Noninvasive Ophthalmic Imaging Measures Retinal Degeneration and Vision Deficits in Ndufs4^−/− Mouse Model of Mitochondrial Complex I Deficiency

Abstract

Purpose: To characterize postnatal ocular pathology in a Ndufs4^−/− mouse model of complex I deficiency using noninvasive retinal imaging and visual testing.

Methods: Ndufs4^−/− mice and wild-type (WT) littermates were analyzed at 3, 5, and 7 weeks postnatal. Retinal morphology was visualized by optical coherence tomography (OCT). OCT images were analyzed for changes in retinal thickness and reflectivity profiles. Visual function was assessed by electroretinogram (ERG) and optomotor reflex (OMR).

Results: Ndufs4^−/− animals have normal OCT morphology at weaning and develop inner plexiform layer atrophy over weeks 5 to 7. Outer retinal layers show hyporeflectivity of the external limiting membrane (ELM) and photoreceptor ellipsoid zone (EZ). Retinal function is impaired at 3 weeks, with profound deficits in b-wave, a-wave, and oscillatory potential amplitudes. The b-wave and oscillatory potential implicit times are delayed, but the a-wave implicit time is unaffected. Ndufs4^−/− animals have normal OMR at 3 weeks and present with increasing acuity and contrast OMR deficits at 5 and 7 weeks. Physiological thinning of inner retinal layers, attenuation of ELM reflectivity, and attenuation of ERG b- and a-wave amplitudes occur in WT C57BL/6 littermates between weeks 3 and 7.

Conclusions: Noninvasive ocular imaging captures early-onset retinal degeneration in Ndufs4^−/− mice and is a tractable approach for investigating retinal pathology subsequent to complex I deficiency.

Translational Relevance: Ophthalmic imaging captures clinically relevant measures of retinal disease in a fast-progressing mouse model of complex I deficiency consistent with human Leigh syndrome.

Introduction

Mitochondrial complex I disorders are a heterogeneous group of rare inherited diseases caused by mutations affecting proteins in the mitochondrial electron transport chain. Mutations disrupting complex I function account for up to 30% of inherited mitochondrial disorders,¹ including >95% of Leber's hereditary optic neuropathy² and >50% of Leigh syndrome cases.³ Disease presentation and symptom severity can vary widely, from lethal neonatal disease (e.g., Leigh syndrome) to aging-related neurodegenerative disorders such as Parkinson's disease^4,5 and amyotrophic lateral sclerosis.⁶ However, neurological and ophthalmic pathologies are common to many forms of complex I deficiency, due to the high energy demands of central nervous system tissues. 

Mammalian complex I (also known as NADH:ubiquinone oxidoreductase) is the first and largest of five mitochondrial complexes in the electron transport chain,⁷ which generates cellular energy via oxidative phosphorylation of adenosine triphosphate (Fig. 1). The complex is composed of 44 different subunits, which are organized into three functional modules (N module, Q module, and P module).⁸ The N module mediates binding and oxidizing of NADH, the Q module catalyzes the transfer of electrons from NADH to coenzyme Q10, and the P module helps establish the mitochondrial membrane potential by pumping H⁺ ions across the inner mitochondrial membrane. Complex I deficiencies destabilize mitochondrial oxidative phosphorylation, resulting in devastating energy production deficits.⁹ 

Figure 1.

 

Ndufs4^−/− model of complex I deficiency. Schematic shows how Ndufs4 deficiency impairs complex I assembly and mitochondrial electron transport chain function. Ndufs4 is a subunit of complex I, which catalyzes the transfer of electrons from NADH to coenzyme Q10 and helps establish the mitochondrial membrane potential by pumping H⁺ ions across the inner mitochondrial membrane. In Ndufs4^−/− cells, the N module failed to assemble, and complex I remained errantly associated with assembly factors NDUFAF1 (NADH:ubiquinone oxidoreductase complex assembly factor 1) and NDUFAF2 (NADH:ubiquinone oxidoreductase complex assembly factor 2). Figure made with BioRender.com.

NADH:ubiquinone oxidoreductase Fe-S protein 4 (Ndufs4) is a nuclear-encoded protein essential for complex I assembly and activity.^10,11 In cells lacking Ndufs4 expression, complex I fails to assemble the N module (Fig. 1).^12,13 Patients with Ndufs4 loss-of-function mutations develop Leigh syndrome, a fatal pediatric disorder characterized by severe mitochondrial energy generation deficits.^14–17 Patients typically present in infancy with elevated lactate in cerebral spinal fluid and serum, hypotonia, loss of vision, and progressive neurological dysfunction due to diminished complex I activity. Vision problems may include nystagmus, strabismus, ophthalmoplegia, and optic atrophy. Similarly, Ndufs4 deficiency in mice results in diminished growth rate, hypothermia, ataxic signs, and elevated serum lactate.¹⁰ Knockout animals develop progressive encephalomyopathy by around 5 weeks of age and die by 50 to 60 days postnatal. Ocular pathology in Ndufs4^−/− animals recapitulates the neurological and ophthalmic features of human complex I deficiency, such as retinal neurodegeneration, inflammation, and visual function deficits. Due to the rapid and severe course of disease progression, the Ndufs4^−/− mouse is an attractive model for testing therapeutic interventions that target retinal pathologies resulting from complex I deficiency.^11,18–21 

The timing of onset, rate of progression, and variability of ocular pathology in the Ndufs4^−/− mouse remain poorly understood, in part due to conflicting reports from inadequately powered studies.^10,11,22 Although some retinal abnormalities have been observed in 2- to 3-week-old animals,²³ most investigations into Ndufs4^−/− neurodegenerative ocular phenotypes use animals 5 to 7 weeks old. Ophthalmic live imaging by optical coherence tomography (OCT) and retinal functional testing by electroretinogram (ERG) enable noninvasive analysis of retinal pathology in living animals. These tools enable direct comparison of ophthalmic clinical readouts from human patients with relevant experimental animal models.²⁴ Here, we present a panel of noninvasive opthalmic measures of retinal pathology and visual dysfunction in Ndufs4^−/− mice. 

Methods

Animals

Ndufs4 ^−/− animals (B6.129S4-Ndufs4tm1.1Rpa/J) are on a C57BL/6 background. Breeding Ndufs4^+/− pairs were purchased from The Jackson Laboratory (Bar Harbor, ME). Male and female offspring from heterozygote breeding pairs were genotyped to select wild-type (WT) and knockout (Ndufs4^−/−) littermates for this study. Genotyping protocol was performed as directed by The Jackson Laboratory (primers: AGT CAG CAA CAT TTT GGC AGT, GAG CTT GCC TAG GAG GAG GT, AGG GGA CTG GAC TAA CAG CA). Three different cohorts of animals were weighed (Fig. 2A), and analyzed at 3 weeks postnatal (p21–p22), 5 weeks postnatal (p35–p36), or 7 weeks postnatal (p49–p50). Knockout morbidity and mortality precluded evaluation of animals older than 7 weeks. All investigations were performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Rodent experiments were approved and monitored by the Institutional Animal Care and Use Committee (IACUC) of Columbia University. 

Figure 2.

 

Animal weights and ophthalmic imaging parameters. (A) Weight (in grams) of Ndufs4^−/− (n for each genotype is indicated in the legend at 3, 5, and 7 weeks). Data are presented as mean ± SEM. Differences between groups were analyzed by two-way ANOVA, followed by Tukey's multiple comparisons test. (B) Representative H&E staining of ocular sections from 7-week-old WT and Ndufs4^−/− littermates. Scale bars: 250 µm (main image) and 50 µm (inset). Schema demonstrate ocular dimension measurements L_y and L_x. (C) Quantification of ocular dimensions (measured as L_x × L_y) of WT (n = 12) and Ndufs4^−/− (n = 10) eyes at week 7. Differences between groups were evaluated by unpaired t-test. (D) Representative view of ocular positioning and 1.5-mm spot size for ERG recording. (E) Representative fundus and OCT imaging demonstrates the positioning of four OCT scans in the same eye. The location of each OCT scan is indicated by a green line on the associated fundus image. (F) Representative image of excessive corneal opacity that precludes OCT layer thickness analysis. (G) Representative image of corneal opacity that is acceptable for OCT measurement of retinal layer thicknesses but precludes evaluation of outer retinal reflectivity.

Histology and Eyecup Measurements

Mice were euthanized with intraperitoneal (IP) administration of overdose of ketamine (160–200 mg/kg) plus xylazine (10–20 mg/kg) and perfused with saline, followed by fixation with 4% paraformaldehyde. Eyes were enucleated, embedded in optimal cutting temperature compound, and cryosectioned at 20-µm per section. For hematoxylin and eosin (H&E) staining, slides were submitted to the Columbia University Medical Center Molecular Pathology Shared Resource Histology Service. Eye size measurements of 7-week-old mice were obtained from microscopy images of retinal sections from the medial horizontal plane of the eye. Ocular sections were imaged with a Nikon (Tokyo, Japan) microscope and SPOT digital camera (SPOT Imaging, Sterling Heights, MI). The diameter of each eyecup (L_x and L_y) was measured as depicted in Figure 2B, using FIJI image processing software. The measurements from six sections on each slide were averaged. Sections that greatly deviated from circularity were excluded from measurement. 

Electroretinography

Animals were dark adapted for 2 to 16 hours and handled under deep-red light-emitting diode (LED) lighting to preserve dark adaptation. Animals received tropicamide and phenylephrine chloride eyedrops to dilate the pupil (pupil size was not measured) and were anesthetized with IP administration of ketamine (80–100 mg/kg) and xylazine (5–10 mg/kg). Focal ERGs were recorded with the MICRON IV Image-Guided Focal ERG System (Phoenix Research Industries, Duluth, GA), using a flash spot size of 1.5 mm, centered on the optic nerve head, and 10-ms white light LED stimulus intensities of −0.7, 1.1, and 2.3 log (cd·s/m²). The 1.5-mm flash spot size was selected as it is the largest diameter of stimulus available for the MICRON IV Focal ERG System. A 650-nm deep-red light filter was used to preserve dark adaptation while positioning the eye in contact with the corneal electrode and centering the flash stimulus directly on the retina (Fig. 2D). At each stimulus intensity, 10 ERG traces were recorded, with a 5-second break between each recording and a 30-second break before each increase in stimulus intensity. A sham reading (no light stimulus) was recorded immediately prior to exposure to light stimulus. Amplitudes and implicit times of the a- and b-waves and oscillatory potentials (OPs) were calculated using LabScribe 3 ERG automated software (iWorx Systems, Dover, NH). OPs were derived using a 30- to 300-Hz filter, and the sum of the first six OPs was used to calculate the sum OP amplitude. Waveform amplitude measurements derived from sham recordings were subtracted from all amplitudes calculated by LabScribe 3 ERG for that eye, to account for background noise in focal ERG traces. 

Optical Coherence Tomography

OCT imaging was performed using the Phoenix MICRON IV Image-Guided OCT System. Animals received tropicamide and phenylephrine chloride eyedrops to dilate the pupil and were anesthetized with IP administration of ketamine (80–100 mg/kg) and xylazine (5–10 mg/kg). For each eye, two vertical and two horizontal OCT scans (longitudinal resolution, 1.8 µm; transverse resolution, 3 µm; scan length, 1625 µm; average of 30 frames/scan) were captured approximately 400 µm from the optic nerve (Fig. 2E). Imaging was performed by investigators masked to genotype, and animals with congenital retinal defects or corneal opacities that impair OCT acquisition (Figs. 2F, 2G) were eliminated from evaluation. 

Four OCT images per eye were averaged to generate mean retinal thickness values. Segmentation of individual retinal layers was generated using Phoenix MICRON OCT Segmentation software, and average layer thicknesses were calculated in Excel (Microsoft Corporation, Redmond, WA). Specific retinal layers were identified as indicated in Figure 3A. Total retinal thickness was measured from the ganglion cell layer (GCL) to the retinal pigment epithelium (RPE). 

Figure 3.

 

OCT imaging reveals IPL atrophy in 7-week-old Ndufs4^−/− mice. (A) Schematic diagram (made with Biorender.com) depicting neuronal distribution in different retinal layers and representative OCT images of WT and Ndufs4^−/− littermates at 3, 5, and 7 weeks. (B–H) The total retinal (B), GCL (C), IPL (D), INL (E), OPL (F), outer nuclear layer (ONL) (G), and RPE (H) thickness in WT and Ndufs4^−/− eyes at 3, 5, and 7 weeks. Data are presented as mean ± SEM (n at each timepoint is indicated on the graphs). Differences between groups were analyzed by two-way ANOVA, followed by Tukey's multiple comparisons test. Multiple comparisons plots are shown in Supplementary Figure S1. IS/OS, inner segment/outer segment.

For reflectivity analysis, a 200 × 200-µm region of interest selection centered on the RPE layer was analyzed using FIJI Analyze>Plot Profile to generate a numerical reflectivity profile. Peak reflectivity values of each outer retinal band were normalized against the reflectivity of the outer nuclear layer, and quantifications from four OCT images were averaged for each eye. All OCT image processing was performed by an investigator who was masked to the genotype of the animals analyzed. 

Optomotor Reflex

Contrast sensitivity and visual acuity were measured using the OptoDrum automated optomotor reaction system (Striatech GmbH, Tübingen, Germany).²⁵ Animals were handled by a trained investigator to avoid overactivity and monitored for attentiveness during the imaging sessions. Optomotor reflexes were recorded by an automated tracking system, and limits of visual perception were determined by increasing the spatial frequency (acuity measure) or decreasing the contrast (contrast measure) of the stimulus until an animal failed to respond to the stimulus on two consecutive trials. 

Statistics

Data were analyzed using Excel and Prism (GraphPad Software, San Diego, CA). Statistical tests, n values, and P values are described in the figures and legends. Data are presented as mean ± SEM. The n value at each timepoint is indicated on the graphs. Differences between groups were analyzed by two-way analysis of variance (ANOVA), followed by Tukey's multiple comparisons test. Correlation analysis was performed by linear regression. Significance was defined as P < 0.05. No statistical methods were used to predetermine sample size. 

Results

Retinal Thickness Changes in Postnatal WT and Ndufs4^−/− Animals

Ndufs4 ^−/− mice have weights similar to those of their WT littermates at 3 weeks of age (Fig. 2A). From 3 to 7 weeks of age, the weight of WT male and female mice increases by 200%, but the weight of Ndufs4^−/− animals increases by only 50%. By 5 weeks, Ndufs4^−/− animals are significantly smaller than their WT littermates. H&E staining of ocular sections revealed typical retinal layer morphology in 7-week-old animals (Fig. 2B), and measurements of eye-cup diameters showed no differences between WT and Ndufs4^−/− eyes (Fig. 2C). 

OCT imaging showed normal retinal morphology and no genotype-dependent differences in retinal layer thicknesses in 3- to 5-week-old knockout animals (Fig. 3, Supplementary Fig. S1). At 7 weeks, the inner plexiform layer (IPL) of Ndufs4^−/− animals was 14.1% thinner (95% confidence interval [CI], 9.2–19.0) compared to WT littermates (Figs. 3A, 3D). We also observed a modestly thicker inner nuclear layer (INL) in Ndufs4^−/− mice compared to WT (P = 0.018, ANOVA) (Fig. 3E); however, the effect did not retain statistical significance by multiple comparisons testing of age-matched animals (Supplementary Fig. S1). Postnatal retinal remodeling of the GCL, IPL, and INL results in a 6.3% decrease (95% CI, 4.1–8.9) in total retinal thickness in both WT and knockout animals over weeks 3 to 7 (Figs. 3A–3E). The outer retina and RPE did not show any genotype-dependent or age-dependent thickness changes during this time period (Figs. 3F–3H). 

Outer Retinal Reflectivity Abnormalities in Ndufs4^−/− Mice

The outer retinal region of the OCT contains four high-reflectivity bands, corresponding to the external limiting membrane (ELM), the inner/outer segments of the photoreceptors (ellipsoid zone [EZ]), the RPE, and the choriocapillaris (CC) (Fig. 4A). Quantitative evaluation of outer retinal band reflectivity requires high-quality OCT imaging to determine the resolution of these bands; corneal opacities such as those shown in Figure 2G result in reduced OCT image quality that is not qualified for reflectivity analysis. Although reflectivity of inner retinal layers was highly variable (Fig. 2E), outer retinal reflectivity was generally consistent among different OCT scans. ELM reflectivity was brightest in 3-week-old animals; by 7 weeks of age, ELM reflectivity decreased by 8.0% (95% CI, 1.3–14.8) in WT animals and by 11.4% (95% CI, 2.6–20.2) in Ndufs4^−/− animals (Fig. 4B, Supplementary Fig. S2). Ndufs4^−/− animals had significantly reduced ELM reflectivity compared to WT animals (P = 0.0003, ANOVA). Reflectivity differences between knockout and WT animals became more apparent at 5 weeks (Table); however, differences between groups did not reach statistical significance by multiple comparisons testing at specific timepoints. 

Figure 4.

 

Outer retinal hyporeflectivity in Ndufs4^−/− mice. (A) Representative OCT images and reflectivity profiles of hyperreflective outer retinal bands in WT and Ndufs4^−/− littermates. (B–E) The ELM (B), EZ (C), RPE (D), and CC (E) reflectivity in WT and Ndufs4^−/− eyes at 3, 5, and 7 weeks (n at each timepoint is indicated on the graphs). Data are presented as mean ± SEM. Differences between groups were analyzed by two-way ANOVA, followed by Tukey's multiple comparisons test. One W5 WT outlier removed by Grubbs’ outlier test is shown in (B). Multiple comparisons plots are shown in Supplementary Figure S2. (F) Linear regression between inner retinal thickness and ELM reflectivity in WT and Ndufs4^−/− retinas. (G) Linear regression between IPL thickness and ELM reflectivity in WT and Ndufs4^−/− retinas. (H) Linear regression between inner retinal thickness and EZ reflectivity in WT and Ndufs4^−/− retinas. (I) Linear regression between IPL thickness and EZ reflectivity in WT and Ndufs4^−/− retinas.

Table.

 

Summary of Observed Ndufs4^−/− Ocular Phenotypes and Power Analysis

 

EZ reflectivity was also markedly reduced in 5- to 7-week-old Ndufs4^−/− animals (Fig. 4C). Ndufs4^−/− mice had 20.3% (95% CI, 4.5–36.2) and 23.7% (95% CI, 8.5–38.9) lower EZ reflectivity compared to WT littermates at 5 and 7 weeks of age, respectively. EZ reflectivity did not show significant age-dependent changes (Supplementary Fig. S2); however, differences between genotypes were more apparent in older animals compared to 3-week-old mice. These reflectivity abnormalities are specific to the ELM and EZ bands; we did not observe any age-dependent or genotype-dependent differences in the reflectivity of the RPE or CC bands (Figs. 4D, 4E). Because the reflectivity of outer retinal bands on OCT could be influenced by age-dependent changes in inner retinal thickness, we measured the correlation between ELM or EZ reflectivity and the thickness of the inner retinal layers (retinal ganglion cell [RGC] + IPL + INL) or IPL specifically. ELM reflectivity declines were significantly correlated with inner retinal (Fig. 4F) and IPL (Fig. 4G) thinning in both WT and Ndufs4^−/− animals. Ndufs4^−/− ELM reflectivity deficits were still apparent after controlling for genotype-specific differences in retinal thickness. Conversely, EZ bands, which do not exhibit age-dependent changes in reflectivity, were also not correlated with changes in retinal thickness (Figs. 4H, 4I). 

Ndufs4^−/− ERG Deficits Are Present at Weaning

The two principal measures of the ERG waveform are the a-wave amplitude, which reflects phototransduction by photoreceptors, and the b-wave amplitude, which is generated by cells postsynaptic to the photoreceptors,²⁶ with contributions from bipolar cells, Müller glia, horizontal cells, and amacrine cells.²⁷ OPs are the oscillations appearing on the rising phase of the b-wave, and they are thought to arise from interactions among bipolar cells, amacrine cells, and ganglion neurons²⁷ (Fig. 5A). Although some Ndufs4^−/− mice had profound ERG deficits from weaning, the penetrance of these deficits was more variable in younger animals (Fig. 5, Supplementary Fig. S3). Some 3-week-old Ndufs4^−/− animals had normal ERG readings, whereas others were severely impaired. On average, 3-week-old knockout mice had 43.5% weaker (95% CI, 8.5–78.6) b-wave amplitudes compared to WT mice (Fig. 5A, B). Although there was no statistically significant age-dependent decrease in Ndufs4^−/− ERG function, genotype-specific differences were better resolved after week 5. By 7 weeks, Ndufs4^−/− mice had 56.2% weaker (95% CI, 22.3–90.2) a-wave amplitudes and 62.6% weaker (95% CI, 31.1–93.9) b-wave amplitudes compared to WT mice (Figs. 5A–5C). 

Figure 5.

 

ERG deficits measure functional impairment in Ndufs4^−/− mice. (A) Representative ERG traces (top) and OP traces (bottom) in 3-, 5-, and 7-week-old littermates showed attenuated a-wave, b-wave, and OP amplitudes and delayed b-wave implicit times in Ndufs4^−/− mice. (B–I) The b-wave amplitude (B), a-wave amplitude (C), b/a ratio (D), OP₁ amplitude (E), sum oscillatory potentials amplitude (F), a-wave implicit time (G), OP₁ implicit time (H), and b-wave implicit time (I) measurements in WT and Ndufs4^−/− littermates at 3, 5, and 7 weeks (n at each timepoint is indicated on the graphs). Data are presented as mean ± SEM. Differences between groups were analyzed by two-way ANOVA, followed by Tukey's multiple comparisons test. One W3 Ndufs4^−/− outlier removed by Grubbs’ outlier test is shown in (G). Multiple comparisons plots are shown in Supplementary Figure S3.

The b-wave was generally more affected than the a-wave, as evidenced by the lower b/a ratio in the Ndufs4^−/− animals (Fig. 5D). At 5 weeks of age, 37% of Ndufs4^−/− animals had nearly complete absence of ERG b-wave response (b-wave amplitude < 25 µV), and 25% of Ndufs4^−/− animals had no measurable a-wave. Likewise, the amplitudes of the OPs were substantially weaker in Ndufs4^−/− animals compared to WT animals. The first OP positive inflection (OP₁) was 79.9% lower (95% CI, 29.8–100.0) in 5-week-old Ndufs4^−/− animals compared to WT animals (Fig. 5E). Meanwhile, the sum of the first six oscillatory potential amplitudes (Sum6OP) was 76.8% lower (95% CI, 12.4–100.0) in Ndufs4^−/− animals compared to WT (Fig. 5F). 

Delays in OP and b-wave implicit times reveal compounding defects in retinal electrophysiology downstream of photoreceptor phototransduction. These deficits were genotype dependent and not age dependent. The a-wave implicit time was unaffected (Fig. 5G). The OP₁ implicit time was modestly delayed (P = 0.002, ANOVA), although these differences were not significant by multiple comparisons testing of age-matched animals (Fig. 5H, Supplementary Fig. S3). Meanwhile, the b-wave implicit was delayed by 46.3 ms (95% CI, 21.5–71.1) in 5-week-old Ndufs4^−/− animals (Fig. 4I). 

ERG Deficits in Ndufs4^−/− Mice Correlate With Inner Retinal Atrophy

We next evaluated the relationships between different ERG parameters by linear regression. The b- and a-wave amplitudes were highly correlated, consistent with the method of b-wave quantification, which measures b-wave amplitude from peak negative a-wave deflection (Fig. 6a). The lower y-intercept for Ndufs4^−/− linear regression compared to WT indicates that b-wave amplitudes are more severely affected than a-wave amplitudes, consistent with the reduced b/a amplitude ratio quantifications (Fig. 5D). Because Ndufs4^−/− b-wave amplitudes and implicit time annotations can be confounded by their diminished oscillatory potentials, we investigated the correlation between b-wave amplitude and OP₃ (the third OP, which overlaps the peak of the b-wave) (Fig. 6B). The b-wave amplitudes were significantly correlated with OP₃ measurements, and the slopes of the linear regressions for WT and Ndufs4^−/− mice were not significantly different. If the primary reason for diminished b-wave amplitude in Ndufs4^−/− mice was due to a confounding effect from a diminished OP response, we would expect the two linear regression lines to be closely superimposed. Instead, the y-intercept for Ndufs4^−/− mice was significantly lower than the y-intercept for WT mice (P < 0.0001), indicating that the profound deficit in Ndufs4^−/− OPs is not sufficient to explain the observed b-wave amplitude deficits. Measurement of b-wave implicit time was also significantly correlated with OP₃ (Fig. 6C). The b-wave amplitudes were also significantly correlated with b-wave implicit times, with extreme delays (implicit times >100 ms) exclusively occurring in animals with severely impaired b-wave (Fig. 6D). The slopes of linear regressions between b-wave amplitude and b-wave implicit time were significantly steeper for Ndufs4^−/− compared to WT animals (P < 0.0001), indicating that Ndufs4 deficiency resulted in more severe b-wave implicit time deficits than expected from the attenuated b-wave or OP₃ amplitudes. Because quantification of b-wave implicit times depends on accurate detection of a b-wave peak, delays in b-wave implicit times may be confounded by poorly defined b-wave responses resulting from either b-wave or OP deficits. 

Figure 6.

 

ERG correlates and OMR measures of contrast and acuity perception deficits. (A) Linear regression between a- and b-wave amplitudes in WT and Ndufs4^−/− littermates. (B) Linear regression between OP₃ and b-wave amplitudes in WT and Ndufs4^−/− littermates. (C) Linear regression between OP₃ and b-wave implicit time in WT and Ndufs4^−/− littermates. (D) Linear regression between b-wave and b-wave implicit time in WT and Ndufs4^−/− littermates. (E–H) Linear regressions between IPL thickness and a-wave amplitude (E), OP amplitude (F), b-wave amplitude (G), and b-wave implicit time (H). (I) Schematic demonstrating grating patterns with increasing difficulty of acuity or contrast perception. (J) Acuity sensitivity threshold (cycles/degree) in WT and Ndufs4^−/− mice. (K) Contrast sensitivity threshold (% contrast) in WT and Ndufs4^−/− mice (n at each timepoint is indicated on the graphs). Data are presented as mean ± SEM. Differences between groups were analyzed by two-way ANOVA, followed by Tukey's multiple comparisons test. Multiple comparisons plots are shown in Supplementary Figure S4. Linear regression between IPL thickness (WT, n = 18; Ndufs4^−/−, n = 18) and acuity (L) or contrast sensitivity (M) at 7 weeks. Linear regression between EZ reflectivity (WT, n = 18; Ndufs4^−/−, n = 18) and acuity (N) or contrast sensitivity (O) at 7 weeks.

We next investigated whether ERG deficits were correlated with IPL atrophy in 7-week-old animals. IPL atrophy in 7-week-old Ndufs4^−/− animals was not significantly correlated with a-wave amplitude (Fig. 6E) and was moderately correlated with OP amplitude (Fig. 6F). Meanwhile, both b-wave amplitudes and b-wave implicit times were strongly correlated with IPL atrophy at 7 weeks (Figs. 6G, 6H). Because ERG deficits precede the onset of IPL atrophy, these correlations imply that pathology underlying ERG b-wave dysfunction is related to the drivers of IPL atrophy. 

Optomotor Reflex Measure of Acuity and Contrast Sensitivity

Visual acuity and contrast sensitivities were stable in WT animals 3 to 7 weeks of age (Figs. 6I–6K, Supplementary Fig. S4). Ndufs4^−/− littermates had normal visual function at 3 weeks of age and displayed increasing deficits in both acuity and contrast detection at later ages. In 7-week-old Ndufs4^−/− animals, acuity sensitivity was 46.6% lower (95% CI, 25.1–68.0) compared to WT littermates, and 56% (five out of nine animals tested) failed to detect any visual stimuli presented at <100% contrast. Because IPL atrophy and visual function deficits appeared on similar timelines, we investigated whether these measures were correlated by linear regression (Figs. 6L, 6M). We found no significant correlations between these readouts. We further investigated if optomotor reflex (OMR) deficits were correlated with EZ hyporeflectivity, as EZ disruption is associated with visual acuity deficits in patients with retinal disease.^28,29 Curiously, acuity and contrast sensitivity were both weakly correlated with EZ reflectivity in WT animals (Figs. 6N, 6O); however, this relationship was absent in Ndufs4^−/− littermates. 

Discussion

The Ndufs4^−/− mouse model of complex I deficiency exhibits progressive retinal pathology measurable by OCT, ERG, and OMR. These findings build upon existing literature describing the biochemical and cellular underpinnings of retinal pathology in the Ndufs4^−/− mouse (Fig. 7). Concurrently, physiological retinal remodeling in 3- to 7-week-old WT littermates produces thinning of inner retinal layers, reduction of ELM reflectivity, and attenuation of ERG response. 

Figure 7.

 

Integrated summary of retinal pathology in Ndufs4^−/− mouse. Schematic summarizing timeline of retinal pathologies observed in Ndufs4^−/− mice in published literature and in the present study. ERG impairment secondary to energetic deficits occurred throughout the entire study period (weeks 3–7), and acuity and contrast deficits appeared in 5- to 7-week-old mice. The early phase (week 3) of retinal pathology in Ndufs4^−/− mice featured RGC dendritic atrophy,²³ amacrine and bipolar cell loss,^19,23 and ELM and EZ hyporeflectivity. At 5 weeks, Ndufs4^−/− mice exhibited RGC firing deficits,¹⁹ RGC cell loss,¹⁹ and microgliosis.^18,19 During the terminal phase of disease progression (weeks 6–7), Ndufs4^−/− mice displayed astrogliosis,¹¹ loss of GAD67 immunoreactivity in IPL,¹⁹ and IPL atrophy. Figure made with Biorender.com.

This temporal profile captures both physiological and pathological retinal changes in recently weaned mice. Murine postnatal retinal development is most pronounced during the first 3 weeks of life; however, physiological and morphological changes persist in 3- to 7-week-old animals.^30,31 These temporal changes require careful age matching of WT and knockout animals to avoid conflating genotype-specific phenotypes with physiological retinal development. 

We used both male and female animals and did not observe any overt sex dependence in phenotype; however, many of the observed phenotypes are highly variable, and our studies were not powered to evaluate the effect of sex on Ndufs4^−/− pathology. Our measurements captured the variability and severity of ocular pathology in differently aged cohorts of Ndufs4^−/− and WT littermate mice. Additional subtle differences between knockout and WT littermates may be revealed by longitudinal evaluation of individual animals. Further study should evaluate whether differences in the severity of retinal pathology during late stage of disease can be predicted from ocular readouts in younger animals to help reduce the necessary sample sizes for interventional experiments. Our measurements confirm that ERG deficits are present in Ndufs4^−/− mice at weaning and do not reflect progressive pathology in this model. Conversely, OCT measures of IPL thickness and OMR measures of acuity and contrast sensitivity identified an accelerated period of retinal degeneration in Ndufs4^−/− animals during weeks 5 to 7. These measures represent reliable, noninvasive tools for interrogating retinal pathology caused by complex I insufficiency. 

Our analysis helps explain discrepancies among previously reported ERG phenotypes in Ndufs4^−/− animals.^10,11,22 We observed profound ERG deficits in 3-week-old mice, with only minimal progression of impairment in older animals. The highly variable penetrance of ERG deficits, particularly in younger animals, helps explain why some studies failed to detect an ERG deficit in younger mice. Gospe et al.²² reported that Ndufs4^−/− photoreceptors are fundamentally capable of normal light-evoked signaling and that deficits in ERG response reflect energetic deficits, rather than a neurodegenerative phenotype. Our findings support this interpretation. 

Oscillatory potentials are thought to reflect signaling among bipolar cells, amacrine cells, and ganglion cells.²⁷ Degeneration of these neuronal populations has been reported in 3-week-old Ndufs4^−/− retinas.^19,23 It remains unclear whether OP deficits reflect early degeneration of inner retinal neurons or if impaired OP amplitudes also primarily reflect energy deficits. 

Amplitude and implicit time measures from ERG waveforms were highly interrelated. ERG b-wave amplitude is influenced by the magnitude of the a-wave negative inflection, and overlapping of OPs with the low-frequency components of the ERG waveform may influence the detection of b-wave peaks.³² We sought to address some of the potential confounders among different ERG measures by applying correlation analysis. However, there are several limitations to the present study. First, we relied on automated detection of ERG b-wave amplitude and implicit times. This approach did not enable us to eliminate the contribution of differential OP timing and amplitude from the quantification of ERG b-wave parameters. Second, we relied on the image-guided functionality of the MICRON IV Focal ERG System to ensure appropriate illumination of the central retinal fundus. Pupil dilation was not measured, and equal retinal illumination was assumed between groups and ages. Differences in pupil dilation or corneal clarity may result in reduced retinal illuminance and subsequently reduced ERG components and slower implicit times. Third, to support the larger sample sizes evaluated in this study, ERG responses were measured at three distinct illumination intensities, and comparisons between groups were analyzed at the highest intensity measured. This ERG protocol precluded calculation of a luminance–response function, which requires measurements across a broader range of stimulus intensities. These limitations invite future studies to conduct a more thorough examination of ERG deficits in Ndufs4^−/− mice. 

Thinning of the IPL in 5- to 7-week-old Ndufs4^−/− mice likely reflects atrophy of RGCs, amacrine cells, and bipolar cells.^11,19,20,22 OCT imaging provides a noninvasive and highly reproducible measure of inner retinal degeneration, consistent with reported H&E cross-sections from 7-week-old Ndufs4^−/− mice.¹¹ Because IPL atrophy begins at least 2 weeks after weaning, this measure of retinal disease progression is tractable for evaluating pharmacological interventions. 

We also observed outer retinal OCT hyporeflectivity in Ndufs4^−/− mice. Optical density of OCT imaging is highly sensitive to refractive index changes between retinal layers, and the reflectivity of outer retinal bands on OCT imaging is thought to indicate changes in mitochondrial morphology and localization.^33–36 In patients with outer retinal pathologies, EZ hyporeflectivity correlates with visual acuity deficits.^28,29 Similar reflectivity phenotypes have been observed in animal models of photoreceptor light injury^37,38 and retinal oxidative damage.^39,40 Our analysis found correlation between EZ reflectivity and OMR response in WT mice, supporting a relationship between EZ reflectivity and visual function. However, this effect was not observed in Ndufs4^−/− mice. In hibernating ground squirrels, EZ hyporeflectivity is associated with mitochondrial remodeling during torpor and loss of mitochondria from photoreceptor inner segment region.³³ Thus, it is plausible that outer retinal hyporeflectivity in the Ndufs4^−/− mouse may reflect abnormalities in mitochondrial distribution and morphology. However, without direct characterization of Ndufs4^−/− photoreceptor mitochondria, it is challenging to draw a conclusive interpretation of the OCT hyporeflectivity phenotype. 

OMR measures progressive visual function deficits in the Ndufs4^−/− mouse. The optomotor reflex prompts an animal to turn its head in response to a moving visual stimulus. Contrast sensitivity can specifically highlight deficits in retinal function,^41,42 and measures of visual acuity provide a global readout of visual processing by the retina and visual cortex. WT mice 3 to 7 weeks of age had stable acuity and contrast sensitivity scores, whereas Ndufs4^−/− littermates exhibited marked declines in visual function at weeks 5 to 7. This timeline is consistent with the study by Yu et al.,¹⁸ who reported visual cliff test and slow angled-descent forepaw grasping test deficits in 5-week-old Ndufs4^−/− mice. Visual function testing by OMR may be less physically demanding compared to alternative tests such as the Morris water maze and the visual cliff test (which demand whole-body movements in response to visual cues). OMR response requires equal levels of minimal physical exertion (head turning) to react to visual stimuli of different difficulties. Thus, OMR testing enables visual function assessment in older, physically weakened animals, allowing monitoring of visual function impairments later into the disease process. Activity levels were not quantified between WT and Ndufs4^−/− littermate mice and may present a confounding contributor to poor response to OMR stimulus. It should be noted, that 22% (two of nine) of 7-week-old Ndufs4^−/− mice failed to respond to any visual stimuli during OMR testing. Complete failure to exhibit an OMR response may indicate either complete loss of visual ability or severe neurological or motor function deficits. 

Although deletion of Ndufs4 is most directly related to human Leigh syndrome, deficits in mitochondrial complex I contribute to optic atrophy in several retinal pathologies, including Leber's hereditary optic neuropathy,² autosomal dominant optic atrophy,⁴³ and primary open-angle glaucoma.⁴⁴ The Ndufs4^−/− mouse offers a robust model of ocular pathology from mitochondrial complex I deficiency. The Table summarizes key retinal deficits in Ndufs4^−/− mice at 3, 5, and 7 weeks of age. Although many pathological features could be detected in younger mice, differences between the WT and knockout groups were more clearly evident in 5- to 7-week-old mice. Due to the severity of retinal pathology, knockout phenotypes can be identified using relatively small sample sizes. Evaluation of potential therapeutic strategies, however, may require significantly larger sample sizes due to variable degrees of impairment among knockout animals. OMR acuity and contrast deficits are highly variable among Ndufs4^−/− mice and may be challenging readouts for evaluating potential vision-restorative therapies. ELM and EZ reflectivity deficits were similarly highly variable and require further mechanistic study to identify the significance of this phenotype in Ndufs4^−/− animals. For interventions expected to directly correct complex I activity, ERG responses can be used to detect improvement of retinal phototransduction with sample sizes that are typical for in vivo study designs. Alternatively, OCT morphology has low inter-animal variability and is well suited to evaluate neuroprotective interventions aimed at reducing inner retinal atrophy. In sum, ERG, OCT, and OMR measures offer a panel of translatable noninvasive tools for evaluating retinal pathology in the Ndufs4^−/− model of complex I deficiency. 

Acknowledgments

The authors thank Svitlana Avrutsky for helpful feedback on image data analysis. 

Supported by grants from the National Institute of Neurological Disorders and Stroke, National Institutes of Health (RO1 NS081333 and R03 NS099920 to CMT) and by the Department of Defense Army Research Office/Air Force Office of Research Defense University Research Instrumentation Program (CMT). 

Disclosure: M.I. Avrutsky, None; J.M. Lawson, None; J.E. Smart, None; C.W. Chen, None; C.M. Troy, None