We thank Dr. Weinreb and the IGR editorial board for selecting our work for discussion. We are grateful to all of the reviewers, each a respected clinician scientist, for their time and thoughtful comments. This works extends literature on mitochondria in glaucoma by showing that mitochondrial dysfunction is an early abnormality within retinal ganglion cells in vivo in an inherited, ocular hypertensive glaucoma model. Furthermore, this work shows that levels of NAD or its precursor nicotinamide impact vulnerability to glaucoma. The growing literature linking mitochondrial susceptibility to glaucoma significantly raises excitement for the possibility that our findings will have importance for human glaucoma. Neuroprotective strategies against glaucoma are of great therapeutic need. However, we fully accept that, despite representing a large body of work, our paper is an early step with limitations. We agree that further animal studies and ultimately human trials are needed to assess the safety and efficacy of nicotinamide in glaucoma, and to allow firm conclusions about the general relevance of our findings to human glaucoma. We hope that our paper will serve as a stimulus for such studies.
To elucidate both the intraocular pressure- (IOP) and age- dependent mechanisms that drive retinal ganglion cell vulnerability in glaucoma, we utilized the DBA/2J (D2) mouse model of glaucoma. Our views on the appropriateness of this model for our studies and on some other comments respectfully differ to those from Dr. Quigley and are discussed in further detail below. We had attempted to give adequate methodology for all of the approaches in the paper with supporting references providing more details. The D2 mouse is arguably one of the most well characterized models of glaucoma. Although the greater damage in the anterior segment of D2 eyes (including the mild subclinical inflammation) must be considered and is important in the aetiology of IOP elevation, we are not aware of any robust reported evidence indicating that the mechanisms of neural degeneration are very different in D2 mice compared to other glaucomas. It remains possible that common pathways of neurodegeneration are induced by harmfully high IOP with the relative importance of specific processes varying between individual patients.
Importantly, the D2 model has various other features that are similar to human glaucoma including the variability of the IOP and neural phenotypes1, 2, the topographic pattern of retinal ganglion cell loss3, and the location of a critical insult to retinal ganglion cell axons in the optic nerve head.3 Various pathways and molecules change in both D2 mice and in human glaucoma. Changes in the complement pathway, endothelin pathway, and mitochondrial dysfunction are all reported in human patients and in the D2 model.1,4-11 The progressive variable nature of the IOP phenotype in D2 mice more closely resembles chronic human glaucomas than the suddenly induced IOP elevation in experimentally induced models. As IOP profiles and the timing between harmful IOP and retinal ganglion cell degeneration are also variable in the human disease, traumatically insulting the eye with lasers or microbeads is not necessarily better than the natural D2 disease. However, we fully agree that all of these models are complementary and that studies in a variety of models are needed. Induced models have potential to provide valuable experimental control for further studies, though in our hands we have found substantial variability in both the induction of IOP following the administered insult and the neurodegeneration in these models too.
Regarding elevated IOP, D2 mice are responsive to IOP lowering treatments that lessen the neurodegeneration (without changing the iris disease). Thus, the neural dysfunction and neurodegeneration in D2 eyes appears IOP-dependent.12-16 Importantly, the genes conferring the anterior segment disease have been transferred to another mouse strain. The resultant mice develop an anterior segment disease that is indistinguishable in timing and severity to that of DBA/2J mice, but they do not develop high IOP or glaucoma. This demonstrates that the anterior segment disease does not cause glaucoma by itself 17, 18. As mice of this second background develop glaucoma when IOP is elevated by other means (inherited or experimentally induced 19, 20), these collective data strongly argue for a dominant role of high IOP in D2 glaucoma.
Dr Quigley states that we did not test age as a variable. We used RNA-sequencing of RNA from retinal ganglion cells to elucidate early mechanisms that underlie retinal ganglion cell vulnerability in glaucoma. We used samples from D2 and no glaucoma control mice (D2-Gpnmb+) at two ages to elucidate both age- and IOP-dependant transcriptomic changes. Age was a key variable in this genomic study. Guided by our RNA-sequencing analysis we discovered an age-dependant decline in NAD(t) (NAD total; NAD+ and NADH). We used nicotinamide (an NAD precursor) treatment to restore NAD(t) levels and prevent glaucoma in appropriately aged animals. Decreased NAD levels are increasingly becoming implicated in a broad array of aging deficits and age-related diseases.21-24 Supported by this rapidly growing literature on NAD in aging and metabolic health, the decreased NAD levels in aged mice were hypothesized to increase vulnerability to mitochondrial dysfunction and glaucoma following periods of elevated IOP. We agree that we did not assess young and old animals with the same degree of IOP exposure. This is a complementary approach that controls for IOP exposure but does not invalidate our approach. Additionally, we agree that we did not test the efficacy of nicotinamide treatment in both young and old animals, in part because the decline of NAD(t) has not occurred in the young mice. This would be an approach to assessing the efficacy at both young and older ages (a different issue to the one that we addressed). Instead, we set out to therapeutically target the age-related decrease of NAD(t) levels in animals at ages when the glaucoma develops and this robustly protected these older mice. Due to its multiple effects, NAM may well protect at younger ages too, but such a result would not indicate that NAD(t) changes are unimportant for the age-related disease. Many groups studying aging use approaches that are similar to ours, and we prefer to use a natural disease model. In the future, it will be valuable to apply similar transcriptomic approaches to panels of improved, human-relevant, mouse models having human mutations in various human glaucoma genes (humanized models of glaucoma).
Regarding Dr. Quigley's methodological comments in relation to our use of the term disease progression (first comment). In this context, we had defined progression as based on progressive transcriptomic changes. The goal of these experiments was to identify gene expression changes that occur during disease initiation and very early molecular progression. This required the study of eyes that are not yet distinguishable by conventional optic nerve or retinal ganglion cell analyses, as such early molecular changes occur prior to neurodegeneration. To accomplish this, we used unsupervised hierarchical clustering (to identify molecularly defined groups based on transcript expression). The study design required closely age-matched mice and used a large number of samples that were individually analyzed by RNA-sequencing (initially 72 samples, plus an additional 10 samples from nicotinamide treated mice). This strategy identified 4 molecularly distinct groups or stages of early progression. These groups were ordered in terms of their increasing molecular distance from the age-, sex- and genetically-matched no glaucoma controls that were housed in the same environment. With this increasing distance from controls, the number of differentially expressed genes increased. Over the past decade, similar hierarchical clustering designs have transformed understanding of molecular processes contributing to a variety of biological processes and diseases. For example, they have provided key information about new molecular subtypes of cancer.25,26 In the near future, such powerful genomic approaches will likely be adopted to improve clinical trials by allowing molecular stratification of disease types as well as stratification by molecular responses to treatments. There is ample evidence that such approaches identify groups with biologically meaningful molecular differences. We have previously used this hierarchical clustering approach to characterize and order early, pre-degenerative stages of glaucoma.4,27 In the first of these studies, we included no glaucoma controls as well as mice at different stages of glaucoma (including pre-degenerative stages and stages with moderate or severe degeneration as determined by the degree of retinal ganglion cell axon and soma loss). Importantly, the unsupervised computational process sorted all of these samples into meaningful groups as evidenced by the fact that previously known groups consisting of no glaucoma controls, moderate or severe glaucoma were successfully recreated using the unsupervised process.4 This unsupervised but meaningful grouping of samples is clearly evident in our subsequent studies.11,27 The recreation of biologically meaningful groups by the unsupervised hierarchical clustering, along with data that targeting the molecular processes identified through this grouping strongly protect from glaucoma4,6,10,11, validates our approach for grouping samples as a valuable way to provide new mechanistic knowledge about glaucoma.
To target declining NAD(t) we chose the precursor nicotinamide (the amide of vitamin B3). Nicotinamide effectively raises NAD(t) levels in other systems, and was potent at raising NAD(t) levels in D2 retinas (by ~3-fold). For these experiments, we used 2 doses of nicotinamide, 550 mg/kg/d and 2000 mg/kg/d. In response to Dr. Pasquale's comments, an equal number of male and female mice were used for the drug studies, with sexes balanced in both treatments and control groups. We have not assessed lower doses. The human dose equivalents of our doses are 2.7 g/d and 9.8 g/d respectively for a 60 kg human.28 Doses of 3-9 g/d of nicotinamide or nicotinic acid have been used long-term in humans for other conditions with few adverse effects (in some cases for up to 5 years). Doses of 3 g/d and below are generally considered to be safe. There are some cases of individual susceptibility to hepatotoxicity (reported on higher doses e.g. 9 g/d) but they are reversible and rare. Overall nicotinamide has a good risk to benefit ratio and has fewer unpleasant side effects than nicotinic acid. (For a review on nicotinamide safety see29) To follow up further on Dr. Pasquale's points, macular edema is a rare complication of nicotinic acid treatment (0.67% patients treated for hyperlipidemia) and is reversed by stopping treatment.30 In the literature, the term niacin initially referred to nicotinic acid but sometimes refers to a mix of nicotinic acid and nicotinamide. It is unknown whether nicotinamide itself causes macular edema. Given the differing biochemical properties of these molecules including differences in their side effects, we propose the further testing of nicotinamide in glaucoma.
Nicotinamide treatment was profoundly protective in D2 mice. It also protected against TNF?-induced retinal ganglion death in vivo as well as against retinal ganglion cell death in an axotomy culture system. Regarding Dr. Quigley's comments on our optic nerve and retinal ganglion cell analyses in D2 mice, axon counts have been performed confirming the protection in treated mice. These data were not shown in the original Science paper but have now been included in a subsequent manuscript that is under consideration. We included data for our morphological determination of the degree of optic nerve damage using PPD stained nerves in the Science paper (Figure 2B). We included this data, as a large number of eyes were assessed using this approach, and as this technique has been validated against axon counting in various publications with high reproducibility between masked investigators, e.g.3,4 PPD stains the myelin sheathes of axons darkly but more lightly stains other nerve components in healthy nerves. Its great value and sensitivity is in detecting individual stressed or damaged axons among a sea of healthy axons. It allows this by differentially staining the axoplasm of such stressed or damaged axons more darkly than that of healthy axons. It detects stress/damage in axons that otherwise appear healthy and have an intact myelin sheath. As the severity of injury to an axon increases it becomes very darkly stained. Since axon number is variable in mouse nerves, and since PPD allows sensitive detection of damaged axons, we argue that this is a very sensitive approach for detecting disease especially at early stages of disease. Importantly, our method not only considers the number of healthy axons (light staining axoplasm) and damaged axons (darkly staining axoplasm) but also the cross sectional area of the nerve and the degree of gliosis. In the discussed study, >50% of the untreated control mice had very obvious and severe damage with massive axon loss and profound gliosis while this phenotype was only present in 3% of mice treated with the higher dose of nicotinamide. Dr. Quigley discusses numbers of eyes that need to be compared and that we only counted retinal ganglion cells for 8 eyes in each experimental group (shown in Figure 2C). We only counted retinal ganglion cells in 8 eyes per group, as this was only to confirm that the protection of retinal ganglion cell axons and retinal ganglion cell bodies were not uncoupled as distinct molecular processes participate in the degeneration of these neuronal compartments31,32 (i.e. that cell bodies were saved in eyes with protected axons but lost in eyes with axon loss). Our optic nerve analysis (Figure 2B) shows the robust nature of the nicotinamide protection. Figure 2B contained data for >50 eyes per group. The sample size is listed in the legend for Figure 2B, panel F as the nerve images are shown in F. The same data is plotted as a histogram in Figure 2B. We apologize that this may not have been adequately clear. In total, Figure 2B shows data for 191 nicotinamide treated nerves and 136 untreated controls. Adding confidence of the importance of NAD increasing treatments for glaucoma, the protection has proven robust for >620 treated DBA/2J eyes that we have now published. (Breakdown: 274, nicotinamide treatment alone; 349, Wlds gene, Nmnat1 gene therapy, or combination of gene therapy and nicotinamide (see below); 315, untreated controls3,7,10,11). Regarding axon transport, we used cholera toxin B with a fluorescent tag as a marker of axoplasmic transport. Although not empirically quantified, labeling was absent from all brains with severe optic nerve disease while there was qualitatively normal cholera toxin B labeling in the superior colliculus and lateral geniculate nucleus of protected mice with no optic nerve disease. Combined with pattern electroretinography analysis this demonstrates that in nicotinamide treated mice retinal ganglion cells and their axons are both histologically and functionally protected from glaucoma.
We next tested a gene therapy, over-expressing Nmnat1 (a terminal enzyme in NAD production). NMNATs has been shown to be protective in other axon degeneration systems and in induced glaucomas.21,33-35 Gene therapy of Nmnat1 was protective against glaucoma in D2 eyes and showed additional protection when combined with nicotinamide administration (at the lower dose of 550 mg/kg/d). We agree with Dr. Martin's comment on a somal/ nuclear versus axonal site of action for this treatment. In this gene therapy experiment, the entire gene of interest (Nmnat1) was overexpressed. The nuclear localization signal (NLS) was intact, and so the protein product (NMNAT1) should be localized predominantly at the nucleus within the soma. Poly ADP ribose polymerase (PARP) increases in retinal ganglion cells during pre-degenerative stages of D2 glaucoma. Since PARP depletes nuclear NAD and the NMNAT1 NLS was intact, it seems most likely that NAD-mediated protection is mediated within the nucleus of the cell body; however, a role within axons cannot be ruled out.
Dr. Welsbie raises an important point regarding potential independent effects of nicotinamide and NMNAT1. Given the roles of NAD in aging and metabolism, increasing NAD(t) levels are likely to be a key component of the protection. In this context NMNATs are likely to be the main effector. We have previously demonstrated that WLDS can protect D2 eyes from glaucoma.3 In a recent publication, we demonstrated that WLDS increases NAD(t) levels in the retina, and that the combination of WLDS plus nicotinamide offers near-complete protection from glaucoma (94% eyes have no detectable glaucoma).10 This further supports a key role for increased NAD(t) levels, but other NAD independent effects of WLDS cannot be ruled out as at least partially contributing to the protection. Dr. Welsbie also points out that NMNAT1 has an inhibitory effect on SARM - an important NAD degrading molecule during axon degeneration. In addition to being an NAD precursor, nicotinamide also has other importance molecular properties that might contribute to its protective effect in glaucoma including inhibition of the NAD consumers CD3836, PARPs37, and histone deacetylases23,38, as well as ADP-ribosyl cylase39-41 which impacts vascular tone and calcium signalling. This combination of functions may explain why nicotinamide is so protective against D2 glaucoma.
To conclude, we thank the commentators for their thoughtful discussion and insightful comments. We look forward to working with the glaucoma research community to continue exploring the role of mitochondrial biology, NAD, and NAD-based treatments in glaucoma.