Amino acid sequences of Japanese anchovy SDF-1a (WKC57597.1), SDF-1b (WKC57598.1), CXCR4a (WKF24609.1), CXCR4b (WKF24610.1), human (Homo sapiens) SDF-1 (NP_000600.1), CXCR4 (NP_003458.1), and zebrafish (Danio rerio) CXCR4a (NP_571957.2), CXCR4b (NP_571909.1), SDF-1a (NP_840092.1), SDF-1b (NP_001307343.1) were obtained from NCBI. Amino acid sequence alignment was performed in CLUSTAL W 2.1 with default setting (Slow/Accurate), and motif search was performed in MOTIF Search with default setting in the Pfam database.

For protein 3D model, an online server, Colab Fold, based on Alpha Fold2 (Colab Fold v1.5.3: Alpha Fold2 using MMseqs2) was used for simulation, with default parameters. The resulting PDB files were edited and visualized in UCSF Chimera (Pettersen et al., 2004) and subsequently used for secondary structure analysis. Cavity searches using CB Dock (Liu et al., 2022; Yang et al., 2022) for CXCR4 were also performed, and the cavity size of the receptor site was shown. These results of CXCR4 were also used for antagonist binding analysis that used auto blind docking of CB Dock. AMD3100 (ab120718, abcam, USA), AMD3465 (ab120809, abcam, USA), WZ811 (S2912, Selleck Chemicals, USA), and LY2510924 (S8505, Selleck Chemicals, USA) were used in this study. The 3D Conformer models (SDF file) used for docking simulations were obtained from PubChem, and PubChem CIDs were AMD3100 (65015), AMD3465 (483559), and WZ811 (11565518). On the other hand, LY2510924, which has a large molecular size, could not be used in the simulation analysis due to the large number of rotatable parts. The results obtained for the ligands are shown with surface, while the receptors are shown in cartoon form and colored according to hydrophobicity. Each binding site (amino acid residues) are listed in Supplementary Figure S1. Human CXCR4 (PDB ID: 3OE9 (Wu et al., 2010), 4RWS (Qin et al., 2015), N- terminus binding, PDB ID: 2N55 (Ziarek et al., 2017)) and SDF-1 (PDBID: monomer 2KEE (Veldkamp et al., 2009), dimer 4UAI (Smith et al., 2014)) protein were retrieved from available literature.

The Japanese anchovies used for experiment in this study, were breed inhouse, at least for two generations, from a wild stock caught in Kagoshima Bay, Japan. Approximately 300 adult fish were kept in a 4-ton circular FRP tank with 15x daily water exchange. The water temperature and photoperiod were respectively maintained at 20°C–23°C and 11L:13D. Natural breeding occurred between 4 and 6 h after dark and eggs were collected from the outlet of tank. Good quality fertilized eggs (floating and transparent) were transferred to filter sterilized sea water and used for antagonist tests. PGC visualization and microinjection of Japanese anchovy followed previously established protocol (Yahiro et al., 2023; submitted elsewhere). Briefly, the fertilized eggs were collected and placed on a 1.5% agar gel with grooves measuring 0.47–0.57 mm width, filled with seawater to prevent desiccation. Microinjection needles were prepared using GD-1 (Narishige, Japan) micro glass capillaries, pulled and grind to achieve a tip diameter of 1 µm (Ohga et al., 2023). Visualization of PGC was performed by injecting egfp-nanos3 3′UTR mRNA (egfp mRNA, 100 ng/μL). The injection solution was mixed with phenol red (0.05% of final volume, Sigma-Aldrich, USA), adjusted to the required concentration using DPBS(-) (Gibco, USA), and injected into the cytoplasm of 1∼8 cell stage of embryos. After injection, eggs were incubated at 20°C with daily water exchange. All fluorescence observations were made using a stereo microscope MZ10F (Leica, Germany) and photomicrographs were taken with a Flex Cam C1 camera (Leica, Germany).

To ensure direct entry and reliable dosing of antagonists, egfp mRNA solutions containing antagonists were microinjected. Stock solutions of antagonist were prepared in DPBS(-) (Gibco, USA) or DMSO (Dimethyl sulfoxide, Nacalai tesque, Japan), as specified by manufacturers, and diluted to required concentration before use. The concentrations of the various antagonists were selected based on IC50/EC50 from previous studies: AMD3100 13 nM (Schols et al., 1997), AMD3465 10 nM (Hatse et al., 2005), WZ811 1.2 nM (Zhan et al., 2007), LY2510924 0.26 nM (Peng et al., 2015), and higher concentrations were tested to overcome interspecies differences, if any. The solutions used in microinjection contained egfp mRNA, 0.05% phenol red, and various antagonists: AMD3100 (1.3, 13, 130, 1.3 × 10³, 1.3 × 10⁴, 1.25 × 10⁵, 6.0 × 10⁵ nM), AMD3465 (10, 100, 1.0 × 10³, 1.0 × 10⁴, 1.0 × 10⁵ nM), WZ811 (1, 10, 100, 1.0 × 10³nM) LY2510924 (1, 10, 100, 1.0 × 10³, 1.0 × 10⁴, 1.0 × 10⁵ nM), with final dilution adjustment using DPBS(-). Microinjected eggs were placed in sterile seawater (0.22 μm filter sterile) and incubated at 20°C until 48 h after fertilization (hpf). Solvent-control injected, and non-injected eggs were similarly incubated.

Immersion treatment was performed using AMD3465, WZ811, and LY2510924. Higher concentrations were used than in the microinjection method to compensate any loss during penetration. 1, 10, and 50 μM treatment water was prepared just before use, by adding required amount of antagonist in filter sterilized sea water. Seventeen fertilized egfp mRNA microinjected eggs with were placed in 5 mL of treatment water with regular water change at 24 h interval. Hatched larvae were anesthetized and observed under microscope. PGCs visualizations were performed as mentioned above. Solvent-control samples were similarly maintained and observed.

To easily quantify the effect of antagonists on PGC migration, ImageJ Fiji were used (Schindelin et al., 2012). Distance between the farthest PGCs (the most distant PGCs in the body) in each of 6 randomly sampled individuals/experimental group were measured. The distances in pixels were converted to mm units, and used for analysis.

Each in vivo experiment was conducted for 3 times. All data were graphed using Prism 9 (GraphPad, USA) and significance tests were performed using one-way ANOVA and multiple post hoc tests (Tukey test). Error bars in all figures show SEM.

Specific genome duplications have been found in many bony fishes (Wittbrodt et al., 1998; Hoegg et al., 2004; Glasauer and Neuhauss, 2014) including Japanese anchovy. We have isolated two SDF-1 and CXCR4 paralogs (EJ SDF-1a/b, EJ CXCR4a/b) and deposited in NCBI. Based on this, we compared the amino acid sequence of human SDF-1/CXCR4 (HS SDF-1/CXCR4) (Figure 1) and found similarity scores of more than 60 for CXCR4 (Figure 1A) and 40 for SDF-1 (Figure 1B) paralogs. Motif searches suggested that CXCR4 Chemokine receptor N terminal domain (PF12109), required for SDF-1 binding is preserved in EJ CXCR4a, while a deletion was suspected in CXCR4b (Figure 1A), that is same as in zebrafish CXCR4 paralogs (DR CXCR4a/b) (Supplementary Figure S2A). The 7 transmembrane receptors (rhodopsin family, PF00001), characteristic of G-protein coupled receptors (Jacoby et al., 2006), were conserved in both EJ CXCR4a and EJ CXCR4b. Additionally, EJ CXCR4a showed to have more domains similar to other receptors such as TAS2R (bitter taste receptors), Anoctamin (chloride channel), and Cox2 (Cyclooxygenase2). Furthermore, large differences were observed in the amino acid sequence of SDF-1 between Japanese anchovy and humans. The putative RFFESH domain, which is essential for CXCR4 binding, exhibited high similarity between EJ SDF-1a and b, and consistent with zebrafish SDF-1a/b (DR SDF-1a/b) (Supplementary Figure S3), but differed from HS SDF-1. On the other hand, it is noteworthy that the presence of four functional cysteines and the IL8 motif (PF00048, Figure 1B), crucial for CXCR4 binding, remained highly conserved.

We used Colab Fold (Mirdita et al., 2022) to construct 3D models of each protein (Figure 2). Result of EJ CXCR4a/b, HS CXCR4, and DR CXCR4a/b protein conformations (Figure 2A, Supplementary Figure S2A) suggest high similarity across these three species, with a typical 7-transmembrane α-helix and a CXCR4 N-terminal loop toward the outer cell membrane. Cavity searches for EJ CXCR4a/b and HS SDF-1 revealed that the size of the presumably functional cavities differed significantly, with EJ CXCR4a (6,307 Å³) being nearly five times larger than EJ CXCR4b (1,241 Å³) and HS CXCR4 (1722 Å³). Zebrafish depicted slightly different picutre, with both DR CXCR4a/b having larger cavities (Supplementary Figure S2A). All the examined SDF-1 protein 3D structures exhibited a distinctive binding site (RFFESH Domain) at the center of the 3D model (Figure 2B, Supplementary Figure S2B), signifying domain conservation. However, ligand (EJ SDF-1a/b) - receptor (EJ CXCR4a/b) binding models (Supplementary Figure S4) were largely different from human counterpart (Phanourios and Christodoulos et al., 2014) and require further investigation.

To select antagonists, a simple simulation binding model was created (Figure 3) using CB Dock. For LY2510429 (MW = 1,189.45 Dalton), CB Dock could not be performed because of its large structure, while AMD3100, AMD3465, and WZ811 possessed a size of 794.48, 896.08 and 290.36 Da, respectively. Results of auto blind docking for three remaining antagonists show relatively higher binding capacities. For EJ CXCR4a, the lowest Vina Score (lower energy required for binding, an indication of easier binding) was AMD3100 (−9.6), followed by AMD3465 (−8.8) and WZ811 (−8.4). On the other hand, EJ CXCR4b demonstrated the strongest binding affinity for AMD3465 (Vina Score −8.9) followed by AMD3100 (−8.6) and WZ811 (−7.9). The location of the binding of AMD3100 and AMD3465, which are close in molecular weight size, shows that they bind well to the 7 TM motif, and could be confirmed with each primordial binding to each TM, regardless of CXCR4a or CXCR4b. On the other hand, WZ811 showed the highest Vina Score, and seemed to have space in the binding site with the 7 TM motif, perhaps due to its smaller size. In fact, no binding was observed (Figure 3; red background) for the 1 TM in CXCR4a and for the 1 and 7 TM in CXCR4b, suggesting that the effect may be weak or easy to remove.

To directly introduce the antagonist into the eggs and determine its effect on PGC, we microinjected the antagonist into fertilized eggs, measured the hatchability at 48 hpf (Figure 4), and compared with the solvent injected (egfp control) and non-injected controls. The results showed that hatchability decreased in both injected groups compared to their respective non-injected controls. Lowest hatchability was observed in WZ811 treatment groups and DMSO (solvent control of WZ811) treated individuals. However, no significant difference was recorded in antagonist-treated groups compared to their egfp counterpart, across different concentration (except for AMD3465, 10 nM). Apart from WZ811 treated and DMSO treated fish, no external malformation was observed. These results suggest that none of the antagonists are biotoxic, and the reduced hatchability and malformations observed in the WZ811 group are due to solvent (DMSO) effects. Furthermore, fluorescence microscopy depicted that the PGCs were clustered in the putative gonadal ridge of egfp group larvae while varying degree of PGC dispersion (characteristics of PGC mismigration) were observed in antagonist-treated individuals (Figure 5). Significantly higher mismigration, i.e., DFP (Figure 6), was observed in AMD3465 treated fish at 10, 100 and 1 × 10⁵ nM concentration. This result corresponds with our in silico low Vina score prediction data. Further, as predicted in our in silico analysis, significantly different DFP was not observed WZ811 treated fish. Contrastingly, despite low vina score prediction, relative PGC mismigration remained unaffected in AMD3100 treated fish across various doses. Additionally, about LY2510924 treatment groups, a dose dependent escalation in DFP, except at 1 × 10⁵nM, was observed.

Microinjection method ensures direct delivery into the eggs, but significantly damages the egg membranes, resulting in poor survival rates. It also demands a great deal of technician skill and time, greatly limiting the number of eggs that can be processed. Therefore, a simpler and more practical method is needed for mass scale treatment in actual aquaculture. We employed immersion treatment using AMD3465, WZ811 and LY2510924 only. Similar to direct delivery microinjection experiment, no significant decrease in hatchability (Figure 7) and in any of the antagonist groups was observed. However, an inverse dose dependent trend was observed in WZ811 treatment group due to the different amounts of solvents used. As seen with microinjection, at 48 hpf PGCs were largely dispersed (Figure 8) in antagonist treated individuals with highest DFP (Figure 9) average of 0.71 mm followed by 0.62 mm and 0.50 mm in 50 µM-WZ811, 50 μM-AMD3465 and 10 μM-AMD3465 treatment groups, respectively. On the other hand, the LY2510924 group showed a reduced DFP. Overall, unlike microinjection, DFP analysis showed no significant differences among all immersion treatment groups.