All animal protocols were approved by the Animal Care Committee of Wenzhou Medical University. The YFPH line of transgenic mice was obtained from the Jackson Laboratory (strain: B6. Cg-Tg (Thy1-YFPH) 2Jrs/J; Bar Harbor, Maine). All mice were given ad libitum access to food and water under a 12 h light/dark cycle with 50–60% humidity. The day of birth was counted as postnatal day 0 (P0).

The littermates of the Thy1-YFPH mice were randomly divided into two groups. Pups received intraperitoneal (IP) injections of the A_2AR antagonist KW6002 (10 mg/kg body weight, freshly prepared in dimethyl sulfoxide (DMSO, Sigma), ethoxylated castor oil (Sigma), and phosphate-buffered saline (PBS) with a proportion of 15%:15%:70% (Miao et al., 2021)) every day from P4 to P6. The control group was administered the corresponding vehicle in the same volume. The neonatal inflammation was induced in Thy1-YFPH mice by an intraperitoneal injection of lipopolysaccharide (LPS, 1 mg/kg, E. coli 055: B5; Sigma) 4 min after KW6002 treatment at P4.

Immunohistochemistry experiments were carried out as previously described (Gao et al., 2018). Briefly, after the Thy1-YFPH mice were anesthetized, the eyes were enucleated on P21 and fixed in 4% paraformaldehyde (PFA) for 30 min. The retinas were isolated from eyeballs, fixed in 4% PFA for additional 10 min, and incubated with 3% H₂O₂ for 20 min. After being blocked in a blocking solution (5% normal donkey serum plus 1% BSA, 0.2% glycine, 0.2% lysine, and 0.3% Triton X-100) for 2 h at room temperature, retinas were incubated with goat polyclonal antibodies against GFP (1:500, NB100-1770, Novus Biologicals) for 2 days at 4°C. Then the retinas were sequentially incubated with the biotinylated donkey anti-goat antibodies, the avidin-biotin complex (Vectastain ABC Elite Kit; Vector Laboratories, USA), the 3,3′-diaminobenzidine (DAB tablets, Sigma), and finally flat-mounted on glass slides with aqueous mounting medium (IMMCO Diagnostics, Inc).

The fully and strongly stained YFP-positive RGCs (that achieve the standard for the detailed morphological analyses) were randomly selected and reconstructed, while those with faint staining and uncompleted structure were discarded. The morphology of RGCs was reconstructed by using Neurolucida system (MicroBrightField Inc., USA) and a bright-field light microscope (Zeiss, Germany) at a magnification of ×63, as previously described (Gao et al., 2018). A battery of morphological parameters were extracted from 251 fully reconstructed RGCs by the NeuroExplorer (MicroBrightField Inc., USA). Sholl analysis on dendrites of RGCs was also performed using “Sholl Analysis” in Neuroexplorer (Sholl, 1953). The spatial distributions of dendritic intersection with the concentric circles were quantified in terms of stepped distance circle regions (10 μm) from the soma.

Results were expressed as mean ± standard error of mean (SEM). Kruskal-Wallis one-way ANOVA (k samples), independent Student’s t-test and chi-square test were performed by SPSS 26. The significant level was set at p < 0.05 for all comparisons.

Since the RGCs are rather heterogeneous in the retina, we take advantage of the Thy-1 YFPH transgenic line, which expresses the yellow fluorescent protein (YFP) only in a fraction of RGCs (Barnstable and Dräger, 1984; Feng et al., 2000), to study the effect of A_2AR on the development of RGCs. During retinal development, Thy-1 YFPH neonates received intraperitoneal injections of the A_2AR antagonist KW6002 from P4 to P6 and were sacrificed at P21 ( Figure 1A). 3D reconstruction of well-stained YFP⁺ cells (n = 120 cells) for detailed morphometric analyses was performed in the retina at P21 ( Figure 1B). As described previously, we classified these Thy1-positive RGCs labeled with YFP into three major morphological classes (Type I, II and III), based on the morphological features of the dendritic field and dendrite density (Gao et al., 2018). Type I had a small dendritic field area and high dendrite density, whereas Type III had a large dendritic field area but low dendrite density. Type II was just between Type I and III ( Figure 1C). The quantitative analysis further confirmed the significant differences among the three RGC types (**p < 0.01, ***p < 0.001, Figures 1D,E). Compositions of the three RGC types were similar between the KW6002-treated and control groups (Type I: control, n = 12, 22.22% vs. KW6002, n = 22, 33.33%; Type II: control, n = 26, 48.15% vs. KW6002, n = 30, 45.45%; Type III: control, n = 16, 29.63% vs. KW6002, n = 14, 21.21%; p > 0.05; Figure 1F). Therefore, KW6002 had no effect on the proportion of these RGC morphological types during retinal development.

The RGC somata have different shapes, such as triangular, round, and oval. RGCs with different morphological shapes are used to the different parameters. We firstly studied the somatic development of RGCs and found that KW6002 significantly increased the soma area by 18.69% in Type I (control, 244.04 ± 15.10 μm² vs. KW6002, 289.66 ± 13.04 μm²; *p < 0.05; Figures 2A,B and Supplementary Table S1), while no significant effect was found on Type II and III (p > 0.05; Figures 2A,B). Meanwhile, no significant difference was found in the soma perimeter of the three morphological types of RGCs by KW6002 (p > 0.05; Figures 2A,C). These results indicate that A_2ARs can differentially modulate the somatic development of RGCs during retinal development.

We further compared the morphological features of dendrites, such as dendritic field area, dendrite density, segment number, length, surface area and volume etc, between the two groups (Figure 3). The dendritic field area was diminished by 10.13% and 10.99%, respectively, in Type II (control, 78,610.04 ± 2295.94 μm² vs. KW6002, 70,646.83 ± 2472.02 μm²; *p < 0.05; Figures 3A,B) and Type III RGCs (control, 123,035.42 ± 4269.09 μm² vs. KW6002, 109,513.94 ± 2438.7 μm²; **p < 0.01; Figures 3A,B) after KW6002 treatment, whereas no significant effect was found in Type I (p > 0.05; Figure 3B). Compared to the control group, KW6002 attenuated the dendrite density by 12.00% (control, 9.50 ± 0.48 (×10^–2, 1/µm) vs. KW6002, 8.36 ± 0.28 (×10^–2, 1/µm); *p < 0.05; Figure 3C) in Type I RGCs, but induced 19.62% enhancement of the dendrite density (control, 3.72 ± 0.14 (×10^–2, 1/µm) vs. KW6002, 4.45 ± 0.24 (×10^–2, 1/µm); *p < 0.05; Figure 3C) in Type III RGCs. KW6002 didn’t change the dendrite segment number and total dendrite length of all three RGC types during normal development (p > 0.05; Figures 3D,E). As to the dendrite surface area and volume, no significant effect was found in each RGC type after KW6002 treatment during normal development (p > 0.05; Figures 3F,G). These results implied that A_2ARs can reorganize the dendritic architecture of RGCs in the retina, which is dependent on the RGC types.

To further investigate the effect of KW6002 on the spatial distribution of dendritic morphology, Sholl analyses were performed on quantifications of the distribution of dendritic intersections and revealed that KW6002 had a dual effect on Type I and III RGCs. KW6002 significantly decreased the dendritic intersection of Type I RGCs at 30–50 μm, but increased them at 110–160 μm from the soma (*p < 0.05 or **p < 0.01; Figure 4A) On the contrary, as to the Type III RGCs, the dendritic intersections were significantly increased at 50–70 μm and 100 μm but decreased at 180–200 μm from the soma (*p < 0.05 or **p < 0.01; Figure 4C) after KW6002 treatment. Meanwhile, KW6002 significantly decreased the dendritic intersection of Type II RGCs mainly at 120–170 μm, which is far from the soma (*p < 0.05 or **p < 0.01; Figure 4B). These results indicate the fine-tune effect of A_2ARs on the dendritic development of RGCs during the retinal development.

Apart from its physiological role, we further studied the effects of A_2AR on the development of RGCs after the neonatal inflammation. To induce the neonatal inflammation, neonates received a single intraperitoneal injection of LPS immediately after KW6002 treatment at P4. Then they were administrated with KW6002 in the same manner as that in the normal condition and were sacrificed at P21 (Figure 5A). The well-stained YFP⁺ cells (n = 131 cells) in the retina were 3D reconstructed for detailed morphometric analyses at P21. We found that after neonatal LPS exposure, the compositions of Type I and Type II RGCs significantly increased, but Type III decreased in the KW6002-treated groups, compared to the control groups (Type I: control, n = 9, 14.29% vs. KW6002, n = 16, 23.53%; Type II: control, n = 37, 58.73% vs. KW6002, n = 45, 66.18%; Type III: control, n = 17, 26.98% vs. KW6002, n = 7, 10.29%; *p < 0.05; Figure 5B). These results indicate that A_2ARs altered the compositions of the three RGC types after neonatal inflammation.

After LPS treatment, KW6002 significantly enlarged the soma area by 19.10% (control, 303.54 ± 16.96 μm² vs. KW6002, 361.51 ± 18.69 μm²; *p < 0.05; Figures 6A,B and Supplementary Table S2) and the soma perimeter by 10.73% (control, 69.64 ± 1.79 μm vs. KW6002, 77.11 ± 2.01 μm; **p < 0.01; Figure 6C) in Type II. However, KW6002 had no significant effect on the soma area and perimeter of Type I and III after neonatal LPS exposure(p > 0.05; Figures 6B,C). Beside the soma, we also examined the dendritic morphology and found that the dendritic field area and total dendrite length of the three morphological types of RGCs were not affected by KW6002 after neonatal LPS exposure (p > 0.05; Figures 7A, B,E). Interestingly, KW6002 significantly augmented the dendrite density (control, 5.73 ± 0.17 (×10^–2, 1/µm) vs. KW6002, 6.32 ± 0.19 (×10^–2, 1/µm); *p < 0.05; Figure 7C) and segment number (control, 79.16 ± 3.50 vs. KW6002, 92.29 ± 4.03; *p < 0.05; Figure 7D) of Type II RGCs, but didn’t affect those of Type I and Type III RGCs after neonatal LPS exposure. Moreover, KW6002 significantly enhanced the dendrite surface area and volume of Type I RGCs (control, 3861.30 ± 227.05 μm² vs. KW6002, 4700.80 ± 283.86 μm²; *p < 0.05; Figure 7F; control, 474.82 ± 31.88 μm³ vs. KW6002, 629.75 ± 40.26 μm³; **p < 0.01; Figure 7G) after neonatal exposure to LPS, while no significant change was found in the other two RGC types. These results suggest that A_2ARs induced differential alterations in the soma and dendritic architecture of RGCs after neonatal inflammation.

We further performed Sholl analyses to study the effect of KW6002 on the spatial distribution of dendritic morphology after neonatal LPS exposure. We found that KW6002 only significantly decline dendritic intersections at 110 and 130 μm from the soma in Type III RGCs (*p < 0.05 or **p < 0.01; Figure 8C), while no significant difference was found either in Type I or Type II RGCs (p > 0.05; Figures 8A,B). These results suggest that A_2ARs only had slight effect on dendritic complexity after neonatal LPS exposure.