C57BL/6J WT female mice (7–9 weeks old) were purchased from Jackson Laboratories (Bar Harbor, ME, United States) and housed in standard cages on a 12-h light–dark cycle. All experimental procedures were performed in compliance with the animal protocols approved by the IACUC at Stanford University School of Medicine.

The detailed procedure of AAV production has been described previously (Yang et al., 2014, 2016; Miao et al., 2016; Wang et al., 2020). Briefly, the AAV titers were determined by real-time PCR and diluted to 1.5 × 10¹² vector genome (vg)/ml. For intravitreal injection, mice were anesthetized by xylazine and ketamine based on their body weight (0.01 mg xylazine/g + 0.08 mg ketamine/g). A pulled and polished microcapillary needle was inserted into the peripheral retina just behind the ora serrata. Approximately 2 μl of the vitreous was removed to allow injection of 2 μl of AAV into the vitreous chamber to achieve 3 × 10⁹ vg/retina.

Two weeks after AAV intravitreal injection, EAE/optic neuritis was induced with myelin oligodendrocyte glycoprotein (MOG_33–55, GenScript, Cas No.163913-87-9) peptide and pertussis toxin (PTX, List Biological Laboratories, Cambell, CA, United States; Cat. No.181) in 9-week-old female mice with a procedure modified from the established protocol (Hasselmann et al., 2017; Huang et al., 2017). Briefly, 225 μg of MOG_33–55 peptide emulsified with incomplete Freund’s adjuvant (IFA, BD Difco^TM, 263910) and 2.5 mg/ml mycobacterium tuberculosis (BD Difco^TM, 231141) were injected subcutaneously for immunization twice at day 0 and day 7, and 500 ng of PTX in PBS (2.5 μg/ml) was intraperitoneally injected twice at day 0 and day 2 to break down the blood–brain barrier. Mice injected with the same volume of Freund’s adjuvant emulsion without MOG_33–55 were used as sham controls. The behavioral deficits of these mice were assessed daily with a five-point scale: no disease = 0; partial tail paralysis = 0.5; tail paralysis or waddling gait = 1.0; partial tail paralysis and waddling gait = 1.5; tail paralysis and waddling gait = 2.0; partial limb paralysis = 2.5; paralysis of one limb = 3.0; paralysis of one limb and partial paralysis of another = 3.5; paralysis of two limbs = 4.0; moribund state = 4.5; and death = 5.0.

The detailed procedures have been published before (Huang et al., 2017; Zhang et al., 2019; Li et al., 2020; Wang et al., 2020; Fang et al., 2021). Briefly, after perfusion fixation with 4% PFA in PBS, mice eyeballs and ONs were dissected out and post-fixed with 4% PFA for 2 h at room temperature. Sucrose (30%) was then used for cryoprotection of the tissues. Retinas were dissected out for wholemount retina immunostaining. For cross sections of retina or longitudinal sections of ONs, the eyeballs or ONs were embedded in tissue-tek OCT on dry ice for subsequent cryo-section with a Leica cryostat. The primary antibodies used for immunostaining were: anti-RBPMS (custom made at ProSci Inc.), anti-NMNAT2 (Santa Cruz, Cat. No. sc-515206), anti-HA (Roche, 11867423001), anti-Tuj1 (Biolegend, 845502), anti-MBP (Covance, 808401), anti-CD3 (Santa Cruz Biotechnology, sc-18843), and anti-Iba1 (Wako, 019-19741). For quantification of cells in ONs, three to four fields were randomly sampled from each ON by using a Zeiss confocal microscope with × 40 oil immersion objective lens. Positive cells were counted by NIH ImageJ’s cell counter plugin. For quantification of MBP⁺ area in ONs, we calculated the percentage of the MBP⁺ area against the total area. Positive MBP area was identified based on a fluorescence intensity greater than the baseline intensity threshold and the percentage of positive area was measured by NIH ImageJ.

Mice were sacrificed by cervical dislocation. ONs were immediately dissected out and rinsed in pre-cooled PBS. Then 80 μl of RIPA buffer (Thermo Fisher Scientific, 89900) with protease and phosphatase inhibitor (Thermo Fisher Scientific, 78441) was added to each sample and incubated on ice for 30 min after homogenization and cleared by centrifugation at 20,000 g for 15 min. The protein concentration of the tissue lysate was measured with the BCA protein assay kit (Thermo Fisher Scientific, 23227) following the protocol of the manufacturer. Then 30 μg of protein from each sample was loaded for Western blotting analysis. The primary antibodies used were anti-HA (Roche, 11867423001) and anti-GAPDH (Cell Signaling, Cat. No. 2118).

For RGC counting, the wholemount retinas were imaged with the × 20 objective lens of a Keyence fluorescence microscope tiled by 13 × 13 to cover the entire area of the retina. Concentric circle plugin of NIH Image J was used to divide each retina to eight parts by eight circles in order to define the peripheral area (Supplementary Figure 1E). Ten percent of the peripheral area was stereologically sampled with 100 × 100μm counting frames. All surviving RGCs in the sampled areas were manually identified and counted by ImageJ’s cell counter plugin. The investigators who counted the RGCs were masked to the treatment of the samples.

ONs were post-fixed in situ with 2% glutaraldehyde and 2% PFA in 0.1 M PBS. Semi-thin (1 μm) cross sections of the ON 2 mm distal to the eye (globe) were collected. After PPD staining, the semi-thin sections of the ON were imaged through a × 100 oil immersion objective of a Keyence bright field microscope tiled by 13 × 13 to cover the entire area of the ON. The stereological sub-sampling plugin AxonCounter of NIH ImageJ was used to sample 5–10% of the entire ON area with 10 × 10 μm counting frames (Koschade et al., 2019). All surviving axons in the sampled areas were manually identified and counted by using ImageJ’s cell counter plugin. The investigators who counted the axons were masked to the treatment of the samples.

To prepare ultrathin sections of ON for TEM, one-micron-epoxy sections were examined under light microscope to determine the precise location to cut ultrathin sections. Ultrathin sections (70 nm) were collected onto formvar-coated copper grids and dried overnight. Sections were stained next day with uranyl acetate for 30 min, washed in PBS, and then stained with lead citrate for 7 min. Sections were again washed and dried before observing under TEM. TEM imaging was performed at the Cell Sciences Imaging Facility at the Beckman Center, Stanford University. The cross sections of the entire ON were examined and imaged randomly without overlap at × 4,000 with 11.6 μm × 11.6 μm frames on a JEOL JEM-1400 TEM microscope (JEOL United States, Inc., Peabody, MA, United States). For each ON, 25–45 images were taken to cover the whole area of the ON. Both myelinated and unmyelinated axons were counted for axon density calculation.

The NAD⁺ levels of the ONs were measured according to the protocol of the manufacturer protocol of the NAD⁺/NADH assay kit (Abnova, KA1657). Two weeks after intravitreal injection of AAV2-mSncg-3HA-NMNAT2Δex6 in the left eyes and control AAV2 in the right eyes, mice were sacrificed by cervical dislocation and the ONs were collected gently and quickly. Six mice were used in total, two optic nerves were pooled together as one measure, thus three measures were performed for NMNAT2 overexpression group and control group, respectively. Samples were homogenized in the NAD⁺ extraction buffer and then heated at 60°C for 5 min. The homogenate added with the assay buffer was centrifuged at 14,000 rpm for 5 min to remove cellular debris. After adding reagents to 40 μl of supernatants and standard solutions, the absorbance was determined at 565 nm by TECAN SPARK plate reader (Tecan, Switzerland). The results were normalized to one microgram of protein concentration.

The detailed procedure has been published previously (Zhang et al., 2019; Li et al., 2020; Fang et al., 2021). Briefly, after anesthetization and pupil dilation, the retina fundus images were captured with the Heidelberg Spectralis SLO/OCT system (Heidelberg Engineering, Germany). The mouse retina was scanned with the ring scan mode centered by the optic nerve head under high-resolution mode (each B-scan consisted of 1,536 A scans). The ganglion cell complex (GCC) includes the retinal nerve fiber layer (RNFL), ganglion cell layer (GCL), and inner plexiform layer (IPL). The average thickness of the GCC around the optic nerve head was measured manually with the aid of Heidelberg software. The investigators who measured the thickness of GCC were masked to the treatment of the samples.

The detailed procedure has been published previously (Zhang et al., 2019; Li et al., 2020; Fang et al., 2021). Briefly, after anesthetization and pupil dilation, PERG of both eyes was recorded simultaneously with the Miami PERG system (Intelligent Hearing Systems, Miami, FL, United States) according to the instructions of the manufacturer. Two consecutive recordings of 200 traces were averaged to achieve one readout; each trace recorded up to 1,020 ms. The first positive peak in the waveform was designated as P1 and the second negative peak as N2. The amplitude was measured from P1 to N2.

The detailed procedure has been published previously (Zhang et al., 2019). Briefly, mice were placed on a platform in the center of four 17-inch LCD computer monitors (Dell, Phoenix, AZ, United States), with a video camera above the platform to capture the movement of the mouse. A rotating cylinder with vertical sine wave grating was computed and projected to the four monitors by OptoMotry software (Cerebral Mechanics Inc., Lethbridge, Alberta, Canada). The sine wave grating, settled at 100% contrast and speed of 12°/s, provides a virtual-reality environment to measure the spatial acuity (cycle/degree) of the left eye when rotated clockwise and the right eye when rotated counterclockwise. The maximum frequency (cycle/degree) that the mouse could perform tracking was identified and recorded by investigators masked to the treatment.

GraphPad Prism 7 was used to generate graphs and used for statistical analyses. Data are presented as means ± SEM. Student’s t-test was used for two groups comparison and one-way ANOVA with post hoc test was used for multiple comparisons.

To determine whether the neuronal autonomous effect of NMNAT2 provides RGC and ON protection in EAE/optic neuritis, we first confirmed that NMNAT2 can be expressed in mouse RGCs in vivo (Figure 1). NMNAT2Δex6 (soluble forms of NMNAT2, lacking of exon6) is more stable and has greater axon protective capacity than wild-type NMNAT2 (Milde et al., 2013b). Therefore, we drove HA-tagged NMNAT2Δex6 with the RGC-specific mSncg promoter (Wang et al., 2020) in AAV2 vector (Figure 1A). Intravitreal AAV injection achieved RGC-specific expression of HA-NMNAT2Δex6 in the RGC somata in the retina and axons in ON, examined at 2 weeks after AAV injection (Figures 1B–D). Consistent with our previous report (Wang et al., 2020), about 86% RGCs expressed HA-NMNAT2 (Figures 1B,E). Importantly, biochemical assays confirmed the overexpression of 3HA-NMNAT2Δex6 and increase of the NAD⁺ level in the ON 2 weeks after the AAV injection (Figures 1F,G). In summary, we established RGC-specific upregulation of NMNAT2, which enabled us to evaluate the autonomous effect of neuronal NMNAT2 modulation on EAE/optic neuritis neuroprotection.

A previous study reported that the exogenous NMNAT2 is toxic to primary cortical neurons (Mayer et al., 2010), and a recent study showed that overexpression of NMNAT2 in the CA1 area in vivo enhanced seizure susceptibility and caused neuronal loss (Wu et al., 2021). Therefore, we next investigated the effect of NMNAT2Δex6 overexpression on the viability and visual function of the naïve RGCs. To determine the visual function of living animals, we measured the general electrophysiological activities of the RGCs in response to pattern visual stimuli by PERG (Chou et al., 2014; Porciatti, 2015) and visual acuity by the OKR (Prusky et al., 2004; Douglas et al., 2005). Both procedures are established in our lab (Zhang et al., 2019; Li et al., 2020; Fang et al., 2021). Compared with the contralateral control eyes injected with AAV2 vectors expressing AAV2 capsid itself, there were no significant changes in the P1–N2 amplitude or visual acuity in the NMNAT2-overexpressed eyes 10 weeks after AAV intravitreal injection (Supplementary Figures 1A,B). Consistent with these analyses of the RGC function, there was no significant difference in the RGC morphology between the control and NMNAT2 overexpression eyes: thickness of ganglion cell complex (GCC) in living animals was comparable (Supplementary Figures 1C,D), and histological analysis of post-mortem retina and ON consistently showed no significant RGC somata loss or axon degeneration 10 weeks after AAV injection (Supplementary Figures 1E,F). Therefore, RGC-specific expression of NMNAT2 is non-toxic, at least in the experimental window that we tested.

Two weeks after intravitreal injection of AAVs, we induced EAE in female mice with myelin oligodendrocyte glycoprotein (MOG_33–55) peptide and pertussis toxin (PTX) according to the established protocol (Hasselmann et al., 2017; Huang et al., 2017). We performed histological studies of the ON inflammation and demyelination at 3 weeks post MOG immunization (3 wpi) (Figure 2A). Symptoms of weakness began to develop around day 11 after MOG immunization, peaked around day 19, and plateaued for the next several weeks (Figure 2B). The sham group of mice injected with the same reagent mixture except replacing MOG_33–55 with PBS showed no signs of neurological disease during the experimental course. ON demyelination and optic neuritis were obvious at 3 wpi, evidenced by significant loss of myelin basic protein (MBP) (Figure 2C) and infiltration of inflammatory T-cells and macrophage (Figure 2D). These findings in the EAE mice, but not in the sham mice, were consistent with the time course of EAE/optic neuritis observed in our previous study of this model (Huang et al., 2017). We found no immune cell infiltration in the retinas of these mice (Supplementary Figure 2). Importantly, RGC-specific NMNAT2 expression did not affect the demyelination or inflammation in the ONs of the EAE mice (Figures 2C,D).

After confirming the RGC-specific expression of NMNAT2, we investigated its effect on the retina morphology and visual function in mice with EAE/optic neuritis. The GCC thickness measured by OCT in living animals was relatively normal at 3 wpi but decreased progressively at 5 and 8 wpi (Figures 3A,B), consistent with the previous time course of RGC degeneration described previously in EAE mice (Huang et al., 2017). However, there was no significant difference between eyes with NMNAT2 overexpression and contralateral control eyes injected with control AAVs (Figures 3A,B), suggesting no neuroprotection by NMNAT2 overexpression. PERG and OKR measured at different time points after immunization also revealed no significant difference in the P1–N2 amplitude (PERG) or visual acuity (OKR) between NMNAT2 overexpressing eyes and contralateral control eyes, despite the obvious visual deficits in the EAE mice at 3–8 wpi (Figures 3C,D). We sacrificed half of the animals at 5 wpi for histological studies and saved the other animals until 8 wpi. We want to note that the severe paralysis of the EAE mice at 8 wpi precludes reliable measurement with OKR, which relies on body movement in response to visual stimuli. At 5 wpi, the mean visual acuity of NMNAT2 overexpressing eyes (0.024 ± 0.05) was even lower than that of contralateral control eyes (0.118 ± 0.17); however, the difference was not statistically significant (p-value, 0.098). Taken together, the RGC-specific NMNAT2Δex6 overexpression did not protect visual function in the EAE mice.

Quantification of the surviving RGC somata in wholemount retinas (Figures 4A,B) and axons in the ON semi-thin and ultrathin cross-sections images revealed significant RGC soma and axon loss at 5 and 8 wpi (Figures 4C–F). However, RGC-specific NMNAT2Δex6 expression alleviated neither RGC soma nor axon degeneration in mice with EAE/optic neuritis. Taken together, the present results showed that NMNAT2 overexpression in RGCs does not achieve neuroprotection or preserve visual function in EAE/optic neuritis.