Beyond Neovascularization: Light Response Deficits in Oxygen-Induced Retinopathy Models Authors: Silke Becker Published: 2026 ABSTRACT Retinopathy of prematurity (ROP) has traditionally been viewed as a retinal vascular disease. However, even after vascular recovery, many patients experience retinal light response deficits for years, possibly lifelong. Despite major treatment advances that target abnormal retinal neovascularization, ROP remains a major cause of vision loss in premature infants. The mouse and rat oxygen-induced retinopathy (OIR) models are widely used to study ROP and have been instrumental in developing anti-VEGF treatments for retinal neovascular diseases. In contrast, the mechanisms underlying retinal light response deficits in ROP remain poorly understood—and this limits the development of therapies targeting the neuronal component of ROP. This review summarizes current knowledge of retinal light response deficits in OIR across different retinal cell populations. Available evidence suggests that deficits originate in photoreceptors and propagate through downstream retinal neurons, resulting in cumulative impairment across the retinal circuit. By reviewing existing findings, discussing underlying mechanisms, and identifying key knowledge gaps, this review evaluates OIR as a valuable platform to investigate retinal light response deficits and to guide the development of strategies to improve long-term vision outcomes for patients with ROP. Many children with retinopathy of prematurity (ROP) achieve regression of retinal neovascularization—yet they often lack normal visual acuity later in life1–3 and experience lower vision-related quality of life.4 Historically regarded as a disease of the retinal vasculature, current ROP treatments prevent or regress abnormal neovascularization and decrease the risk of complete blindness.5 Persistent vision problems in patients with a history of ROP remain common,6,7 and clinical studies demonstrate impaired retinal light responses,8–13 which can continue at least into adolescence10,12,14 and may contribute to long-term vision loss. ROP affects infants born before 30 to 32 weeks of gestation with incomplete retinal vascular development, and ROP risk increases as gestational age and birth weight decrease.15 Blood vessels begin to form in the human central retina at 14 to 15 weeks of gestation16,17 and the superficial plexus reaches the ora serrata by around 32 weeks; the deep plexus forms from 25 to 26 weeks and continues to mature until term.17 Therefore, infants born at the current limit of viability (22–23 weeks of gestation) have an incomplete superficial plexus and lack the deep plexus, and the peripheral retina appears avascular on clinical examination. Prematurely born infants experience different oxygen concentrations than in utero: Therapeutic oxygen supplementation and fluctuating systemic blood oxygenation delay physiological retinal vascular development and thus increase the risk of ROP.15 Oxygen-induced retinopathy (OIR) recapitulates these conditions: In mice, hyperoxia (typically 75% oxygen from postnatal day [P]7–P12) regresses retinal blood vessels.18 The avascular retina becomes hypoxic after the animals return to room air,19 and neovascularization peaks around P17. In rats, oxygen fluctuates between 50% and 10% from P0 to P14, followed by room air exposure until neovascularization peaks around P19.20 Although other paradigms exist, these are the most used, and this review refers to them as the mouse OIR and the rat 50/10 OIR models. Both models have been instrumental in developing anti-VEGF therapies, which are currently being tested for the treatment of ROP.21–23 Despite the widespread use of OIR to study ROP, research has largely focused on its vascular pathology; retinal neural deficits have received much less attention. Our ability to measure retinal light responses offers a direct window into these neural changes. This review evaluates retinal light response deficits in OIR, identifies potential underlying mechanisms and critical knowledge gaps, and provides context from studies in ROP (albeit without providing a comprehensive analysis of light responses in ROP, which have been reviewed elsewhere14) to argue that OIR provides a valuable platform to study retinal light response deficits in ROP. Measuring Retinal Light Responses Light generates electrical signals across the retina that can be recorded as field potentials using the electroretinogram (ERG). In its standard form, the ERG reflects the summed electrical activity from multiple retinal cells types: The initial, negative-going a-wave primarily reflects photoreceptor activity, whereas the subsequent, positive-going b-wave is dominated by ON-bipolar cell (ON-BPC) responses. High-frequency wavelets on the ascending edge of the b-wave (oscillatory potentials [OPs]), arise from the inner retinal circuitry, including amacrine cells (Fig. 1). Scotopic and photopic recording conditions can distinguish between rod- and cone-mediated pathways.24 Additional ERG components, including the RPE-derived c-wave,25 the retinal ganglion cell (RGC)- and amacrine cell–dominated photopic negative response,26,27 or the OFF-BPC–derived d-wave,25,28 require specialized stimulation and recording parameters (which have so far not been applied to OIR models). Light response components of the in vivo ERG and their cellular origins. Adapted from Becker et al.24 (A) The a-wave represents the rod or cone photoreceptor (PR) response, the b-wave represents the rod or cone ON-BPC response to light in the in vivo ERG. (B) OPs, shown after filtering for clarity, are oscillatory wavelets superimposed on the b-wave and are derived predominantly from amacrine cells (ACs). Response amplitudes are measured from the baseline to the a-wave trough (photoreceptor amplitude), from the a-wave trough to the b-wave peak (ON-BPC amplitude), and from the trough to the peak of an OP (OP amplitude). Implicit time (IT) is defined as the interval from stimulus onset to the response peak or trough. The in vivo ERG records light responses in intact animals (or human subjects) using electrodes placed on the cornea or near the eye. By its nature, the in vivo ERG can be influenced by systemic factors, such as oxygen and nutrient supply, circulating factors, and anesthesia drugs. Depending on the experimental question, this can represent either a strength or a limitation: Although the in vivo ERG captures retinal function in a living organism, it cannot readily distinguish retina-intrinsic effects from systemic influences. In addition, the contributions of individual cell types are difficult to resolve because the in vivo ERG waveform represents the summed activity from all retinal cell types and the RPE. This limitation becomes particularly apparent when components of the ERG are altered in amplitude and/or timing: The b-wave overlays the a-wave, so that even subtle shifts in photoreceptor and/or ON-BPC response kinetics can alter the a-wave amplitude and implicit time (Fig. 2) and vice versa. The overlap of a- and b-waves can complicate waveform interpretation. A reduced and delayed b-wave (in the red trace) can make the underlying a-wave appear larger and delay the implicit time (IT) compared with normal responses (blue trace). The photoreceptor component that underlies the ON-BPC response is depicted as a dotted line. Ex vivo ERG overcomes several limitations of the in vivo ERG by isolating the retina (generally without the RPE) from the eye and recording light responses in a specimen holder, where the extracellular environment can be precisely manipulated.29,30 This strategy allows the recording of retina-intrinsic functional changes under optimal recording conditions while minimizing or removing systemic confounders.31,32 Furthermore, pharmacological agents can be applied to isolate contributions from specific cell populations, namely photoreceptors, ON-BPCs, and Müller glia (Fig. 3).29,30 However, in contrast with in vivo recordings, ex vivo ERG fails to capture the influence of systemic differences. Ex vivo ERG (as well as other, advanced electrophysiological techniques, such as suction electrode and patch clamp recordings) remain very much underused in published OIR studies; a broader use of ex vivo ERG could, for example, help to define the impact of OIR on photoreceptor sensitivity, as well as on the activation, deactivation, and recovery kinetics under controlled metabolic conditions, but also determine ON-BPC light responses in isolation. Ex vivo ERG to study light responses in isolated retinas. (A–C) Retinas are carefully removed from the eye cups of the animals and mounted on the dome of a specialized specimen holder. Samples are superfused with heated, oxygenated Ames' media in the absence or presence of pharmacological inhibitors. Light stimuli are applied through the objective of a microscope and potential differences are recorded by electrodes connected to either side of the retina. Isolated ON-BPC or PII responses (F) can be calculated by subtraction of photoreceptor (PR) or fast PIII responses ([E], recorded in the presence of BaCl2 to inhibit K+ channels in Müller glia and DL-AP4 to block signal transmission to ON-BPC) from combined fast PIII and PII responses in the presence of BaCl2 alone (D). Reproduced from Becker et al.24 The greatest strength of in vivo and ex vivo ERG, therefore, would lie in their complementary use, by exploiting light responses under physiological conditions with in vivo ERG and dissecting retina-intrinsic mechanisms at the level of individual cell types with ex vivo ERG. Most studies in OIR animals (or, indeed, in patients with ROP) have relied on in vivo ERG33–47; ex vivo ERG has only rarely been used,46 and future studies should more fully harness its potential. The following sections discuss the current evidence for light response deficits in OIR, primarily with the more widely used in vivo ERG. Reduced Photoreceptor Light Responses in OIR and Sources of Variability Multiple studies report that OIR reduces a-wave amplitudes,33–41,47,48 delays implicit times,34,39,45,48 and reduces maximal photoreceptor amplitude and sensitivity43–45 in mice and rats (Table 1). These findings parallel clinical observations, where increasing ROP severity is associated with progressively reduced sensitivity and maximal photoreceptor amplitude in infancy and beyond.10,12,49 Although reduced photoreceptor light response amplitudes have also been reported in prematurely born children without ROP,50 potentially reflecting nutritional51 or other, prematurity-associated factors, deficits remain when comparing prematurely born children with and without ROP.10,14 Table 1.a- and b-Waves in OIRAgePR ampPR ITS, RmaxON-BPC ampON-BPC ITσ, VmaxCommentsMouse OIR model (75% oxygen, P7–P12) Martini et al., J Neurochem (2011)37P17↓*,†Not reportedNot reported↓†Not reportedσ ↓‡Vmax ↓†– Cammalleri et al., Frontier Cell Neurosci (2017)38P17↓†Not reportedNot reported↓†Not reportedNot reported– Vessey et al., The Journal of Comparative Neurology (2011)35P18Mixed ↓‡Rods↓*Not reported=Not reportedMixed ↓‡Rods ↓† Cones ↓‡Rods =Cones ↑‡Not reportedTwin flash paradigm to isolate rod a-wave, unable to isolate cone photoreceptor responses Fu et al., Graefes Arch Clin Exp Ophthalmol (2015)34P30↓*↑*Not reported↓†=Not reported– Wang et al., IOVS (2014)58P36=Not reportedNot reported↓ (P = 0.05)Not reportedNot reported– Schmitz et al., Cells (2025)468–9 weeks post OIRPhotopic ↓‡Scotopic =Not reportedNot reportedScotopic ↓‡Photopic ↓‡,*Not reportedNot reportedEx vivo ERG a-wave Calzi et al., JCI Insight (2019)41P17P24P31P17P24P31Scotopic↑†↓‡=Not reportedNot reportedScotopic=↓*=Photopic=↓†=Not reportedNot reportedReported light intensity unclear Yang et al., J Transl Med, 202648P17P25P17P25Scotopic↓*,†,‡↓‡Photopic↓†↓*↑†,‡=↑†↑*Not reportedScotopic↓*,†,‡↓‡Photopic↓†↓†↑*,†,‡=↑†↑†Not reported– Mezu-Ndubuisi et al., Mol Vis (2020)39P19P24P32P47==↓* at one intensity=↑*===Not reported==↓* at one intensity↓*,‡↑‡ at one intensity===Not reported77% O2 from P7 to P12, differences at selected light intensities Mowat et al., AJP (2012)40P26P60↓†↓‡Not reportedNot reported↓†↓†Not reportedNot reportedb/a-wave ratio ↓† at P26, ↓‡ at P60; photopic 10 Hz flicker ↓† at P26, ↓* at P60 Villacampa et al., PLoS (2017)36P26–P27P60↓‡=Not reportedNot reported↓†↓†Not reportedNot reported– Nakamura et al., PLoS (2012)334 weeks6 weeks8 weeks↓*‡==Not reportedNot reported↓*,‡ ↓‡at one intensity =Not reportedNot reportedRat OIR model (50/10, P0–P14) Zhang et al., IOVS (2013)43P55Not reportedNot reportedS ↓‡ Rmax ↓‡Not reportedNot reportedσ ↓‡Vmax ↓‡– Liu et al., IOVS (2006)59P18, P31Not reportedNot reportedS ↓‡, Rmax ↓*Not reportedNot reportedσ ↓*Vmax ↓*– Vähätupa et al., JoVE (2020)47Not reported↓*,†,‡Not reportedNot reported↓†Not reportedNot reported–Other mouse OIR models Villacampa et al., 2017 36P26–P27P60==Not reportedNot reported↓†↓†Not reportedNot reported85% O2 from P8 to P11Other rat OIR models Reynaud et al., 199542P13P18Not reportedNot reportedRmax ↓*S ↓*Rmax =S ↓*Not reportedNot reportedVmax =σ =Vmax ↓*σ ↓*Recordable responses in 5/8 OIR ratsRecordable responses in all OIR rats Dorfmann et al., 200844P15–P17, 24P19, 30, 60=↓‡Not reportedRmaxP15–P16 =P17–P60↓‡Rod–coneP15–P16 =P17–P60 ↓‡ConeP15 =P16–P60 ↓‡Not reportedNot reported80% O2 with 3.0 × 0.5 hours room air, delayed rod response maturation Dembinska et al., 200145P30P60↑‡ for P0–P14 O2 exposure↑‡ for P0–P14 O2 exposureRmax ↓‡,S ↓‡ for P0–P14 O2Rmax ↓‡,S ↓‡ for P0–P14 O2 exposure↑‡ for P0–P14 O2 exposure=Vmax ↓‡,σ = for P0–P14 O2Vmax ↓‡,σ ↓‡ for P0–P14 O2 exposure80% O2 with 3.0 × 0.5 hours room air greater effect with longer O2 exposureIT, implicit time; PR, photoreceptor; Rmax, maximal photoreceptor amplitude; S, photoreceptor sensitivity.* P < 0.01.† P < 0.001.‡ P < 0.05. a- and b-Waves in OIR IT, implicit time; PR, photoreceptor; Rmax, maximal photoreceptor amplitude; S, photoreceptor sensitivity. P < 0.01. P < 0.001. P < 0.05. Despite numerous studies examining photoreceptor light responses in OIR, few have directly investigated whether rods or cones are more susceptible. Vessey et al.35 reported reduced rod light response amplitudes in OIR mice by subtracting cone from mixed rod–cone waveforms. However, their approach failed to directly isolate cone photoreceptor responses and relied on photopic b-waves as a surrogate. These findings are supported by recently reported deficits in both (scotopic) rod and (photopic) cone amplitudes and implicit times.48 In contrast, Schmitz et al.46 found reduced cone, but not rod, photoreceptor responses with ex vivo ERG in OIR mice at a time point when vascular OIR features have resolved. Reconciling these differing findings requires additional studies, particularly at earlier time points and using ex vivo ERG that isolates photoreceptor from postreceptoral responses. The limited OIR literature on rod and cone light response deficits is also complemented by clinical findings: In patients with a history of ROP, both pathways are affected, although cones appear to be more resilient.52 Within the cone pathway, the central retina may be more susceptible than the peripheral retina,13 potentially reflecting macular abnormalities,53 although (less severe) regional differences have also been reported in prematurely born children without ROP.54,55 Although OIR models in rodents (which lack a macula) cannot directly recapitulate macular pathologies, future studies should explore whether regional differences in photoreceptor vulnerabilities also occur in OIR, for example, in the avascular vs. vascularized retina. Focal or multifocal ERG would provide valuable tools for these investigations, yet neither has yet been applied to examine regional differences in OIR retinas. Photoreceptor light responses deficits in OIR are not universal, however, and upon closer examination several sources of variability emerge. First, the age at testing matters, because photoreceptor function can improve over time.33,36,48 Mouse OIR studies (which are more numerous than those in rats and thus allow a comparison of different time points) more often report reduced photoreceptor light response amplitudes during or shortly after active neovascularization (P17–P19 and P24–P28) than at later time points (P30–P32, P42–P47, and P56–P63) (Fig. 4). This finding suggests improvements through currently unknown mechanisms during retinal maturation. Photoreceptor light response amplitudes in OIR mice across developmental time points compared with room air controls. Findings from published mouse OIR studies are organized by the developmental time point at which photoreceptor light response amplitudes were recorded. ↓ denotes reduced, ↑ increased, and = unchanged a-wave amplitudes in OIR mice compared with room air controls. Superscript numbers indicate citations. DA, dark adapted; LA, light adapted. Second, discrepancies may reflect differences between oxygen exposure paradigms. OIR studies have used multiple oxygen exposure paradigms (Table 1), which differ in the retinal developmental stage, vascular maturity, and mechanisms underlying vascular pathology (i.e., blood vessel regression in the mouse vs. delayed vascular development in the rat 50/10 model). The current literature does not report mechanistic differences in vascular pathology between the standard mouse and the rat 50/10 OIR models, and both paradigms more consistently impair photoreceptor function than alternative oxygen exposure paradigms, which can spare photoreceptor light response amplitudes entirely or allow rapid functional recovery36,42,44,45 (Table 1). Additional factors influence at least the vascular pathology in OIR: strain-specific susceptibility and genetic background,56 and even differences between vendors.57 Whether these factors similarly affect photoreceptor light responses (or other aspects of retinal neuronal deficits) currently remains unclear. The observed variability may represent different manifestations of retinal injury rather than conflicting results. Elucidating how experimental paradigms, developmental stage at injury, and recovery interval influence photoreceptor function may help to identify the mechanisms responsible for retinal dysfunction in OIR and ROP. Mechanisms of Photoreceptor Impairment in OIR The combination of reduced photoreceptor sensitivity and maximal response amplitude suggests that OIR affects both the efficiency of phototransduction (i.e., the conversion of light into an electrical signal) and the maximal photoreceptor output. However, the relative contributions of potentially reversible physiological dysfunction, structural remodeling, and permanent photoreceptor loss remain unclear. OIR provides the opportunity to investigate these mechanisms experimentally, which is difficult in patients due to our inability to perform cellular and molecular analyses. Evidence for widespread photoreceptor loss in OIR remains limited: Outer nuclear layer thickness is preserved in many33,35,39,60 (but not in all61) studies in OIR mice and rats. Similarly, total rhodopsin levels remain unchanged in OIR rats,62 arguing against substantial photoreceptor loss as the primary driver of light response deficits. Instead, structural abnormalities of photoreceptors appear to be likely. In OIR mice and rats, photoreceptor outer segments can be shortened and disorganized,35 particularly during retinal hypoxia (after the return to room air).62 These findings are consistent with altered photoreceptor ultrastructure, although their extent, persistence, and physiological relevance, particularly in cones, remain incompletely characterized. Recent work has also begun to unravel the molecular mechanisms that impair phototransduction in OIR. Mouse OIR photoreceptors express less guanylate cyclase activator 1A and, consequently, contain less cyclic guanosine monophosphate.48 Reduced cyclic guanosine monophosphate would decrease cyclic nucleotide–gated channel opening, lowering the photoreceptor dark current and reducing baseline photoreceptor excitability. This mechanism provides a plausible explanation for reduced a-wave amplitudes across stimulus intensities and reduced maximum response amplitudes observed in many OIR studies. Additional effects on other components of the activation, inactivation, and recovery of the phototransduction cascade remain possible, but are not yet defined. Metabolic changes represent another potential mechanism of photoreceptor impairment in OIR. Mitochondrial dysfunction has been demonstrated in OIR retinas,63 albeit without distinguishing different retinal cell types. Since photoreceptors have an exceptionally high metabolic demand and limited mitochondrial reserve capacity,64 even modest mitochondrial dysfunction could plausibly compromise phototransduction. Given that photoreceptors constitute a large proportion of retinal cells and account for a substantial fraction of retinal energy consumption, whole-retina metabolic deficits may (at least in part) reflect impaired photoreceptor metabolism. Oxidative stress also contributes to OIR,65 arising from multiple sources, including mitochondria and nicotinamide adenine dinucleotide phosphate oxidase.66,67 Oxidative stress can damage photoreceptor membranes and has been implicated in models of photoreceptor degenerative diseases.68 Although antioxidant strategies can preserve photoreceptors and their function in retinal degenerative diseases,69 a direct causal link between ROS modulation and photoreceptor light responses in OIR has yet to be established, especially in the likely absence of extensive photoreceptor death in OIR. Finally, reduced oxygen and nutrient delivery from the choroid may contribute to photoreceptor light response deficits in OIR. Photoreceptors are primarily supported by the choriocapillaris,70 and changes in choroidal perfusion or structure selectively impair photoreceptor light responses.71 (Inner retinal neurons are largely supplied by the inner retinal circulation72 and thus their function is not affected by choroidal thinning.71) Whether the suggested thinning of the choroid73 significantly contributes to photoreceptor light response deficits currently remains uncertain and warrants further investigation. Overall, multiple mechanisms may contribute to photoreceptor light response deficits in OIR, with the best documented evidence for structural changes to photoreceptors and impaired phototransduction. Studying the relative contributions and interactions of potential mechanisms of damage would dramatically improve our understanding of photoreceptor deficits in OIR. ON-BPC Light Responses in OIR ON-BPCs exhibit more severe and persistent functional deficits after OIR compared with photoreceptors. Nearly all studies in OIR mice and rats report impaired postreceptoral response parameters: reduced b-wave amplitudes, maximal postreceptoral amplitude (Vmax), and postreceptoral sensitivity (σ).33–38,40,48,58 Importantly, these deficits occur across different oxygen exposure paradigms and developmental time points, including those that leave a-wave amplitudes relatively preserved33,36,44 (Table 1). Consistent with ON-BPC dysfunction, Mowat et al.40 reported a reduced b-/a-wave ratio, a metric that compares postreceptoral and photoreceptor function, but has been reported surprisingly rarely. Collectively, the available evidence suggests that ON-BPC dysfunction likely represents a primary and persistent feature of OIR. Few studies have separated rod- and cone-mediated ON-BPC pathways in OIR. Vessey et al.35 reported reduced amplitudes in both pathways, whereas delayed implicit times were confined to the cone pathway of OIR mice. Greater improvements of the scotopic compared with the photopic b-wave also suggest that damage to cone ON-BPC cells may persist for longer.48 Overall, however, pathway-specific vulnerabilities of ON-BPCs remain poorly defined. Clinical studies similarly support ON-BPC involvement: Patients with a history of ROP show abnormal postreceptoral light response parameters.8,9,12,74 Notably, although OIR rodents frequently exhibit reduced Vmax and σ, clinical studies more commonly report reduced σ with a relatively normal Vmax. Whether this discrepancy reflects substantially longer recovery periods in human subjects, species-specific differences, or methodological differences between clinical and experimental studies currently remains unclear. Mechanisms of ON-BPC Impairment in OIR The combined reduction in Vmax and σ indicates that OIR impairs both the maximal output of postreceptoral signaling and sensitivity. However, the underlying mechanisms, including the relative contributions of ON-BPC loss, impaired synaptic connectivity, and intrinsic ON-BPC dysfunction have yet to be determined. Several OIR studies report cell death in the inner nuclear layer (where ON-BPCs reside),58,75 as well as inner nuclear layer thinning that can be patchy or eccentricity-dependent.39,40,60,76 Although these data are consistent with reduced ON-BPCs in OIR, the cell number has not yet been quantified. However, the preserved Vmax reported in many clinical ROP studies argues against widespread ON-BPC loss as the sole explanation for postreceptoral dysfunction. Whether the discrepancy between animal models and human subjects reflects differences between species, longer recovery times in patients, regional rather than global ON-BPC loss, or compensatory mechanisms that preserve maximal ON-BPC responses despite reduced cell numbers remains unknown. Impaired synaptic connectivity is widely supported: Multiple studies demonstrate thinning and disruption of the outer and inner plexiform layers (i.e., the synaptic strata between photoreceptors and BPCs and between BPCs and RGCs, respectively) in OIR mice and rats.33,35,40,60,76 These structural changes are consistent with synaptic retraction,75 which could reduce the efficiency of signal transmission. Reduced gain of photoreceptor–to–ON-BPC signaling may arise from several non-exclusive mechanisms, which currently remain speculative: (1) alterations at the photoreceptor–to–ON-BPC synapse, including reduced presynaptic glutamate release from the photoreceptor ribbon, dendritic retraction of ON-BPCs, or misalignment of presynaptic and postsynaptic components; (2) dysfunction of key proteins in the ON-BPC signal transduction cascade, including mGluR6, TRPM1, Gαo, and nyctalopin; (3) a decreased number of rods converging onto each rod ON-BPC due to loss or remodeling of rod terminals; and (4) broader retinal circuit remodeling involving, for example, amacrine cell connectivity and gap junction coupling. At present, direct evidence for these mechanisms remains elusive. Taken together, current evidence supports that reduced ON-BPC light responses in OIR may reflect ON-BPC loss, impaired synaptic connectivity, and reduced signaling gain. However, direct measurements of synaptic transmission and ON-BPC signaling remain outstanding. Third-Order Neuronal Damage and Cell Loss in OIR Not all OIR studies report OPs; those that do consistently demonstrate markedly reduced OPs in both mouse and rat OIR models33–35,37,40,42,43,46 (Table 2). Because OPs primarily reflect activity within inner retinal circuits and are thought to depend heavily on amacrine cell function, these findings support substantial inner retinal dysfunction. Loss of specific amacrine cell populations—particularly of dopaminergic, glycinergic, and AII amacrine cells77,78—likely contribute to these deficits. Notably, surviving amacrine cells can retain their normal dendritic fields and exhibit spontaneous and light-evoked electrical activity.77 Given the central role of amacrine cells in shaping temporal processing, contrast sensitivity, edge detection, and motion detection, as well as other aspects of visual signaling, even selective disruption of specific amacrine cell populations could have profound consequences for inner retinal function. Table 2.OPs in OIRAgeAmplitudeITCommentsMouse OIR model (75% oxygen, P7–P12) Martini et al., J Neurochem37P17Summed OP2-4 ↓*,†Not reported– Cammalleri et al., Front Cell Neurosci (2017)38P17Summed OP2-4 ↓‡Not reported– Vessey et al., J Comp Neurol (2011)35P18OP2= OP3-4 ↓‡Not reported– Mowat et al., AJP (2012)40P26Summed OP2-5 ↓†Not reported–P60Summed OP2-5 ↓† Fu et al., Graefes Arch Clin Exp Ophthalmol (2015)34P30Summed OP2-4 ↓†=– Nakamura et al., PLoS (2012)334 weeksOP2-OP5 ↓*Not reported–6 weeksOP2-OP5 ↓‡,*8 weeksOP3-OP5 ↓‡ Schmitz et al., Cells (2025)468-9 weeks post OIROPs↓Not reportedNo stats or averagesRat OIR model (50/10, P0–P14) Zhang et al., 201343P55Saturating amplitude ↓‡Not reportedSensitivity ↓ (P < 0.05)Other rat OIR models– Reynaud et al., 199542P18OP2, OP4 =OP3 ↓*==OPs in P13 control rats not reliableIT, implicit time.* P < 0.01.† P < 0.001.‡ P < 0.05. OPs in OIR IT, implicit time. P < 0.01. P < 0.001. P < 0.05. Similarly, light-evoked RGC deficits46 have been demonstrated in the mouse OIR model. These abnormalities may arise from reduced synaptic input from upstream retinal neurons, RGC loss, and intrinsic alterations within surviving RGCs. Evidence for RGC loss is relatively strong: Multiple studies have reported thinning of the nerve fiber and ganglion cell layers40,76 (findings that are mirrored in clinical studies)79; reduced RGC density, particularly in the central retina80 (including loss of photosensitive ipRGCs)46,81; and increased markers of cell death within the ganglion cell layer,58 although these findings are not universal.33 Surviving RGCs also appear to undergo remodeling of their dendritic arborization,80 suggesting that OIR causes a combination of RGC loss and altered retinal circuitry remodeling. Important questions remain, including the vulnerability of specific RGC subtypes, changes in their receptive fields, and the impact of OIR on spontaneous and light-evoked RGC activity. Hyperoxia vs. Hypoxia The mechanisms driving light response deficits in OIR remain poorly understood. Surprisingly, this lack includes the relative contributions of hyperoxia and hypoxia. The mouse OIR model offers a clear advantage in that it clearly separates hyperoxia (75% O2 from P7 to P12) and retinal hypoxia (in the avascular area upon return to room air).18,19 However, hypoxia cannot be studied in isolation in this or other OIR models, because it arises only after hyperoxia or O2 fluctuations have disrupted normal vascular development. Most mouse OIR studies examine vascular outcomes at the end of hyperoxia and at the peak of retinal hypoxia, but they typically assess light responses later—either at the peak of hypoxia or after the vasculature has recovered. Technical constraints have largely driven this delay because in vivo ERG requires eye opening (beginning around P11–P12 in mice),82 which can be delayed by OIR. As a result, early neuronal effects of hyperoxia remain largely unexplored. Importantly, hyperoxia and hypoxia may differently affect retinal neurons: For example, dopaminergic amacrine cells in OIR mice are preserved during hyperoxia but selectively lost during hypoxia.77 Ex vivo ERG circumvents the requirement for eye opening by recording light responses from isolated retinas; nevertheless, photoreceptor and ON-BPC light responses immediately after hyperoxia in mouse OIR have not yet been systematically studied with this approach. This area represents a critical gap in our understanding of how early oxygen exposure impacts retinal light responses. Systemic Influences in OIR Several factors affect the vascular features of OIR, such as genetic background, nutritional deficits, and poor postnatal weight gain.56,83,84 To reduce variability, experiments are commonly performed in C57BL6/J mice or Sprague–Dawley rats, litter sizes are standardized, and pup weights are monitored. Whether these factors similarly modify retinal neuronal function has not been systematically investigated, but limited information suggests they may: Synaptic connectivity in the outer and inner plexiform layers—between photoreceptors and ON-BPCs and between ON-BPCs and RGCs, respectively—is more disrupted in Long Evans compared with Sprague–Dawley rats.75 Poor postnatal weight gain in OIR mice has also been associated with deficits in specific aspects of light responses—a reduced saturating amplitude of the BPC response and decreased 8-Hz flicker responses at P70.84 Nevertheless, the temporal progression of these functional changes remains largely unexplored. Growth factors associated with ROP risk and poor postnatal weight gain, like insulin-like growth factor (IGF-1) and erythropoietin (Epo), may provide a mechanistic link between systemic development and retinal light response deficits. IGF-1 levels sharply decline after preterm birth85 and are lower in preterm infants than in healthy in utero fetuses.86 Although intrauterine growth restriction increases IGF-1 in rat pups, OIR outweighs this effect.87 Retinal neurons express the IGF-1 receptor, which supports cell survival and enhances light response amplitudes.88 Taken together, it is conceivable that reduced IGF-1 in premature infants may compromise retinal neuronal survival and function and thus link premature birth to persistent visual deficits in ROP. Premature infants also produce less endogenous Epo than fetuses of the same gestational age. In OIR mice, Epo supplementation temporarily improves photoreceptor and ON-BPC light responses, but these improvements are lost if Epo levels are not maintained.40 It is possible that Epo administered to extremely and very premature children may improve light responses years later,89 when endogenous Epo production has presumably recovered. However, because the clinical use of Epo, as well as the patient population, remain complex, evidence regarding increased ROP risk and severity remains mixed.90–92 Treatments Targeting VEGF VEGF represents another potential link between vascular pathology and neuronal dysfunction. VEGF drives pathological neovascularization in OIR and, as an important neurotrophic factor, supports the survival of photoreceptors,93 RGCs,94 and cells in the inner nuclear layer95—likely BPCs and/or Müller glia. This dual role raises the important clinical question: Could anti-VEGF treatments in infants with ROP inadvertently disrupt retinal neuronal development and maturation? Indeed, broad VEGF inhibition potentially supports this concern. High doses of the VEGF trap aflibercept in OIR mice cause retinal thinning and disrupt the outer plexiform layer, leading to reduced ON-BPC light responses.96 Similar structural and light response deficits in adult diabetic mice treated with anti-VEGF drugs suggest that these effects are not limited to the developing retina.97 However, clinical comparisons found no significant difference between anti-VEGF treatment and photocoagulation.98 More targeted approaches—such as selectively reducing pathological VEGF production in Müller glia99—may mitigate risks and preserve retinal light responses in OIR rats.100 Together, these findings highlight the need to balance the antiangiogenic benefits of anti-VEGF treatments with the preservation of retinal structure and light responses in the management of ROP. Translational Implications and Limitations of OIR Models Studies in patients with ROP remain challenging due to methodological constraints when working in vivo and the lack of available postmortem tissue—these factors limit structural, biochemical, and functional analyses. As described elsewhere in this article, OIR parallels photoreceptor and ON-BPC light response deficits in patients with ROP and supports the concept that ROP is not solely a retinal vascular, but also a neuronal disease. Although retinal neuronal damage is less well understood in patients, OIR models provide a valuable platform to study mechanisms and bridge our knowledge gaps. Despite the insights provided by rodent OIR models, several important questions remain unresolved: What causes persistent photoreceptor and ON-BPC dysfunction after vascular recovery? Does impaired neurovascular coupling contribute to lasting visual deficits? Can neuroprotective interventions improve retinal function independent of vascular outcomes? OIR studies hint at potential approaches that could complement current neovascular therapies: Preserving or restoring photoreceptor health, synaptic integrity, nutrient and oxygen delivery, cellular metabolism and mitochondrial function, and oxidative balance could conceivably improve long-term retinal light responses independent of vascular outcomes. Given that ROP and OIR involve both vascular and neuronal pathology, it is tempting to speculate whether restoring neurovascular coupling could simultaneously improve vascular recovery and retinal light responsiveness. Despite its possibilities, OIR incompletely recapitulates several important aspects of human ROP. Most notably, although retinal development remains incomplete in mouse and rat pups, these animals are born at term rather than prematurely. In addition, OIR oxygen exposure paradigms are simplified compared with the highly variable oxygen environments experienced by preterm infants in neonatal intensive care units. OIR models attempt to incorporate certain key risk factors associated with ROP. For example, nutritional deficits can be induced by using large litter sizes, and models of intrauterine growth restriction have been combined with both the mouse101 and rat 50/10 OIR models.87 However, OIR lacks many systemic contributors to ROP pathogenesis, including infection, pulmonary immaturity, and other developmental comorbidities of premature infants. Structural differences between rodent and human retinas further limit direct translation, because rodents lack a fovea and macula and possess a rod-dominant retina. Thus, although OIR remains highly valuable for mechanistic studies of neurovascular injury and retinal dysfunction, caution is warranted when extrapolating therapeutic efficacy directly to the human disease. Given the limitations of mechanistic studies in infants with ROP, future OIR studies should incorporate additional, clinically relevant risk factors and comorbidities to more accurately model the complexity of human ROP. Conclusions Retinal light response deficits can persist in patients with ROP long after vascular abnormalities have resolved. Several additional factors can contribute to impaired vision, including nystagmus, amblyopia, myopia, and strabismus. However, retinal light response deficits may importantly contribute to long-term vision impairment based on their consistent presence and association with increasing ROP severity. Current ROP therapies effectively target retinal neovascularization and prevent retinal detachment; they do not address retinal light response deficits—largely because the underlying mechanisms remain poorly understood. As a valuable experimental tool, OIR will accelerate our understanding of retinal light response deficits and guide the development of future therapies in ROP. As outlined in this article, OIR reproduces retinal light response deficits throughout the retinal circuit, beginning in photoreceptors and cumulatively propagating through downstream synaptic connections and inner retinal neurons. The contributions of hyperoxia and hypoxia to retinal light response deficits are among the most important knowledge gaps. Most previous studies could not address this question by relying on in vivo ERG. Future studies with ex vivo ERG in the mouse OIR model will be well positioned to distinguish between these phases. Because ERG records the summed light responses in the entire retina, the function of individual cells and their synaptic connectivity remains impossible to resolve with this technique alone. Addressing this question will require single-cell recordings, including suction electrode, patch clamp, and multielectrode array recordings, combined with intracellular calcium imaging or measurements of neurotransmitter release. Additional important questions that these techniques will address include the relative vulnerability of rod and cone photoreceptors and their downstream pathways. OIR as a tool to study retinal light response deficits in ROP will undoubtedly help to guide the development of therapies to improve vision outcomes for increasingly premature infants who survive into adulthood. Crucially, future strategies should target both vascular and neuronal components of the disease to fully realize the potential for preserving or restoring vision in ROP.