance-shrunk logarithmized values of multimodal morphometrical features for reduction of the asymmetrical influence of non-standardized morphological features on cluster segregation. With logarithmized morphometric values, we compared the morphologies of astrocytes from the hippocampus of long-distance migratory birds with contrasting migratory flights (C. semipalmatus, C. pusilla, and A. macularius) with each other, and with a non-migratory species (C. collaris). Except for the hippocampal formation of A. macularius where we found four morphotypes, all other species showed three hippocampal astrocytes morphotypes. Thus, in this work, we expanded the studies of morphometry of astrocytes to C. collaris and A. macularius hippocampal formations indicating the occurrence of three main morphological types of astrocytes in C. pusilla, C. semipalmatus, and C. collaris and 4 morphotypes in A. macularius. Morphological complexity and convex hull volume were the variables that best distinguished the morphological families suggested by hierarchical cluster analysis. These morphological families of astrocytes were validated by comparing the mean values of morphological variables for type and species studied using PERMANOVA and canonical discriminant analysis. Type I was used to name the cell population with the highest average values of morphological complexity. Type II, Type III, and Type IV named astrocytes with decreased morphological complexity. We suggest that the differences found between migratory birds’ hippocampal astrocyte morphologies might be related to the adaptive response imposed by the contrasting long-distance migratory flights on learning and memory for recognition of olfactory, geomagnetic, and visual cues during migration or to recognize local cues for both migratory birds and non-migratory C. collaris (Mouritsen et al., 2016; Magalhaes et al., 2017; Carvalho-Paulo et al., 2018; da Costa et al., 2020; Henrique et al., 2020; Guerreiro et al., 2021).

Notably, a phylogenetically independent contrast (PIC) approach (Felsenstein, 1985) showed that most morphometric differences in astrocytes found in the different species were not influenced by phylogenetic differences (Henrique et al., 2020). The PIC approach used DNA sequences of intron 7 of beta-fibrinogen (fib 7) gene, recombination activating gene 1(RAG1), cytochrome oxidase c subunit 1 (COI) gene and cytochrome b (cyt b) gene obtained from GeneBank for C. pusilla, C. semipalmatus, A. macularius, and C. collaris (Henrique et al., 2020).

In the present report, however, we were expecting that A. macularius and C. semipalmatus, that rely more on remembering visual cues during overland migration than C. pusilla, which migrates via a long-distance non-stop flight over the Atlantic Ocean (Diniz et al., 2016) would show higher mean values of morphological complexity. We also hypothesized that C. collaris, a non-migrant species, as compared with the migrant species would show the lowest morphological complexity. We found that the non-migratory species C. collaris and C. semipalmatus showed similar mean values for type I (higher complexity) and III (lower complexity) and that type II (median complexity) mean values in C. collaris were greater than that of C. semipalmatus. Similar mean values were found for morphological complexity of type I and II in A. macularius and C. pusilla but these values were significantly higher than those of correspondently named astrocytes of C. collaris and C. semipalmatus. In addition, A. macularius showed greater type III mean values than those of C. pusilla correspondent morphotype.

From these findings, it is not possible to anticipate the mechanisms underneath the hippocampal astrocyte morphological differences between wintering migrating birds with contrasting migratory flights and non-migrating birds. However, in a previous study it clearly emerged that the glial morphologies of birds collected in August in the Bay of Fundy (Canada) and September to March on Isla Canela (Bragança, Brazil) are clearly different, both in C. pusilla and C. semipalmatus (Henrique et al., 2020). These findings were interpreted as due to contrasting migratory routes (non-stop transatlantic flight vs mostly overland migration). However, many other possibilities were pointed out as possible contributors to the morphological differences. For example, different capture dates (Carvalho-Paulo et al., 2018) with differential implications for hormones and receptors for stress and pre- and post-breeding conditions (Riou et al., 2010; Eikenaar et al., 2015; Deviche et al., 2016; Surbhi Rastogi et al., 2016). Here, we also raise the hypothesis that distinct morphotypes, may have differential physiological roles in different species and this may be at least part of the underlying hippocampal circuitry adaptive response for behavioral changes in those species (Hwang et al., 2021).

Calidris pusilla, C. semipalmatus, and A. macularius migratory journeys between the northern breeding sites of United States and Canada and wintering grounds in the Amazon basin of northern South America are among the longest migratory routes of shorebirds. As previously indicated, significant association was observed between the size of the simple sequence repeats (SSR) and migration distance in these three migrant species. Consistent with this, differences in food intake and glucocorticoid effects during overland and transatlantic flights may differentially affect migrating and wintering birds’ metabolic pathways, with significant influences on astrocyte morphologies. For example, during the 5-day non-stop flight of C. pusilla a short supply of glucose and a high demand for lipids occurs, inducing the brain to increase ketone body metabolism to support the transoceanic flight (Achanta and Rae, 2017). This uninterrupted flight of C. pusilla compared with the multiple stopover flights of C. semipalmatus and A. macularius, may impose differential demand for PACAP, which is synthesized by ADCYAP1.

Because SSR in the regulatory region 3’UTR of ADCYAP1 may modify gene function and post-transcriptional processes (Riley and Krieger, 2009; Steinmeyer et al., 2009) it is reasonable to expect that any changes in SSR found in A. macularius, C. pusilla, and C. semipalmatus, may benefit migratory behavior of these species. In agreement with this expectation, it has been demonstrated that in the CNS, PACAP is involved in the rhythmicity of melatonin production and in the increase of cAMP in birds (Nakahara et al., 2002; Nowak and Zawilska, 2003), as well as acting as a co-transmitter with glutamate to shift the phase of the CNS circadian rhythm in a similar way to light (Michel et al., 2006; Vaudry et al., 2009). In addition, previous data show that PACAP plays important role in controlling astroglial functions by regulating cell proliferation and glycogen metabolism (Magistretti et al., 1998; Masmoudi-Kouki et al., 2007; Nakamachi et al., 2011).

Moreover, PACAP is affected by diet and fasting (Iwasa et al., 2016; Nakata et al., 2016) with important implications for regulation of food intake (Kataoka et al., 2013; Nguyen et al., 2020) and energy homeostasis (Chang et al., 2021) and this seems to include birds (Tachibana et al., 2015; Simon et al., 2017). PACAP is also affected by photoperiodic light changes (Saino et al., 2015; Bazzi et al., 2016a; Adamska et al., 2018; Haraguchi et al., 2019), reproductive conditions (Prisco et al., 2019; Winters and Moore, 2020) and stress (Agarwal et al., 2005; Norrholm et al., 2005; Hammack et al., 2010). Due to differential pre- and post-breeding physiological conditions, distinct metabolic demands and diets imposed by migration with multiple stopovers and fasting uninterrupted transoceanic flight, along with differential stress levels along the migratory journey of migrant species compared to the non-migrant C. collaris, it may be possible that at least part of the astrocyte morphological differences between species may be associated with interspecific differential expression of ADCYAP1. Indeed, PACAP mimicked effects of forskolin, a direct activator of adenylate cyclase, on the actin cytoskeleton of astrocytes with resultant astrocyte morphological changes (Perez et al., 2005). In addition, PACAP is essential for lactate production and secretion in astrocytes, a central step in the neuronal physiology of hippocampal learning and memory (Kambe et al., 2021) the center for integrative information for familiar landmarks and landscape features in homing pigeons (Bingman and Ewry, 2020; Gagliardo et al., 2021), and for avian long-distance migratory journeys (Frost and Mouritsen, 2006; Barkan et al., 2016; Bingman and MacDougall-Shackleton, 2017).

Migratory phenotypes and ADCYAP1 and CLOCK genes have been tested in other species including migratory distance, wing morphology and spring migration arrival of Sylvia atricapilla (Mueller et al., 2011; Mettler et al., 2015); migratory restlessness in the songbird genus Junco and S. atricapilla (Mueller et al., 2011; Peterson et al., 2013); time of C. pusilla migration (Bazzi et al., 2016b); and activation of CLOCK and other circadian genes in Gallus gallus (Nagy and Csernus, 2007).

In line with these findings, we found significant differences in all comparisons of allele mean length between migrant species (A. macularius, C. pusilla, and C. semipalmatus), and the non-migratory species (C. collaris), with the lower mean value in the non-migratory C. collaris. This may suggest that different migratory behaviors are associated with size differences of ADCYAP1 microsatellites, and this may be acting through PACAP to induce morphological and functional changes in the astrocytes of the hippocampal formation (Vaudry et al., 2009). It is reasonable to assume as well, that in C. collaris, ADCYAP1 microsatellites are not under positive selection for migratory behavior.

Because we did not track individuals during the migratory flights we estimated the distance between breeding sites and the wintering places of capture (see section “Materials and Methods”). We followed previous descriptions of suggested trajectories between stop overs for each species (Williams and Williams, 1978; Brown, 2014; Anderson et al., 2019), as indicated in Figure 2. Because C. collaris is not a migrant species we arbitrarily attributed 0 km as distance traveled for this species and because this species is not sedentary, this is a limitation of the present report. Migrant birds also move during the wintering periods and their local movements were ignored as well. Because local displacements of all individuals during the wintering period may not be similar in different species, their potential influence on astrocytes morphologies could not be assessed, and this is limitation to be explored in future studies.

To measure possible influence of capture dates on hippocampal astrocyte morphological complexities of wintering birds, in a previous report dedicated to C. pusilla, we compared astrocyte morphologies of birds captured at different time points of the wintering period using the same statistical analysis (Carvalho-Paulo et al., 2018). We found smaller changes in the mean values of hippocampal astrocytes morphological complexity which did not significantly affect the results (Carvalho-Paulo et al., 2018). However, recent findings from morphological analysis of hippocampal astrocytes of Arenaria interpres, differed significantly in the morphological complexity of hippocampal astrocytes of autumnal recently arrived migrant birds (captured in September/October) and spring premigratory individuals (captured in April/May) suggesting that as wintering period progresses, significant changes in hippocampal circuitry occur (da Costa et al., 2020). Thus, it is reasonable to expect that astrocytes have their morphology changed during the wintering period.

Because there is no information in the literature about potential influence of sex and age on hippocampal astrocyte morphology in long-distance migratory birds, and we did not measure the age of individuals in our sample due to technical limitations, it is difficult to discuss these potential influences in detail. However, experience and sex are important variables that have been previously demonstrated to influence hippocampal-dependent tasks in birds (Astié et al., 2015; Rensel et al., 2015; Guigueno et al., 2016; Bingman and MacDougall-Shackleton, 2017), and migratory behavior is accompanied by hippocampal morphological changes including volume, and neurogenesis (Barkan et al., 2016, 2017; de Morais Magalhaes et al., 2017) which should be considered in future studies of hippocampal astrocyte morphologies in long-distance migratory birds.

Although the correlational analysis between the length of ADCYAP1 microsatellites, distance of migration, and differences in the morphologies of astrocytes of the hippocampal formation seems to be coherent, the story could be different if another higher order brain area, less involved in migration, exhibited similar astrocytes differences and this is a potential limitation of the present study, that could be avoided if another area was explored (Carvalho-Paulo et al., 2018; Henrique et al., 2020).

It is not uncommon to find contradictory results in comparative studies due to ambiguities in the definition of the objects and areas of interest, variations in histological procedures, or in the case of 3D reconstructions, dissimilarities in the sampling approach to select cells for reconstruction (West, 2002). In this report all samples were obtained with the same tissue processing protocols (perfusion, antigen retrieval, immunoreaction, dehydration, counterstaining, and clearing) and the specificity of the immunohistochemical pattern was confirmed using a control reaction that omitted the primary antibody (Saper and Sawchenko, 2003; Schmitt et al., 2004; Fritschy, 2008). To obtain sufficient contrast between foreground and background we improved the signal/noise ratio with glucose-oxidase-DAB-Nickel peroxidase amplification method (Shu et al., 1988).

It has been also demonstrated that the z-axis (perpendicular to the cutting surface), shrinks by approximately 75% of the cut thickness after dehydration and clearing (Carlo and Stevens, 2011). Based on those findings, all astrocytes’ reconstructions were corrected for z-axis shrinkage. No corrections were applied to X/Y axes and used the same software and hardware approaches for sampling, reconstruction, and analysis. These procedures guarantee similar, systematic and random sampling selection of astrocytes across all regions of the areas of interest. Finally, to detect possible variations in the criteria for identifying and including only complete astrocyte arbors inside the area of interest, we undertook checking procedures of the results by having different investigators reconstruct astrocytes in the same regions using the same GFAP antibody as a marker for selective labeling. Thus, we expected to reduce possible sources of non-biological variation.