This study was approved by the Animal Care Committee and the Ethics committee of Chang Gung Memorial Hospital in Kaohsiung. Ten to twelve Sprague-Dawley rat pups per dam were used and housed with a 12/12 h light/dark schedule in a temperature- and humidity-controlled colony room. The pups were housed with their dams until weaning at postnatal (P) day 21 and then housed in groups of 4–5 per cage.

At P7, the animals were anesthetized with 2.5% halothane (balance, room air), and the right common carotid artery was surgically exposed and permanently ligated. After surgery, the pups were returned to the dam for 1 h, then placed in air-tight containers through which humidified 3 L/min 8% oxygen (balance, nitrogen) was maintained for 2 h (Huang et al., 2012). The sham controls underwent anesthesia and surgical exposure but did not receive artery ligation and were not placed in a hypoxic chamber.

Two hours before and 18 h after the induction of HI, the rat pups were injected intraperitoneally with either DHF (5 mg/kg; Tokyo Chemical Industry Co., Tokyo, Japan) or dimethyl sulfoxide (DMSO; 10%; Sigma-Aldrich Corp., St. Louis, MO, United States).

The Rat Neurogenesis RT² Profiler ^TM PCR Array (Qiagen, Catalog # PARN-404Z, MD, United States), which consisted of primers for 84 genes related to neurogenesis and neural stem cells was used for the gene expression analysis. The rats treated with either DHF or DMSO were sacrificed, and total RNA was prepared from the retinas at P10 and immediately frozen at −70°C. Aliquots of 1 μg RNA per retina were reverse-transcribed using a RT² First Strand Kit (Qiagen). The complementary DNA (cDNA) was mixed with SYBR Green (Qiagen) into the array plates, and cycling was performed according to the manufacturer’s protocol. The data obtained from the array were normalized using multiple housekeeping genes and analyzed by comparing 2^–ΔCt of the normalized data. Fold changes were calculated relative to the retinal extracts from HI animals treated with DHF and DMSO. The results were confirmed using a quantitative real-time reverse transcriptase polymerase chain reaction analysis on the individual samples for genes that showed the strongest upregulation and downregulation.

Retinas were dissected and ground with a mortar and pestle in liquid nitrogen under RNase-free conditions. Total RNA was extracted using TRIzol reagent (Invitrogen, Carlsbad, CA, United States). Aliquots of 5 μg total RNA were reverse-transcribed to cDNA using SuperScript III Reverse Transcriptase (Invitrogen). The cDNA was amplified by PCR using the following gene-specific primers: ARTN, 5′-CAGAGCCTGGAAAGATGACC-3′ (forward) and 5′-AGAGCTGGGATCCATGAACA-3′ (reverse); and glyceraldehyde 3-phos-phate dehydrogenase (GAPDH), 5′-TCTTGTGCAGTGCCAGCCTC-3′ (forward) and 5′-GTCACAAGAGAAGGCAGCCCTGG-3′ (reverse). The template was amplified at 95°C for 5 min, followed by 45 cycles of PCR at 95°C for 10 s, 60°C for 20 s, and 72°C for 20 s using the LightCycler^® 480 SYBR Green I Master (Roche, Indianapolis, IN, United States) and LightCycler^® 480 instrument (Roche) for analyzing the ARTN and GAPDH. The C _t assigned as the beginning of the logarithmic amplification was computed using the equipment’s software program (Roche). The relative expression level was defined as 2^–ΔCt, where ΔC_t = C_{t target gene} – C_{t β –actin}. The fold changes in mRNA expression were defined as 2^–ΔΔCt, where ΔΔC_t = ΔC_{t treatment} – ΔC_{t vehicle}.

An evaluation of whether the protective effect of DHF comes from upregulating endogenous ARTN, intravitreal injection of either the ARTN-neutralizing antibody (ARTN Ab, 1 μg; R&D systems, Minneapolis, MN, United States) (DeBerry et al., 2015), or phosphate buffered saline (PBS) was performed at post-HI 24 h, which was 6 h after the DHF treatment. To assess the effect of exogenous ARTN in HI injured retinas, either ARTN (1 μg, Peprotech, Rocky Hill, NJ, United States) (Omodaka et al., 2014) or H₂O was administered by intravitreal injection at post-HI 24 h. The animals received intraperitoneal injection of 5-bromo-2′-deoxyuridine (BrdU; 100 mg/kg; Sigma-Aldrich Corp.) for three consecutive days from P8 to P10 for the purpose of identifying cell proliferation.

At P22 and P29, full-field scotopic flash electroretinograms (ERGs) (RETIport ERG; Roland Consult, Brandenburg, Germany), were recorded from both eyes of the rat pups, as previously described (Huang et al., 2012). Briefly, the pupils were topically dilated with 1% Tropicamide (Mydriacyl, Alcon, Puurs, Belgium) and 1% cyclopentolate (Cyclogyl, Alcon, Puurs, Belgium), then the eyes were dark-adapted for 1 h before performing ERG. The animals were sedated using intramuscular injections of a mixture of Rompun (10 mg/kg; Bayer Korea, Seoul, South Korea) and Zoletil-50 (25 mg/kg; Virbac, Carros, France); then, a standard white flash on a dark background scotopic 0-dB ERG was recorded. The stimulus luminance was 3 cds/m² with a duration of 10 ms. Responses from 20 identical flashes applied at 10-s intervals were averaged (Huang et al., 2015).

Paraffin sections of the retina were dewaxed, hydrated through graded concentrations of alcohol, and placed in phosphate-buffered saline. Cryosections were prepared after fixation in 4% paraformaldehyde and dehydration in a sucrose gradient. Eye blocks were cut at 10 μm. Two sections per retina were randomly selected for hematoxylin and eosin staining. Images were acquired using a light microscope (Nikon, Tokyo, Japan). Retinal damage was quantified within central retinal areas (100–200 μm from the optic disk), as described in our prior study (Huang et al., 2012) by assigning different grades: grade 0, preserved retinal ganglion cell (RGC) and all retinal layers comparable to the sham control; grade 1, moderate decrease in RGC counts and thickness of the inner plexiform layer (IPL); grade 2, complete loss of RGCs and IPL (Supplementary Figure 1).

After antigen unmasking and blocking of nonspecific sites, the sections were incubated overnight at 4°C with primary antibodies against ARTN (1:10; R&D systems), phosphorylated (p)RET (1:10; Abcam, Cambridge, United Kingdom), ED1 (1:100; Biosource, Camarillo, CA, United States), antiglial fibrillary acidic protein (GFAP; 1:200; Millipore, Temecula, CA, United States), and BrdU (1:100; Novocastra, Newcastle upon Tyne, United Kingdom) and then subsequently incubated with secondary antibodies for 60 min at room temperature. The immunoreactivity of ARTN was evaluated at a 200× magnification by calculating the integrated optical density (IOD) with ImagePro Plus 6.0 software (Huang et al., 2018). The number of ED1⁺ and Brdu⁺ cells were counted in an area of 400 × 100 μm at 200× magnification. The ameboid ED1⁺ cells were defined as reactive microglial cells. GFAP immunoreactivity was quantified by assigning different grades: grade 1, immunoreactivity in the nerve fiber layer (NFL) and around vessels; grade 2, immunoreactivity in the NFL in an outward tentacle-like pattern, extending toward the inner nuclear layer (INL); grade 3, showing occasional and grade 4, showing extensive GFAP immunoreactivity extending from the NFL to the outer nuclear layer (ONL; Huang et al., 2018).

Retinas were homogenized, and 40 μg samples were resolved using a 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis, after which they were blotted to nitrocellulose membranes. Membranes were blocked with 5% non-fat dry milk, incubated with primary antibodies and horseradish-conjugated secondary antibodies, and the signal was visualized with enhanced chemiluminescence. The following primary antibodies were used: anti-ERK (1:10,000; Cell Signaling Technology, Danvers, MA, United States), anti-pERK (1:2000; Millipore), anti- c-Jun N-terminal kinase (JNK, 1:2,000; Cell Signaling Technology), anti-pJNK (1:1,000; Cell Signaling Technology), anti-p38 (1:5,000; Abcam), anti-pp38 (1:10,000; Abcam), anti-Akt (1:10,000; Cell Signaling Technology), and anti-pAkt (1:1,000; Cell Signaling Technology). After the densitometric analysis, data were normalized against GAPDH (Millipore), and the ratio of protein expression in the treated eyes to the sham controls was calculated.

Statistical analyses were performed using a 1-way ANOVA or a Kruskal–Wallis test using GraphPad Prism 4 software (GraphPad, San Diego, CA, United States). Data were presented as mean ± standard error, and p values of <0.05 were considered statistically significant.

Previously we showed that systemic DHF treatment was able to protect the immature retina against HI injury (Huang et al., 2018). To identify the mediators for the HI protective effects of DHF treatment in the immature retina, a PCR array was performed at P10, and it was found that the expression of ARTN gene in the DHF-treated HI group was 2 times higher than that in the DMSO-treated HI group. In contrast, other neurotrophic factors, including glial cell line-derived neurotrophic factor (GDNF) and BDNF, were not significantly elevated in the DHF-treated HI group (Supplementary Table). Using a RT-PCR, we confirmed that the ARTN mRNA levels in the DHF-treated HI group were significantly higher than those in the DMSO-treated HI group and the sham controls 24 h after HI (Figure 1A). The immunohistochemical stain showed that ARTN protein was expressed in the RGC, IPL, and INL of the sham controls and the DHF-treated HI group at P8 but was decreased in the DMSO-treated HI group (Figure 1B and Supplementary Figure 2). The ARTN immunoreactivity was significantly lower in the INL of the DMSO-treated HI group compared to the DHF-treated HI group and the sham group at P10 (p < 0.05; Figure 1C).

After HI injury, endogenous ARTN was sequestered using intravitreal injections of artemin-neutralizing antibody (ARTN Ab) to determine whether the neuroprotective effects of DHF in the immature retina were mediated by artemin (Figure 2A). ERG performed at P22 and P29 showed that the a-wave (associated with rod photoreceptor activity) and b-wave (Müller glial and bipolar cells activity) amplitudes in the DHF-PBS group were relatively preserved, while in the DMSO-treated group and the DHF-ARTN Ab group, the amplitudes of the a-wave and b-wave were markedly decreased (Figure 2B). Group data showed that the b-wave amplitude in the DHF-ARTN Ab group was significantly lower (p < 0.05) than that in the DHF-PBS group but was similar to that in the DMSO-treated group (Figure 2B). There were no significant among-group differences in the a-wave amplitudes of the DHF-PBS, DHF-ARTN Ab, and DMSO-treated HI groups (data not shown).

Compatible with the functional alterations, there was almost a complete loss of the inner retinal layers, including the RGC, IPL, and INL, in addition to a partial loss of the outer retina layers, including the outer plexiform layer (OPL) and the ONL in the DMSO-treated HI group at P16 and P29 (Figure 2C). Although in general thinner than the controls, the retinal layers in the DHF-PBS group were relatively preserved after HI. In contrast, the DHF-ARTN Ab group showed severe inner retinal damages after HI, especially at P29. Semiquantitative data showed that there were no significant between-group differences in the severity of retinal damage in the DMSO-, DHF-PBS- and DHF-ARTN Ab-treated HI groups at P16. However, at P29, the DMSO- and the DHF-ARTN Ab-treated HI groups showed significantly more severe retinal damage than the DHF-PBS group (p < 0.05; Figure 2C). There were no significant between-group differences in the DMSO- and DHF-ARTN Ab-treated HI groups. These data suggest that the neuroprotective effects of DHF are largely mediated by ARTN and that ARTN was involved in the long-term neuroprotective effects of DHF against HI retinal injury.

In our previous study, we also showed that the long-term neuroprotective effects of DHF were related to increased neurogenesis and decreased astrogliosis (Huang et al., 2018). In the current study, immunohistochemical staining showed that at P17, both the DHF-PBS and DHF-ARTN Ab groups had significantly increased Brdu⁺ cells in the inner retina as compared to in the DMSO-treated HI groups (p < 0.05; Figure 3A and Supplementary Figure 3A), suggesting that DHF promotes cell proliferation in the inner retina, but its treatment effects are not affected by the blockade ARTN protein.

The GFAP immunoreactivity, indicating the presence of astrogliosis, in the DHF-PBS group was significantly lower than that in the DMSO-treated HI group at P29. However, the DHF-ARTN Ab group had extensive GFAP immunostaining throughout all of the retinal layers, similar to the DMSO-treated HI group, which was higher than the DHF-PBS group (p < 0.05; Figure 3B and Supplementary Figure 3B).

At P17, reactive microglial cells, or active ED1⁺ cells, were present in the inner retina of all of the groups other than the control. However, the number of ED1⁺ cells was similar between the DMSO-treated and the DHF-ARTN Ab HI groups and was significantly higher than the number in the DHF-PBS group (p < 0.05; Figure 3C and Supplementary Figure 3C). These data suggest that the neuroprotective mechanisms of DHF mediated by ARTN involves modulation of astrogliosis and neuroinflammation, but not cell proliferation.

To verify whether ARTN provides protection against HI injury, the rat pups received intravitreal injections of either ARTN or H₂O 24 h after HI injury. The ERG at P22 and P29 showed that the a-wave and b-wave in the ARTN-treated HI group were relatively preserved, with b-wave amplitudes that were significantly higher than those in the H₂O-treated HI group (both p < 0.05; Figure 4A). In line with the functional alterations, despite the presence of damaged inner retinal layers in the ARTN-treated HI group, the damage still occurred to a lesser degree than that in the H₂O-treated HI group and was statistically significant at both P16 and P29 (both p < 0.05; Figure 4B).

Immunohistochemical staining showed that at P17, the number of proliferating or Brdu⁺ cells were not significantly different between the ARTN- and H₂O-treated HI groups (Figure 5A and Supplementary Figure 4A). However, GFAP immunostaining at P29 demonstrated that ARTN treatment significantly decreased astrogliosis after HI (p < 0.05; Figure 5B and Supplementary Figure 4B). Both the ARTN- and H₂O-treated HI groups showed significantly greater neuroinflammation, with increased active ED1⁺ cells, as compared to the sham controls at P10 (p < 0.001; Figure 5C and Supplementary Figure 4C). However, at P17, the number of active ED1⁺ cells in the ARTN-treated HI group was significantly decreased compared to the H₂O-treated HI group (p < 0.05; Figure 5C and Supplementary Figure 4C), which was similar to the sham controls. These data suggest that ARTN protects the immature retina by decreasing astrogliosis and late neuroinflammation after HI.

The ARTN-treated HI group had prominent ARTN immunostaining in the RGC and INL at P10, which indicates that ARTN continued to be expressed in the inner retina two days after the treatment. There were significant differences between the ARTN-treated HI group and the H₂O-treated HI group and the sham controls (p < 0.01; Figure 6A and Supplementary Figure 5A). Immunostaining showed that the ARTN treatment after HI induces significantly increased RET phosphorylation (pRET) immunoreactivity in the RGC, IPL, and OPL (p < 0.05; Figure 6B and Supplementary Figure 5B).

To investigate the downstream signaling pathways of ARTN/RET, western blots were performed at P10. There were no significant differences in the total and phosphorylated p38 or AKT between the ARTN- and H₂O-treated HI groups at P10 (Figures 6C,D). However, the ARTN-treated group exhibited significantly increased phosphorylation of ERK and JNK as compared to the H₂O -treated HI group (both p < 0.05; Figures 6E,F). There were no significant differences in the total ERK and JNK expression between the ARTN- and H₂O-treated HI groups.